Manual of Aquatic Viral Ecology

Manual of Aquatic Viral Ecology ASLO's first eBook publication is the Manual of Aquatic Viral Ecology (MAVE), edited by Steven

Wilhelm, Markus Weinbauer and Curtis Suttle. It contains 19 chapters reflecting state-of-the-art opinions on approaches to studying viruses in aquatic systems. Topics range from the enumeration of viruses to molecular techniques designed to dissect and query individual virus populations as well as communities of viruses. The content of this e-book was selected in consultation with the Scientific Committee for Oceanographic Research’s working group on marine viruses, and its publication has

been supported by the Gordon and Betty Moore Foundation.

S.W. Wilhelm, M.G. Weinbauer, and C.A. Suttle [eds.] 2010. Manual of Aquatic Viral Ecology. Waco, TX: American Society of Limnology and Oceanography. doi:10.4319/mave.2010.978-0-9845591-0-7

Table of Contents

Markus G. Weinbauer, Janet M. Rowe, and Steven W. Wilhelm Determining rates of virus production in aquatic systems by the virus reduction approach Chapter 1, pp 1-8

Ruth-Anne Sandaa, Steven M. Short, and Declan C. Schroeder Fingerprinting aquatic virus communities Chapter 2, pp 9-18

André M. Comeau and Rachel T. Noble Preparation and application of fluorescently labeled virus particles Chapter 3, pp 19-29

John H. Paul and Markus Weinbauer Detection of lysogeny in marine environments Chapter 4, pp 30-33

Michael J. Allen, Bela Tiwari, Matthias E. Futschik, and Debbie Lindell Construction of microarrays and their application to virus analysis Chapter 5, pp 34-56

Kenneth M. Stedman, Kate Porter, and Mike L. Dyall-Smith The isolation of viruses infecting Archaea Chapter 6, pp 57-64

Susan A. Kimmance and Corina P. D. Brussaard Estimation of viral-induced phytoplankton mortality using the modified dilution method Chapter 7, pp 65-73

Roberto Danovaro and Mathias Middelboe Separation of free virus particles from sediments in aquatic systems Chapter 8, pp 74-81

Steven M. Short, Feng Chen, and Steven W. Wilhelm The construction and analysis of marker gene libraries Chapter 9, pp 82-91

Keizo Nagasaki and Gunnar Bratbak Isolation of viruses infecting photosynthetic and nonphotosynthetic protists Chapter 10, pp 92-101

Corina P.D. Brussaard, Jérôme P. Payet, Christian Winter, and Markus G. Weinbauer Quantification of aquatic viruses by flow cytometry Chapter 11, pp 102-109

K. Eric Wommack, Télesphore Sime-Ngando, Danielle M. Winget, Sanchita Jamindar, and Rebekah R. Helton Filtration-based methods for the collection of viral concentrates from large water samples Chapter 12, pp 110-117

Mathias Middelboe, Amy M. Chan, and Sif K. Bertelsen Isolation and life cycle characterization of lytic viruses infecting heterotrophic bacteria and cyanobacteria Chapter 13, pp 118-133

William H. Wilson and Declan Schroeder Sequencing and characterization of virus genomes Chapter 14, pp 134-144

Curtis A. Suttle and Jed A. Fuhrman Enumeration of virus particles in aquatic or sediment samples by epifluorescence microscopy Chapter 15, pp 145-153

Grieg F. Steward and Alexander I. Culley Extraction and purification of nucleic acids from viruses Chapter 16, pp 154-165

Janice E. Lawrence and Grieg F. Steward Purification of viruses by centrifugation Chapter 17, pp 166-181

Hans-W. Ackermann and Mikal Heldal Basic electron microscopy of aquatic viruses Chapter 18, pp 182-192

Alexander I. Culley, Curtis A. Suttle, and Grieg F. Steward Characterization of the diversity of marine RNA viruses Chapter 19, pp 193-201

1

Introduction

Since the rediscovery of the importance of viruses in marineenvironments (Bergh et al. 1989; Proctor and Fuhrman 1990;Suttle et al. 1990), researchers have worked to try and determinethe quantitative nature of virus effects on marine microbialfood webs. Originally documented in aquatic systems almost100 years ago (Duckworth 1976) the implications of virus activ-ity have remained elusive. Beginning in the early 1990s how-ever, efforts on several fronts began to quantify the rate at

which virus particles were produced and “turned-over” in pri-marily pelagic aquatic environments. Whereas many of thesemethods have not been set aside, an appreciation of the differ-ent options available to the aquatic viral ecologist is necessary.

Prior to understanding the methods that are available toestimate virus production rates in aquatic systems, it is per-haps best to understand how the information is importantand will be used (as the intended fate of the information may,in part, dictate the manner of its collection). Virus productionrates are most commonly used to infer the losses of primary orsecondary production in aquatic systems due to the activity ofviruses. In the case of direct estimates of particle productionrate, knowledge concerning the number of viruses producedper lytic event (the burst size) allows for one to estimate thenumber of host cells destroyed by the activity of viruses. Assuch, estimates of virus activity need to be made over timesscales that are on the same temporal order as the turnover rateof the host population.

TEM assessments of microbial mortality—One of the earliestattempts to estimate the mortality inferred on microbial com-munities was the percentage of visibly infected cells approach(Proctor et al. 1993). The approach is based on the assumptionthat intact virus particles are visible in infected cells for a cer-tain percentage of the lytic cycle. By undertaking controlledinfections within a lab setting, Proctor and colleagues (1993)were able to estimate the percentage of the lytic cycle thatviruses were visible within infected cells. By extrapolating thisrelationship to microbial communities, estimates of the per-centage of microbial cells that carried a visible virus infectioncould be made using a transmission electron microscope set to

Determining rates of virus production in aquatic systems by thevirus reduction approachMarkus G. Weinbauer1*, Janet M. Rowe2, and Steven W. Wilhelm2*1CNRS UMR 7093, Laboratoire d’Océanographie de Villefranche, 06234 Villefranche-sur-Mer Cedex, France; Université Pierre etMarie Curie, Paris 6, Laboratoire d’Océanographie de Villefranche, 06230 Villefranche-sur-Mer, France2Department of Microbiology, The University of Tennessee, Knoxville, TN 37922, USA

AbstractThe reduction approach to assess virus production and the prokaryotic mortality by viral lysis stops new

infection by reducing total virus abundance (and thus virus–host contacts). This allows for easy enumeration ofviruses that originate from lysis of already infected cells due to the decreased abundance of free virus particles.This reoccurrence can be quantified and used to assess production and cell lysis rates. Several modifications ofthe method are presented and compared. The approaches have great potential for elucidating trends in virusproduction rates as well as for making generalized estimates of the quantitative effects of viruses on marinemicrobial communities.

*Corresponding authors: E-mail: [email protected] or [email protected]

AcknowledgmentsPublication costs for the Manual of Aquatic Viral Ecology were pro-

vided by the Gordon and Betty Moore Foundation. This document isbased on work partially supported by the U.S. National ScienceFoundation (NSF) to the Scientific Committee for OceanographicResearch under Grant OCE-0608600. Any opinions, findings, and con-clusions or recommendations expressed in this material are those of theauthors and do not necessarily reflect the views of the NSF.

The authors would like to thank colleagues and particularly previousstudents who have helped work out the bugs of the described tech-niques. The authors also acknowledge the support of the AgenceNational de la Recherché (ANR Aquaphage 07 BDIV 015 01 MGW) andthe National Science Foundation (NSF-OCE 0452409 SWW) for the sup-port of their research programs that lead to the development of theseideas, and the Scientific Committee for Oceanographic Research for sup-porting working group 126 (marine virus ecology).

ISBN 978-0-9845591-0-7, DOI 10.4319/mave.2010.978-0-9845591-0-7.1Suggested citation format: Weinbauer, M. G., J. M. Rowe, and S. W. Wilhelm. 2010. Determiningrates of virus production in aquatic systems by the virus reduction approach, p. 1–8. In S. W.Wilhelm, M. G. Weinbauer, and C. A. Suttle [eds.], Manual of Aquatic Virus Ecology. ASLO.

MAVE Chapter 1, 2010, 1–8© 2010, by the American Society of Limnology and Oceanography, Inc.

MANUALof

AQUATIC VIRAL ECOLOGY

a high accelerating voltage. Scoring a sufficient number of cells,a researcher can then make estimates of the percentage of a pop-ulation carrying a virus burden. This approach was appliedusing thin-sectioning (Proctor et al. 1993) as well as a whole cellapproach with (Bratbak et al. 1992) or without “lysis-from-with-out” by streptomycin (Weinbauer and Peduzzi 1994).

Radioactive incorporation—Popular for assaying the produc-tion rate of bacteria in aquatic environments, the incorpora-tion rate of radiotracers to estimate virus production wasdeveloped and proposed as a method in the early 1990s (Stew-ard et al. 1992a, 1992b). In brief, the method mimics bacterialproduction assays by estimating the incorporate of a 3H-, 32P-,or 14C-labeled radiotracer (thymidine or leucine) into virusparticles. As such, the technique is highly dependent on theability of the researcher to separate intact virus particles fromboth whole and lysed bacterial and algal materials. Mechani-cal separation (by filtration) is typically the method of choicefor this approach because a high throughput rate of samples isneeded for significant replicates to be processed. However, thesize-range that viruses occupy (~50–750 nm) overlaps with thesize-range of the operational exclusion range (>0.2 µm) lead-ing to the loss of some portion of some samples. Moreover, fil-ter “breakthrough” (the passage of particles greater than theoperational cut off of the filters into samples) quickly con-taminates this assay. This filtration step in the assay is critical,as even minor amounts of contamination from a couple ofbacterial or algal cells can result in a significant error in theestimates of the amount of viral DNA or protein that is pro-duced. As such, both the variance and opportunity for errorassociated with this approach reduce its attractiveness.

Indirect methods: virus decay rates—One approach to deter-mine the rate of production of virus particles is to examinetheir loss rates from the water column. Given that virus parti-cle abundance is static (a tenuous assumption in some cases),then the loss rate of virus particles should be balanced by theproduction rate. Several variations on this approach exist,including the use of tracer particles (Garza and Suttle 1998;Suttle and Chen 1992; Wilhelm et al. 1998a), the use of natu-ral communities, and the arrest of virus production by addingpoisons (Heldal and Bratbak 1991), and the addition of fluo-rescently labeled particles that can be tracked as a percentageof the population (Noble and Fuhrman 2000). In all cases,these approaches provide information on specific groups innatural samples, although the information comes at a cost ofsome tractability for the system in question.

What do we want from a virus production method?—Ulti-mately, the estimation of virus production rates should be asnoninvasive as possible, and be able to provide informationconcerning the production rate of either total virus particles orspecific groups within a sample. To this end, many labs nowfavor the dilution and reoccurrence approach that has been inuse for the last 6 y (Weinbauer et al. 2002; Wilhelm et al.2002). It has been suggested to use the name “virus reductionapproach” (VRA) (McDaniel et al. 2002), because a dilution

approach has been used for a long time for grazers (Landryand Hassett 1982) and has been recently applied to viruses(Baudoux et al. 2007; Evans et al. 2003).

In a comparison with the radiotracer incorporation and thefluorescently labeled viruses approach, Helton et al. (2005)conclude that the VRA should be the most widely applicablemethod because it is the least difficult and the most efficientmethod. In brief, the method involves the removal of freevirus particles from a sample, and then documents their reoc-currence over time. The rate of this reoccurrence, when cor-rected for the relative abundance of potential host cells in asample, allows for an estimate of the production rate of parti-cles in the sample by direct counts using epifluorescencemicroscopy or flow cytometry. In addition, it is possible toestimate the percentage of infected cells (PIC) in the initialpopulation. Adaptations, including the enumeration of spe-cific particles (by qPCR quantification) or infectious particles(via plaque assays or MPN assays), allow for multiple compo-nents of the virus community to be assayed from individualexperiments.

Materials and proceduresGeneral remarks—The reduction and reoccurrence method

for estimating virus production has become the new “goldstandard” by which virus production rates have been mea-sured. This approach has been tested in a number of environ-ments and in different seasons. Whereas the approach itself isrelatively simple, several different adaptations of the approachnow exist. These adaptations are discussed below, each withtheir own variations. The areas of this process can be parti-tioned into the following areas: 1) methods to reduce theabundance of viruses, 2) incubation and sampling of samples,and 3) data processing and interpretation.

Methods to reduce the abundance of viruses—The major differ-ence between all published approaches to measure virus pro-duction by the reduction and reoccurrence method is theprocess of reducing the abundance of free virus particles. Beforecollecting the host community, prefiltration can be used toavoid loss of newly produced viruses by attachment to largeparticles or grazing on infected cells. Like any filtration step,this also has the potential to lead to loss of hosts or viruses. Assuch, if prefiltration of samples is going to occur, it needs to becompleted in a manner appropriate for the samples in ques-tion. As well, separation of the microbial (i.e., host) commu-nity from free viruses also requires filtration, which can lead tosignificant losses or changes in the efficacy of the approach. Tothis end, the choice of membrane material is important, andwhile some membranes (e.g., low protein binding-matrices)may be more expensive that others (e.g., glass fiber or cellulosenitrate) they offer advantages in reduced analytical variancesthat are well worth the extra expense.

In this paper, three different approaches to reduce theabundance of free virus particles are discussed. Whereas eachmethod has its benefit and drawback, it is incumbent on the

Weinbauer et al. Estimating virus production

2

users to understand these as well as to choose the methodmost appropriate for their question of interest.

Approach 1: Over filter virus reduction with continuous cellresuspension (Wilhelm et al. 2002)—In this approach, themicrobial host community (~300 mL) is gently (vacuum pres-sures of <200 mmHg) collected over a 0.2-µm nominal pore-size low protein-binding filter (e.g., Durapore, Millipore Cor-poration) while virus-free (ultrafiltrate, UF) water is added tomaintain the approximate sample volume. After three pas-sages of sample volume through the filter, the retained micro-bial community is distributed (n ≥ 3) for incubation (seebelow). During the filtration process, bacteria are gently andcontinually resuspended from the filter surface using a trans-fer pipette to resuspend cells that may become trapped on themembrane. Since the original approach for this assay, a num-ber of adaptations have been made: these include the use of atube and peristaltic pump to keep cells in suspension (Heltonet al. 2005).

Approach 2: Tangential flow filtration (TFF) based concentra-tion and resuspension of cells in virus-free water (Weinbauer et al.2002)—Bacteria in a 200-300 mL water sample are concen-trated using a 0.2-µm pore-size tangential flow filtration sys-tem (e.g., a Vivaflow 50 cartridge, 0.2-µm pore size, polysul-fone; Vivascience operated by a peristaltic pump). Thebacterial concentrate (ca. 10-15 mL; i.e., the retentate) is keptand the filtrate (permeate) containing the viruses is passedthrough a 30- or 100-kDa filter unit to generate virus freewater. Note that some concentrate is in the cartridge and tubesbut can be collected by removing the feed tube and pumpingthe concentrate into the retentate container. The bacterialconcentrate is then mixed with the UF, and samples are dis-tributed in triplicate into incubation tubes.

Approach 3: TFF virus reduction and continuous cell resuspen-sion (Winget et al. 2005)—This approach is similar to approach2, however, UF is made before and fed into the bacterial reten-tate to keep the volume constant. Filtered volumes are as inapproach 1. One caveat is that passages of the sample volumethrough the filter have been found to marginally improveviral reduction over use of 3 passages of the sample volume(Winget et al. 2005).

Comments on microbial community collection and virusreduction—Ultrafiltered water can be made by a variety of car-tridges that are available from several providers. Either 30 kDaor 100 kDa exclusion cartridges are typically used as they arein the generation of virus concentrates (Wilhelm and Poorvin2001). In practice the 100 kDa should remove less dissolvedorganic matter and, as such, lead to fewer changes in dissolvedsolute concentrations. However, the 100 kDa cartridges mightnot retain very small viruses, such as some RNA viruses.

In all three approaches, the goal is to maintain the hostpopulation while reducing the abundance of free viruses. Typ-ically viral abundance is reduced to ~10%–20% of the initialconcentration, while bacterial abundance is reduced to ~50%.However, recovery efficiency can vary strongly. One would

expect that the recovery efficiency differs among environ-ments, but this has been not studied systematically. While notideal, the reduction in host abundance reduces virus–hostcontact rates and the frequency with which new infectionsoccur during the incubation stage. For approaches 1 and 3, theprocedures require the separate generation of virus-free waterprior to experimental set-up, and this can be time consumingas the virus-free water should be generated from the specificstation where the incubation sample is collected. In practice,this time lag can be reduced by using a larger scale concentra-tion system (e.g., the Amicon M12 system, Millipore), whichcan more rapidly generate virus-free water. One advantage ofapproach 2 is that the virus-free water can be generated in par-allel with the collection of the microbial host community,allowing for more rapid pre-processing and experimental set-up (and as such allowing for multiple samples to be processedin parallel). However, this approach carries with it the caveatthat cells are concentration up to 10-fold beyond their in situabundances for a short period, and this increased cell densitymay have unknown effects on microbial metabolism (e.g.,activation of quorum sensing pathways).

Experiment incubation and sample collection—To determinethe rate of virus production, each of the above approachesrequires that samples containing the reduced virus commu-nity be incubated under in situ conditions so that the micro-bial metabolism can proceed and viruses continue the lyticcycle. Several options are available here, including the use ofenvironmental chambers that can control temperature. In thefield, one of the most common approaches is to use flowinglake/seawater incubators. In this case, water is pumped fromthe sea surface (often exploiting existing equipment if on aresearch vessel, i.e., the ship’s deck water or fire systems) intoan on deck box incubator, and then allowed to return over-board by means of an overflow system. Care must be taken inthese cases to ensure that the volumes and flushing rates ofthe incubators are sufficient to allow for complete incubationof sample bottles while cycling the flow-through fast enoughto maintain surface temperatures (i.e., to avoid heating in thesun). One other question commonly raised concerns whetherto carry out the incubations at in situ light levels or in dark-ness. To date most studies have focused on the heterotrophicbacterial community, and as such, have used darkened bottlesor incubators for this step. Incubation under in situ light con-ditions can be completed and may favor virus production inphotoheterotrophs or alga, but comes with the caveat of virusloss due to light effects. Please see the Assessment section formore details on the impacts of light versus dark incubations.

To determine the rate of virus production in the experi-mental sample, subsamples are collected from the incubationbottles at increments appropriate for the system being studied.In environments where the microbial community is rapidlyturning over, this may be on the order of every 1.5 h, whereasin environments where microbial growth is slow this may beon the order of every 4-6 h. Typically, subsampling is best


3

completed at 2.5-3 h intervals over a period of 10-12 h,although in environments of low trophic status/growth rateexperiments can run 18-24 h. It is critical here that the precisetime of sampling is noted, as this information is required todetermine the rates of virus production within the samples.

Subsamples, once collected, need to be quickly processed orpreserved for enumeration of the virus community. To date,the only published information using any of these approachesinvolves the enumeration of the total virus community withinsamples. Ongoing research, however, is focusing on the reduc-tion and reoccurrence approach to enumerate the rates of pro-duction of individual virus groups (e.g., by plaque assay orquantitative PCR).

Data processing and interpretation—The processing andanalysis of the data collected by the above experimentaldesigns is as important as the choice of method to set up theexperiment. In each case, the results of the enumerationsresult in 3 independent rates of virus production. These ratesare determined from the slopes of plots of virus abundanceversus time for the independent incubations. These in situexperimental production rates must then be corrected for thebacterial losses during sample set up: to do this one simplytakes the ratio of in situ bacterial abundance to experimental(T = 0) bacterial abundance and multiplies this by the produc-tion rate (Table 1, Eq. 1). It is critical to determine these ratesfrom the individual incubations and not from the mean of thevirus abundance in the 3 separate samples, as the independentrates can be used to calculate a mean rate and an estimate ofvariance (the first standard deviation) for that measure.

Once the rate (and variance) of virus production is deter-mined, a number of secondary calculations become availableto the researcher beyond the variations in virus productionrates under different environmental conditions or spatio- temporally. It is important at this juncture to note that eachof these calculations comes with the caveats of not only thismethod, but also of the method used to determine the com-panion parameters discussed below.

The most basic calculation typically completed from thevirus production data is to develop an estimate of the hostcells lost. This estimate is calculated from the rate at whichviruses are produced and an empirically (preferably) deter-

mined or estimated burst size (Table 1, Eq. 2). This calculationmakes the assumption that the viruses produced within a sam-ple are produced primarily from the lysis of heterotrophic bac-teria. While this may not be completely correct, it is generallyconsidered a safe assumption that aquatic viruses in most sam-ples (>90%) are produced this way (Weinbauer 2004).

To estimate the percentage of the microbial communitythat was infected at the beginning of the experiment (%infected cells, PIC), the abundance of viruses produced duringthe observation is divided by the burst size to estimate thenumber of bacterial cells that were lysed (Table 1, Eq. 3). Thisrepresents a conservative estimate of the cells carrying a virus-burden at the onset of the experiment, as some cells in theearly stages of the lytic cycle and with long lytic cycle timesmay not yet have lysed. The PIC is then calculated as 100 × thenumber of lysed cells divided by bacterial abundance at T = 0.

Furthermore, virus production can be related to viral-mediated mortality of bacterioplankton in several ways. Formore detailed calculations, see also http://www.univie.ac.at/nuhag-php/vipcal/ (Luef et al. 2009). Virus production can bedivided by the burst size and bacterial abundance at T = 0 toobtain a lysis rate of the standing stock: for example, as % ofbacterial abundance per day. Using burst size estimates, virallysis rate can also be compared with bacterial production andexpressed as % mortality in the sense of % of productionlysed. In the latter case, it is important to either correct forlosses of bacterial abundance or measure bacterial productionat T = 0 of the incubations.

The PIC can also be related to bacterial mortality usingmodels. Two models have been used (Binder 1999; Proctor etal. 1993) to make these estimates from transmission electronmicroscopy measures. These models are predicated on theassumption that in steady state one of the two daughter cellsoriginating from cell division is lost. Thus, in the model ofProctor et al. (1993) the percentage of infected cells is multi-plied by two to obtain that (“factor-of-two rule”). Binder(1999) developed a more elaborate model that including graz-ing on infected cells to estimate the fraction of mortality fromviral lysis. Note that in those studies the authors chose (webelieve incorrectly) the term frequency instead of percentage,but the calculations are the same.


4

Table 1. Formulae for inferring the production, turnover, and effects of viruses on marine microbial communities

Equation Parameter Units Formula

1. In situ virus production rates (VPR) Particles per volume per time Experimental virus production = (in situ BA/BAex, T=0)

2. Virus-inferred bacterial lysis Bacteria per volume per time Bacterial lysed = VPR/BS

3. Number of lysed cells Bacteria per volume Number of lysed cells = Maximum minus minimum

viral abundance/average burst size

4. Percentage of infected cells Percentage PIC = Number of lysed cell divided by bacterial abundance

5. Virus remobilized nutrients Nutrients per time Nutrients = Virus-inferred bacterial lysis

× nutrient quota per cell

VPR, in situ virus production rate; BA, in situ bacterial abundance; BAex, T=0, experimental bacterial abundance at T = 0; BS, burst size; VA, in situvirus abundance.

A final calculation that has become very relevant as of lateis the production of estimates of nutrients “recycled” due tovirus-mediated cell lysis. In both marine and freshwater envi-ronments, some knowledge of the biochemical impacts ofviruses is desired to better develop models of geochemicalbudgets and cycles. In the current case, the abundance of cellslysed by viruses can be used to estimate carbon and nutrientregeneration rates by multiplying cells lysed by the cellularquota for the nutrient in question (Poorvin et al. 2004). Onecaveat to this calculation is that the fate of elements releasedby virus-mediated cell lysis remains unsure, as only a few stud-ies (Gobler et al. 1997; Middelboe and Jörgensen 2006; Mid-delboe and Lyck 2002; Mioni et al. 2005; Poorvin et al. 2004)have carefully addressed this issue. That said, the role ofviruses within these cycles is no doubt critical (Brussaard et al.2008; Suttle 2007; Wilhelm and Suttle 1999), and potentiallya fruitful area of future research.

AssessmentA series of factors to consider when choosing the

approach that is most appropriate for a lab is given inTable 2. One of the problems with the virus reductionmethod is that the manipulation of the sample could influ-ence virus production. For example, the loss of cells andrelease of organic compounds due to stress or cell breakageduring filtration (Nagata and Kirchman 1990) could influ-ence rates. This alteration could affect bacterial productionand ultimately affect the burst size (Parada et al. 2006).While no changes in bacterial production were seen in earlytrials of the virus reduction approach (Wilhelm and Suttle,unpubl. data), this problem suggests that bacterial produc-tion rates should be measured at the start of the experimentand either during or at the end of incubations to determineif virus production is related to losses of heterotrophic pro-duction. In a previous virus-reduction type assay where thiswas tested, the burst size did not differ between in situ and

mitomycin C–treated samples (Weinbauer and Suttle 1996).This is, so far, the only indication that the VRA does notinfluence burst size, however, it has to be noted that burstsize was not checked in the untreated controls. Anotherpotential problem is that many protistan grazers of prokary-otes are destroyed or inactivated by excessive handling.Because protists can ingest viruses (Gonzalez and Suttle1993) and could preferentially graze on infected cells (Wein-bauer and Peduzzi 1995), such losses could result inincreased variance in the estimates of viral productionalthough preliminary observations (unpublished) suggestthat exclusion of grazers at the beginning of the experimentdoes not affect rates. Finally, viral decay rate is usually notmeasured during the incubations, although it can be impor-tant (Winter et al. 2004).

Only a few studies have compared different approaches toassess prokaryotic mortality by viruses (for example, summa-rized in Weinbauer 2004). Some studies have also comparedvarious approaches of the VRA. Weinbauer et al. (2002) foundno consistent differences between Approach 1 and 2 for fivesamples from coastal and offshore Mediterranean water whencalculating FIC. Another comparison was done usingapproach 1 and 3 (Winget et al. 2005). In three experiments,one of the two methods yielded negative and the othermethod positive values. For the two where positive valueswere obtained, there was no significant difference between thetwo samples.

As part of the current assessment, a comparison ofapproaches 1 and 3 were completed during cruise transects inthe Pacific and Atlantic Oceans (Fig. 1). Across 8 different sta-tions, only one station (occupied in the North Atlantic)showed a significant difference in estimated virus productionrates using these two techniques. That station, which was partof a larger survey of the North Atlantic described elsewhere(Rowe et al. 2008), was a general statistical outlier for a num-ber of parameters.


5

Table 2. A comparison of the pros and cons of the three virus reduction assay approaches

Approach Advantages Disadvantages

Over filter concentration approach • Cells are not concentrated • UF has to be made before the start of the

(Wilhelm et al. 2002) • High reduction efficiency (75 – 80% +) incubations (adds 0.5–1 h to processing time)

• Limited material requirements • Weak recovery of bacteria

TFF Concentration and resuspension • Parallel sampling processing is easy • Bacteria are concentrated, which might increase infection

(Weinbauer et al. 2002) (only one pump needed) and affect performance of cells and physiology

• Most rapid approach • Low reduction efficiency

• Volume needed: 200 mL • Requires 30 or 100 kDa filters to generate ultrafiltered water

• Good recovery of bacteria

TFF concentration with continual resuspension • Cells are not concentrated • UF has to be made before the start of the incubations

(Winget et al. 2005) • High reduction efficiency (75 – 80% +) (adds 0.5–1 h to processing time)

• Good recovery of bacteria • Multiple UF filters needed (0.2 and 30 or 100 kDA)

Another area that has yet to be assessed in terms of esti-mates of virus production rates is the effect of light levels dur-ing the incubation process. While exposure to light may causea loss of virus particles integrity or particle infectivity (Wil-helm et al. 2003; Wilhelm et al. 1998b), exposure to levels ofphotosynthetically active radiation may enhance host pro-duction rates or drive the photoreactivation of viruses thathave experienced DNA damage (Weinbauer et al. 1997; Wein-bauer et al. 1999). To date most studies have focused on esti-mating the production rates of infecting heterotrophic bacte-ria, so the incubation step has been completed with darkenedbottles or incubators. To examine the effects of ambient lightexposure, seven comparisons (using the over-filter approach)were completed during a transect from Hawaii to Australia.The assays were completed in a Plexiglas incubator (therebyreducing UV wavelengths) in bottles where the light field wasreduced to 30% ambient (light) or completely darkened. Asshown in Fig. 2, no significant difference was seen betweenthe light and dark incubations in terms of estimated virus pro-duction rates. While not an exhaustive survey (and absent ofinformation on the richness and evenness of viruses withinthe samples), the results demonstrate that virus productionestimates appear to be independent of light field. One caveatto this is that changes in virus community structure were notexamined in this study: it is possible that some virus popula-tions increased in production whereas others were lost in thecontrasting light and dark incubations.

Discussion

As new estimates of virus production rates appear, oneoverarching observation is that the production rates oftenseem to be too high to be sustainable by estimated bacterialproduction rates. The immediate effect of this is the genera-tion of problems when extrapolating to food web or biogeo-chemical models. However, it has to be noted, that preciselyquantifying both the production and mortality rates is diffi-cult for microorganisms in general. This observation also illus-trates the critical point that microbial communities are doubt-less never in “steady-state” and as such (relatively)near-instantaneous observations of rates do not neatlydescribe community function (Hutchinson 1961). Moreover,this observation also suggests that the production of virusesfrom the lysis of phototrophs and/or protists may be impor-tant in some situations.

Given the above caveat, there remains an opportunity toemploy the information generated by these measures toexamine the effects of viruses on biogeochemical processesand food web interactions. For example, the virus reductionapproach in its various forms has revealed ecologically rele-vant trends of viral infection such as diel cycles (Winter etal. 2004), seasonal variations, and changes along trophicgradients (Winget et al. 2005) or fronts (Wilhelm et al.2002). Carbon and Fe release have also been estimated usingthis approach (Poorvin et al. 2004; Strzepek et al. 2005).


6

Fig. 1. Side-by-side comparisons of the over-filter method reductionapproach (gray circles, Wilhelm et al. 2002) and the tangential flow fil-tration (TFF) method for the reduction of free virus particles (white cir-cles, Winget et al. 2005). Samples were collected in the southeasternPacific Ocean (January 2007) as well as the Sargasso Sea and the NorthAtlantic during May–June 2005 (Rowe et al. 2008). Experiments werecompleted at sea using the described protocols (n = 3, ±SD), andresults are displayed as viruses produced (log scale). No significant dif-ferences were seen (Student t-test, 2-tailed, P < 0.05) except for station11 (far right).

Fig. 2. Side-by-side comparisons of virus production rates determinedusing the over-filter method reduction approach (Wilhelm et al. 2002)with incubation stages completed in the dark (black circles) or reducedsunlight (white circles). For light exposed experiments, samples wereincubated at ambient temperatures in a continuous flow incubator withsolar intensity reduced to 30% using neutral density screening. Sampleswere collected in the southeastern Pacific Ocean (January 2007). Experi-ments were completed at sea using the described protocols (n = 3, ±SD)and results are displayed as viruses produced (log scale). No significantdifferences were seen (Student t-test, 2-tailed, P < 0.05).

Comparisons to other mortality processes, such as grazing,can also be made to gauge how environmental parametersinfluence mortality mechanisms (Gobler et al. 2008; Wein-bauer and Höfle 1998). In all, the availability of a method toestimate virus production rates provides researchers with anopportunity to begin to develop quantitative estimates ofthe effect of viruses on marine microbial communities.

Comments and recommendationsThere are now many adaptations appearing for the above

experimental approaches, and the reader is encouraged toreview the literature prior to undertaking these experiments.New applications, including the use of infection assays andquantitative polymerase chain reaction (qPCR) estimates ofvirus abundance are now being used to allow researchers tofocus on the production rate of specific viruses within wholecommunity populations. As well, the virus reductionapproach has also been employed to estimate virus turnoverrates in marine sediments (Hewson et al. 2001; Mei andDanovaro 2004). This requires alterations to the protocolsdescribed above, and researchers are encouraged to seek outthose references prior to attempting such a study.

ReferencesBaudoux, A. C., M. J. W. Veldhuis, H. J. Witte, and C. P. D.

Brussaard. 2007. Viruses as mortality agents of picophyto-plankton in the deep chlorophyll maximum layer duringIRONAGES III. Limnol. Oceanogr. 52:2519-2529.

Bergh, Ø., K. Y. Børsheim, G. Bratbak, and M. Heldal. 1989.High abundance of viruses found in aquatic environments.Nature 340:467-468.

Binder, B. 1999. Reconsidering the relationship betweenvirally induced bacterial mortality and frequency ofinfected cells. Aquat. Microb. Ecol. 18:207-215.

Bratbak, G., M. Heldal, T. F. Thingstad, B. Riemann, and O. H.Haslund. 1992. Incorporation of viruses into the budget ofmicrobial C-transfer. A first approach. Mar. Ecol. Progr. Ser.83:273-280.

Brussaard, C. P. D., and others. 2008. Global-scale processeswith a nanoscale drive: the role of marine viruses. ISME J.2:575-578.

Duckworth, D. H. 1976. Who discovered bacteriophage? Bac-teriol. Rev. 40:793-802.

Evans, C., S. D. Archer, S. Jacquet, and W. H. Wilson. 2003.Direct estimates of the ocntribution of viral lysis and micro-zooplankton grazing to the decline of a Micromonas spp.population. Aquat. Microb. Ecol. 30:207-219.

Garza, D. R., and C. A. Suttle. 1998. The effect of cyanophageson the mortality of Synechococcus spp. and selection for UVresistant viral communities. Microb. Ecol. 36:281-292.

Gobler, C. J., D. A. Hutchins, N. S. Fisher, E. M. Cosper, and S.A. Sãnudo-Wilhelmy. 1997. Release and bioavailability ofC, N, P, Se, and Fe following viral lysis of a marine chryso-phyte. Limnol. Oceanogr. 42:1492-1504.

———, and others. 2008. Grazing and viral induced mortalityof microbial populations before and during the onset ofannual hypoxia in Lake Erie. Aquat. Microb. Ecol. 51:117-128.

Gonzalez, J. M., and C. A. Suttle. 1993. Grazing by marinenanoflagellates on viruses and virus-sized particles: inges-tion and digestion. Mar. Ecol. Progr. Ser. 94:1-10.

Heldal, M., and G. Bratbak. 1991. Production and decay ofviruses in aquatic environments. Mar. Ecol. Progr. Ser.72:205-212.

Helton, R. R., M. T. Cottrell, D. L. Kirchman, and K. E. Wom-mack. 2005. Evaluation of incubation-based methods forestimating virioplankton production in estuaries. Aquat.Microb. Ecol. 41:209-219.

Hewson, I., J. M. O’Neil, J. A. Fuhrman, and W. C. Dennison.2001. Virus-like particle distribution and abundance in sed-iments and overlying waters along eutrophication gradi-ents in two subtropical estuaries. Limnol. Oceanogr.46:1734-1746.

Hutchinson, G. E. 1961. The paradox of the plankton. Amer.Nat. 95:137-145.

Landry, M. R., and R. P. Hassett. 1982. Estimating the grazingimpact of marine micro-zooplankton. Mar. Biol. 67:283-288.

Luef, B., F. Luef, and P. Peduzzi. 2009. Online program VIPCALfor calculating lytic viral production and lysogenic cellsfrom a viral reduction approach. Environ. Microbiol. Rep.1:78-85.

McDaniel, L. D., L. Houchin, S. Williamson, and J. H. Paul.2002. Lysogeny in marine Synechococcus. Nature 415:496.

Mei, M. L., and R. Danovaro. 2004. Virus production and lifestrategies in aquatic sediments. Limnol. Oceanogr.49:459-470.

Middelboe, M., and P. Lyck. 2002. Regeneration of dissolvedorganic matter by viral lysis in marine microbial communi-ties. Aquat. Microb. Ecol. 27:187-194.

———, and N. O. G. Jörgensen. 2006. Viral lysis of bacteria: animportant source of dissolved amino acids and cell wallcompounds. J. Mar. Biol. Assoc. U. K. 86:605-612.

Mioni, C. E., L. Poorvin, and S. W. Wilhelm. 2005. Virus andsiderophore-mediated transfer of available Fe between het-erotrophic bacteria: characterization using a Fe-specificbioreporter. Aquat. Microb. Ecol. 41:233-245.

Nagata, T., and D. L. Kirchman. 1990. Filtration-inducedrelease of dissolved free amino acids: application to culturesof marine protozoa. Mar. Ecol. Progr. Ser. 68:1-5.

Noble, R. T., and J. A. Fuhrman. 2000. Rapid virus productionand removal as measured with fluorescently labeled virusesas tracers. Appl. Environ. Microbiol. 66:3790-3797.

Parada, V., G. Herndl, and M. G. Weinbauer. 2006. Viral burstsize of heterotrophic prokaryotes in aquatic systems. J. Mar.Biol. Assoc. U. K. 86:613-621.

Poorvin, L., J. M. Rinta-Kanto, D. A. Hutchins, and S. W. Wil-helm. 2004. Viral release of Fe and its bioavailability tomarine plankton. Limnol. Oceanogr. 49:1734-1741.


7

Proctor, L. M., and J. A. Fuhrman. 1990. Viral mortality ofmarine bacteria and cyanobacteria. Nature 343:60-62.

———, A. Okubo, and J. A. Fuhrman. 1993. Calibrating esti-mates of phage-induced mortality in marine bacteria:Ultrastructural studies of marine bacteriophage develop-ment from one-step growth experiments. Microb. Ecol.25:161-182.

Rowe, J. M., and others. 2008. Constraints on virus productionin the Sargasso Sea and North Atlantic. Aquat. Microb. Ecol.52:233-244.

Steward, G. F., J. Wikner, W. P. Cochlan, D. C. Smith, and F.Azam. 1992a. Estimation of virus production in the sea: 2.Field results. Mar. Microb. Food Webs 6:79-90.

———, ———, D. C. Smith, W. P. Cochlan, and F. Azam.1992b. Estimation of virus production in the sea: 1.Method development. Mar. Microb. Food Webs 6:57-78.

Strzepek, R. F., M. T. Maldonado, J. L. Higgins, J. Hall, K. Safi,S. W. Wilhelm, and P. W. Boyd. 2005. Spinning the “FerrousWheel”: The importance of the microbial community in aniron budget during the FeCycle experiment. Global Bio-geochem. Cycles 19:GB4S26 [doi: 10.1029/2005GB002490].

Suttle, C. A. 2007. Marine viruses – major players in the globalecosystem. Nat. Rev. Microbiol. 5:801-812.

———, A. M. Chan, and M. T. Cottrell. 1990. Infection of phy-toplankton by viruses and reduction of primary productiv-ity. Nature 347:467-469.

———, and F. Chen. 1992. Mechanisms and rates of decay ofmarine viruses in seawater. Appl. Environ. Microbiol.58:3721-3729.

Weinbauer, M. G. 1995. Significance of viruses verses het-erotrophic nanoflagellates for controlling bacterial abun-dances in the northern Adriatic Sea. J. Plank. Res. 17:1851-1856.

———. 2004. Ecology of prokaryotic viruses. FEMS Microbiol.Rev. 28:127-181.

———, and P. Peduzzi. 1994. Frequency, size and distributionof bacteriophages in different marine morphotypes. Mar.Ecol. Progr. Ser. 108:11-20.

———, and C. A. Suttle. 1996. Potential significance oflysogeny to bacteriophage production and bacterial mortal-ity in coastal waters of the Gulf of Mexico. Appl. Environ.Microbiol. 62:4374-4380.

———, S. W. Wilhelm, C. A. Suttle, and D. R. Garza. 1997.

Photoreactivation compensates for UV damage and restoresinfectivity to natural marine viral communities. Appl. Env-iron. Microbiol. 63:2200-2205.

———, and M. G. Höfle. 1998. Significance of viral lysis andflagellate grazing as controlling factors of bacterioplanktonproduction in an eutrophic lake. Appl. Environ. Microbiol.64:431-438.

———, S. W. Wilhelm, C. A. Suttle, R. J. Pledger, and D. L.Mitchell. 1999. Sunlight-induced DNA damage and resist-ance in natural viral communities. Aquat. Microb. Ecol.17:111-120.

———, C. Winter, and M. G. Höfle. 2002. Reconsideringtransmission electron microscopy based estimates of viralinfection of bacterioplankton using conversion factorsderived from natural communities. Aquat. Microb. Ecol.27:103-110.

Wilhelm, S. W., M. G. Weinbauer, C. A. Suttle, and W. H. Jef-frey. 1998a. The role of sunlight in the removal and repairof viruses in the sea. Limnol. Oceanogr. 43:586-592.

———, ———, ———, R. J. Pledger, and D. L. Mitchell. 1998b.Measurements of DNA damage and photoreactivationimply that most viruses in marine surface waters are infec-tive. Aquat. Microb. Ecol. 14:215-222.

———, and C. A. Suttle. 1999. Viruses and nutrient cycles inthe sea. Bioscience 49:781-788.

———, and L. Poorvin. 2001. Quantification of algal viruses inmarine samples, p. 53-66. In J. H. Paul [ed.], Marine Micro-biology. Academic Press.

———, S. M. Brigden, and C. A. Suttle. 2002. A dilution tech-nique for the direct measurement of viral production: acomparison in stratified and tidally mixed coastal waters.Microb. Ecol. 43:168-173.

———, W. H. Jeffrey, A. L. Dean, J. Meador, J. D. Pakulski, andD. L. Mitchell. 2003. UV radiation induced DNA damage inmarine viruses along a latitudinal gradient in the south-eastern Pacific Ocean. Aquat. Microb. Ecol. 31:1-8.

Winget, D. M., K. E. Williamson, R. R. Helton, and K. E. Wom-mack. 2005. Tangential flow diafiltration: an improvedtechnique for estimation of virioplankton production.Aquat. Microb. Ecol. 41:221-232.

Winter, C., G. J. Herndl, and M. G. Weinbauer. 2004. Dielcycles in viral infection of bacterioplankton in the NorthSea. Aquat. Microb. Ecol. 35:207-216.


8

9

Introduction

To circumvent the limitations of cultivation-based studies ofcomplex microbial communities, gene-based molecular tech-niques were developed, and the discipline molecular microbialecology was born. The earliest studies in this new discipline cre-ated a wealth of information about the species richness ofmicrobial communities. Because these studies were based on rel-atively labor-intensive cloning and sequencing approaches,however, they were often limited to investigations of only a fewselect samples. The application of genetic fingerprinting tech-niques such as pulsed field gel electrophoresis (PFGE) and dena-turing gradient gel electrophoresis (DGGE) overcame these lim-itations by permitting community composition comparisons ofmultiple samples. DGGE, a PCR-based separation technique,

gave microbiologists the tools to examine genetic diversity ofpreviously uncharacterized mixed microbial populations(Muyzer et al. 1993). PFGE, initially developed to separate yeastchromosomes (Schwartz and Cantor 1984), was successfullyexploited to characterize microbial populations based on theirgenome sizes. Over the last decade or so, aquatic virologistshave applied these two fingerprinting techniques to great effect.

PFGE studies explored the community dynamics of dsDNAviruses in natural aquatic environments (Jiang et al. 2004, Rie-mann and Middelboe 2002, Sandaa and Larsen 2006, Sandaaet al. 2003, Steward et al. 2000, Wommack et al. 1999). Themethod was first used to characterize bacteriophages in sheeprumen, and it was demonstrated that one PFGE band con-sisted of DNA from a single phage genotype (Swain et al.1996). In the marine environment, DNA from virus-like parti-cles were found to range in size from 15 to 630 kb, with mostof the DNA present in the range of 20–80 kb (Sandaa 2008,Steward et al. 2000). Use of PFGE to study marine dsDNA viraldiversity has shown substantial temporal and spatial changesin genome sizes, and the results emphasize that the highlydynamic viral communities are tightly linked to the dynamicsof the host populations (Larsen et al. 2001, 2008; Øvreås et al.2003; Steward et al. 2000; Wommack et al. 1999). In meso-cosm studies, PFGE was used together with analytic flowcytometry to demonstrate a distinct dynamic link betweentwo specific large dsDNA viruses, EhV and CeV, and theirrespective hosts Emiliania huxleyi and Crysochromulina ericina(Castberg et al. 2001; Larsen et al. 2001, 2008). Further, this

Fingerprinting aquatic virus communitiesRuth-Anne Sandaa,1* Steven M. Short,2 and Declan C. Schroeder3

1University of Bergen, Department of Biology, Bergen, Norway2University of Toronto Mississauga, Department of Biology, 3359 Mississauga Road North, Mississauga, Ontario, Canada3Marine Biological Association of the UK, Citadel Hill, Plymouth, UK

AbstractTo circumvent the limitations of cultivation-based studies of complex microbial communities, molecular fin-

gerprinting techniques such as pulsed field gel electrophoresis (PFGE) and denaturing gradient gel elec-trophoresis (DGGE) have been used to examine their richness, diversity, and dynamics. PFGE is based on theelectrophoretic separation of extremely large DNA, raising the upper size limit from 50 kb (standard agarose sep-aration) to well over 10 Mb. This technique has been used to separate aquatic virus genomes ranging in sizefrom tens to hundreds of kilo base pairs (kb); aquatic virus genomes range from 15 to 630 kb, with the majori-ty between 20 and 80 kb. DGGE, on the other hand, is based on the electrophoretic separation of PCR- amplifiedgene fragments of similar sizes, but differing in base composition or sequence. In this chapter, we provide a briefoverview of each of these methods and their application to the study of aquatic viruses. We describe some ofthe common equipment, reagents, and procedures involved, and conclude by briefly considering some of thestrengths and weaknesses of each method.

*Corresponding author: E-mail: *[email protected]



ISBN 978-0-9845591-0-7, DOI 10.4319/mave.2010.978-0-9845591-0-7.9Suggested citation format: Sandaa, R.-A., S. M. Short, and D. C. Schroeder. 2010. Fingerprintingaquatic virus communities, p. 9–18. In S. W. Wilhelm, M. G. Weinbauer, and C. A. Suttle [eds.],Manual of Aquatic Viral Ecology. ASLO.


MANUALof


method was used in conjunction with DGGE to investigatethe relationship between bacteriophages and their hosts innutrient-manipulated mesocosms and the effect of both top-down and bottom-up control of the bacterial community(Øvreås et al. 2003).

Among the first reports of applying DGGE as a tool foraquatic virus research were studies of marine cyanophagediversity. Wilson et al. (2000) initiated this work by exam-ining the diversity of a viral capsid protein (g20) in marineenvironments. Later, Frederickson et al. (2003) adopted thisDGGE-based technique and demonstrated that patterns incyanophage communities were related to the physical struc-ture of the water column in marine environments. Otherstudies that were also based on this DGGE approachrevealed that cyanophage diversity was temporally variablein Lake Bourget, France (Dorigo et al. 2004), and that highlysimilar phage sequences could be obtained from a variety ofmarine and freshwater samples (Short and Suttle 2005).More recently, DGGE was used to study the diversity of g20in Norwegian coastal waters, revealing surprising geneticdiversity and seasonal shifts in cyanophage communities(Sandaa and Larsen 2006). Around the same time that thefirst reports of DGGE analysis of cyanophage genesappeared in the literature, other investigators were usingDGGE to examine the diversity of DNA polymerase genesbelonging to viruses that infect eukaryotic algae, i.e., thephycodnaviruses (Short and Suttle 1999). Eventually, theseresearchers used DGGE-based methods to show that somephycodnaviruses were widely distributed in nature (Shortand Suttle 2002), and some persisted in British Columbiacoastal waters throughout most of a year-long study (Shortand Suttle 2003). Similarly, to study the population dynam-ics of viruses that infect the bloom-forming alga E. huxleyi,Schroeder et al. (2003) developed a DGGE method to exam-ine the diversity of the major capsid protein of E. hux-leyi–specific viruses and demonstrated that only a few mem-bers of the E. huxleyi virus community actually caused thedemise of a bloom observed in a mesocosm experiment.This observation was made again 3 years later at the samestudy site (Martinez et al. 2007).

Although the use of DGGE/PFGE to directly study aquaticviruses is limited to a relatively small number of reports, sev-eral studies have used DGGE to link changes in host commu-nity composition to changes in virus abundance and/or virusdiversity as inferred from PFGE analysis of virus genome sizes(e.g., Castberg et al. 2001, Goddard et al. 2005, Larsen et al.2001, Riemann and Middelboe 2002, Simek et al. 2001, VanHannen et al. 1999b, Weinbauer et al. 2007, Zhang et al.2007). Together, the body of literature describingDGGE/PFGE-based investigations of aquatic virus communi-ties demonstrates the importance of these methods for aquaticvirologists and convincingly demonstrates that fingerprint-based investigations have had a profound impact on currentunderstanding of virus community ecology.

Materials and procedures

An overview of the methods involved in fingerprintedaquatic virus communities is shown in Fig. 1. Because the twomethods considered in this chapter (PFGE and DGGE) are verydifferent, each will be considered independently. Because theanalysis of the fingerprint patterns generated by each methoddoes not differ, this section will conclude with a brief com-ment on fingerprint analysis. Furthermore, in this section wehave provided the formulation for many of the reagents usedfor PFGE and DGGE; nevertheless, more detailed recipes formany of them (e.g., electrophoresis buffers) can be found inmost molecular biology laboratory manuals (e.g., Ausubel etal. 2002, Sambrook et al. 1989).

PFGE—The viral concentrate used for PFGE analysis mustbe molded into plugs, followed by lysis of the virus particles torelease their DNA. It is possible to run solution-based prepara-tion of viral DNA for PFGE (Steward 2001); however, largeDNA molecules (>100 kb) are extremely sensitive to mechani-cal shearing in aqueous solution (Bouchez and Camilleri1997). The consensus is that lysis inside viral plugs preventsmechanical shearing of the DNA, resulting in more discretePFGE bands. Intact viral genomes are then separated by size byPFGE. After separation, the banding pattern is visualized bystaining with a fluorescent DNA stain. This banding patternprovides a visual record of the genome size distribution thatcan be used for qualitative and quantitative comparisonsbetween samples.

Equipment and reagents:• Pulsed field gel electrophoresis system. We recommend

the CHEF-DR II (Bio-Rad);• Casting stand, comes in 14 × 13 cm (small gel, 10 wells)

or 21 × 14 cm (large gel, 15 wells) frame and platform;• Combination comb holder;• Combs, 1.5 mm thick;• Plug molds (Bio-Rad), each well holds 80 µL;• Screened caps (Bio-Rad);• DNA molecular weight standards, e.g., lambda ladder

(available in blocks that have to be sliced into smallerpieces before use) and 5-kb ladder (e.g., Bio-Rad);

• Dilution or storage buffer: SM buffer (0.1 M NaCl, 8mM MgSO4

•7H2O, 50 mM Tris-HCl, 0.005% (wt/vol)glycerin);

• Lysis buffer: must be freshly made (250 mM EDTA, pH 8.0,1% SDS, 1 mg/mL Proteinase K); 5 mL lysis buffer isneeded per sample;

• 1.5% PFGE-grade agarose (e.g., SeaKem GTG agarose;Cambrex) for preparing plugs dissolved in TE 10:1 (10mM Tris, 1 mM EDTA, pH 8.0); the agarose can be storedat 4°C in between uses;

• Washing buffer: TE 10:1 (10 mM Tris, 1 mM EDTA, pH8.0), 150 mL per sample;

• Storage buffer: TE 20:50 (20 mM Tris, 50 mM EDTA, pH8.0), 20 mL per sample;

Sandaa et al. PFGE and DGGE for describing aquatic virus communities

10

• Running buffer: 2 L of 1 × TBE (10 × TBE stock: 108 g Tris,55 g boric acid, 40 mL 0.5 M EDTA pH 8.0);

• Gel buffer: 1 × TBE, 100 mL (small gel) or 150 mL (large gel);• 1% PFGE-grade agarose for gel electrophoresis (e.g.,

SeaKem GTG agarose; Cambrex);

• Fluorescent DNA strain, e.g., SYBR I or SYBR gold (Molec-ular probes);

• UV transilluminator and gel documentation system.Molding of PFGE viral plugs: For long-term stability of

viruses before PFGE and to avoid downstream interference


11

Fig. 1. Flowchart of methods involved in fingerprinting aquatic virus communities.

with nucleic acid extraction or electrophoretic properties ofPFGE, we recommend buffering of the virus concentrate in SMbuffer, prepared via either ultracentrifugation or dialysis(Wommack et al. 2010, this volume). Samples can be storedfor 2–3 days in SM buffer, but we recommend you use it imme-diately. Each of the agarose plugs should represent the sameamount of sample volume; alternatively, number of VLPs.

Prepare the 1.5% agarose and the lysis buffer. Incubate theagarose at 80°C until further use. Dispense 5 mL of the freshlymade lysis buffer into 50-mL Falcon tubes, one for each sam-ple. Combine equal volumes (200 µL) of virus concentrate andmolten 1.5% agarose, mix briefly, and dispense the mixtureinto plug molds with a pipette. Avoid bubbles in the plugs.Place the plug molds in the freezer (–20°C) for at least 2 minto set. Remove the tape from the bottom of the plug moldsand push the plugs out from the molds into 5 mL lysis buffer.Make sure that the entire plugs are submerged in the lysissolution. Incubate the plugs in lysis buffer overnight in thedark at 30°C.

The next day, decant the lysis buffer using a plastic sieve(screened cap) that can be attached at the top of the Falcontube. Be sure that no plugs are stuck in the cap before movingon to decant the next sample. Wash the plugs three times, 30min each, in TE buffer 10:1 at room temperature. The plugscan be stored at 4°C in TE 20:50 for several month before fur-ther processing; nevertheless, we recommend running thesamples as soon as possible, because degradation of the viralDNA will occur and result in less discrete bands.

Gel preparation: Set up the gel rig. Be sure that the combsits evenly along its entire length. Prepare a 1% agarose gel in1× TBE buffer. Melt until the agarose is completely dissolved.Place the warm agarose at 60°C for 10 min before pouring intothe gel rig and allowing it to cool. Avoid air bubbles in the gel.Keep ~5 mL agarose at 60°C for later use to seal the wells.When the agarose is set, pour 50 mL of 1× TBE running bufferon the top of the gel and place it in the refrigerator for at least20 min or overnight. Place molecular weight standards (slicesof ~5 mm) on either side of the gel. Place the samples betweenthe markers using a sterile loop. Be sure that no air bubbles aretrapped in the well. Overlay the wells with leftover molten 1%agarose. Remove the gel from the pouring rig and remove anyextraneous agarose from the bottom and edges.

Electrophoresis: Prepare the 1× TBE (running buffer) andplace at 14°C until further use. Place the gel into the elec-trophoresis chamber and carefully pour the cooled 1× TBErunning buffer into the chamber. Run the gel at 6 V cm–1 withpulse ramps from 20 to 40 s (for example) at 14°C for 22 h.Size markers should be encompassed to facilitate size determi-nation for all the different PFGE viral bands. A number of sizemarkers for PFGE are commercially available. These condi-tions result in runs that make a good starting point for furtheranalysis. To separate different viral genome size classes, eachsample could be run three times: (1) 1–5 s switch time with 20h run time for separation of small genome sizes (0–130 kb); (2)

8–30 s switch time with 20 h run time for separation ofmedium genome sizes (130–300 kb); (3) 20–40 s switch timewith 22 h run time for separation of large genome sizes(300–600 kb). Stain the gel for 30 min or overnight in fluores-cent stain (according to manufacturer’s instructions) and viewon a UV transilluminator.

DGGE—Using DGGE, similar-sized PCR products that differin nucleotide composition (sequence) can be separated indenaturing gradient gels. Denaturing gradient gels are createdusing acrylamide (structural material) solutions that containdifferent amounts of denaturants (urea and formamide), suchthat the highest concentration of denaturants is at the bottomof the gel and the lowest concentration is at the top. AsdsDNA fragments differing in sequence migrate into the gelduring electrophoresis, they encounter increasing concentra-tions of denaturants, and each fragment partially melts (i.e.,double-stranded regions dissociate into single-stranded) at adifferent place in the gel depending on its sequence. Becausethe electrophoretic mobility of partially melted DNA frag-ments is greatly reduced compared to dsDNA, same-sized DNAfragments with different sequences focus at different positionsin these gels. Because PCR with universal primers can be usedto amplify related, but different, DNA sequences, and differentsequences focus at different positions in a denaturing gradientgel, DGGE can be used to produce a unique banding pattern,or fingerprint, for each PCR product amplified from differentmicrobial communities.

Equipment and reagents:• DGGE apparatus, e.g., Hoefer Scientific SE600, Bio-Rad

DCode system (includes gradient former), Ingeny, CBSScientific;

• Peristaltic pump and gradient maker for casting gradientgels, e.g., SG Gradient Maker (GE Healthcare);

• Power supply for electrophoresis systems;• Denaturing gel solutions: Solution A (for 250 mL of an

8% gel solution with 0% denaturant), 50 mL 40% acry-lamide/bis stock solution (37:5:1 acrylamide:bis-acry-lamide solution), 2.5 mL 50× TAE, adjust to 250 mL withsterile distilled water (sdH2O); and Solution B (for 250 mLof an 8% gel solution with 100% denaturant; i.e., 7 Murea and 40% formamide), 50 mL 40% acrylamide/bisstock solution (37:5:1 acrylamide:bis-acrylamide solu-tion), 2.5 mL 50× TAE, 105 g urea, 100 mL deionized for-mamide, adjust to 250 mL with sdH2O;

• 10% wt/vol ammonium persulfate solution in sdH2O(APS); this reagent should be prepared fresh each time itis used;

• TEMED (N,N,N′,N′ -tetramethylenediamine);• 1× TAE running buffer (from 50× buffer stock solution: 20

mM Tris-acetate (pH 7.4), 10 mM sodium acetate, 0.5 mMEDTA);

• 6× loading buffer (e.g., 60% glycerol, 10 mM Tris-HCl, pH7.6, 60 mM EDTA, 0.03% bromophenol blue, 0.03%xylene cyanol FF);


12

• Fluorescent DNA strain (e.g., SYBR Gold or SYBR II gelstain; Molecular Probes);

• UV transilluminator and gel documentation system.Gel preparation: Using lint-free tissues, wash glass plates,

spacers, and combs thoroughly with 70% ethanol. Do not usesoap or harsh abrasive cleaning materials to clean any of theequipment. If the materials are cleaned diligently, there is noneed to use any detergents; a simple water rinsing followed by70% ethanol will suffice. The following instructions will varydepending on the apparatus used; refer to the manufacturerfor specific instruction pertaining to their system. Assemblethe gel sandwich by placing the small glass plate on top of thelarge plate, being sure to correctly place a 1-mm spacer alongeach edge of the plate assembly. To prevent current leakageand the resultant “smiles” in the bands near the edges of thegel, grease both sides of the spacers with as little as possible sil-icon grease to cover the full length of the spacer but only aquarter of the spacer width. Attach the plate clamps and placethe entire assembly into the casting stand. Inspect the plateassembly to ensure that the two glass plates and the spacersform a flush surface along the sides, and ensure that all gasketsadequately seal the plate assembly. Breaches in the seal of theplate assembly with the bottom of the pouring stand willresult in leakage during gel polymerization. Check the gradi-ent maker and flush with sdH2O. Empty pump tubing andattach pipette tip at the outlet tube to the top-middle of thegel chamber. Although the reagents listed above specify theformulation of an 8% gel, the percentage of acrylamide in thegel depends on the size of the PCR products to be resolved;i.e., 6% gel is recommended for 300–1000 bp, 8% for 200–400bp, and 10% for 100–300 bp (BioRad manual, DCode Univer-sal Detection System).

To optimize the gradient conditions for a new DGGE exper-iment (new primer sets, new sample type/habitat, etc.), we usu-ally start with a relatively broad gradient (20% to 80% denatu-rant). We then focus the gradient around the area of interest toinclude the highest and lowest bands in different samples.Table 1 can be used to determine the appropriate compositionof the denaturing gradient gel (16 × 16 cm) that has a total vol-ume of 29 mL. The size and volume may vary between the dif-ferent apparatuses. Make up two solutions of 14.5 mL each, alow denaturant concentration solution and a high denaturantconcentration solution. For example, if you wish to make a30% to 55% gradient, then you would make a 30% (low) solu-tion and a 55% (high) solution based on the reagent volumesin the table. Mix the solutions A and B to the desired percent-age. Alternatively, once desired gradient conditions are empir-ically determined, each denaturing solution can be madedirectly using the reagent volumes and amounts shown inTable 2. After preparing the denaturing gel solutions, degas for15 min and filter through a 0.45-mm syringe filter. Immedi-ately before casting the gel, add 145 µL 10% APS and 7.25 µLTEMED into each solution and swirl gently to mix. Thesereagents begin the polymerization of the acrylamide. At this

point, you will have approximately 10 min to pour the gel. Weuse the SG gradient maker attached to a peristaltic pump. Inour hands, this system has produced more consistent gradientsthan other systems. Make sure the pump is off and the gradi-ent maker-channel is closed (handle up). Pour solution withthe highest denaturant in the right leg of the gradient maker(at the pump side) and the solution with the lowest denaturantin the left leg. Turn the magnetic stirrer on, while simultane-ously starting the pump (5 mL/min). Simultaneously, start thepump (5 mL/min) and move the handle of the gradient makerto horizontal position (channel open). The gel chamber fillsslowly. Use approximately 4–5 min to fill the gel. It is impor-tant to avoid bubbles in the gel, as this will stop the productsfrom migration. Empty the tubing and flush thoroughly withsdH2O. Insert the comb, flat or straight side down, making surethat there are no bubbles under the comb. This is to ensure asmooth, even finish when you come to create your wells afteryour gel has set. Different combs (16 or 20 wells) are available,depending on the number of samples that you want to run.Cover gels with cling film and allow ~2 h for the gel to poly-merize. The gel can be kept at 4°C until the next day.

Electrophoresis: Different DGGE systems require differentvolumes of running buffer. The following procedure is basedon the DCode system from Bio-Rad. Prepare approximately 7L of 1× TAE and fill the buffer chamber. Put about 0.5 L asidefor later use. To enhance the circulation of the running buffer,place the tank on a magnet stirrer and add a magnetic stirrerbar in the bottom of the tank. Preheat the buffer in the DCodeapparatus to 60°C; this will take about 2 h. Attach the gelplates to the core assembly. Loosen the clamps a quarter-turncounterclockwise to prevent breaking of the sandwich clamps(due to heat expansion). Then place the core assembly intothe heated buffer in the tank. Note, the following procedurescan be carried out either while the gel is standing at the benchor when the core assembly is loaded in the tank. Switch off themagnetic stirrer (if loading in tank). Flush each well withbuffer using syringe with needle to remove any unpolymer-ized acrylamide and excess urea. Failure to do this might resultin uneven well floors and unresolved bands. Flush each wellwith buffer again before loading approximately 10–50 µL ofPCR products mixed with loading dye into each well. The vol-ume loaded depends on yield and the expected diversity ofthe PCR products. For quantitative comparisons among sam-ples, equal quantities of DNA must be loaded in each lane. Itis also recommended to use standard markers on the gels toallow gel-to-gel comparisons. The standard marker should becomposed of fragments covering a range of denaturant con-centrations. To quantify PCR products, we recommend gelquantification using a DNA mass standard (e.g., Low DNAMass ladder; Invitrogen) and commercially available gelquantification software such as Quantity One (Bio-Rad) orfree software such as Image J (available for download athttp://rsbweb.nih.gov/ij/download.html). In the DGGE gel,load a marker on each side of the gel adjacent to the samples


13

(markers can be custom made for each DGGE applicationusing known PCR products, or common molecular weightmarkers can be used) for determination of band positions, orcomparisons of different gels. Apply a loading voltage of 200V for 5 min before starting the pump to circulate the buffer,then turn on the magnetic stirrer. The length of the run andthe running voltage depend on the size of the PCR productsand the percentage of acrylamide/bis in the gel. A good start-ing point is to run the gel at 60 V (about 20 mA for one gel)for 19 h. Optimal run times and conditions, however, shouldbe empirically determined for each type of fragment.

When the electrophoresis is complete, take apart the appa-ratus and remove the glass plates from the gel clamps. Care-fully separate the plates, leaving the gel exposed on the largeplate. Use the edge of the small plate to trim the well walls,but be sure to leave the leftmost wall slightly higher than theothers for use as a gel orientation reference. For easy manipu-lation, the gel can either be stained on the large plate or trans-ferred to, stained on, and transported on a plastic sheet. Stainthe gel for 30 min in 50–500 mL fluorescent gel stain (depend-ing on the container, and according to manufacturer’s instruc-tions). Destain the gel for 30 min in 1× TAE (not always nec-essary). Remember, the fluorescent dye binds to nucleic acids;therefore, it is important to minimize contact with skin, sogloves (powder-free) should be worn. If staining in a con-tainer, use plastic and not glass, as the fluorescent dyes accu-mulate over time on glass surfaces. Slide the gel off of the plas-tic sheet or large plate onto a UV transilluminator and viewthe gel. As an alternative to staining gels with fluorescentdyes, some researchers have used fluorescently labeled PCR

primers for DGGE analyses (Neufeld and Mohn 2005).Analysis of PFGE and DGGE fingerprints—DGGE and PFGE

fingerprints can be analyzed using a variety of commerciallyavailable gel analysis software products (e.g., GelCompar II,Applied Maths; BioNumerics 5.1, Applied Maths; QuantityOne, Bio-Rad). A common method to analyze DGGE/PFGEfingerprints involves creating a binary matrix representing thebands occurring in a set of DGGE/PFGE patterns. The presenceor absence of bands in a sample is simply scored in a binarymanner as 1 (present) or 0 (absent), relative to all of the bandsdetected in a set of DGGE/PFGE patterns. The binary data canthen be presented in a dendrogram where the differences infingerprint patters are represented in a graphical format or asa dendrogram using a distance-based cluster analysis tech-niques such as unweighted pairwise grouping with mathe-matical averages (UPGMA). Another possibility is to use mul-tidimensional scaling (MDS) to reduce a complex fingerprintpattern to a point in a two-dimensional space (Van Hannen etal. 1999a). It is important to note that these types of analysesdepend on consistency when detecting bands, and subjectivedetermination of presence or absence should be avoided.Luckily, most commercially available gel analysis software pro-grams allow researchers to use automated band detectionparameters (e.g., band width and intensity), or even set theirown thresholds for each parameter. As an alternative to com-parisons based on presence or absence, the overall pattern ofgel lanes can be compared directly using densitometry profiles(i.e., the pixel intensities at discrete positions in the gel). Thistype of analysis is based on pairwise correlations of profilesand can be used to avoid biases associated with band detection


14

Table 1. Formulation for DGGE gels using 0% and 100% denaturing solutions.

% denaturant of desired gel solution Volume of solution A (0% denaturants) Volume of solution B (100% denaturants)0 14.50 0.00

15 12.32 2.18

20 11.60 2.90

25 10.87 3.63

30 10.15 4.35

40 8.70 5.80

45 7.97 6.53

50 7.25 7.25

55 6.54 7.96

60 5.80 8.70

65 5.07 9.43

70 4.35 10.15

Table 2. Denaturant amounts for various denaturing gel solutions.a

Denaturant concentration 10% 20% 30% 40% 50% 60% 70% 80% 90%Formamide, mL 4 8 12 16 20 24 28 32 36

Urea, g 4.2 8.4 12.6 16.8 21.0 25.2 29.4 33.6 37.8aFor 8% acrylamide gels, add 20 mL of a 40% acrylamide/bis (37.5:1) solution, 2 mL 50× TAE, and bring the volume up to 100 mL using sdH2O.

or loading unequal amounts of DNA. For densitometry profileanalyses, each lane profile is compared to the others, and theresulting matrix of correlation values correspond to lane sim-ilarities. The correlation values can then be transformed to dis-similarity values (1 – similarity) that can be used for clusteranalysis via UPGMA.

Quantitative analysis of DGGE and PFGE gels is also possi-ble. For DGGE, only semiquantitative analysis is possible,since individual band intensities cannot be used to infer targetabundances in natural samples because differences in bandintensity can arise from variable amplification efficiencies fordifferent targets, and/or differences in background DNA. Thusfingerprint patterns can be compared, but individual bandintensities should not be used to infer target abundance innatural samples. For PFGE, quantitative analysis involvesmeasuring the relative fluorescence of each band. Based onthis information, it is possible to obtain values for richnessand abundance that can be used to calculate diversity indices.This type of analysis can be conducted using the commercialsoftware noted above. Of course, it should be noted that bothDGGE and PFGE are subject to a number of confoundingerrors (see “Assessment”). These sources of error should becarefully considered when deciding what types of analyses orcomparisons should be conducted and not the least when theoutcome of the analysis is interpreted.

AssessmentThe PFGE fingerprinting technique provides a PCR-based

independent approach that can be used to characterize viralassemblages from diverse environments. The technique iscapable of electrophoretic separation of DNA fragments rang-ing from <10 to >1000 kb in length, by applying a pulsed elec-tric field with alternating orientation (Birren and Lai 1993,1994). The pulsed electrical field forces pieces of DNA to movein different directions. Although the predominant direction istoward the bottom of the gel, the back-and-forth motionallows the large DNA fragments to migrate their way throughthe gel. Thus, PFGE analysis can differentiate between themany different viral genome sizes found in the aquatic envi-ronments, such as sizes ranging from the giant virus genomesbelonging to the family Phycodnaviridae (>150 kb) to thesmaller single-stranded DNA chp1-like microphage (<10 kb).As this technique makes use of an extraction and separationmethod that places minimal shear on the genomes, the resultof the analysis is a banding pattern that provides a visualrecord of the genome size distribution within a given sample.In practice, from a single sample 10–30 different bands (viralpopulations) can be resolved. In the marine environment, ithas been shown that the most abundant viral populations arethose with the smallest genomes, between 20 and 80 kb; lessabundant are viruses with middle-sized genomes, between 80and 280 kb, and viruses with the largest genomes, between280 and 500 kb, are the least abundant group (Sandaa 2008).In general, the distribution of environmental genome size cor-

responds with the genomic size range of viruses in culture andthe abundance of the respective host community in themarine environment (Sandaa 2008, Steward et al. 2000).

The PFGE assay will detect only the most abundant viralpopulations in the sample. In mesocosm studies, Sandaa andcolleagues have been able to detect between 5 to 11% of thetotal viral community using PFGE. The lowest detectable virusnumber was approximately 102 particles mL–1 and was foundfor a viral population with a genome size of 487 kb (R.-A. San-daa, unpublished results). The first critical step of the PFGEanalysis is to obtain a high-titer viral concentrate (Wommacket al. 2010, this volume). This concentration step is importantfor the sensitivity of the assay, and all biases in this step willbe reflected in the PFGE analysis. The sample volume requiredcan vary greatly depending on the initial concentration ofviruses. Nevertheless, there will be a lower threshold levelwere the bands no longer can be detected on the gel. Theintensity of a given band on a PFGE gel will in additiondepend on the size of the virus genome. It follows from thisthat a virus population with a genome size of 19 kb have to bein a concentration of 2.6 × 103 particles mL–1 to yield the sameband intensity as a virus population with a genome size of 487kb and 102 particles mL–1. Quantitative analysis of bandingpatterns are possible; however, several factors should be keptin mind when analyzing these gels. Different viruses mighthave the same genome size and therefore be indistinguishableon a PFGE gel. This means that the band number is a mini-mum estimate of the total viral diversity. In addition, there issome difficulty in resolving different viral genomes withinnarrow size ranges. This might occur as a diffuse band on thegel. To reduce the latter uncertainty, each sample can be runseveral times, with focus on separation in different genomesize ranges (for more details about limitations of PFGE for viralcommunity studies, see Steward 2001). Another concern thatshould be considered is that some viruses do harbor severalgenomes of different sizes; thus one band does not necessarilymean one viral population (Holmfeldt et al. 2007). Taking allthe limitations of the method into consideration, PFGE mightnot catch all shifts in the viral community; however, themethod has been proven to be useful for observing thedynamics of viral communities in a variety of aquatic envi-ronments (Riemann and Middelboe 2002, Sandaa and Larsen2006, Steward et al. 2000, Wommack et al. 1999).

DGGE separates PCR-generated DNA fragments by usingdifferent sets of oligonucleotide primers designed to target dif-ferent genes. PCR of environmental DNA can generate ampli-cons with different DNA sequences that represent many geno-types. Because PCR products from a given reaction are ofsimilar size (bp), however, conventional separation by agarosegel electrophoresis results in only a single DNA band that doesnot provide any indication of the number of differentsequences that were amplified. DGGE can overcome this lim-itation by separating PCR products with different sequencesbased on the differential denaturing characteristics of the indi-


15

vidual DNA fragments. The technique exploits (among otherfactors) the different stability of G–C pairing (three hydrogenbonds per pairing) as opposed to A–T pairing (two hydrogenbonds). A mixture of DNA fragments of different sequence canbe separated in an acrylamide gel containing a gradient ofincreasing DNA denaturants. In general, DNA fragmentsricher in GC will be more stable and remain double-strandedat higher denaturant concentrations than fragments withlower GC content. Because double-stranded DNA fragmentsmigrate faster in acrylamide gels than partially denatured mol-ecules, partially denatured molecules are effectively larger, andtheir electrophoretic mobility is greatly reduced. Thus, DNAfragments of differing sequence (i.e., G + C content) can beseparated in an acrylamide gel containing a linear gradient ofchemical denaturants. Theoretically, each band in a DGGE gelrepresents a different microbial population present in thecommunity, and therefore, each PCR of an environmentalsample can produce a unique fingerprint that reflects the com-position of the community amplified. Hence, DGGE finger-prints can be used to compare the virus community composi-tion of different aquatic environments.

In addition, the increased stability of G–C pairing can beexploited to increase the resolution of DGGE. Early in thedevelopment of DGGE, it was demonstrated that attachmentof a 40-base-pair G + C–rich sequence (i.e., GC-clamp) to oneend of PCR fragments prevented complete denaturation ofshort amplicons and provided better resolution of very similarsequences. Attachment of GC-clamps to PCR fragments is eas-ily achieved by adding a GC-clamp to the 5′ end of one of thePCR primers (Sheffield et al. 1989, 1992). An important caveatfor the use of GC-clamp primers is that they can compromisethe efficacy of the original primer set. However, this problemcan be easily overcome by conducting two stages of PCR. Dur-ing this two-stage procedure, the first round of PCR is con-ducted with the original primers to amplify specific DNA frag-ments from natural samples, and the second round isconducted using first-round amplicons as templates (e.g., gel-purified PCR products) and a modified GC-clamp primer.

Like any molecular technology, DGGE is subject to certainspecific limitations, and researchers using this techniqueshould be aware of these limitations and interpret their DGGEfingerprints with caution. Perhaps the most significant caveatresearchers should be aware of when using DGGE is that it ispossible that some bands appearing in a fingerprint are theresult of amplification or electrophoretic artifacts. Obviously,the inclusion of artifact bands in community analyses couldconfound similarity comparisons and lead to erroneous con-clusions. Documented examples of these artifacts have beenpublished. For example, Janse et al. (2003) used DGGE to dif-ferentiate different strains of cyanobacteria by amplifying the16S to 23S internal transcribed spacer (ITS) region of rRNAgenes. Although these workers were able to discriminateclosely related cyanobacterial strains using this DGGEmethod, they also noted that ITS rDNA amplification from

several strains produced multiple bands in DGGE gels. In theirreport, they concluded that some cyanobacteria may havemore than one rRNA-ITS operon, and cautioned that the pres-ence of multiple operons in a single organism, or PCR (e.g.,chimera and heteroduplex formation) and electrophoreticartifacts (e.g., comigration of different DNA fragments), couldconfound studies of complex microbial consortia (Janse et al.2003). Eventually, these same researchers demonstrated thatincreasing the final elongation step in their PCR helped alle-viate artifacts arising from the production of multiple bandsfrom a single target molecule (Janse et al. 2004). It should benoted that previous studies have also documented the comi-gration of different DNA fragments in DGGE gels (e.g.,Sekiguchi et al. 2001), and that band re-amplification andsequencing from very different regions in a DGGE gel can pro-duce identical sequences (Nikolausz et al. 2005). Based onthese observations, Nikolausz et al. (2005) concluded that “thebanding pattern in a DGGE gel may not be simply the resultof the separation of different amplicons according to theirmelting behavior, but a consequence of complex interactionsamong different DNA structures.” Thus, it is important to con-sider these potential artifacts when interpreting DGGE-basedcommunity analyses.

DiscussionThat there is no single universal gene present in all viruses

has made it difficult to make inferences regarding the totalviral diversity in natural viral communities. However, genomesize is a universal phenotypic characteristic of sufficient vari-ability to be used as a proxy for classification of the mostabundant populations of dsDNA viruses. PFGE provides awhole-genome fingerprinting of the most abundant and prob-ably most active viral populations in a sample. This informa-tion can be used to reveal variability in the viral compositionover space and time. In addition, the availability of quantita-tive analysis of viruses belonging to the different genome sizepopulations makes it possible to use the assay for diversitystudies (Larsen et al. 2004, Sandaa et al. 2003). Further, as themethod is based on intact viral genomes, it is possible to iden-tify any band of interest by, for instance, specific viral primeror whole-genome sequencing (Sandaa et al. 2008, Santos et al.2007). PFGE can thus serve as a starting point for moredetailed studies of the viral community composition and canbe used to study the ecology of important viral groups in theocean without the need of a cultivable host.

Despite the fact that there are inherent limitations associ-ated with DGGE, the positive aspects of the technique (it isfast, inexpensive, and relatively simple to use and the resultsare reproducible) outweigh its limitations. Although identicalor nearly identical DGGE fingerprints can be produced fromthe amplification of different targets, different fingerprintscannot be produced from the same target given identicalexperimental conditions. In a similar way, PCR is a “unidirec-tional” method that permits the conclusion that a target is


16

present, but not that it is absent from a sample; obviously, thisconstraint on PCR-based experiments has not limited its wide-spread use throughout the life sciences! Thus, DGGE shouldbe considered an effective tool that, for example, can be usedto guide other components of a research program. As just oneexample of the many potential uses of DGGE, the method canbe used to focus clone library construction and sequencingefforts on samples that produce different fingerprint patterns.In this way, the recovery of different sequence types can bemaximized while circumventing the need to clone andsequence every sample in an experimental series. Given anappreciation for its limitations and pitfalls, DGGE remains anextremely useful tool for studies of the molecular ecology ofaquatic virus communities.

Comments and recommendationsAlthough PFGE and DGGE each have their own specific

limitations, they have been widely used to study aquaticmicrobial ecology. Because PFGE is based on the analysis ofintact virus genomes, it can be used to provide a snapshot, orfingerprint, of the richness and dynamics of entire aquaticvirus communities. DGGE, on the other hand, is based on theanalysis of PCR-amplified gene fragments and can be used tofingerprint the richness and dynamics of specific groups ofaquatic viruses. As noted above, both of these methods havedefinite pitfalls, and researchers are encouraged to be cautiouswhen interpreting their fingerprint data. Nonetheless, after aninitial setup cost for the equipment, both of these methods arerelatively inexpensive, quick, and easy, and the results arereproducible. Thus, it is likely that these methods will con-tinue to be useful tools for the study of aquatic viruses foryears to come.

ReferencesAusubel, F. M., and others [eds]. 2002. Short protocols in

molecular biology: A compendium of methods from cur-rent protocols in molecular biology, 5th ed. Wiley.

Birren, B., and E. Lai. 1993. Pulsed field gel electrophoresis, apractical guide. Academic.

———, and ———. 1994. Rapid pulsed-field separation of DNAmolecules up to 250 kb. Nucleic Acids Res. 22:5366-5370.

Bouchez, D., and C. Camilleri. 1998. High molecular weightDNA extraction from Arabidopsis, p. 61-70. In J. Martinez-Zapater and J. Salinas [eds.], Arabidopsis protocols. HumanaPress.

Castberg, T., and others. 2001. Microbial population dynamicsand diversity during a bloom of the marine coccol-ithophorid Emiliania huxleyi (Haptophyta). Mar. Ecol. Prog.Ser. 221:39-46.

Dorigo, U., S. Jacquet, and J. F. Humbert. 2004. Cyanophagediversity, inferred from g20 gene analyses, in the largestnatural lake in France, Lake Bourget. Appl. Environ. Micro-biol. 70:1017-1022.

Frederickson, C. M., S. M. Short, and C. A. Suttle. 2003. The

physical environment affects cyanophage communities inBritish Columbia inlets. Microb. Ecol. 46:348-357.

Goddard, V. J., and others. 2005. Temporal distribution ofviruses, bacteria and phytoplankton throughout the watercolumn in a freshwater hypereutrophic lake. Aquat.Microb. Ecol. 39:211-223.

Holmfeldt, K., M. Middelboe, O. Nybroe, and L. Riemann.2007. Large variabilities in host strain susceptibility andphage host range govern interactions between lytic marinephages and their Flavobacterium hosts. Appl. Environ.Microbiol. 73:6730-6739.

Janse, I., M. Meima, W. E. A. Kardinaal, and G. Zwart. 2003.High-resolution differentiation of cyanobacteria by usingrRNA-internal transcribed spacer denaturing gradient gelelectrophoresis. Appl. Environ. Microbiol. 69:6634-6643.

———, J. Bok, and G. Zwart. 2004. A simple remedy againstartifactual double bands in denaturing gradient gel elec-trophoresis. J. Microbiol. Methods 57:279-281.

Jiang, S., G. Steward, R. Jellison, W. Chu, and S. Choi. 2004.Abundance, distribution, and diversity of viruses in alka-line, hypersaline Mono Lake, California. Microb. Ecol. 47:9-17.

Larsen, A., and others. 2001. Population dynamics and diver-sity of phytoplankton, bacteria and viruses in a seawaterenclosure. Mar. Ecol. Prog. Ser. 221:47-57.

———, and others. 2004. Spring phytoplankton bloomdynamics in Norwegian coastal waters: Microbial commu-nity succession and diversity. Limnol. Oceanogr. 49:180-190.

Larsen, J. B., A. Larsen, R. Thyrhaug, G. Bratbak, and R. A. San-daa. 2008. Response of marine viral populations to a nutri-ent induced phytoplankton bloom at different pCO2 levels.Biogeosciences 5:523-533.

Martinez, J. M., D. C. Schroeder, A. Larsen, G. Bratbak, and W.H. Wilson. 2007. Molecular dynamics of Emiliania huxleyiand cooccurring viruses during two separate mesocosmstudies. Appl. Environ. Microbiol. 73:554-562.

Muyzer, G., De Waal, E. C., and Uitterlinden, A. G. 1993. Pro-filing of complex microbial populations by denaturing gra-dient gel electrophoresis analysis of polymerase chain reac-tion-amplified genes coding for 16S rRNA. Appl. Environ.Microbiol. 59:695-700.

Neufeld, J. D., and W. W. Mohn, 2005. Fluorophore-labeledprimers improve the sensitivity, versatility, and normaliza-tion of denaturing gradient gel electrophoresis. Appl. Envi-ron. Microbiol. 71:4893-4896.

Nikolausz, M., R. Sipos, S. Revesz, A. Szekely, and K. Mari-aligeti. 2005. Observation of bias associated with re-ampli-fication of DNA isolated from denaturing gradient gels.FEMS Microbiol. Lett. 244:385-390.

Øvreås, L., and others. 2003. Response of bacterial and viralcommunities to nutrient manipulations in seawater meso-cosms. Aquat. Microb. Ecol. 31:109-121.

Riemann, L., and M. Middelboe. 2002. Stability of bacterial


17

and viral community compositions in Danish coastalwaters as depicted by DNA fingerprinting techniques.Aquat. Microb. Ecol. 27:219-232.

Sambrook, J., E. F. Fritsch, and T. Maniatis. 1989. Molecularcloning: A laboratory manual, 2nd ed. Cold Spring HarborLaboratory Press.

Sandaa, R.-A. 2008. Burden or benefit? Virus-host interactionsin the marine environment. Res. Microbiol. 159:374-381.

———, E. F. Skjoldal, and G. Bratbak. 2003. Virioplanktoncommunity structure along a salinity gradient in a solarsaltern. Extremophiles 7:347-351.

———, and A. Larsen. 2006. Seasonal variations in virus-hostpopulations in Norwegian coastal waters: Focusing on thecyanophage community infecting marine Synechococcusspp. Appl. Environ. Microbiol. 72:4610-4618.

———, M. Clokie, and N. H. Mann. 2008. Photosynthetic genesin viral populations with a large genomic size range fromNorwegian coastal waters. FEMS Microbiol. Ecol. 63:2-11.

Santos, F., A. Meyerdierks, A. Pena, R. Rossello-Mora, R.Amann, and J. Anton. 2007. Metagenomic approach to thestudy of halophages: The environmental halophage 1. Env-iron. Microbiol. 9:1711-1723.

Schroeder, D. C., J. Oke, M. Hall, G. Malin, and W. H. Wilson.2003. Virus succession observed during an Emiliania huxleyibloom. Appl. Environ. Microbiol. 69:2484-2490.

Schwartz, D. C., C. R. Cantor. 1984. Separation of yeast chro-mosome-sized DNAs by pulsed field gradient gel elec-trophoresis. Cell 37:67–75.

Sekiguchi, H., N. Tomioka, T. Nakahara, and H. Uchiyama.2001. A single band does not always represent single bacte-rial strains in denaturing gradient gel electrophoresis analy-sis. Biotechnol. Lett. 23:1205-1208.

Sheffield, V. C., D. R. Cox, L. S. Lerman, and R.M. Meyers,1989. Attachment of a 40-base-pair G + C rich sequence(GC-clamp) to genomic DNA fragments by the polymerasechain reaction results in improved detection of single-basechanges. Proc. Natl. Acad. Sci. USA 86:232-236.

———, J. S. Beck, and E. M. Stone, 1992. A simple efficientmethod for attachment of a 40-base pair, GC-rich sequenceto PCR-amplified DNA. BioTechniques 12:386-387.

Short, C. M., and C. A. Suttle. 2005. Nearly identical bacterio-phage structural gene sequences are widely distributed inboth marine and freshwater environments. Appl. Environ.Microbiol. 71:480-486.

Short, S. M., and C. A. Suttle. 1999. Use of the polymerasechain reaction and denaturing gradient gel electrophoresisto study diversity in natural virus communities. Hydrobi-ologia 401:19-32.

———, and ———. 2002. Sequence analysis of marine viruscommunities reveals that groups of related algal viruses arewidely distributed in nature. Appl. Environ. Microbiol.68:1290-1296.

———, and ———. 2003. Temporal dynamics of natural com-munities of marine algal viruses and eukaryotes. Aquat.Microb. Ecol. 32:107-119.

Simek, K., and others. 2001. Changes in bacterial communitycomposition and dynamics and viral mortality rates associ-ated with enhanced flagellate grazing in a mesoeutrophicreservoir. Appl. Environ. Microbiol. 67:2723-2733.

Steward, G. F. 2001. Fingerprinting viral assemblages by pulsedfield gel electrophoresis (PFGE), p. 85-103. In J. H. Paul[ed.], Marine microbiology. Academic Press.

———, J. L. Montiel, and F. Azam. 2000. Genome size distri-butions indicate variability and similarities among marineviral assemblages from diverse environments. Limnol.Oceanogr. 45:1697-1706.

Swain, R. A., J. V. Nolan, and A. V. Klieve. 1996. Natural vari-ability and diurnal fluctuations within the bacteriophagepopulation of the rumen. Appl. Environ. Microbiol. 62:994-997.

Van Hannen, E. J., M. Veninga, J. Bloem, H. J. Gons, and H. J.Laanbroek. 1999a. Genetic changes in the bacterial com-munity structure associated with protistan grazers. ArchivFur Hydrobiologie 145:25-38.

———, G. Zwart, M. P. Van Agterveld, H. J. Gons, J. Ebert, andH. J. Laanbroek. 1999b. Changes in bacterial and eukary-otic community structure after mass lysis of filamentouscyanobacteria associated with viruses. Appl. Environ.Microbiol. 65:795-801.

Weinbauer, M. G., K. Hornak, J. Jezbera, J. Nedoma, J. R. Dolan,and K. Simek. 2007. Synergistic and antagonistic effects ofviral lysis and protistan grazing on bacterial biomass, pro-duction and diversity. Environ. Microbiol. 9:777-788.

Wilson, W. H., N. J. Fuller, I. R. Joint, and N. H. Mann. 2000.Analysis of cyanophage diversity in the marine environ-ment using denaturing gradient gel electrophoresis. In C. R.Bell, M. Brylinsky, and P. Johnson-Green [eds.], Microbialbiosystems: New frontiers: Proceedings of the 8th Interna-tional Symposium on Microbial Ecology, Halifax, Canada,Augus 9-14, 1998. Kentville, NS: Atlantic Canada Societyfor Microbial Ecology. p. 565-570.

Wommack, K. E., J. Ravel, R. T. Hill, J. S. Chun, and R. R. Col-well. 1999. Population dynamics of Chesapeake Bay virio-plankton: Total community analysis by pulsed-field gelelectrophoresis. Appl. Environ. Microbiol. 65:231-240.

———, T. Sime-Ngando, D. M. Winget, S. Jamindar, and R. R.Helton. 2010. Filtration-based methods for the collection ofviral concentrates from large water samples, p. 110-117. InS. W. Wilhelm, M. G. Weinbauer, and C. A. Suttle [eds.],Manual of Aquatic Viral Ecology. ASLO.

Zhang, R., M. G. Weinbauer, and P. Y. Qian. 2007. Viruses andflagellates sustain apparent richness and reduce biomassaccumulation of bacterioplankton in coastal marine waters.Environ. Microbiol. 9:3008-3018.


18

19

Introduction

Almost two decades ago, pelagic marine viruses were firstreported to exist in high numbers in the marine environ-ment, exceeding the typical abundance of bacteria (Bergh etal. 1989, Proctor and Fuhrman 1990). Since then, they have

been demonstrated to be agents of significant mortality ofheterotrophic bacteria, cyanobacteria, and phytoplankton(Fuhrman 1999, Wilhelm and Suttle 1999, Suttle 2007).Specifically, it has been shown by a variety of researchers thatviruses are capable of causing up to 50% of the bacterial mor-tality in a range of aquatic environments (e.g., Fuhrman andNoble 1995, Steward et al. 1996, Guixa-Boixareu et al. 1996,Weinbauer and Höfle 1998). The variability of the impact thatviruses have on bacterial assemblages can be high, even overshort periods in the same study area (Bratbak et al. 1996,Steward et al. 1996, Bongiorni et al. 2005). With their influ-ence on aquatic microbial populations, viruses appear to havethe potential to affect the flow of energy and matter inmarine ecosystems. For example, viral infection of bacterialcells, and subsequent cell lysis, has been suggested to result ina “short circuit” in the microbial loop, where recycling fuelsbacterial production and respiration and reduces the amountof organic matter available to macroorganisms (Fuhrman1992, Thingstad et al. 1993). Accurate measurements of viralproduction and turnover are important to accurately assesstheir impacts on microbial food webs, carbon cycling, andtrophic dynamics.

Epifluorescence microscopy, combined with the use of fluo-rescent stains such as DAPI, SYBR Green I, and Yo-Pro I, is a well-documented approach for the enumeration of bacteria andviruses (Hennes and Suttle 1995, Weinbauer and Suttle 1997,

Preparation and application of fluorescently labeled virusparticlesAndré M. Comeau1 and Rachel T. Noble2*1Department of Biology, Université Laval, Québec, QC, Canada2University of North Carolina–Chapel Hill, Institute of Marine Sciences, Morehead City, NC, USA

AbstractTracing the fate of individual host cells and viruses is a challenging problem for microbial ecology. The

emergence of a new assay, using fluorescently labeled viruses (FLVs), offers the promise of a quick and easymethod for monitoring host dynamics and virus decay. Using FLVs as probes (FLVPs) for host cells, an assaywas optimized for use with SYBR Green I–stained phages and was shown to be an efficient and reliable methodfor detection of a strain of Vibrio sp. Various microcosm experiments were conducted that demonstrated theutility of the FLVP assay in resolving ecological interactions at the community level. The assay was also usedto show that FLVPs, at high enough multiplicities of infection, can directly inhibit viral infection by “titeringout” or “coating” the host’s cell surface receptors. FLVs can also be used as tracers for studies of virus produc-tion and decay. The approach is mathematically similar to the isotope dilution technique, employed in thepast to simultaneously measure the release and uptake of ammonium and amino acids. The method can beused to determine rates of viral degradation, production, and turnover for investigations of microbial foodwebs in aquatic systems.

*Corresponding author: E-mail: [email protected]



The authors thank Mya Breitbart at the University of South Florida forhelpful comments and critical reading of the chapter. A. M. Comeau wassupported by the National Sciences and Engineering Research Council ofCanada (NSREC), a Fondation des Treilles scientific prize, and the FrenchCentre National de la Recherche Scientifique (CNRS). R. T. Noble acknowl-edges the support provided by the National Science Foundation and theNational Institutes of Health through the Ecology of Infectious DiseaseProgram grant OCE-03-27056 and an associated REU supplement grant.

ISBN 978-0-9845591-0-7, DOI 10.4319/mave.2010.978-0-9845591-0-7.19Suggested citation format: Comeau, A. M., and R. T. Noble. 2010. Preparation and application offluorescently labeled virus particles, p. 19–29. In S. W. Wilhelm, M. G. Weinbauer, and C. A.Suttle [eds.], Manual of Aquatic Viral Ecology. ASLO.


MANUALof


Noble and Fuhrman 1998, Weinbauer et al. 1998, Patel et al.2007). Tracing the fate of individual host cells and viruses, how-ever, has historically been a challenging problem for microbialecology. Fluorescently labeled viruses (FLVs) can be used as bothprobes of host cells of interest and tracers of viral dynamics. Inthis chapter, we detail both uses of FLVs, using examples of howthe methods can be applied where appropriate.

Using FLVs as probes (FLVPs)—The use of fluorescent label-ing in aquatic ecology developed rapidly around the early1990s, especially in studies on the dynamics of grazing andbacterivory using fluorescently labeled bacteria, algae,viruses, and even synthetic microspheres (Rublee and Galle-gos 1989, Sherr et al. 1991, González and Suttle 1993, Epsteinand Rossel 1995). FLVs were originally created using the first-generation nucleic acid–binding fluorochromes, such as acri-dine orange and DAPI, which had been previously used tovisualize individual viruses (Daley and Hobbie 1975, Cole-man et al. 1981, Hara et al. 1991). Shortly thereafter, Henneset al. (1995) described a new use for FLVs, using second-gen-eration cyanine-based dyes, as specific probes (FLVPs) for sin-gle species of prokaryotes. The fluorescent labeling did nothinder the protein–protein interactions required for viralattachment to the host, and the species-specific nature ofviruses allowed for the in situ detection of the host and thecharacterization of abundances/dynamics in mixed microbialassemblages. They also showed that multiple FLVPs of vary-ing colors could be used to detect multiple species in thesame samples. Since the introduction of this method, FLVPshave been used to trace different bacteria in natural settings,such as phage-sensitive bacteria in biofilms (Doolittle et al.1996) and the polyphosphate-accumulating bacteria Microlu-natus in sludge samples (Lee et al. 2006). Significant interestin FLVPs has also been shown for tracking human pathogenssuch as Salmonella (Mosier-Boss et al. 2003) and Escherichiacoli (Goodridge et al. 1999, Tanji et al. 2004, Kenzaka et al.2006). The focus of the present chapter is to describe the FLVPmethod, using a slightly simplified protocol with a newer flu-orochrome, and to expand on the work of Hennes et al.(1995) by further illustrating the use of FLVPs in microcosmstudies.

Using FLVs as tracers—Rates of virus production and removalcan be determined using calculations previously used for theisotope dilution technique to measure rates of release anduptake of amino acids or dissolved ammonium by usingradioisotopes or stable isotopes, respectively, as tracers (Black-burn 1979, Glibert et al. 1982, Fuhrman 1987). The initial useof this approach is discussed at length in Noble and Fuhrman(2000). The FLVs are similar to labeled molecules (e.g., radioiso-topes) when used as tracers and when added to water samplesat low levels (<10% of the ambient viral concentration).Processes of virus decay and clearance decrease the number ofFLVs and unstained viruses in relative proportion to the totalvirus abundance. New virus production produces only unla-beled viruses, however, thereby diluting the initial pool of FLVs.

Using the rate of change of both labeled and unlabeled virusesover time, rates of virus production and removal can be calcu-lated. The method is particularly useful for measuring virus pro-duction and removal in oligotrophic areas, where radiolabelingapproaches such as that outlined by Steward et al. (1992a,1992b) are difficult. The method permits simultaneous deter-mination of rates of virus production and removal using epiflu-orescence microscopy. The data generated from the approachcan be used to populate conceptual and numerical models ofvirus production and decay. This approach can be effectivelyused in oligotrophic and mesotrophic aquatic systems, but isnot intended for use in highly productive aquatic systems (suchas eutrophic estuaries) where high concentrations of total sus-pended solids and detritus are found (see “Assessment”).

Materials and proceduresFLVP assays—Viral and bacterial isolates, propagation and

preparation of stocks: The virus–host system used for theFLVP experiments was Vibrio alginolyticus strain PWH3a, amarine heterotrophic bacterium, and its phage PWH3a-P1,a species-specific dsDNA virus of the Myoviridae family. Bothstrains were originally isolated from the coastal waters of theGulf of Mexico (Suttle and Chen 1992). The host was main-tained as a –80°C glycerol stock to minimize any long-termculturing effects and was cultivated at 30°C with agitation(120–200 rpm) using Marine Luria-Bertani broth (MLB)(0.5 g L–1 each of casamino acids, peptone, and yeast extract,0.3% vol/vol glycerol, in 25 psu ultrafiltrate base). The viruswas amplified using the plate lysate/liquid elution method(Suttle 1993), substituting ultrafiltered (virus-free) seawaterfor sterile media as the eluting agent. The eluant from multi-ple plates was pooled into a 50-mL centrifuge tube and spunat ~4000g for 20 min to remove large debris. The supernatantwas then collected and filtered through a 47-mm–diameter,0.22-µm–pore-size Durapore (Millipore) membrane to removeany remaining host cells. The final viral stock was titered byplaque assay (Suttle 1993) and kept at 4°C in the dark untilneeded.

The following materials and equipment are required for thepreparation of FLVPs:

• amplified, concentrated virus stock (preferably ≥1010

viruses mL–1); calculate amount of stock needed using thefollowing guide: 1 tube (1.7 mL) of stock = 50 µL of FLVPs→ 1 µL FLVPs/slide = 50 slides;

• Screw cap 1.7-mL microcentrifuge tubes (often listed as1.5 mL);

• 0.02-µm filtered water or appropriate filtered seawatermedium for resuspensions; use the latter if the virus isdestroyed by freshwater;

• SYBR Green I dye working stock (Molecular Probes);• RC80 Beckmann (or similar) ultracentrifuge with an SW40

(or similar) swinging-bucket rotor; alternatively, use arotor capable of directly accepting microcentrifuge tubesin a centrifuge capable of reaching >50,000g ;

Comeau and Noble FLV applications

20

• epifluorescence microscope equipped with a 100× oil-immersion objective and a blue-light excitation filter(such as the Olympus U-MWB2/U-MWIB2 filters).

The following procedure is an adaptation of the originalFLVP protocol from Hennes et al. (1995) for use with SYBRGreen I stain. Normally, this protocol will take 2 days giventhat the softening of the viral pellet (see below) is an overnightstep. However, if the viral pellet redissolves quickly, it can beaccomplished in 1 day.1. Pipette 1.7 mL amplified 0.22-µm filtered virus stock into

1.7-mL screw cap tubes. Float the microcentrifuge tubes inlong ultracentrifuge tubes (14 × 95 mm) using water untilthey are just flush with the top of the ultracentrifugetubes; balance them to within 1 g. Most standard brands ofO-ring screw cap microcentrifuge tubes (10-mm–diameterbodies, 12-mm–diameter screw caps) fit snugly into thelarger ultracentrifuge tubes. Alternatively, use a rotor capa-ble of directly accepting microcentrifuge tubes in a cen-trifuge capable of reaching >50,000g to pellet the viruses.

2. Load the ultracentrifuge tubes into the SW40 rotor andspin them at 133,000g for 1 h. Alternatively, spin at lowerspeeds in a microcentrifuge-accepting rotor for an equiva-lent duration (i.e., 2.5 h at 50,000g ). A small, whitish pel-let should be visible in the microcentrifuge tubes after cen-trifugation.

3. Recover the microcentrifuge tubes. Remove the super-natant and resuspend the pellet as follows: pipette off 1.5mL using a P1000; switch to a P100 and gently removenearly all of the supernatant (remainder usually totals ~10µL); add 40 µL water or seawater medium to bring the vol-ume to 50 µL (remember freshwater versus seawater choicefor viral isolate); gently vortex the tubes to disrupt the pel-lets and place them at 4°C overnight to soften.

Perform all subsequent steps under subdued light since thestain will fade if exposed.4. Thaw the SYBR Green I, then add 1 µL stain to each con-

centrated virus tube and incubate for 15 min in the dark.5. Verify the staining (and monodispersal) of the viruses by

pipetting 1 µL of the suspension onto a microscope slide.Add an 18 × 18 mm coverslip and observe the slide underthe epifluorescence microscope. A veritable “sea” of FLVPsshould be visible, which will fade nearly instantaneouslyas you scan from field to field due to the lack of antifade.If you wish to observe the FLVPs for longer periods, add 1µL of antifade (0.1% p-phenylenediamine) to the FLVPsbefore adding the coverslip.

6. Add 1.65 mL water or seawater medium to each tube andrespin as above. The resulting pellet should be slightlyorange in the case of SYBR Green I.

7. Remove the supernatant and resuspend as above (com-pletes first wash out of stain).

8. Repeat steps 6 and 7 (completes second wash out of stain).9. Repeat steps 6 and 7 again (completes third wash out of

stain and ends protocol).

Ensure that the resulting FLVPs are monodispersed (asabove) after the three centrifugation wash steps. If not, gentlyvortex the tubes to disrupt the pellets and place them at 4°Covernight to soften. The FLVPs can then be used the followingday (day 3) or stored at 4°C for a significant amount of time(see “Assessment”).

FLVP assay: The following materials and equipment arerequired to perform the FLVP assay:

• FLVP stock or 1:100 working stock (diluted in 0.02-µm fil-tered water or seawater medium);

• 1.5-mL microcentrifuge tubes;• appropriate host diluent (e.g., sterile seawater medium);• 0.2-µm Anodisc (Whatman) and 0.45-µm cellulose (back-

ing) filters;• microscope slides and coverslips (25 × 25 mm);• mounting medium containing antifade (50:50 phosphate-

buffered saline:glycerol with 0.1% p-phenylenediamine);• epifluorescence microscope capable of 1000× magnifica-

tion and equipped with a blue-light excitation filter (suchas Olympus U-MWB2/U-MWIB2 filters).

The following continued adaptation of Hennes et al. (1995)is a result of FLVP assay optimizations for the Vibrio alginolyticusPWH3a-phage P1 phage-host system (PHS) for use with SYBRGreen I stain. Novices to microscopy would benefit from read-ing Wen et al. (2004) for details on preparing and storing slidesusing SYBR Green I stain. This protocol is a modification of thestandard slide preparation techniques for bacteria and viruses inaquatic samples (see Suttle and Fuhrman 2010, this volume):1. Prepare a 10-fold dilution series of the culture or natural

sample to be enumerated so that ~105 cells mL–1 of the tar-get (FLVP-specific species) will be obtained. This will resultin a multiplicity of infection (MOI) of at least 3000 virusesper host cell (with an FLVP stock prepared as above from a≥1010 viruses mL–1 initial phage stock). When mixing yourhost of choice with natural samples (e.g., for microcosmstudies), also prepare a background control to check forFLVP attachment to natural cells.

Perform all subsequent steps under subdued light, since thestain will fade if exposed.2. Add 0.1 mL FLVP working stock to 0.9 mL sample in a

microcentrifuge tube for each slide to be prepared and vor-tex to mix. Conversely, use 1 µL concentrated FLVP stockin 1 mL sample if not using the diluted working stock.

3. Allow up to 30 min for adsorption of FLVPs to target cells.Approximately 15 min was adequate for the PWH3a-P1PHS, given that the adsorption kinetics of this phage aresimilar to typical coliphages such as T4; the time dependson the adsorption kinetics of your particular virus–hostsystem and will have to be modified as such.

4. Filter each 1 mL sample onto a 0.2-µm Anodisc filter usinga 0.45-µm HA filter for backing.

5. Pipette 10 µL mounting medium onto the surface of aslide, place the filter over the drop, pipette 10 µL mount-ing medium onto the surface of the filter, and place a 25 ×


21

25 mm coverslip over the filter. (See “FLV tracer experi-ments” below for a discussion of mounting media.)

6. Observe the slides under blue-light excitation and countthe cells with a fluorescent halo. If a problem with non-specific staining (excess, unwashed stain or leakage fromFLVPs) occurs, it will be visible here as diffuse, whole-cellstaining instead of the trademark halos.

FLVP microcosm experiments: Seawater samples for themicrocosm experiments were obtained from either a seawaterholding tank at the University of British Columbia (49°16’N,123°15′W) or a station in Vancouver Harbor (49°18’N,123°06’W). All samples, before use in experiments, werescreened for the endogenous presence of PWH3a using theFLVP assay.

In the first experiment, where the added (exogenous) bac-terium did not dominate the system, PWH3a was added to afinal concentration of 6.3 × 106 cells mL–1 in 500 mL of a natu-ral background of prokaryotes at a final concentration of 2.1 ×106 cells mL–1 (~3:1 final ratio PWH3a:natural). The microcosmwas enriched with MLB (10% final concentration) and incu-bated at 20°C with a 14:10 h light:dark cycle. Subsamples weretaken at ~24-h intervals for prokaryotic cell counts (see Suttleand Fuhrman 2010, this volume) and FLVP assays. Shortly afterthe 72-h time point, PWH3a-P1 virus was added to the micro-cosm at a final concentration of ~105–106 viruses mL–1.

In the second microcosm, where the added (exogenous)bacterium did dominate the system, the PWH3a final concen-tration was increased to 1.9 × 107 cells mL–1 in 50 mL of a fur-ther reduced natural prokaryotic background of 4.4 × 105 cellsmL–1 (~40:1 final ratio PWH3a:natural). The microcosm wasenriched as before and maintained at the same light level asthe previous microcosm, yet the incubation was at the highertemperature of 30°C with aeration. Subsamples were taken asbefore, and infection of the microcosm was performed shortlyafter the 48-h time point with the same concentration ofPWH3a-P1 virus as above.

FLVP receptor titration experiment: Four flasks of 10 mLsterile MLB were inoculated with PWH3a at a final concen-tration of 1.4 × 106 cells mL–1. FLVPs were added to three ofthe flasks at the following MOIs: 0 (uncoated control),1000, and 10, to attempt to coat the PWH3a bacterium’scell surface receptors and render it resistant to subsequentinfection. PWH3a-P1 virus was then added to the threeflasks at an MOI of 0.01–0.1. The fourth flask did not con-tain any FLVPs or PWH3a-P1 and served as the control. Sub-samples were taken from the four flasks at 2-h intervals forFLVP assays.

FLV tracer assays—Concentration of viruses and preparationof FLVs for tracer assays: The following steps describe how toprepare each virus concentrate. There are multiple options formany of the steps; in the case where there is more than oneoption they are noted by capital letters.

Note: All virus concentration steps should be performedeither on ice or in a centrifuge held at <10°C, so as to mini-

mize degradation of virus particles during the concentrationsteps.1. Collect up to 20 L using either four 5-L Niskin bottles (or

other permutation) or by a triple acid- and sample-rinsedbucket into an acid-rinsed 20-L low-density polyethylenecarboy.

2A.The sample should be filtered at 5 kPa through a 142-mm–diameter, 0.22-µm–pore-size Durapore filter to removebacteria and protists. The virus-sized fraction (materialbetween 0.22 µm and 30 kDa) is concentrated to ~150 mLusing a spiral cartridge concentration system (Suttle et al.1991). Further concentration should be conducted usingCentriprep-30 centrifugal concentration units (Millipore)to a final volume of ~5 mL.

2B. Alternatively, the sample can be directly concentratedusing a tangential flow filtration spiral cartridge concen-tration system with either a 30- or 100-kDa cutoff (bothhave shown excellent recovery rates for marine viruses inpast experimental procedures; e.g., Suttle et al. [1991], Bre-itbart et al. [2002]; GE Healthcare, Inc.) and then filteredusing a 0.2-µm Sterivex-type filter (Millipore) to removeunwanted protists and prokaryotes. If desired, further con-centration should be conducted using Centriprep-30 orsimilar centrifugal concentration units (Millipore) to afinal volume of ~5 mL.

3A.To each of the virus concentrates, SYBR Green I should beadded at a final concentration of 2.5% vol/vol and incu-bated in the dark for at least 8 h at 4°C.

3B. To each of the virus concentrates, SYBR Gold (MolecularProbes, Inc.) should be added at a final concentration of2.5% vol/vol and incubated in the dark for at least 4 h at4°C.

4A.After the staining period, the unbound stain can berinsed away by adding an equal volume of 0.02-µm fil-tered seawater (prepared by filtering fresh seawater fromthe same location through an acid-rinsed, autoclavedNalgene filtration unit housing a 47-mm, 0.02-µmAnodisc filter) to the concentrate and centrifuging it inCentriprep-30 ultraconcentration units at 3,000g for 15min. This rinse is done three times. Each time, thelabeled virus particles are resuspended in a total of 5 mLof 0.02-µm filtered seawater while reusing the sameCentriprep-30 unit.

4B. After labeling, the FLVs can be diluted into 1 L of 100-kDafiltrate from the sample site and reconcentrated using tan-gential flow filtration (TFF). This process is repeated threetimes to ensure removal of all stain.

5. The final concentrates should be resuspended in a total of5 mL of 0.02-µm filtered seawater. To determine the con-centration of viruses in the concentrate, 10 µL concentrateis diluted to a final volume of 2 mL with 0.02-µm filteredseawater, filtered through a 0.02-µm Anodisc, and countedby epifluorescence microscopy under blue excitation(Noble and Fuhrman 1998, Patel et al. 2007).


22

FLV tracer experiments:1. Collect seawater samples from the desired location. The

FLV concentrates should be freshly prepared at each newsite and for each new experiment. After determining theconcentration of the FLVs in the concentrate, and theambient concentration of viruses in the seawater, theproper amount of FLV concentrate should be added attracer levels (<10% of original ambient virus concentra-tion) into sample volumes of no less than 400 mL (the rec-ommended sample volume is 1 L).

2. Designate a control treatment. Several approaches can beused for controls. Formalin-treated (FT) killed controls con-sist of 0.02-µm filtered formalin added at a final concentra-tion of 2%. Heat-treated (HT) controls are seawater boiled for10 min and then cooled to ambient seawater temperature.The heat treatment denatures active proteins and enzymesand kills most vegetative bacteria (Karner and Rassoulzade-gan 1995). If using TFF, the filtrate can also be used as a con-trol treatment. Any measurable rate of disappearance of FLVsin FT, TFF-filtered, or HT treatments is subtracted from thatseen in the untreated bottles. Because SYBR Green I stainfades quickly in sunlight, the samples should be incubated atambient seawater temperatures in the dark. Experiments canbe started at dusk so that the beginning of the experiment isdone under simulated in situ conditions.

3. At each time point, total viral abundance and FLV num-bers should be determined in duplicate from small volumesubsamples (5–30 mL) taken into sterile, 15- or 50-mLpolyethylene tubes. The volumes of the subsamplesdepend on the final concentration of the FLVs, theexpected concentration of the native viruses in the sample,and the type of microscope to be used for enumeration.Subsamples are immediately fixed with 1% to 2% (finalconcentration) 0.02-µm filtered formalin and stored at4°C. A suggested framework for the experimental approachmight be to sample at time 0 h and at 4, 8, 12, and 16 hafter initiation of the experiment.

4. Slides should be prepared immediately after sample collectionfor best results, particularly to avoid fading of the FLV signal.Slides should be prepared according to Noble and Fuhrman(1998) or Patel et al. (2007). Attention should be paid to themounting solution chosen, as it has been observed that different mounting solutions (p-phenylenediamine, ascorbicacid, ProLong) perform differently in different water sampletypes and different environments (R. T. Noble, data not pub-lished). Breitbart et al. (2004) suggested that samples can beheld without adverse fading for up to 2 weeks; however, wedo not advocate sample storage for longer than a few daysunless absolutely necessary.

Note: It is recommended that rates of bacterial production bemeasured simultaneous to all time points for virus mea-surements. There are two reasons for this: (1) if measured simul-taneously, the researcher can estimate virus productionthroughout the experiment, rather than relying simply on bac-

terial production estimates from time zero; and (2) bottle effectsare common in small-volume experiments such as these. Mea-surements of bacterial production throughout the experimentwill help the researcher to identify times when bacterial pro-duction is heightened (or reduced) due to bottle effects.

Calculation of virus production and removal rates: Produc-tion and removal rates are calculated from the equations ofGlibert et al. (1982) and Fuhrman (1987). The decay constant,k, is calculated as

(1)

where t is the incubation time and R0 and Rt are the ratios oflabeled to unlabeled viruses at time 0 and time t, respectively.The first two time points in this experiment were t0 and t1. Forexample, R0 is FLV0, divided by the number of total number ofvirus particles (stained and unstained), C0, at time 0.

The mean specific activity, R–, is then calculated as

(2)

The viral decay or removal rate, Dv, is calculated as

(3)

where FLV0 and FLVt are the concentrations of FLV at t0 and attime t , respectively.

The viral production rate, Pv , is calculated as

(4)

where C0 and Ct are the concentrations of virus particles at t0

and time t , respectively.If the virus abundance does not change over time, then

the removal rate is equal to the production rate (and theequation is not used). For each experiment, initial rates (usingthe first two time points, t0 and t1) and overall rates (using theentire time course) of production and decay are calculated. Ini-tial rates of decay/production are closest to in situ rates, as allof the experiments can be started at dusk and held underambient natural conditions. Overall rates representdecay/production under natural conditions for ~12 h, butsamples held in the dark the following morning should notbe exposed to natural sunlight.

Estimates of viral-induced bacterial mortality can be calcu-lated using overall rates of virus production, mean viral abun-dance, mean bacterial abundance and growth rates, and eitheran empirically measured or estimated burst size. Briefly, virus

P

R

R

C

Ct

C Cvt

t

t=

⎛

⎝⎜

⎞

⎠⎟

⎛

⎝⎜

⎞

⎠⎟×

× −( )ln

ln

0

0

0

DFLV FLV

R tvt=

−( )×( )

0

RR

k te kt=

×

⎛

⎝⎜

⎞

⎠⎟× −( )−0 1

k

R

R

tt=

⎛

⎝⎜

⎞

⎠⎟ln 0


23

production rates are divided by the estimated burst size (weused a range from 20 to 50) to determine the number of bac-terial cells killed L–1 day–1. The researcher can divide the num-ber of bacterial cells killed L–1 day–1 by the rate of bacterial pro-duction in cells L–1 day–1, to determine the portion of thebacterial community killed due to viral lysis. All of the pro-vided calculations assume steady state.

AssessmentCreation of FLVPs—The optimization of the staining proto-

col using SYBR Green I resulted in high-quality, monodis-persed FLVPs after very short incubation times (~15 min). Theuse of distilled water, versus seawater, during preparationsseemed essential in obtaining particles with the highest fluo-rescent yield. This comparison was done because (1) thecationic interference of some fluorescent dyes had beenreported (Hennes and Suttle 1995) and (2) some marine virusesare unstable in distilled water owing to osmotic stress (Zachary1976, Faruque et al. 2000). In the case of SYBR Green I, Nobleand Fuhrman (1998) reported that the dye was not affected byseawater during the preparation of slides for epifluorescencemicroscopy on natural populations. Our trials suggest thatinterference does take place in the case of some pure virusstocks, however—effects that would not be seen in mixed, nat-ural populations—which may be related to the nature of theparticles themselves. Regarding long-term storage, only a min-imal loss in stain quality was observed in the FLVP stock main-tained at 4°C over 8 months. The effects of storage at–20/–80°C were not investigated, but we assume that FLVPsshould be stable if they are made from a virus stock whichitself is known to be stable during cryopreservation.

Evaluating the FLVP assay—The tagging protocol fromHennes et al. (1995) was modified for SYBR Green I and sim-plified, in part due to the more amenable characteristics ofSYBR Green I compared to the original YOYO-1 and POPO-1dyes. The high-quality FLVPs, at a high MOI, created clearhalos around the host cells (Fig. 1), allowing easy identificationand discrimination in mixed prokaryotic populations. Thestrong halos were an indication that the host-cell receptor forthe PWH3a-P1 virus is in high concentration on the cell sur-face and is evenly distributed. During the course of experi-mentation, the observed receptor pattern did not change,regardless of the physiological status of the host. Nor werethere any cells showing stain having penetrated into the cells(halos becoming diffuse, whole-cell staining), indicating thatthe FLVPs remained non-infectious (no DNA injection) andthat there was no leakage of the stain out of the FLVPs into thecells. This is consistent with observations from other researchgroups, who have found that FLVPs made from a variety ofphages remain non-infectious and the DNA stain is notinjected. Nonspecific binding of the FLVPs to other prokary-otes, eukaryotes, or detritus was not observed. The efficiency ofthe assay was tested by adding a known abundance of PWH3acells, determined by independent acridine orange counts, to a

natural population and subsequently recovering them usingthe assay. Cell counts were not significantly different (data notshown). The sensitivity of the assay depends on the standardstatistical detection limit for epifluorescence microscopy,which is ~200 cells mL–1 (see Suttle and Fuhrman 2010, this vol-ume). However, filtration of a larger sample volume when lowabundances are suspected can increase this detection limit.

The viability of FLVP slides was investigated after storage at–20°C for varying lengths of time. For slides stored up to 54days, the cell counts did not vary significantly (mean change1.1%, n = 6). This observation is in accordance with Wen et al.(2004), who showed that freshly made epifluorescence slidesmaintain their viability for significant periods of time. Theirstudy, however, did highlight significant problems when mak-ing slides from stored samples that had been fixed with alde-hydes. Previous studies have reported incompatibilities of fluo-rescent stains with aldehydes, but SYBR Green I was purportedto not suffer from this disadvantage (Hennes and Suttle 1995,Weinbauer and Suttle 1997, Noble and Fuhrman 1998). Evengiven this advantage, aldehyde fixatives cross-link membraneproteins, and their compatibility with the FLVP assay wasdoubtful owing to the assay’s dependence on functional virusreceptors on the cell surface. We therefore examined the effectsof glutaraldehyde fixation on tagged cells because of the com-mon use of this fixative with marine microbial samples. Asexpected, glutaraldehyde fixation inhibited adsorption of theFLVPs to the cells. In contrast, fixation after adsorption did notdestroy the interaction between the virus and its receptor, nordid it impede enumeration. Finally, as generally observed byWen et al. (2004) for marine virus samples, a significant dropin FLVP-tagged cells was observed when these were stored inglutaraldehyde before slide preparation. Losses of 13% to 33%(n = 7) of the original cell counts were recorded within hoursof initial fixation. The relative intensities of the FLVP-taggedcells remained essentially constant (i.e., FLVPs stayed


24

Fig. 1. Vibrio alginolyticus strain PWH3a tagged with FLVPs. FLVPs werecreated with SYBR Green I stain and demonstrate uniform attachment tothe host cells.

attached), indicating that these drops in counts were actuallosses of individual cells and probably not decay of FLVPs onthe surface of the cells making them uncountable.

FLVP microcosm experiments—Two microcosm experimentswere conducted to demonstrate the capability of the FLVPassay to resolve ecological interactions by following onespecies within natural communities. The first experimentinvolved the addition of the host strain PWH3a to a naturalcommunity under conditions where the former did not dom-inate the system (Fig. 2). PWH3a composed 75% of the totalprokaryotic community at the outset, and high nutrient con-ditions (10% MLB) were used, as it was anticipated that thesewould favor the PWH3a strain given that Vibrio spp. are com-monly copiotrophic. However, the incubation temperature of20°C was not at its optimal, as this strain was originally iso-lated in the Gulf of Mexico and preferentially maintained at30°C. Consequently, other prokaryotes in the communityincreased in concentration to dominate the system whilePWH3a decreased slightly in concentration. Upon visualinspection of the microcosm, a bloom of organisms was appar-ent after the first couple of days. The color of the bloom wasconsistent with PWH3a, but was proven incorrect by the FLVPassay. Last, concentrated PWH3a-P1 virus was added to themicrocosm shortly after 72 h, and a resulting disappearance ofPWH3a was recorded. In the second microcosm, PWH3a wasagain added to a natural community (Fig. 3). This time, theconcentration of PWH3a was altered so that it composed~98% of the total community at the outset, and the incuba-tion temperature was also raised to 30°C in a further effort topromote PWH3a dominance. As the FLVPs reveal, this domi-nance was achieved and maintained until PWH3a-P1 viruswas again added to lyse the cells. Upon removal of PWH3a,other prokaryotes, presumably suppressed earlier, were thenable to dominate the community 24 h later.

FLVP receptor titration experiment—Attachment to the hostreceptor is crucial in viral infection, and receptor quantity issometimes tied to physiological status (differential receptorexpression), such as for the maltoporin receptor of phage λ(Boos and Shuman 1998), or is sometimes simply a matter ofcell surface dimensions, as in the case of fixed receptor com-ponents such as lipopolysaccaride for phage T4 (Thomassen etal. 2003). Theoretically, FLVPs could be used to roughly titerthe number of receptor sites by testing for resistance to infec-tion—host cells with enough attached FLVPs should have theirreceptor sites saturated, preventing adsorption of normal(unstained) infecting phage. We attempted to demonstrate thisby pre-incubating PWH3a cells with two different MOIs ofFLVPs, 1000 and 10, following an attempt to infect them bysubsequently adding normal PWH3a-P1 virus. Infection ofPWH3a coated with the lower MOI of FLVPs did not seem to beaffected, as the cells were lysed at the same rate as the controlcells (no FLVPs; Fig. 4). The PWH3a cells pre-incubated withthe higher MOI of FLVPs, however, did resist infection. Theabundance decreased slightly in the first 4 h after infection,

presumably as some less protected cells (with slightly less than1000 FLVPs cell–1) were successfully infected and lysed. Afterthis time period, however, the PWH3a population growth par-alleled the uninfected control. These observations show thatFLVPs can act as inhibitors to slow down the kinetics of infec-tion. Additionally, carefully adjusted MOIs of FLVPs may beuseful in roughly titering the relative amounts of viral receptormolecules on host cells.

FLVs as tracers—Since the late 1990s, several groups havereported the use of FLVs as tracers (e.g., Breitbart et al. 2004,Helton et al. 2005). In one study, the authors used FLVs todetermine rates of virus production in hot spring environ-ments in the Sierra Mountains of Central California. Breitbartet al. (2004) observed rates of phage production of 1 × 109 to


25

Fig. 2. Addition of Vibrio alginolyticus strain PWH3a to a natural com-munity of prokaryotes, where the exogenous bacterium did not dominatethe system. PWH3a counts and total prokaryote counts are represented,as well as the time point (~72 h) at which PWH3a-P1 virus was added(arrow). Error bars representing the range of duplicate samples are smallerthan the width of the symbols.

Fig. 3. Addition of Vibrio alginolyticus strain PWH3a to a natural com-munity of prokaryotes, where the exogenous bacterium dominates thesystem. PWH3a counts and total prokaryote counts are represented, aswell as the time point (49 h) at which PWH3a-P1 virus was added (arrow).Error bars representing the range of duplicate samples are smaller thanthe width of the symbols.

1.5 × 109 viruses L–1 day–1 at both sites. The turnover timesobserved in this study were similar to those measured in othernear-shore marine and freshwater communities (Breitbart etal. 2004). The authors employed several alterations to the pre-viously published FLV method by Noble and Fuhrman (2000).These alterations are noted in the described protocol here aspossible alternative steps. Another study was conducted tocompare the available incubation-based approaches for meas-uring virus production in estuaries (Helton et al. 2005). In thisstudy, the authors determined that the FLV tracer assay over-estimated virus production in highly eutrophic waters, sincegreater than 100% of the measured bacterial productionwould have been consumed through viral lysis, according totheir measurements. There are a few reasons for the observedresults in the published estuary study. First, estuarine samplesare difficult to work with for FLV additions. After severalattempts at FLV additions to other highly eutrophic estuarineenvironments in eastern North Carolina, we have observed ahigh likelihood for FLVs to attach to large abiotic particles,making accurate enumeration at time 0 particularly patchyand difficult to obtain. Second, it is difficult to enumerate theFLVs in such waters, given the high levels of detritus andcyanobacteria and high eubacterial abundances. Finally, it ishighly possible that over time, FLV numbers are reduced dueto irreversible binding to both abiotic and biotic (suspendedsolids and phytoplankton) particles, processes that have noth-ing to do with the intended measurement of viral production.We suggest that the FLV tracer assay was intended originallyfor use in oceanic or extreme (e.g., hot spring) environments,which is where the protocol was optimized. Dilution-basedapproaches are optimal for measuring virus production inestuarine environments, and this has been empirically shownin several estuarine locations (Helton et al. 2005; R. T. Noble,unpublished data).

DiscussionThe FLVP assay has been demonstrated as a sensitive and

efficient method for enumerating single prokaryotic species.The FLV tracer assay has been used in a range of different envi-ronments, and researchers that have used the method inextreme environments (hot springs) reported excellent successwith this approach. Some of the possible ecological interac-tions that can be resolved with these assays are presentedherein, as well as possible uses of FLVs as inhibitors of infec-tion kinetics or as titrators of viral receptors. Traditional cul-ture methods have been inefficient at these tasks, and alterna-tive modern techniques, such as fluorescent in situhybridization (FISH) with rRNA probes (Amman et al. 1995) orimmunofluorescence (Middelboe et al. 1996), have more com-plex protocols and requirements. The FLV assay’s sole require-ments are (1) a virus for the host of interest (for probes); (2)sufficient receptor sites for the virus on the host cell surface;and (3) an adequate fluorochrome, such as one of the effectivecyanine dyes optimized in this study. Given these facts, theFLV-based assays should be powerful tools in applied settingsand in aquatic microbial ecology.

Comments and recommendationsThe FLV assays described in this chapter should be relatively

robust, and changing fluorescent dyes and/or virus–host sys-tems will probably require only small changes, if any, to theprotocols developed herein, possibly regarding the ratio ofviruses to hosts and/or the adsorption times in the FLVP assay.FLVs have been successfully made from a variety of viruses andusing a multitude of fluorochromes, even as integrated capsidfusions to fluorescent proteins (Tanji et al. 2004, Slootweg et al.2006). The FLVP assay has thus far been limited to bacterio-phages, however, and a potential future use of the assay wouldbe to expand to eukaryotic viruses to track the dynamics ofindividual microalgae. Certain algal viruses, such as the Het-erosigma akashiwo virus HaNIV (Lawrence et al. 2001), haveburst sizes that should be adequate to generate the high titerstocks needed for FLVP preparation. Additionally, Hennes et al.(1995) showed that the autofluorescence of autotrophic cells(in this case Synechococcus cyanobacteria) did not interfere withthe FLVP assay when fluorochromes were appropriatelyselected and that the viral receptor sites remained available forthe assay, even though their expression may be tightly linkedto light exposure. The protocols developed herein are stringentwith regard to protection from light to minimize stain fading;however, they may be relaxed (using subdued light conditionsor short expositions to full light) if phototrophs show amarked, and quick, loss of receptors when placed in dark con-ditions. We suspect that occurrence of this problem should berelatively rare, as incubation with the FLVPs is relatively short(15–30 min) and that many receptors, even for viruses of pho-totrophs, are probably fairly integral components of cell mem-branes (such as lipopolysaccharide) without rapid recycling


26

Fig. 4. Titration of Vibrio alginolyticus strain PWH3a phage receptors bypretreatment with FLVPs. PWH3a counts are represented for the four con-ditions, in order: uncoated (no FLVPs), uninfected control; PWH3a-P1–infected with 1000 FLVPs cell–1 precoating; PWH3a-P1–infected with10 FLVPs cell–1 precoating (under triangles); and PWH3a-P1–infected withno precoating.

permitting immediate receptor deletion. However, it is stillimportant to determine whether there are enough viral recep-tors or whether their distribution on the cell surface allows foradequate enumeration, along with whether they remain avail-able at all times throughout the assay.

It is possible that the future of marine microbial ecologycould include the concept of tagging virus concentrates withquantum dots. This approach has been conducted for humanpathogenic viruses, such as HIV, and has been detailed in avariety of journal articles (e.g., Kampani et al. 2007). Quantumdots can be purchased commercially, can be linked to theviruses in a variety of ways, and should not suffer from thefading problems associated with the SYBR-based dyes usedpreviously in FLV-type studies. It is possible that quantumdot–labeled viruses could be used as probes similar to thoseused in the FLVP assay and could facilitate enumeration andsorting of the tagged host cells through the use of equipmentsuch as flow cytometry, the latter of which has only been usedonce (possible signal-to-noise ratio problems) with standardfluorochrome FLVPs (Goodridge et al. 1999).

Since development of the protocols in the late 1990s, sev-eral groups have made suggested improvements and modifi-cations to conduct FLV tracer experiments under a range ofconditions (i.e., at times when centrifuges are not available,or other equipment items cannot be used). The generalapproach and protocol presented can be modified in a num-ber of ways to accomplish the goal of the researcher. The cre-ation of the virus concentrate could be accomplished using arange of approaches not tested here (e.g., hollow fiber filtra-tion, filter cartridges, filtration-elution). The original conceptwill benefit from further testing in a range of different aquaticenvironments. The FLV tracer assay was originally designedfor use in oceanic environments and is likely an optimalapproach for use in aquatic systems without high concentra-tions of detrital or particulate organic material. As presentedby Helton et al. (2005) and noted in previous experimentsconducted by the authors of this chapter, we suggest that theFLV assay may not be appropriate for highly eutrophic estu-arine environments. Experiments conducted in the NeuseRiver Estuary, North Carolina, have basically shown that FLVtracer studies were not appropriate in waters with high levelsof TSS (>50 mg/L) and high turbidity (R. T. Noble, unpub-lished data). However, we suggest that the FLV tracerapproach is one of several methods that could be used in thecoming years to better understand the roles and function ofviruses in the global oceans.

ReferencesAmann, R., W. Ludwig, and K.-H. Schleifer. 1995. Phylogenetic

identification and in situ detection of individual microbialcells without cultivation. Microbiol. Rev. 59:143-169.

Bergh, O., K. Y. Børsheim, G. Bratbak, and M. Heldal. 1989.High abundance of viruses found in aquatic environments.Nature 340:467-468.

Blackburn, T. H. 1979. Method for measuring rates of NH4+

turnover in anoxic marine-sediments, using a 15N-NH4+

dilution technique. Appl. Environ. Microbiol. 37:760-765.Bongiorni, L., M. Magagnini, M. Armeni, R. Noble, and R.

Danovaro. 2005. Viral production, decay rates, and lifestrategies along a trophic gradient in the North AdriaticSea. Appl. Environ. Microbiol. 71:6644-6650.

Boos, W., and H. Shuman. 1998. Maltose/maltodextrin systemof Escherichia coli: Transport, metabolism, and regulation.Microbiol. Mol. Biol. Rev. 62:204-229.

Bratbak, G., M. Heldal, T. F. Thingstad, and P. Tuomi. 1996.Dynamics of virus abundance in coastal seawater. FEMSMicrobiol. Ecol. 19:263-269.

Breitbart, M., and others. 2002. Genomic analysis of uncul-tured marine viral communities. Proc. Natl. Acad. Sci. USA99:14250–14255.

———, L., Wegley, S. Leeds, T. Schoenfeld, and F. Rohwer.2004. Phage community dynamics in hot springs. Appl.Environ. Microbiol. 70:1633-1640.

Coleman, A. W., M. J. Maguire, and J. R. Coleman. 1981.Mithramycin- and 4′-6-diamidino-2-phenylindole (DAPI)-DNA staining for fluorescence microspectrophotometricmeasurement of DNA in nuclei, plastids and virus particles.J. Histochem. Cytochem. 29:959-968.

Daley, R. J., and J. E. Hobbie. 1975. Direct counts of aquaticbacteria by a modified epifluorescence technique. Limnol.Oceanogr. 20:875-882.

Doolittle, M. M., J. J. Cooney, and D. E. Caldwell. 1996. Trac-ing the interaction of bacteriophage with bacterial biofilmsusing fluorescent and chromogenic probes. J. Ind. Micro-biol. 16:331-341.

Epstein, S. S., and J. Rossel. 1995. Methodology of in situ graz-ing experiments: Evaluation of a new vital dye for prepara-tion of fluorescently labeled bacteria. Mar. Ecol. Prog. Ser.128:143-150.

Faruque, S. M., M. M. Asadulghani Rahman, M. K. Waldor, andD. A. Sack. 2000. Sunlight-induced propagation of the lyso-genic phage encoding cholera toxin. Infect. Immun.68:4795-4801.

Fuhrman, J. A. 1987. Close coupling between release anduptake of dissolved free amino acids in seawater studied byan isotope dilution approach. Mar. Ecol. Prog. Ser. 37:45-52.

———. 1992. Bacterioplankton roles in cycling of organicmatter: The microbial food web, p. 361-383. In Falkowski,P. G., and A. D. Woodhead. [eds.], Primary productivity andbiogeochemical cycles in the sea. Vivirito, K. [technicaled.]. Plenum.

———. 1999. Marine viruses and their biogeochemical andecological effects. Nature 399:541-548.

———, and R. T. Noble. 1995. Viruses and protists cause simi-lar bacterial mortality in coastal seawater. Limnol.Oceanogr. 40:1236-1242.

Glibert, P. M., F. Lipschultz, J. J. McCarthy, and M. A. Altabet.1982. Isotope dilution models of uptake and remineraliza-


27

tion of ammonium by marine plankton. Limnol. Oceanogr.27:639-650.

González, J. M., and C. A. Suttle. 1993. Grazing by marinenanoflagellates on viruses and virus-sized particles: Inges-tion and digestion. Mar. Ecol. Prog. Ser. 94:1-10.

Goodridge, L., J. R. Chen, and M. Griffiths. 1999. Develop-ment and characterization of a fluorescent-bacteriophageassay for detection of Escherichia coli O157:H7. Appl. Envi-ron. Microbiol. 65:1397-1404.

Guixa-Boixareu, N., J. I. Calderón-Paz, M. Heldal, G. Bratbak,and C. Pedrós-Alió. 1996. Viral lysis and bacterivory asprokaryotic loss factors along a salinity gradient. Aquat.Microb. Ecol. 11:215-227.

Hara, S., K. Terauchi, and I. Koike. 1991. Abundance of virusesin marine waters: Assessment by epifluorescence and trans-mission electron microscopy. Appl. Environ. Microbiol.57:2731-2734.

Helton, R. R., M. T. Cottrell, D. L. Kirchman, and K. E. Wom-mack. 2005. Evaluation of incubation-based methods forestimating virus production in estuaries. Aquat. Microb.Ecol. 41:209-219.

Hennes, K. P., and C. A. Suttle. 1995. Direct counts of virusesin natural waters and laboratory cultures by epifluorescencemicroscopy. Limnol. Oceanogr. 40:1050-1055.

———, C. A. Suttle, and A. M. Chan. 1995. Fluorescentlylabeled virus probes show that natural virus populationscan control the structure of marine microbial communities.Appl. Environ. Microbiol. 61:3623-3627.

Kampani, K., K. Quann, J. Ahuja, B. Wigdahl, Z. K. Khan, P.Jain. 2007. A novel high throughput quantum dot-basedfluorescence assay for quantitation of virus binding andattachment. J. Virol. Methods. 141:125-32.

Karner, M., and F. Rassoulzadegan. 1995. Extracellular enzymeactivity: Indications for high short-term variability in acoastal marine ecosystem. Microb. Ecol. 30:143-156.

Kenzaka, T., F. Utrarachkij, O. Suthienkul, and M. Nasu. 2006.Rapid monitoring of Escherichia coli in Southeast Asianurban canals by fluorescent-bacteriophage assay. J. HealthSci. 52:666-671.

Lawrence, J. E., A. M. Chan, and C. A. Suttle. 2001. A novel virus(HaNIV) causes lysis of the toxic bloom-forming alga Hete -rosigma akashiwo (Raphidophyceae). J. Phycol. 37:216-222.

Lee, S. H., M. Onuki, H. Satoh, and T. Mino. 2006. Isolation,characterization of bacteriophages specific to Microlunatusphosphovorus and their application for rapid host detection.Lett. Appl. Microbiol. 42:259-264.

Middelboe, M., N. O. G. Jørgensen, and N. Kroer. 1996. Effectsof viruses on nutrient turnover and growth efficiency ofnon-infected marine bacterioplankton. Appl. Environ.Microbiol. 62:1991-1997.

Mosier-Boss, P. A., S. H. Lieberman, J. M. Andrews, F. L.Rohwer, L. E. Wegley, and M. Breitbart. 2003. Use of fluo-rescently labeled phage in the detection and identificationof bacterial species. Appl. Spectrosc. 57:1138-1144.

Noble, R. T., and J. A. Fuhrman. 1998. Use of SYBR Green I forrapid epifluorescence counts of marine viruses and bacteria.Aquat. Microb. Ecol. 14:113-118.

——— and J. A. Fuhrman. 2000. Rapid virus production andremoval as measured with fluorescently labeled viruses astracers. Appl. Environ. Microbiol. 66:3790-3797.

Patel, A., R. T. Noble, J. Steele, I. Hewson, M. Schwallbach, andJ. A. Fuhrman. 2007. Virus and prokaryote enumerationfrom planktonic marine environments by epifluorescencemicroscopy with SYBR® Green I. Nat. Protoc. 2:269-276.


Rublee, P. A., and C. L. Gallegos. 1989. Use of fluorescentlylabeled algae (FLA) to estimate microzooplankton grazing.Mar. Ecol. Prog. Ser. 51:221-227.

Sherr, E. B., B. F. Sherr, and J. McDaniel. 1991. Clearance ratesof <6 µm fluorescently labeled algae (FLA) by estuarine pro-tozoa: Potential grazing impact of flagellates and ciliates.Mar. Ecol. Prog. Ser. 69:81-92.

Slootweg, E. J., H. J. H. G. Keller, M. A. Hink, J. W. Borst, J.Bakker, and A. Schots. 2006. Fluorescent T7 display phagesobtained by translational frameshift. Nucl. Acids Res.34:e137.

Steward, G. F., J. Wikner, W. P. Cochlan, D. C. Smith, and F.Azam. 1992b. Estimation of virus production in the sea: 2.Field results. Mar. Microb. Food Webs 6:79-90.

———, J. Wikner, D. C. Smith, W. P. Cochlan, and F. Azam.1992a. Estimation of virus production in the sea: I. Methoddevelopment. Mar. Microb. Food Webs 6:57-78.

———, D. C. Smith, and F. Azam. 1996. Abundance and pro-duction of bacteria and viruses in the Bering and ChukchiSeas. Mar. Ecol. Prog. Ser. 131:287-300.

Suttle, C. A. 1993. Enumeration and isolation of viruses, p.121-134. In Kemp, P. F., B. F. Sherr, E. B. Sherr, and J. J. Cole.[eds.], Handbook of methods in aquatic microbial ecology.Lewis Publishers.

———. 2007. Marine viruses: Major players in the globalecosystem. Nature Rev. Microbiol. 5:801-812.

———, A. M. Chan, and M. T. Cottrell. 1991. Use of ultrafil-tration to isolate viruses from seawater which arepathogens of marine phytoplankton. Appl. Environ. Micro-biol. 57:721-726.


———, and J. A. Fuhrman. 2010. Enumeration of virus parti-cles in aquatic or sediment samples by epifluorescencemicroscopy, p. 145-153. In S. W. Wilhelm, M. G. Wein-bauer, and C. A. Suttle [eds.], Manual of Aquatic Viral Ecol-ogy. ASLO.

Tanji, Y., C. Furukawa, S. H. Na, T. Hijikata, K. Miyanaga, andH. Unno. 2004. Escherichia coli detection by GFP-labeledlysozyme-inactivated T4 bacteriophage. J. Biotechnol.114:11-20.


28

Thingstad, T. F., M. Heldal, G. Bratbak, and I. Dundas. 1993.Are viruses important partners in pelagic food webs? TrendsEcol. Evol. 8:209-213.

Thomassen, E., and others. 2003. The structure of the recep-tor-binding domain of the bacteriophage T4 short tail fibrereveals a knitted trimeric metal-binding fold. J. Mol. Biol.331:361-373.

Weinbauer, M. G., and C. A. Suttle. 1997. Comparison of epi-fluorescence and transmission electron microscopy forcounting viruses in natural marine waters. Aquat. Microb.Ecol. 13:225-232.

———, C. Beckmann, and M. G. Höfle. 1998. Utility of green flu-orescent nucleic acid dyes and aluminum oxide membrane

filters for rapid epifluorescence enumeration of soil and sedi-ment bacteria. Appl. Environ. Microbiol. 64:5000-5003.

———, and M. G. Höfle. 1998. Size-specific mortality of lakebacterioplankton by natural virus communities. Aquat.Microb. Ecol. 15:103-113.

Wen, K., A. C. Ortmann, and C. A. Suttle. 2004. Accurate esti-mation of viral abundance by epifluorescence microscopy.Appl. Environ. Microbiol. 70:3862-3867.

Wilhelm, S. W., and C. A. Suttle. 1999. Viruses and nutrientcycles in the sea. BioScience 49:781-788.

Zachary, A. 1976. Physiology and ecology of bacteriophages ofmarine bacterium Beneckea natriegens: Salinity. Appl. Envi-ron. Microbiol. 31:415-422.


29

30

Introduction

Lysogeny occurs when a temperate phage establishes a sta-ble symbiosis with its bacterial host. This is accomplishedmost often by integration of the host genome into one of thehost’s replicons, although prophages that exist as autonomousplasmids have also been described in marine bacteria (Mob-berley et al. 2008). The integrated temperate phage genome istermed a prophage. The prophage usually confers an alteredphenotype to the host. When this is the result of expression ofphage genes, it is termed conversion. A phage infection thatresults in both the production of high phage titers and hostcells is termed pseudolysogeny (Ackermann and DuBow 1987).Historically, lysogeny has been regarded to impart increasedfitness to the lysogenized host compared with the uninfectedhost (Edlin et al. 1975), yet there are few demonstrations of

this in the literature, and none in marine bacteria (Paul 2008).A similar process of viral integration occurs in eukaryotes,termed latency, is outside the scope of this review.

The operational definition of a lysogen is a bacterium thatcontains an inducible prophage particle, most often detectedthrough the use of an inducing agent (Ackermann and DuBow1987). “The most sensitive method is thus induction by mito-mycin C or UV light (or a combination of both) followed bythe spot test in combination with electron microscopic exam-ination” (Ackermann and DuBow 1987). However, not allprophages are inducible with mitomycin C (Ackermann andDuBow 1987). A more stringent definition of a lysogen is abacterium that contains a prophage that is capable of infect-ing other hosts and establishing a lysogenic relationship. Forthis to occur, one needs an uninfected yet sensitive host, andsuch complete systems (temperate phage, lysogen, and unin-fected host) have seldom been described for marine bacteria.

The value of lysogeny can only be inferred from ecologicalstudies in natural populations. Lysogeny seems to occur dur-ing conditions that are unfavorable for rapid vegetative hostgrowth. Such conditions may be manifested in deep sea envi-ronments (Weinbauer et al. 2003), oligotrophic surface waters(Long et al. 2008), Antarctic lakes (Lisle and Priscu 2004), Arc-tic lakes (Laybourne-Perry et al. 2007), or during wintermonths (McDaniel et al. 2002, Williamson et al. 2002). Theseconditions are characterized by low host abundance andgrowth rates, which would result in a low probability of a suc-cessful host encounter and lytic infection. Survival as aprophage also ensures protection from some of the viral inac-tivating factors that free phage particles encounter (i.e., UVinactivation and grazing; Wommack and Colwell 2000).

Detection of lysogeny in marine environmentsJohn H. Paul*1 and Markus Weinbauer2

1College of Marine Science, University of South Florida, 140 Seventh Ave. S, St. Petersburg, FL 337012CNRS and Université Pierre et Marie Curie-Paris 6, Laboratoire d’Océanographie de Villefranche, 06230 Villefranche-sur-Mer, France

AbstractSilent viral infections occur in all forms of life, from bacteria to humans, as indicated from genomic sequenc-

ing. Temperate phages can infect bacteria and establish a symbiotic relationship termed lysogeny, enabling thephage genome to be propagated in host daughter cells. The expression of prophage genes often results in analtered bacterial phenotype, often turning benign bacteria into virulent pathogens. The most widely usedmethod to detect lysogens is to chemically induce their prophages and detect these via microscopy or flowcytometry. Although chemical induction is the gold standard in prophage detection, not all prophages can bedetected by it. This review gives two methods for prophage induction in heterotrophic bacterioplankton, amethod for induction of Synechococcus populations, and a method for isolating temperate phage, as well as asimple method to recognize prophage-like elements in bacterial genomes.




This work was supported by NSF grants OCE-0221763 and EF-0801593 to J. H. Paul.

ISBN 978-0-9845591-0-7, DOI 10.4319/mave.2010.978-0-9845591-0-7.30Suggested citation format: Paul, J. H., and M. Weinbauer. 2010. Detection of lysogeny in marineenvironments, p. 30–33. In S. W. Wilhelm, M. G. Weinbauer, and C. A. Suttle [eds.], Manual ofAquatic Viral Ecology. ASLO.


MANUALof


Paul and Weinbauer Lysogeny in the seas

31

In this review, we provide methods for detection oflysogeny in natural populations of marine bacteria andcyanobacteria, a method for isolating temperate phages frommarine viral concentrates, and a simple method for detectingprophage-like elements in marine bacterial genomes bioinfor-matically.

Natural populations of heterotrophic bacteria, noviral reduction

Materials• Freshly filtered (0.02 µm; Whatman Anodisc) formalde-

hyde solution (formalin; 37%);• Mitomycin C (Sigma; 1 mg/mL stock solution, dissolved

in deionized water [DI]);• Materials for epifluorescence or flow cytometry enumera-

tion of viral particles;• Electron microscopy–grade glutaraldehyde (Sigma-

Aldrich).Prophage induction—For unconcentrated seawater samples,

add 25 mL each to a control or treatment, 50-mL sterile, con-ical centrifuge tubes. If many inducing agents are to be inves-tigated, increase the number of treatment tubes accordingly.Take an additional sample (25 mL) and fix with 1% 0.02-µmfiltered formalin. For the treatment samples, add 1 µg/mL mit-omycin C (or 0.5 µg/mL in oligotrophic environments). Ifother mutagens are to be used, it is a good idea to include amitomycin C treatment as a positive control. Mutagens can beadded at any concentration desired, but this can be limited bythe solubility of the mutagen (e.g., polynuclear aromatichydrocarbons; Jiang and Paul 1996). The samples are incu-bated for 16–24 h at room temperature and fixed with either2% glutaraldehyde (for TEM), 1% formalin (epifluorescencemicroscopy), or 1% formalin/0.5% glutaraldehyde (flowcytometry [FCM]).

Samples for enumeration by epifluorescence microscopyshould be counted within 24 h of collection or stored asfrozen slides stained with SYBR Gold (Chen et al. 2001; seeDanovaro and Middelboe 2010, this volume). Count both bac-teria and viruses in control and treated samples. For inductionto have occurred, viral counts in the treatment must exceedthose in the control (i.e., be statistically different).

Calculate the % lysogenic bacteria as follows:

% lysogens = [(VDC T – VDCC)/BZ]/BDC T=0,

where VDCT is the viral direct counts (in viruses/mL) in thetreatment, VDCC is the viral direct counts in the control, BZ isthe average burst size, and BDCT=0 is the bacterial counts at theset up of the experiment (T = 0). The average burst size can bederived by TEM observation of bacterial bursts (i.e., whenviruses become visible in the cell at the end of the latentperiod; Ackermann/Heldal, this volume). We have found anaverage for our samples from the Gulf of Mexico of 30, whereastaking an average of the literature from a recent review (Wom-mack and Colwell 2000) indicates a value of 53.5 ± 48.

Natural populations of heterotrophic bacteria, viralreduction

Materials• Freshly filtered (0.02 µm) formaldehyde;• Mitomycin C (Sigma);• Materials for epifluorescence or flow cytometry enumera-

tion of viral particles;• Cartridge (30- or 100-kDa cutoff) to make virus-free water;• Filtration (0.2 µm pore size) to reduce viral abundance

(see also Weinbauer et al., this volume).Prophage induction—The rationale of the virus reduction

approach is to avoid new infection by reducing the number ofviruses and, thus, the encounter rates with hosts (Weinbauerand Suttle 1996). This can be accomplished by several meth-ods (see Weinbauer et al., this volume). Prokaryotic cells withreduced viral abundance (25–50 mL) are incubated at in situtemperature in triplicates with or without inducing agent C(see also above). Samples for enumeration of prokaryotes andviruses are taken periodically and fixed as described above.Calculation of induced viral production and the percentage ofcells containing a prophage (% lysogens) is calculated asdescribed above.

Prophage induction in marine SynechococcusMaterials• Sterile 96 well microtiter plates• Indicator host culture (i.e., Synechococcus WH7803)Prophage induction—The samples for prophage induction

are pretreated by the technique of viral reduction (Weinbauerand Suttle 1996). Each sample is filtered through a 0.2-µm fil-ter to a volume of approximately 5 mL to remove most of theambient viruses. Virus-free (0.02-µm filtered) water preparedfrom the same sample is added and the volume reduced a sec-ond time. The retentate is then returned to its original volumeby addition of virus-free seawater, divided into aliquots, andincubated with and without inducing agent. Treated samplesare amended with the inducing agent mitomycin C at a con-centration of 1 µg/mL or with the inducing agent of choice.

To enumerate the cyanophage population, the most proba-ble number (MPN) method is employed (Suttle and Chan1994). By this method, a one- to five-dilution series of theenvironmental or prophage induction treatment sample isprepared using 96-well microtiter plates (Costar, CorningInc.). A susceptible Synechococcus host is then freshly diluted1:10 and placed in each well (either Synechococcus isolateWH7803, our own isolate GM9901, or both). Control platesare prepared similarly using sterile SN media in the first col-umn of wells. Three replicate treatment and control plates areprepared from each site. The plates are incubated until goodgrowth of the host organism is evident (10–14 days). Wells arescored as positive for virus if lysis of the host organism is evi-dent as a well clearing. Viral abundance is calculated for eachplate using an MPN program (Hurley and Roscoe 1983).


32

Data analysis—Treatment and control cyanophage andSynechococcus counts are evaluated by paired t test betweensamples using Minitab statistical software. Comparison ofinduction results and environmental parameters are also per-formed using linear regression and χ2 analysis, also usingMinitab.

Isolation of temperate phages by plaque agar overlayThe isolation of temperate phages requires a cultivatable

host and a source of concentrated viruses. For example, we iso-lated a pseudotemperate phage (φHSIC) and its host from thesame bacterial/viral concentrate (Jiang et al. 1998) that weobtained in the Sand Island Channel, Oahu, HI, USA. The hostwas first isolated by standard isolation streaking on marineagar using inoculum from a microbial population concentrate.This protocol uses the conventional plaque agar overlay andlooking for turbid or haloed plaques, a hallmark of temperatephages.

Materials• Standard marine agar (1.5%) plates (i.e., Zobell 2216 or

ASWJP);• Sterile marine broth;• Sterile marine soft agar (1%), 3 mL per 15-mL tube;• Water bath;• Marine host bacterial culture in exponential growth,

20–50 mL;• Viral concentrate.Procedures—Soft agar overlay tubes are melted in boiling

water and placed in the 47°C water bath. The host bacteriumshould be growing exponentially (this can be verified by A600

measurements of about 0.4–0.6). One tube of soft agar isremoved from the water bath (the agar should have cooled to47°C), and 1.0 mL host culture and either 1.0 or 0.1 mL viralconcentrate is added. The contents of the tube is mixed wellby rolling back and forth between two hands, and the tubecontents are immediately emptied onto an agar plate. The topagar is gently spread over the agar surface by sliding the plateon the bench surface using a circular motion. The top agar isallowed to harden by not disturbing the plates for 30 min. Theplates are incubated (top agar side down) overnight to 48 h.

Temperate phage plaques will appear as turbid or cloudyplaques, whereas purely lytic phage will appear as sharplydefined, clear plaques. Plaques may appear haloed (clear areawith a larger turbid halo) and are often the result of pseudotem-perate phages. Turbid plaques can be picked and replaqued topurify the temperate phage (three replaquings are recom-mended). It may also be possible to isolate the lysogenized hostby carefully picking the turbid plaque and using isolationstreaking on marine agar plates. The putative lysogen can bechecked for harboring a prophage by mitomycin C induction.

Identification of prophages in marine bacterial genomesSeveral initiatives have as their goal sequencing of bac-

terial and archaeal genomes (www.moore.org/microgenome;

genome.jgi-psf.org/mic_home.html). The easily availablemarine microbial genomes are ideal for the bioinformatic dis-covery of putative prophage genomes. A computationalapproach to this task has been published (Phagefinder; Fouts2005). However, for this approach to be successful, thegenome must be reduced to one or very few contigs. Manygenomes are now deposited in GenBank in 10’s to more than100 contigs, precluding the Phagefinder approach. J. H. Paulhas adopted a simplified approach to prophage finding inmarine bacterial genomes that requires no sophisticated bioin-formatic software (Paul 2008). Using the NCBI website(www.ncbi.nlm.nih.gov), the genome of the organism inquestion is found using the genome search engine. Once thegenome of the microbe of interest is found, clicking on theaccession number brings up the Genome Results page, a table oflinks to various pages of information. In this table, under theFeatures column, find Protein coding and click on the link(number) of protein coding features. This opens a page of allopen reading frames (ORFs) in order in the genome. Usingyour browser’s “find in this page command” or similar searchfunction, look for phage genes, searching for the term“phage,” “terminase,” “capsid,” “portal,” or other phage term.This will locate a phage-like ORF or at least one whose puta-tive identity matches your search term. Once a phage-like ORFis found, scan 10–15 ORFs on either side of the found ORF foradditional phage-like ORFs. A typical prophage genomic sig-nature is a stretch of “hypothetical proteins” interspersed withphage proteins that extend for 30–50 kb and lack host meta-bolic genes. Many prophages begin with an integrase gene,but assigning a start and endpoint of the prophage is often dif-ficult and can be verified only by experimental procedures likePCR and cloning/sequencing of induced lysates. Once a puta-tive prophage is found, it is recommended to export thesequence to a general bioinformatics software program such asLasergene (DNAStar) or Kodon (Applied Maths). These pro-grams assist in visualizing the prophage gene arrangementand can assist in determining the termini of the prophage.

AssessmentEstimating the occurrence of lysogeny by prophage induc-

tion has been used worldwide, from the Arctic to the Antarc-tic and marine environments in between, with results rangingfrom 0 to >100% of the ambient population being lysogenized(Williamson et al. 2002). A controversial extension of theassay is to treat cultures with mitomycin C for only 30 minfollowed by cell collection by centrifugation and resuspensionin fresh growth media (Chen et al. 2006). This procedure min-imizes the general toxicity of mitomycin C and reportedly hasresulted in greater yields of temperate phages.

The virus-reduction approach has the advantage that thecontrol is likely more reliable, since new infection is largelystopped, and thus, the approach is not affected by interferenceof new infection with the mitomycin C treatment. However,there are also cons with this approach. For example, it


33

involves manipulation of samples. Another problem is thatthe virus reduction cannot be applied to all environments. Forexample, testing mitomycin C in anoxic or suboxic environ-ments without changing the oxygen concentration is feasible(within reasonable constraints) only with the nonreductionapproach.

DiscussionClearly our understanding of lysogeny is only as good as

our methods. Much of the detection of prophages in isolatesor natural populations rests on induction by mitomycin C.Clearly not all lysogens are inducible by mitomycin C. Someprophage-like particles don’t seem to be inducible by anycommon agents, but rather increase in concentration in thegrowth media as the culture reaches late stationary phase(similar to that observed for gene transfer agents [GTAs]).Unfortunately, there are no alternate methods to induction todetect prophages (save for testing other inducing agents suchas UV). There are few conserved lysogeny genes that could bedetected by amplification.

Comments and recommendationsThe approach suggested here to assess prophage induction

heterotrophic prokaryotes can be expanded to specificprokaryotic groups. Examples discussed here are Synechococcusand cyanophages. Other groups can be targeted with thedevelopment of primers for qPCR. Prophage induction assayshave also been applied to marine sediments (Mei andDanovaro 2004). For the modification of the protocols, see thecited literature.

ReferencesAckermann, H. W., and M. S. DuBow. 1987. Viruses of prokary-

otes. V. 2, General properties of bacteriophages. CRC Press.Chen, F., J. R. Lu, B. J. Binder, Y. C. Liu, and R. E. Hodson.

2001. Application of digital image analysis and flow cytom-etry to enumerate marine viruses stained with SYBR Gold.Appl. Environ. Microbiol. 67:539-545.

———, K. Wang, J. Stewart, and B. Belas. 2006. Induction ofmultiple prophages from a marine bacterium: A genomicapproach. Appl. Environ. Microbiol. 72:4995-5001

Danovaro, R., and M. Middelboe. 2010. Separation of freevirus particles from sediments in aquatic systems, p. 74-81.In S. W. Wilhelm, M. G. Weinbauer, and C. A. Suttle [eds.],Manual of Aquatic Viral Ecology. ASLO.

Edlin G., L. Lin, and R. Bitner. 1975. Lambda lysogens ofEscherichia coli reproduce more rapidly than non-lysogens.Nature 255:735-737.

Fouts, D. E. 2005. Phage_Finder: Automated identification andclassification of prophage regions in complete bacterialgenome sequences. Nucleic Acids Res. 34:5839-5851.

Hurley, M. A., and M. E. Roscoe. 1983. Automated statisticalanalysis of microbial enumeration by dilution series. J.Appl. Bacteriol. 55:159-164.

Jiang, S. C., and J. H. Paul. 1996. The abundance of lysogenicbacteria in marine microbial communities as determinedby prophage induction. Mar. Ecol. Prog. Ser. 142:27-38.

———, C. A. Kellogg, and J. H. Paul. 1998. Characterization ofmarine temperate phage-host systems isolated fromMamala Bay, Oahu, Hawaii. Appl. Environ. Microbiol.64:535-542.

Laybourn-Parry, J., W. A. Marshall, and N. J. Madan. 2007.Viral dynamics and patterns of lysogeny in saline Antarcticlakes. Polar Biol. 30:351–358.

Lisle, J. T., and J. C. Priscu. 2004. The occurrence of lysogenicbacteria and microbial aggregates in the lakes of theMcMurdo Dry Valleys, Antarctica. Microb. Ecol. 47:427-439.

Long, A., L. D. McDaniel, J. Mobberly, and J. H. Paul. 2008.Differences in prophage induction between heterotrophicand autotrophic microbial populations in the Gulf of Mex-ico. ISME J 2:132-144.

McDaniel, L., L. Houchin, S. Williamson, and J. H. Paul. 2002.Lysogeny in natural populations of marine Synechococcus.Nature 415:496.

Mei, M. L., and R. Danovaro. 2004. Virus production andlife strategies in aquatic sediments. Limnol. Oceanogr.49:459–470.

Mobberley, J. M, N. Authement, J. H. Paul, J. Koomen, and A.M. Segall. 2008. Complete genome sequence of ΦHAP-1, alinear plasmid-like temperate marine phage of Halomonasaquamarina. J. Virol. 82:6618-6630.

Paul, J. H. 2008. Prophages in marine bacteria: Dangerousmolecular time bombs or the key to survival in the seas?ISME J. 2:579-589.

Suttle, C. A., and A. M. Chan. 1994. Dynamics and distribu-tion of cyanophages and their effect on marine Synechoccusspp. Appl. Environ. Microbiol. 60:3167-3174.

Weinbauer, M. G., and C. A. Suttle. 1996. Potential signifi-cance of lysogeny to bacteriophage production and bacter-ial mortality in coastal waters of the Gulf of Mexico. Appl.Environ. Microbiol. 62:4374-4380.

———, I. Brettar, and M. Hofl. 2003. Lysogeny and virus-induced mortality of bacterioplankton in surface, deep, andanoxic marine waters. Limnol. Oceanogr. 48:1457-1465.

Williamson, S., L. McDaniel, L. Houchin, and J. H. Paul. 2002.Seasonal variation in lysogeny as depicted by prophageinduction in Tampa Bay, Florida. Appl. Environ. Microbiol.68:4307-4314.

Wommack, K. E., and R. R. Colwell. 2000. Virioplankton:Viruses in aquatic ecosystems. Microbiol. Mol. Biol. Rev.64:69-114.

34

Introduction

A DNA microarray is a microscopic collection of DNAprobes (or features) arrayed onto a solid surface. Microarraystake advantage of the highly selective nature of nucleic acid

interactions (either DNA–DNA or DNA–RNA) and are mostcommonly used for expression profiling and comparativegenomic hybridization. Although the technology is cuttingedge, it is based on the basic principles developed in theSouthern (DNA) and northern (RNA) blots (Southern 1975,Alwine et al. 1977). These techniques are still widely used butare usually limited to the study of a handful of targetsequences/genes. Microarrays, on the other hand, provide theopportunity to screen tens to hundreds of thousands of targetssimultaneously. The high cost involved in obtaining genomicinformation and constructing microarrays originally limitedtheir use to model organisms such E. coli, mouse, andDrosophila. Technological developments, however, have cre-ated an explosion in the genomic information available andhave significantly reduced the costs associated with develop-ing arrays for other organisms. This is particularly relevant tothe field of virology, where the genomes studied are smaller(and hence more accessible) and the associated arrays can bedeveloped on a limited budget. To put this into context, thereare approximately 57 million bases of sequence informationfrom complete virus genomes in the current GenBank data-base (as of 13 December 2007). This equates to the amount ofsequence generated from a single sequence run on the newlydeveloped pyrosequencing systems.

Microarrays have been a revelation in the field of molecu-lar biology. The simultaneous analysis of thousands of genes

Construction of microarrays and their application to virusanalysisMichael J. Allen1*, Bela Tiwari2, Matthias E. Futschik3, and Debbie Lindell41Plymouth Marine Laboratory, Prospect Place, Plymouth, UK2NERC Environmental Bioinformatics Centre, NERC Centre for Ecology and Hydrology, Mansfield Road, Oxford, UK3IBB—Institute for Biotechnology and Bioengineering, Centre for Molecular and Structural Biomedicine, University of Algarve,Faro, Portugal4Faculty of Biology, Technion—Israel Institute of Technology, Haifa, Israel

AbstractDNA microarray is the term used to describe a microscopic collection of DNA probes arrayed onto a solid sur-

face. Microarrays take advantage of the highly selective nature of nucleic acid interactions and are commonlyused for expression profiling, for comparative genomic hybridization, to aid genomic annotation, and for detec-tion of mutations within genomes. In this virus-focused chapter, we deal primarily with the use of microarraysfor expression analysis (the most popular usage) of host and virus systems during infection. We examine aspectsrelated to array platform choice (spotted and oligonucleotide arrays), probe and array design considerations,experimental procedures and data analysis, normalization, processing, and curation. We also provide in-depthexamples for the study of viral transcriptome analysis for both spotted long oligonucleotide (coccolithoviruses)and Affymetrix GeneChip (cyanophage) arrays.




M. J. Allen is supported by grants from Natural Environment researchCouncil (NERC) through the Environmental Genomics program(NE/A509332/1 and NE/D001455/1) and Oceans 2025. B. Tiwari is sup-ported by the NERC Environmental Bioinformatics Centre. M. E. Futschikis supported by the Ciência 2007 initiative of the Fundação para aCiência e a Tecnologia, Portugal. D. Lindell is supported by the MorashaProgram of the Israel Science Foundation (grant 1504/06) and a MarieCurie International ReIntegration Grant (MARCYV) within the sixthEuropean Community Framework Programme and is a Shillman Fellow.There are no declared conflicts of interest.

ISBN 978-0-9845591-0-7, DOI 10.4319/mave.2010.978-0-9845591-0-7.34Suggested citation format: Allen, M. J., B. Tiwari, M. E. Futschik, and D. Lindell. 2010.Construction of microarrays and their application to virus analysis, p. 34-56. In S. W. Wilhelm,M. G. Weinbauer, and C. A. Suttle [eds.], Manual of Aquatic Viral Ecology. ASLO.


MANUALof


Allen et al. Construction of microarrays and virus analysis

35

has provided a wealth of information that we are only nowcoming to terms with. Microarrays have allowed the fine-tuned mapping of transcriptional pathways and cascades froma variety of organisms responding to a plethora of environ-mental stimuli such as physical, chemical, and biologicalstressors. Yet microarrays should be thought of as a techniqueto be used in combination with others. Typically, a large groupof genes is analyzed to identify a small group of genesinvolved in a particular process, response, or even phenotype.Of course, assaying huge numbers of genes simultaneouslywill never be as accurate as assaying each gene individually,but microarrays offer a relatively cheap and quick way to assaygenes on a genomewide scale, allowing future focus on thegenes identified; i.e., microarrays can be thought of as amolecular compass to guide research in the right direction.Thus, when performing genomewide transcriptional profiling,it is essential that techniques such as quantitative real-timePCR and northern blotting are used to confirm the findingsfrom a subset of genes from microarray experiments. In thecase of comparative genomic hybridization, an array can pin-point where variation occurs in a genome but cannot deter-mine what the basis of the variation is. In this case, it is impor-tant to verify the results through direct sequencing.

Despite the insight that microarrays can help provide, themarine and aquatic sciences have been relatively slow toembrace this technology. In the past, marine-focused microar-ray work has been limited to a handful of isolated laboratoriesspread around the globe. We hope this review will promotethe use of microarrays to answer important aquatic sciencequestions, in particular in the field of aquatic virology. Thepotential of microarrays for use in studying virus biology isenormous (Allen and Wilson 2008). Relatively small virusgenomes can have profound and dramatic effects on hostglobal gene expression, making microarrays a powerful tool ina molecular biologist’s armory.

In this chapter, we deal primarily with the use of microar-rays for expression analysis of host and virus during infection.We give in-depth description of the most commonly usedplatforms and methodology and discuss issues that must beconsidered when designing and using arrays for aquatic virusstudies. We do not attempt to give a comprehensive overviewof all available technologies and options, but invite the readerto investigate other options using the issues discussed hereinas a guide (Karsten and Geshwind 2002, Liu 2007, Page et al.2007, Sipe and Saha 2007, Bier et al. 2008, Coppee 2008, Gre-sham et al. 2008, Simon 2008, Dufva 2009a, 2009b). Microar-rays are ideal for determining the transcriptional program ofviruses during infection (as well as the host’s transcriptionalresponse to this infection) because of their ability to receiveinformation on the transcription of all genes from bothgenomes from the same sample. This requires that probes forboth host and viral genomes are designed on the one array. Inthese experiments, where temporal dynamics of transcriptionduring infection are being investigated, it is important to

ensure synchronous infection. The time scale for such experi-ments is generally the length of the latent period. Wheninvestigating host responses, it is also important to ensure thatthe vast majority of cells are infected so that transcriptionalchanges in the infected cells are not masked by expression lev-els in uninfected cells. The contamination status of cultures isalso worth considering; it is best to work with axenic culturesif possible. With environmentally relevant host–virus systems,actual host infection levels at a particular multiplicity of infec-tion (MOI) do not always match those based on theoreticalconsiderations from Poisson distribution. Therefore, a numberof preliminary experiments are essential before embarking ona microarray experiment beyond determining infectionparameters (such as the length of the lytic cycle and latentperiod). These experiments include determining the condi-tions for maximizing the percent of infected cells and theactual degree of infection, as well as conditions for synchro-nous infection by assessing various cell and virus concentra-tions in addition to the appropriate MOI. It is also very usefulto determine the timing of different stages of the infectioncycle to which transcriptional data can be compared. Forexample, viral genome replication, production of structuralproteins, host genome degradation, and changes in host cellmorphology can all help place the transcriptional informationin the wider context of the infection process.

Materials and proceduresMicroarray platforms—At every stage of microarray design

and construction, there are various options that can be takendepending on the available budget, the number of probes onthe array, and the local infrastructure and facilities available(Fig. 1). It is crucial that before microarray design begins, thescientific questions of interest are considered and defined, sothat a microarray fit for the purpose is constructed. For exam-ple: Can the virus system be accurately studied on a microar-ray independent of the host system? Are there any genes com-mon to both virus and host genomes? How much virusmessage will be present in relation to host message at variousstages of infection? Will any amplification of message beneeded? The answers to these sorts of questions can have aprofound impact on the nature of the microarray developed.Other questions that may affect your microarray designinclude the following: How many microarray experiments willbe run? Will the microarray be used by just one researchgroup, or will it be made available to other interested parties?What local microarray infrastructure is available?

Here we focus primarily on two array platforms: spottedarrays and Affymetrix GeneChip arrays. We also occasionallycomment on Agilent arrays when relevant. Other array typesexist, but these three are common platforms that provide agood overview of the different approaches currentlyemployed. Each array platform offers its own particular advan-tages and disadvantages, which we discuss. We start offdescribing high-density custom oligonucleotide systems and


36

then move on to the custom spotted microarrays. One issuewe will not touch on in this review is specific costs, which arehighly subjective and prone to changes in the future; here, wediscuss only the current costs of each system relative to otheroptions presently available. With respect to cost calculations,the stripping and reuse of previously used microarrays is tech-nically feasible for many types of microarray system, but westrongly advise against this practice as it can introduce uncon-trollable variability.

High-density oligonucleotide microarrays—High-density oligo -nucleotide microarray systems (such as the Affymetrix and Agi-lent systems) usually offer the best reliability, reproducibility,and coverage. Many high-density microarrays are constructedusing a process known as photolithography (Affymetrix andNimblegen), whereby light is used to stimulate in situ DNA syn-thesis in defined positions; others use inkjet technology (suchas Agilent) (Table 1). In both these cases, single-strandedoligonucleotides are sequentially synthesized base by base,directly on the solid surface of the array. Whereas Affymetrixtechnology uses a number of short 25-mer probes per gene (gen-erally 8–11), Agilent arrays consist of a single 60-mer probe foreach gene. The manufacturing process is an extremely robustprocedure with no noticeable differences between arrays. Apowerful application of high-density microarrays is to producewhat is referred to as a tiling array. Tiling arrays are designedindependently of annotation data, cover entire stretches ofgenomic sequence (usually total genomes) in an unbiased man-

ner (e.g., 25-mer probes designed with a space of approximately50 bases in between, along the length of the whole genome irre-spective of annotation), and allow the identification of noveltranscribed sequences (often unannotated) as well as regulatoryelements. By determining which probes generate positive sig-nals and their intensity relative to neighboring probes, regionsof the genome that are transcribed (i.e., the genes) or are regu-latory can be easily identified. Tiling arrays, developed by com-panies including Affymetrix, Nimblegen, and Agilent, areincredibly powerful but are commonly restricted to modelorganisms for which there is a large commercial market.

Affymetrix expression arrays can produce highly repro-ducible data. These arrays are designed and manufactured byAffymetrix based on empirical but proprietary informationand therefore are the easiest for the researcher to “design,”although also the most costly. After RNA is extracted, it can belabeled by the researcher or, for a cost, at an Affymetrix arrayfacility. Hybridization of the arrays is carried out at a special-ized array facility, generally by facility personnel. This makesthis procedure relatively simple, especially for researchers notso familiar with RNA work, but requires that the researcherfind a reliable facility. The greatest disadvantage of thesearrays is the high cost incurred for the custom design neces-sary for nonmodel systems. Cost per array is also quite high(with a minimum order of 90 arrays).

Agilent arrays—both probes and array layout—can bedesigned for free (using their Web program at earray.chem.agilent.com/earray) or can be designed by Agilent for areasonable fee. A major advantage of the Agilent platform isthe flexibility for probe and array design it provides, being fea-sible to order any number of arrays at a time (even a singlearray) and redesign probes for subsequent arrays. RNA labelingcan be carried out by the researcher or at an array facility;however, hybridization and scanning are generally carried outat an array facility, again making the process quite simple forthe researcher. The biggest disadvantage of the Agilent arraysis the cost per array, which is considerably higher than for theother platforms. It is possible, however, to hybridize multiplesamples in different compartments on one slide if the numberof probes is small enough (for example, if investigating theexpression dynamics of the viral genome alone), making theprice more reasonable per sample.

Thus, for high-density oligonucleotide arrays, Affymetrix isthe platform of choice for systems when a large number ofexperiments will be carried out, although the cost for arraydesign is quite high. Agilent is the platform of choice whenhigh design flexibility is desired and few samples will be inves-tigated, although the cost per array is quite high.

Custom spotted microarrays—The costs associated withdeveloping high-density microarrays can make them finan-cially prohibitive. The development of a custom spotted arrayis an economical alternative, although there is a price to payin reduced array reproducibility. A spotted microarray is alsothe platform of choice for large-volume experiments when

Fig. 1. Microarray work flow.


37

high design flexibility is required. The researcher has theadded advantage of having complete control over array designand total flexibility in its use. For spotted microarrays, themost common method is to use a glass slide as the basis forprinting. Glass slide arrays offer researchers great control: theycan generate their own labeled samples, perform their ownhybridizations, and scan using their own equipment. Depend-ing on the infrastructure, once microarrays have been fabri-cated, all that is needed is a microarray scanner and somebasic laboratory equipment. Probes can be immobilized ontoglass slides by a variety of techniques. Initially, this was doneby physical contact between robotically controlled pins andthe slide surface. It is now more common to use a techniqueknown as piezoelectric printing, which is akin to ink-jet print-ing and provides greater control over spotting quality andquantity. As the print head moves across the array, electricalstimulation causes the DNA to be delivered onto the surfacevia tiny jets in a noncontact process. The process of “slideprinting” is time consuming, technically demanding, andrequires expensive robotic machinery. Often this is beyondthe budget of most research laboratories; however, microarrayprinting facilities are now commonplace and offer a cheapand reliable method of fabricating microarrays.

Perhaps the greatest advantage of glass slide microarraysover their high-density counterparts is the flexibility theyoffer in the nature of the material printed on the slide. Manydifferent types of material, including PCR products, plasmidlibraries (cDNA/Expressed Sequence Tag [EST], plasmid, shot-gun [randomly fragmented DNA]), or presynthesized oligonu-cleotides have been printed on glass slide microarrays depend-ing on what was available and most cost-effective for a givenproject. We strongly recommend the use of long, presynthe-sized oligonucleotides when sequence information is avail-able. Presynthesized oligonucleotides provide high specificityand allow design of probes with similar hybridization charac-teristics. A length between 50 and 70 bases generally providesa good balance between sensitivity, specificity, cost, and con-sideration of the decreased coupling efficiency during synthe-sis with increasing probe length. As the price of generatinglonger probes has decreased over recent years, we recommendprobes of approximately 70 base pairs (although in theory

they can be synthesized to any size required). Long oligonu-cleotide probes have the added advantages of not needing cer-tain quality checks required for other material types, for exam-ple, amplifying and verifying the quality of PCR fragments, aswell as avoiding the problems associated with hybridizing toprobes of different length and different annealing tempera-tures, which arise when using PCR and plasmid probes.

For those who do not have access to preexisting sequencedata, probes can be generated from plasmid libraries. Here,researchers can hone in on a small number of unknownprobes using a microarray, and sequence only relevant plas-mids of direct interest to them toward the end of their exper-imental work. This used to be a popular use of microarrays;however, the price of sequencing and of generating syntheticoligonucleotides have dropped considerably in recent years.Thus, for most purposes, it is preferable to generate, sequence,and design long oligonucleotides, rather than generate physi-cal materials such as plasmid libraries or PCR products forspotting onto arrays. High-throughput sequencing methodshave already been used to provide sequence for use in design-ing transcriptomic arrays (Vera et al. 2008), and this is likelyto become a popular avenue in the future.

For long oligonucleotide arrays, researchers must choosehow much of the process they want to undertake locally ver-sus the cost in time and money. Designing long oligonu-cleotide probes is within the abilities of most researchers,thanks to the ready availability of probe design software, bothcommercial and free. However, one should not underestimatethe time required to design an array layout and to qualitycheck this array before any experimentation can begin. Formany researchers, the services offered by a company, alongwith a quality guarantee for what they provide and the timescale within which they will provide it, may make the initiallylarger financial outlay worth it in the end.

When engaging a company, the normal rules stand: Ensurethat you lay out clearly what you need and expect, and thatyou read the terms and conditions of their service in detail.Many companies and facilities have much experience workingwith model organisms and the types of chip designs onemight desire for studying such organisms. The needs of theviral community can be somewhat different, however; for

Table 1. Commercial suppliers of high-density microarrays.

Company Web site Description

Affymetrix http://www.affymetrix.com High-density chips can be designed with up to 1.3 million features,

with 25mer probes. Features are 11 µm in size.

Agilent www.chem.agilent.com High-density arrays with up to 243,504 features per array printed on standard

glass slides. Features are 65 µm in size. Standard probe length is a 60mer,

but any length between 25 and 60 bases possible.

CombiMatrix www.combimatrix.com Array chips featuring 12,000 35-40mer probes, feature size 44 µm.

Nimblegen www.nimblegen.com High-density arrays with up to 2.1 million features, 50–75mer probes printed

on standard glass slides. Features are 13 µm in size.


38

example, the requirement to spot multiple, unrelatedgenomes (e.g., virus and host) on a single chip, where thecharacteristics of these genomes can be quite different. Willthe company print probes of different lengths? If not, how dothey plan to deal with the likelihood of different melting tem-perature profiles among genes in each organism? If you aredoing a diversity study, or screening for new genotypes, willthey help you design appropriate probes? How many probesdo you anticipate spotting? What is the minimum number ofarrays you can have made? (For example, study the implica-tions on choice if you are only going to run a few experi-ments/hybridizations with this design.) Could/should thecompany do the hybridizations for you?

If you are designing your own long oligonucleotide spottedarray, then you must take into account the physical character-istics of the probes to include on the array, what types ofsequences to represent, how many probes per gene (or region)to include, whether replicate spots will be printed, and whatcontrols to print on the array. Some of these issues are com-mon to both high-density and spotted arrays, although manyare issues specific to just spotted arrays. Each of these topics iscovered briefly below, as well as a brief comment on availablemicroarray probe design software. In general, unless youalready have the facilities locally, we recommend that youfind a microarray facility to work with and undertake adetailed conversation with them about the needs of yourexperimental system.

Array design considerations—Regardless of which array plat-form will be used, it is imperative that the researcher decideshow the RNA will be labeled, as this determines whether thearray should contain probes that are identical to or complemen-tary to mRNA—termed a sense or antisense array, respectively,by Affymetrix. For example, a protocol that carries out reversetranscription to produce cDNA requires an antisense array,whereas some protocols that include an amplification step willproduce DNA that is identical (sense) to the original RNA.

Probes bound to their targets should have approximatelythe same melting temperature (typically 50–60°C) across thearray. Some oligonucleotide design software (see section below)will allow you to design probes with a range of lengths. Thiscan be useful if more than one organism (e.g., host and virus)with different nucleotide characteristics (e.g., GC content) willbe represented on the array. If you decide to design probes ofdifferent lengths, make sure that the company you are order-ing the oligonucleotides from will manufacture them.

Probes need to be specific to the target of interest and sen-sitive enough to detect low levels of that target. Potential forsecondary structure in the probe or the target can affect sensi-tivity; this is particularly relevant to the longer probes on spot-ted arrays. Some oligonucleotide design software include cal-culations for this potential. The specificity of a probe for itstarget will depend on how many mismatches there arebetween them, and also the location and arrangement ofthose mismatches along the probe sequence (Letowski et al.

2004). Gene-specific probes should have little or no sequencesimilarity to nontarget genes that may be present in the sam-ple. Comprehensive studies on the effect of probe–target char-acteristics have provided tables of empirical results for essen-tial design criteria for gene-specific and group-specific probes(He et al. 2005, Karaman et al. 2005). Note that some softwareuse free energy in place of sequence identity as a measure ofoligonucleotide specificity.

For bacterial arrays, where total RNA (i.e., mRNA, rRNA,and tRNA) is labeled, it is important to ensure that none ofthe probes are similar to ribosomal and transfer RNAsequences, as even very low labeling efficiency of these abun-dant RNAs is likely to mask mRNA levels. In addition, probesfor bacteria and bacterial viruses should not be 3′ biased, asrandom hexamers, rather than polyT priming, will be used inthe majority of protocols for making cDNA. Therefore, if asingle probe is designed per gene it should be positionedtoward the 5′ end of genes. If multiple probes are designed, asin Affymetrix arrays, we suggest these be spread across thegene, although they could also be designed toward the 5′ endof the gene.

If there is sufficient room on the array, designing probes inintergenic regions will enable identification of small unanno-tated genes that may have escaped annotation, as well as smallnoncoding RNAs that are often found in these regions(Steglich et al. 2008). The researcher should also considerwhether probes for the detection of antisense RNAs willenhance the utility of the array. Perhaps rather than includingprobes that are antisense to all mRNAs, preliminary experi-ments for the detection of mRNAs, noncoding RNAs, and anti-sense RNAs could be carried out by Solexa- or 454-likesequencing, which would greatly inform on probe design (O.Wurtzel and R. Sorek, pers. comm.). Alternatively, as men-tioned above, a tiling array could be designed across thegenome on both strands.

High-density Affymetrix arrays can contain many probeson each array, making it worth considering including multiplegenomes per array. However, if the genomes of two of theseorganisms will be investigated at the same time, as for hostand viral genomes, probes with little cross-hybridizationbetween the genomes must be designed (see “Probe design”below) and can be empirically tested with DNA from eachorganism. This is particularly important, as today we knowthat viral genomes often include host-like genes (Hughes andFriedman 2005, Moreira and Brochier-Armanet 2008, Monieret al. 2009). If this design is not feasible, and indeed wheneverpotential cross-hybridizing probes are being used, it is best toconfirm the microarray results or carry out single-gene analy-sis from the outset, with a method capable of differentiatingbetween host and viral copies of the genes—for example,using quantitative RT-PCR. Conversely, if the multiplegenomes on the array will be investigated independently, i.e.,separate arrays for each sample type, then there is no need toensure that the probes do not cross-hybridize.


39

The types of probe sequences required on a microarraydesigned to study biodiversity will be different from thoseused on an array to study differential expression. Combina-tions of probes might be used to detect particular species orthe presence of novel genotypes in a sample (Wang et al. 2002,2003a; Rich et al. 2008). A recent review outlines microarraystudies in microbial ecology (Gentry et al. 2006), including anoverview of the types of probes included on arrays used in dif-ferent types of studies. From this point in this review, weassume that the question of interest requires detection of onlyunique genes.

Probe design—Many computer programs are available to aidin the design of long oligonucleotide probes for microarrays.These programs usually determine specificity and the poten-tial for cross-hybridization potential of all probes designed.Some of these programs are commercial, but many are opensource and freely available for academic use (Li et al. 2002,Herold and Rasooly 2003, Nielsen et al. 2003, Rouillard et al.2003, Wang and Seed 2003, Chou et al. 2004, Reymond et al.2004, Chung et al. 2005, Li et al. 2005, Nordberg 2005, Sten-berg et al. 2005, Schretter and Milinkovitch 2006). Whenlooking for a software package, important considerationsinclude on what basis it chooses probes, what type of inputdata it requires, what format the output data will be in, howeasy it is to install, and how easy it is to use. Ideally, therewill be empirical evidence available on how well the softwarehas worked in designing probes for microarrays already inuse. Another key consideration is whether the software isinstalled and runs on a local machine or on a remotemachine (e.g., entering your data via a Web site). Runningprograms locally has the benefit that data are secure and pri-vate, whereas running programs remotely depends on thegood will of someone else, and in essence involves passingyour sequence set onto someone else’s machine for process-ing. On the upside, the machine at the other end may bemore powerful and therefore quicker (or not!) than what isavailable locally, and the maintenance and installation ofthe software is someone else’s responsibility. You also need toensure that you enter your sequence data in the appropriatedirection (see probe direction section above). More detailedoutlines of software considerations, as well as desirable probecharacteristics, can be found in specific reviews of the topic(Millard and Tiwari 2009). As a good starting point, theauthors have found that the Picky and Yoda software pack-ages are both very user friendly. They are available from complex.gdcb.iastate.edu/download/Picky/ index. html andpathport.vbi.vt.edu/YODA, respectively. In addition, manyoligonucleotide software design packages are listed, and somereviewed, at nebc.nox.ac.uk/tools/bioinformatics-docs/other-bioinf/oligo-nucleotide-design.

Controls—Control spots are vital in the assessment of qual-ity, sensitivity, and reliability of microarray experiments. Dif-ferent types of controls can be used to assess various qualityaspects. We strongly recommend the inclusion of spike-in

labeling controls—probes for RNA that will be spiked intothe sample at defined concentrations, before the labelingprocedure. They can be used to estimate the minimumamount of transcript as well as the minimum change in tran-script abundance detectable by the array technology beingused. It is important to include a sufficient number of spike-in controls so that they can be added at varying concentra-tions to cover the signal intensity range of the experimentaltranscripts. Control probes should be placed on the arraysuch that they appear across the spatial dimensions of thearray. If sufficient controls are included, they can be used fordata normalization (covered below). Affymetrix arraysinclude probes for detecting spike-in hybridization controlsas a default, but extra probes for labeling need to berequested. Commercial control probes and their partner tar-gets are available from companies such as Stratagene. Obvi-ously, all spike-in controls should be for organisms otherthan those being investigated and should show no cross-hybridization to the experimental genomes. When workingwith unusual organisms, especially those for which nosequence is available in the public databases, it is worthchecking with the company involved to ensure that theircontrols do not contain sequences similar to anything inyour samples. If you don’t have sequences for your own sam-ples, or you cannot get information on control sequencesfrom the company, you may need to run test hybridizationswithout the labeled control targets added to ensure thatthere is no cross-hybridization. The inclusion of appropriatecontrols at the fabrication stage is particularly pertinent forvirus-focused microarrays, where virus message may or maynot be present at all in the early stages of the experiment(i.e., the uninfected state of a transcriptional profiling exper-iment). This is discussed further in the data normalizationsection below. It is also useful to include empty spots, orprobes for which no spike-ins will be added, which can pro-vide an indication of nonspecific background signal. Forspotted arrays, print-buffer spots are useful to check that nocarryover effects occur during the printing of spotted arrays,as such artifacts may especially compromise the mea-surement of weakly expressed genes.

Genome annotation is often a process in flux, changing asbetter bioinformatics tools are developed and more experi-mental information becomes available. It is therefore useful tobe able to change the annotation of genes on the array. It ispossible for bioinformaticians to re-annotate arrays them-selves for all platforms, but it is not always trivial. We havefound that Affymetrix will support the need to change theappropriate files free of charge for a while, but will eventuallystart charging for this service; we suggest that the number oftimes Affymetrix will carry out this process be negotiated withthem as part of the design contract. Importantly, once fileswith the updated annotation are included, old array data canbe reanalyzed in light of the newly annotated protein codinggenes, ncRNAs, or regulatory elements.


40

Microarray experimental design—As in any scientificendeavor, appropriate experimental design is crucial. Manyaspects of planning microarray experiments are the same asfor other types of experiment requiring statistical analysis ofthe resulting data. For those who are not comfortable with sta-tistics, there is a solution: find a collaborator who is. This isnot a glib statement, it is a serious recommendation. Planningyour experiment with someone who is familiar with microar-ray statistics and experimental design can mean the differencebetween generating data that allows you to address your ques-tions of interest or generating data providing little or no scopefor meaningful analysis.

The design will include defining the type and number ofsamples needed, certain aspects of their preparation and repli-cation, the number of slides to be hybridized, and what sam-ples will be hybridized to the same slides in the case of two-color experiments. Too often, researchers perform an entireexperiment and then provide a block of data to a statisticianto be analyzed, having unknowingly introduced bias or with-out including appropriate replicates. When this occurs, it canmake analysis difficult or impossible, meaning that all yourtime, samples, and money have just been wasted. Below webriefly outline some basic considerations for experimentaldesign of microarray experiments.

Replicates: Different samples of the same type (e.g., samplesof the organism exposed to the same treatment) are referred toas biological replicates. If you measure the same sample twice(e.g., take the same extract and put it on two microarrayslides), this is a technical replicate. Common questions thatarise when planning a microarray experiment are how manybiological and technical replicates are needed? If your aim isto compare gene expression levels between treatments or con-ditions, then measurements from biological replicates areessential. We recommend that the minimum number of bio-logical replicates considered for most situations is three, withfour to six a more desirable number for spotted arrays. Notethat if variability is high, even this number of biological repli-cates will be inadequate for certain types of analyses. Be waryof limiting the number of samples or experimental replicatestoo much based purely on cost or ease of obtaining samples,as this may lead to the inability to derive useful informationfrom the experiment (and the money would have been bettersaved than spent on the microarrays).

Technical slide replicates inform on the signal variation dueto “uninteresting” differences such as different slides, differenthybridization chambers, different researchers carrying outprotocols, etc. Technical replicates allow for certain types ofquality control checks as well as providing greater precisionfor a given measurement. To be able to glean meaningful datafrom such replicates, it is necessary to use sophisticated statis-tical software to set up statistical models that take technicalreplication into account.

A different sort of technical replication common inmicroarray systems is the inclusion of replicate probes on the

arrays. This is especially common with arrays where there areonly a small number of unique probes, as is the case in manyviral–host arrays. Replicate probe spots provide informationabout signal variation related to position on the array. Optionsfor the analysis of replicate spots include taking a simple arith-metic average of the measure for these replicate spots or usingan error term in the statistical model for each gene to accountfor the variation between spots (Smyth et al. 2005). We rec-ommend the latter. It is important to choose software capableof taking replicate spots into consideration and knowing howsuch replicate spots are treated. For example, some softwareassumes that spots with the same name are the same probe,whereas other software assumes that spot replicates are equi-distant across the array.

In broad terms, we recommend as many true biologicalreplicates as possible, with technical replicate spots withinarrays providing useful information about technical variationif the data are handled appropriately. Although technical slidereplicates are important to run during the initial setting up ofyour array-related protocols, they can be of limited use lateron. We generally do not use technical replicate slides once thesystem has been set up, as biological variability is usuallymuch greater than the technical variation between mea-surements, providing little additional knowledge of the bio-logical system. Therefore we suggest adding more biologicalreplication rather than including technical replicates.

Culture infection concerns: For those studying viral sys-tems grown in lab culture, the division between biological andtechnical replicates, and single and pooled samples, is oftennot clear. Every flask of a host–virus system is a pool of organ-isms. Cultures in two flasks from the same exact source sam-ple are similar to a technical replicate: measuring gene expres-sion in these two samples is likely to give you an idea of howtechnical differences (slight temperature or light variation,flask conditions) affect gene expression in each pool ofvirus–host in each flask. We therefore suggest that cultures begrown separately over many generations to get an indicationof the biological variation possible in this host–virus system.Where possible, we also suggest carrying out biological repli-cation at different times (for example 1 month apart) to fur-ther control for differences that may relate to the particularrun of an experiment. Another consideration for virus infec-tion experiments when host gene expression is being investi-gated is carrying out paired experiments where each replicateculture is divided into two subcultures. One of the subculturesis infected with the virus and the second serves as the unin-fected paired control. This can help reduce potentially irrele-vant variability related to culture physiology that is notrelated to the infection process. It is important to note thatstandard statistical tests commonly used in microarray analy-sis often include the assumption of independent, identicallydistributed measurement errors. This means that each mea-surement from a sample you consider a biological replicateshould be an independent measurement from truly different


41

cultures and not from quasitechnical replicates, such as flasksgrown from the same original culture.

Pooling: Pooling biological samples is often considered as away to keep costs down (as a given number of samples can bemeasured using fewer arrays) or to reduce “noisy” variation.However, pooling biological replicates allows you to measureonly a mean value for those samples. This leads to the lack ofa measure of the biological variability in the system—a mea-sure that is essential to determine the significance of changesin expression of one condition relative to another. If you donot know the inherent variability in your expression levels fora given treatment, you cannot determine when expression issignificantly different or within the level of normal variability.In general, we advise against pooling, as it is likely to maskresults of interest, while usually not providing any real bene-fit. If you decide to pool samples anyway, it is vital to take thepooling into account when interpreting your experimentalresults. As a rule, pool samples only if the experiment is purelyexploratory, for example, providing preliminary data, and ifyou intend to screen all candidate genes in each biologicalreplicate by other techniques such as quantitative RT-PCR.

Variability and confounding: Experiments should bedesigned in a manner that ensures that sources of technicalvariability are not aligned, or confounded, with treatmenttypes. Examples of confounding include the use of arrays fromdifferent batches for different conditions and the use of spot-ted arrays printed early in a print run for one condition andthose from late in the run for another condition. To circum-vent these types of problems, array use should be randomized.The order of slide use can be randomized by generating a setof random numbers. An excellent place to get random num-bers for this purpose is www.random.org/sequences.

Wherever possible, one researcher should carry out a par-ticular task for all samples. If this is not feasible, it is impor-tant to build the design so that each researcher looks afterequal numbers of samples from each condition—preferablyfor paired infected and uninfected samples where appropriate.In a similar manner, it is important to ensure that hybridiza-tion of microarrays from one condition are not hybridized onone day and those from the other on a different day. This wayyou will not confound the effect of the researcher or the dayon which they were handled with the condition itself.

Microarray hybridization designs—A variety of experimentalhybridization designs are commonly referred to in themicroarray literature (Kerr and Churchill 2001). The key dis-tinction between the design types is whether they allow director indirect comparison of samples. Direct comparisons refer tothe hybridization of two samples to a single slide, providing aratio indicating the relative expression levels of the two sam-ples. Indirect comparisons refer to taking measurements fromdifferent slides and comparing them. Single-color designs, asused with the Affymetrix platform, are a type of indirectdesign. One sample is applied to each slide, and comparisonstake place by considering biological replicate measurements of

treated versus untreated samples. As such, one-color designsare relatively straightforward to devise and analyze. Two-colordesigns are more complicated. Thus, the rest of this sectionprovides an overview of two-color microarray designs specificto spotted microarrays.

A common indirect design used with two-color microarrayexperiments is the reference design. In a reference designexperiment, each sample is hybridized onto an array alongwith a reference sample (Fig. 2). This reference sample shouldideally have a hybridization signal for all genes of interest dur-ing the course of your experiment, as you will be working withexpression ratios. A common and generally effective choicefor a reference sample is a pooled sample made from aliquotsfrom each of the samples in the experiment (Kerr et al. 2007).Genomic DNA can also be used as a reference, although a dif-ferent labeling strategy (i.e., labeling DNA, not RNA) must beused to generate such a reference sample. Despite this, it pro-vides advantages in that every gene in your organism is repre-sented at the same level above background, and it enablescomparisons across experiments. In reference design experi-ments, the signal ratio is made up of your sample signal com-pared to a reference signal, and the ratios from different slidesare then compared to one another. The indirect nature of ref-erence design experiments makes them somewhat less effi-cient than direct designs, and they require more arrays (seebelow). However, they are relatively more straightforward andflexible than other designs.

An alternative approach for two-color arrays is directdesigns (Fig. 2). These can provide a more accurate estimate ofdifferences between samples, as each sample is hybridized tothe same slide as the sample it is being compared with. Muchof the technical variation is cancelled out when these ratios aretaken. In this model, uninfected samples could be compareddirectly with infected samples taken from the same time point.

A common extension to direct designs is loop designs(sometimes extended to interwoven loop designs) (Kerr andChurchill 2001, Kerr 2003) (Fig. 3). Here again, two samples ofinterest are hybridized to each slide; however, these designsinvolve a combination of direct and indirect comparisons. Bycomparing two conditions through a chain of other condi-tions, samples can be compared directly with other sampleswith a multiple-pairwise methodology (Pirooznia et al. 2008).These designs have the potential to be more efficient than astandard reference design and have stronger statistical power,but are considerably more complex (see Fig. 3). We recommendILOOP, a freely available Web-based program, as a useful tool infinding optimal loop designs for two-color microarray experi-ments (Pirooznia et al. 2008). Be aware, though, that not allanalysis software is capable of handling data from loop designs.

For virus–host studies using two-color arrays where eithervirus or both host and virus gene expression is to be investi-gated, we generally prefer a reference design. This is becauseuninfected samples have no real signal from the virus probes,which makes taking a direct ratio between uninfected and


42

infected samples very problematic, as would be done in adirect or loop design. If, however, only host gene expression isof interest, direct or loop designs can work well. In a two-colorexperiment, each slide will have two samples hybridized to it.One will be labeled with Cy3 dye, and one with Cy5 dye. Adye swap involves labeling samples from a particular condi-tion with the Cy3 on one slide, and with Cy5 on anotherslide. The aim of dye swaps are to avoid identifying genes aspotentially interesting when in fact a relatively strong or weaksignal is due to a bias associated with the dye a sample hasbeen labeled with. A dye swap can be carried out with techni-cal or biological replicates; however, for reasons discussedabove, we recommend biological dye swaps if direct or loopdesigns will be used. For example, a sample pair from thetreated and untreated conditions are labeled with Cy3 and

Cy5, respectively, whereas samples from another biologicalreplicate pair (from the same treated and untreated condi-tions) are labeled instead with Cy5 and Cy3, respectively. Forreference designs, spot ratios biased by the dye used can beavoided by labeling the reference sample with a particular dyeand the experimental samples with the other dye for thewhole experiment.

Sample labeling and microarray hybridization—Once you havedesigned your microarray, determined the experimental plan,and have extracted nucleic acid samples in hand (see “Casestudies” in “Assessment” for in-depth examples), you areready to proceed with the sample labeling and array hybridiza-tion. The method used for labeling your samples depends onthe chemical nature of your sample (i.e., DNA or RNA) and theamount of sample available (low amounts of starting materialmay require an amplification of message step). Labeling ofmRNA requires the use of reverse transcriptase, whereas label-ing DNA requires the use of the Klenow fragment of DNApolymerase (i.e., minus the proofreading activity). Nucleotidemixes for these labeling reactions may require some optimiza-tion depending on the GC content of the systems under study.One needs to decide whether to use a direct or indirect label-ing method (for spotted arrays); random, specific, or oligo dTprimers; an amplification of message step. Some commercialkits using derivatives of the Eberwine (1996) method claim toquantitatively amplify message by up to a millionfold, but arevery expensive. Direct labeling is the cheapest method for

Fig. 2. Examples of reference and direct designs. Samples connected byan arrow are hybridized to the same array. Dye color can be inferred bythe direction of the arrow (arrow head, Cy5 dye; tail, Cy3 dye). Here, wehave also represented this by coloring the head of the arrow in red andthe tail of the array in green. The reference design is shown with twoexperimental sample types (orange and blue), one reference sample (yel-low), three biological replicates (shades of orange and blue), and no dyeswaps (arrow directions) with a total of six slides. The direct design isshown with two experimental sample types (blue and purple), four bio-logical replicates (shades of blue and purple), two pairs of dye swaps(arrow directions) with a total of four slides.

Fig. 3. Example of an interwoven loop design shown with three experi-mental sample types (blue, orange, and purple), three biological repli-cates (shades of blue, orange, and purple), each sample hybridized thesame number of times and with both Cy3 and Cy5 (also the same num-ber of times), with a total of 27 slides.


43

spotted arrays and involves directly incorporating a florescentlabel conjugated to a nucleotide during polymerization of thecomplementary strand. The increased size of these syntheticnucleotides causes an unavoidable decrease in efficiency inthe labeling reaction. Alternatively, indirect labeling involvesthe incorporation of an aminoallyl-modified nucleotide in theinitial step followed by a second step involving the chemicaladdition of a fluorescent dye ester to the aminoallyl-modifiednucleotide. The structural similarity of aminoallyl nucleotidesto normal nucleotides circumvents the lower labeling effi-ciency related to direct labeling and generally gives strongersignal strength. We therefore recommend indirect labelingdespite the fact that it is a more expensive method.

The precise hybridization conditions will also need to beoptimized for each microarray. The optimum temperature forhybridization should be empirically tested with the tempera-ture calculated by probe design software serving as an initialguide. Decreasing or increasing hybridization temperature(leading to a respective increase and decrease in cross-hybridization potential) can have a profound impact on thequality of the data. Volumes and sample buffer (typically 3×SSC, 0.1% SDS) can also be manipulated to optimizehybridization conditions. Once conditions have been opti-mized, they can be kept constant for a particular array andsample type. Indeed, it is essential that hybridization condi-tions are kept identical within a particular experiment. If dif-ferent sample types are to be used with the same array, how-ever, it may be necessary to change hybridization conditionsbetween experiments. For example, for environmental sam-ples you may wish to decrease hybridization temperature tomaximize signal, or alternatively increase temperature toincrease specificity of signal. It is essential for researchers torealize, however, that such differences in conditions will pre-vent a direct comparison between experiments.

Image acquisition and quantification—Once the sample hasbeen hybridized, microarray signal intensities are collected viaimage acquisition with a microarray scanner. Software is thenused to visualize the image, find features (i.e., the spots), andquantify the signal in each feature. The scanner to be used formicroarray image capture depends on your platform and theavailable infrastructure. Affymetrix microarrays require a spe-cialized Affymetrix scanner, and scanning is generally carriedout at an array facility. For other platforms, scanning may bedone with a variety of scanners in an automated, high-throughput fashion using standard settings for laser power,image position, and pixel size or on a slide-by-slide basis withparameters optimized for each microarray. If you are purchas-ing a scanner, your choice will depend on factors such as theresolution required (between 5 and 10 µm is usual for stan-dard spotted microarrays), the number of microarrays youintend to analyze, and the likelihood of requiring more thanthe two lasers. Excellent scanners commonly used include theAxon GenePix series, PerkinElmer ScanArray series, and Agi-lent scanners. Image analysis software is generally supplied

with the scanner, which ensures that the image is in the cor-rect format for processing. For example, the GenePix seriesuse GenePix Pro software, whereas PerkinElmer suggests Sca-nArray Express for their ScanArray series. In principle, how-ever, any scanner can generate images which can be assessedusing any microarray image analysis software. Other suit-able microarray software packages include BlueGnome,ArrayVision (GE Healthcare), and ImaGene (BioDiscovery),all of which perform well. We therefore recommend thatyou try out a number of different software packages (demoformats can generally be downloaded from the Web) andchoose according to your technical requirements and easeof use.

Data processing—Microarray data are inherently noisy andrequire careful processing before statistical analysis. Pre-analy-sis steps include quality control, background correction, andnormalization. If you are working on a system with more thanone probe per gene (e.g., Affymetrix), there will also be a sum-marization step, where a summary measure for a gene is gen-erated from the multiple probes representing that gene. Agood overview of the steps involved in preprocessing microar-ray data are given in chapter 1 of the 2005 Bioconductor book(Huber et al. 2005). Here we outline basic considerations, butdirect the reader to other chapters of the same book that coverthese topics in greater detail.

Quality assessment of the microarray data can include avariety of methods, for example, looking at regeneratedimages of background signal to assess spatial irregularities,plots of log signal value compared to intensity pre- and post-normalization (so-called MA plots), and box plots of slide datapre- and postnormalization and assessing spot quality flaginformation where this has been supplied by the image cap-ture software. Hierarchical clustering (see “Data Analysis”below) of data pre- and postnormalization is very useful tolook for whether the data are grouping according to non-inter-esting factors such as which day an analysis was carried out.For quality assessment methods for Affymetrix arrays, we alsodirect the reader to a white paper on the Affymetrix Web site(www.affymetrix.com/support/technical/whitepapers/exon_arrays_qa_whitepaper.pdf). Image capture software and analy-sis software manuals usually include some information onquality assessment and control. Of note is that different soft-ware programs often tag spots with indicators of measurementquality, and different analysis software will provide differentways of dealing with these tags. We tend to work only withhigh-quality spots, as microarray data are noisy enough with-out including lower-quality spots, even if downweighted, foranalysis purposes.

Background correction methods aim to remove backgroundsignal from spot signal measurements. Background correction isapplied to data before normalization, although it is not alwaysobvious, as it may be included in a software choice that encom-passes a number of steps, such as background correction, nor-malization, and summarization, under a single command.


44

Normalization describes a variety of methods to correctmicroarray data for variation introduced by experimental pro-cedures rather than biological differences between samples. Forexample, differences in detection efficiency, dye labeling, fluo-rescence yields, and total amount of cRNA/cDNA loaded ontothe array can affect the measured signal intensities. If neg-lected, these factors can be erroneously interpreted as changesin expression. It is therefore important to correct for such vari-ability before further data analyses. Choosing an appropriatenormalization method is a crucial step, since it has consider-able influence on the results (Hoffman et al. 2002). Normaliza-tions can be applied to data within each array, to account forintra-array issues such as dye bias, location-dependent bias,and intensity-dependent bias. They can also be applied acrossarrays, usually referred to as between-array normalization, toaddress scale differences between arrays. The aim here is tomake data comparable across arrays. For example, the popularquantile normalization (Bolstad et al. 2003) shifts the distribu-tions of signals on arrays, resulting in the same empirical dis-tribution across arrays and across channels.

Depending on the type of microarray platform used, differ-ent normalization schemes can be applied, with within-arraynormalizations being applied before between-array normaliza-tions. For one-color microarrays, the first data processing step isthe calculation of the summary indices for each gene based onthe corresponding probes. Two methods are widely used for thistask: Microarray Suite (MAS)/GeneChip Operating Software(GCOS) by Affymetrix and Robust Multi-array Average (RMA)introduced by Irizarry and coworkers (Irizarry et al. 2003).MAS/GCOS calculates the summary indices by averaging probesignals for each array individually. In contrast, RMA simultane-ously calculates summary indices for all arrays included in theexperiment. Because it incorporates both probe and array effectsin the calculation, it can correct for systematic difference in sig-nal intensities between arrays and thus provides a first level ofnormalization. Note that this implicitly assumes that total(logged) expression intensity should be equal for the differentarrays. The obtained distributions of summary indices reflectingthe expression levels can be further adjusted subsequently. ForMAS, scaling of the median expression to a chosen level is usu-ally performed, if we can assume that the majority of measuredgenes are not differentially expressed. A common additionaladjustment for RMA-processed data are quantile normalizationtransforming expression levels to have the same distribution fordifferent arrays (Bolstad et al. 2003).

Several comparisons have been conducted so far betweenMAS and RMA processed data, but the results remain incon-clusive. For example, RMA performed favorably for a bench-mark dataset with a small number of spike-in controls, forwhich the concentrations were known (Cope et al. 2004). Inanother study, where a considerably larger number of spikecontrols was used, MAS outperformed RMA (Choe et al. 2005).The latter study, however, has been criticized for the use of aflawed design (Dabney and Storey 2006, Irizarry et al. 2006).

These contrasting results for different datasets indicate thatthe performance of normalization procedures can stronglydepend on the dataset being processed. This is not surprising,since all of the normalization procedures are based on specificassumptions which may not hold for different datasets. Whenassumptions are violated, normalization might fail and canlead to erroneous results. Thus, it is of vital importance thatresearchers carefully check the suitability of assumptions.

To normalize two-color arrays, the signal intensities of thecohybridized samples are used. The most basic method isglobal normalization, which is the linear scaling of the totalintensities in each channel to the same value. Nonlinear effectsin the signal intensities are frequently observed, however—aphenomenon referred to as dye bias (see earlier discussion ofdye swaps). This kind of artifact causes signals to be systemati-cally larger in one channel for the low-intensity range evenafter balancing the average signal intensities of both channels.To cope with such bias, several intensity-dependent normaliza-tion procedures have been introduced. Some of these elaboratemethods can cope with potential spatial artifacts (Yang et al.2002, Yang and Speed 2002, Futschik and Crompton 2004). Itis important to emphasize, however, that such nonlinear meth-ods require that most of the genes assayed are nondifferentiallyexpressed or that the differential expression is symmetrical,i.e., the overall up- and downregulation is balanced. This is notalways the case for virus-derived samples—for example, if over-all host genome expression declines during viral infection or ifthe array is dominated by virus probes only.

A solution is to scale the data to signal from reference probesthat are kept constant across the experiment. The employedreference can be of either endogenous (e.g., so-called house-keeping genes) or exogenous (i.e., spike-in controls) origin.Normalization based on housekeeping genes predates mostother normalization procedures for two-color arrays (DeRisi etal. 1996) and assumes that the genes chosen really do notchange under the experimental conditions. If you cannot guar-antee a set of expressed but nonchanging genes, we recom-mend you use spike-in controls. Validation, for example usingquantitative RT-PCR, can be used to select an optimal normal-ization procedure (Lindell et al. 2007). Reference-based nor-malization methods seem intuitively very attractive, but care-ful checks are necessary to ensure they are working as desired.We have performed comparisons of several normalizationschemes for microarray experiments for virus-infectedProchlorococcus cells, for which we expected that the majority ofassayed genes underwent differential expression (Lindell et al.2007). The microarrays were customized AffymetrixGeneChips including spike-in hybridization controls. Neitherthe normalization by spike-in control nor by rRNA selected ashousekeeping genes yielded superior results compared withother normalization strategies when using qRT-PCR data as thegold standard in an independent comparison (Lindell et al.unpubl. results). Possible reasons for the inferior performanceof spike-in controls could be their limited number and their


45

restricted intensity range on Affymetrix GeneChips. In the caseof normalization based on rRNA, their high expression levelsand their possible saturation on the array might cause difficul-ties in their use as reference. Having said this, we have had sat-isfactory results using reference-based normalizations on two-color arrays. As with all aspects of microarray experimentalplanning, nothing should be taken for granted. The use ofspike-in controls combined with external checks of representa-tive genes (e.g., quantitative RT-PCR), and plotting of data andcontrols pre- and postnormalization, are necessary to haveconfidence in your microarray results.

Data analysis—Clustering is a popular approach to explorelarge microarray data sets. The aim of clustering is to assigngenes or arrays to groups based on their similarity: genes orarrays displaying similar expression profiles, where you definewhat similar means by choosing an appropriate dissimilaritymeasure, should be assigned to the same clusters, whereas genesor arrays displaying distinct expression profiles should beplaced in different clusters. Using cluster analysis, we can detectprominent expression patterns, coexpressed genes, and similarconditions, which can be further examined for their biologicalmeaning. Many different clustering methods have been appliedto microarray data. Generally, two types of clustering exists:hierarchical and partitional (Jain and Dubes 1988). Hierarchicalclustering creates a set of nested clusters, so that clusters on ahigher level are composed of smaller clusters on lower levels.The resulting hierarchy of clusters are conventionally presentedas a treelike structure, the so-called dendrogram. To performhierarchical clustering, we proceed in a sequential manner. Ineach step, we calculate the pairwise distances between all clus-ters and merge the ones with the smallest distance. In contrast,in partitional clustering, all objects are simultaneously assignedto clusters. This type of clustering typically aims to optimize anobjective function for a given number of clusters. A primeexample of this clustering approach is k-means clustering,which seeks to minimize the between-cluster variation in aniterative manner. Hierarchical and partitional clustering bothhave their advantages and disadvantages. One strength of hier-archical clustering is that it defines relations between andwithin clusters. However, the sequential procedure used can besensitive to the high noise level that is frequently contained inmicroarray data. Partitional clustering tends to be more robustto noise, but commonly fails to present within-cluster struc-tures. Notably, some partitional clustering methods can revealinternal cluster structures and are still highly noise-robust. Forinstance, fuzzy clustering assigns genes graded membership toclusters, i.e., it can indicate how strongly a gene is associatedwith a cluster. Thus, genes that are tightly clustered obtain alarge membership (with values close to 1), whereas genes withnoisy expression patterns receive low membership values. Incontrast to conventional clustering methods, fuzzy clusteringeven allows genes to be placed in several clusters (Futschik andCrompton 2004). A variety of software packages have beendeveloped for clustering analysis of microarray data. Popular

standalone software packages for performing and visualizinghierarchical clustering are Cluster 3.0 and Java TreeView, respec-tively (bonsai.ims.u-tokyo.ac.jp/~mdehoon/software/cluster; jtreeview.sourceforge.net). Alternatively, several Web serversenable online cluster analyses (e.g., EBI expression profiler,www.ebi.ac.uk/expressionprofiler).

A difficult question is how many clusters can be reliablyretrieved from the observed expression data. The difficultyarises from the complexity of microarray data and a high noisecomponent. Frequently, different cluster structures are appar-ent depending on the resolution. For example, several mainclusters may exist, but each might display subclusters. Further-more, the noise component can lead to overlapping clusters,for which a separation might be not justified. Despite these dif-ficulties, some tools have been developed to help researcherschoose the accurate number of clusters and judge the reliabil-ity of the results. Classic approaches are based on the so-calledfigures of merit. These measures capture a desired feature thatwe seek to optimize. To obtain an accurate clustering, we seekto optimize the figure of merit. An example is the Dunn index,defined as the ratio between the minimal intracluster and themaximal intercluster distance (Dunn 1974). Used as a figure ofmerit, we aim to minimize the Dunn index to obtain tight clus-ters that are well separated. A common drawback, however, isthat these measures assume non-overlapping clusters, whichare typically not the case for microarray data. An alternativeapproach is based on measuring the stability of clusters withrespect to data perturbations, e.g., through resampling or addi-tion of noise. According to this concept, reliable clusters arethose that are maintained in spite of perturbation (Bittner et al.2000, Levine and Domany 2001). For instance, we could clus-ter genes using only a subset of measurements and examinethe resulting clusters. As reliable clusters should not depend onsingle measurements, they should still be detectable using par-tial data. Finally, the inspection of the functional compositionof clusters can give us clues about reliability. This strategyassumes that genes sharing the same function tend to be coex-pressed and thus should be placed into the same cluster. Byoptimizing the enrichment of functional gene categories, thenumber of clusters can be chosen (Gibbons and Roth 2002).

Despite these tools, assessing the quality of clusters remainschallenging. Researchers are advised to apply several clusteringapproaches to their datasets, as a single method often workswell with some data sets, but may perform poorly in others. Inpractice, we propose that clustering should be seen primarily asexploratory analysis that can then be followed up with morestringent computational and experimental examination. Agood introduction to clustering as applied to microarray analy-sis is given in chapter 7 of Wit and McClure (2004).

Statistical significance of differential expression—A typical aimof microarray experiments is to identify genes that are differen-tially expressed under different conditions. Because of the largenumber of genes measured by microarray technologies and ran-dom fluctuations in gene expression, we expect a number of


46

genes to show differences in expression simply by chance.Therefore, we use stringent statistical approaches to analyzemicroarray data for differential expression. In classic statisticaltesting, we compare a test statistic calculated from our data to adistribution of that test statistic expected under the null hypoth-esis that the gene is not being differentially expressed. The com-parison of our test statistic to this distribution results in a Pvalue. P values indicate how often you would expect to see dataas extreme as that you just observed if the gene is not being dif-ferentially expressed. P values are not direct indicators of theprobability of the gene being or not being differentiallyexpressed. For differential gene expression studies, a small Pvalue (e.g., <0.01 for a single test, see below) indicates that wewould rarely see a test statistic that extreme if the gene mea-surements in one condition really were from the same distribu-tion of expressions as the condition we are comparing them to.For example, with P = 0.01, we would expect to see values thisextreme in about 1 of every 100 samplings if the data we areworking with were sampled from the same distribution as theone we are comparing it to. Another school of statistics, theBayesian school, provides more readily interpretable probabili-ties, but can be harder to apply in practice. P values are closelyrelated to a more readily interpretable value, the false discoveryrate (Benjamini and Hochberg 1995), which is commonly usedto evaluate microarray results and is discussed further below. Fora good introduction to all the aspects of statistical testing fordifferential expression of microarray data mentioned here, werecommend chapter 8 of Wit and McClure (2004).

There are two general categories of statistical tests:1. Parametric tests. These assume that the populations being

compared can be described by particular distributions. Forinstance, certain tests assume that the underlying distribu-tion is normal. For microarray analyses, the distributionbeing referred to is the distribution of expression values fora given gene; it is not about the distribution of expressionvalues across genes. Typically, having sufficient biologicalreplicates for parametric statistical analysis is a challenge.Some available statistical methods, so-called local poolederror methods, aim to decrease the number of replicatesrequired to carry out reliable statistical tests by pooling thesample variation of genes with similar expression inten-sity; i.e., they fit the error with respect to the signal inten-sity based on the observed data (Baldi and Long 2001,Tusher et al. 2001, Jain et al. 2003, Smyth 2004).

2. Nonparametric tests. These do not make assumptions aboutthe underlying distribution. Such tests commonly rank thedata in value order and then carry out tests based on theorder. This is very useful when we do not know the under-lying distribution of what we are measuring. However,nonparametric tests have less power to detect differencesand thus require more replicates to give similar confidencewhen interpreting your data.

Parametric tests involve comparing the test statistics calcu-lated using your data to a standard distribution with particular

parameters. If you have sufficient biological replication in yourexperimental design, you can instead compare your test statis-tic to the expected null distribution of that statistic (i.e., whena gene is not differentially expressed), which you generatethrough bootstrapping, or through permutation, of your owndata set. For example, instead of comparing a t-test statistic toa standard Student t distribution to determine a P value, youwould generate a distribution of the test statistic from resam-pling your own data set in defined ways and compare your teststatistic to that distribution. Issues relating to bootstrappingand permutation are discussed in Wit and McClure (2004).

Until this point, we have really been discussing testing fordifferential expression of a single gene between two conditions.For a comparison yielding a P value of 0.01, we would expect tosee data this extreme in about 1 of every 100 tests if the genewere not differentially expressed. So, if we test 10,000 genes,and 10 genes are really different between conditions, we couldstill expect around 110 genes with P values less than or equal to0.01. Of these, 100 had “extreme” mean expression values dueto chance, with 10 of them truly differentially expressedbetween conditions. We need to try and increase our ability todiscern genes that are truly differentially expressed. In otherwords, we need to adjust our results to take into account thatwe are carrying out multiple testing. Different types of multipletesting corrections are used in microarray studies (Dudoit et al.2003), but arguably the most popular is the false discovery rate(FDR) (Benjamini and Hochberg 1995). The FDR method allowsyou to define the proportion of false positives you would findtolerable in your results. It then returns the largest list of genesclassified as differentially expressed that includes this specified,expected percentage of nondifferentially expressed genes.Numerous software tools can help researchers to assess the sig-nificance of differential expression. Notably, a highly powerfuland flexible platform, also for other aspects of microarray dataanalysis, is the Bioconductor project (www.bioconductor.org).Alternatively, more specialized software solutions such as SAM(www-stat.stanford.edu/~tibs/*SAM*) or BRB Array Tools(linus.nci.nih.gov/BRB-ArrayTools.html) can be applied to cal-culate false discovery rates for gene expression data.

Experiment annotation and data submission—Many journalsrequire microarray experimental data to be submitted to apublic repository as a condition of publication. Indeed, somefunding agencies require researchers to agree to make theirdata publicly available at the end of the project; for manyresearchers, the most sensible way to do this will be by sub-mitting to a known public repository such as the EBI’s Array-Express (www.ebi.ac.uk/arrayexpress) or the NCBI’s GEO(www.ncbi.nlm.nih.gov/geo). Both these databases requireadequate annotation of the experiment, including at least theinformation required by the Minimum Information about aMicroarray Experiment (MIAME) standard (Brazma et al.2001). For researchers engaging in environmental experi-ments, it is worth also referring to the MIAME Env extension(Morrison et al. 2006).


47

The importance of annotating data properly and making itpublicly available has led to many new “minimum informa-tion” checklists for different domains. The Minimum Infor-mation for Biological and Biomedical Investigations (MIBBI)portal (http://mibbi.sourceforge.net) is a good place to lookfor standards lists. Data standards list what the minimumrequirements are for describing a data set. How the data shouldbe described is usually defined by a set of terms or an ontol-ogy. For key microarray experimental concepts, the Microar-ray Gene Expression Data Society (MGED) make the MGEDOntology available. Use of this ontology, sometimes in com-bination with other domain-specific ontologies, is highly rec-ommended when annotating your experiments. The OpenBiomedical Ontologies (OBO) consortium (Smith et al. 2007)Web site (http://obofoundry.org) is probably the best place tosee if ontologies relevant to your area already exist or areunder development.

Public data repositories offer tools to facilitate submission ofdata, and some external tools support export in a format accept-able to the public repositories. A “non-exhaustive list of possi-ble MIAME compliant software” is held on the MGED site athttp://www.mged.org/Workgroups/MIAME/miame_software.html. For researchers in the environmental sciences, the soft-ware maxdLoad2 supports annotation using the MIAME Envextension of the MIAME standard (Hancock et al. 2005).

Summary—Microarrays offer great potential, but the statis-tical analysis needs careful consideration from the outset. Ourgeneral recommendations are as follows:1. If the samples required to address a particular question are

too difficult to generate or collect, you should adjust thequestion you are asking. Your data will not miraculouslyprovide answers to the original question if you do nothave sufficient or appropriate samples.

2. If you are not experienced with statistics, then find a col-laborator who is. Access to shiny software with easy-to-usemenus is not the same thing as statistical knowledge.

3. Know what software you (or your collaborator) are goingto use for the analysis and be sure that it is capable of ana-lyzing data generated under a particular design. It is also agood idea to define how technical replicate spots and tech-nical replicate slides will be handled if these are part ofyour design.

4. Technical replicates provide a different type of informa-tion than biological replicates and can be difficult to han-dle appropriately using some software. For many purposes,more biological replicates is a better option than carryingout technical replicate hybridizations.

AssessmentThe marine sciences have been relatively slow to embrace

microarray technology, primarily because of the historical lackof available genomic sequence. This has changed significantlyover the past few years, and it is now becoming common formarine-focused researchers to develop microarrays to answer

their questions of interest. The influx of molecular biol-ogy–trained researchers to the field of marine virology hasserved to hasten this pace further. To date, microarrays havebeen developed and used to study a number of aquatic virusesincluding cyanophages (see “Case study 1” below; Lindell et al.2007, Millard et al. 2009), coccolithoviruses (see “Case study 2”below; Allen et al. 2006a, Allen and Wilson 2006, Allen et al.2007), and shrimp white spot virus (Dhar et al. 2003). Althoughthe array platforms used for these case studies are different,many aspects are relevant to any platform, especially thoserelating to experimentation, RNA extraction, and data analysis.

Case study 1: Affymetrix microarrays and the cyanophages—Custom-made high-density Affymetrix GeneChip arrays weredesigned for two marine cyanobacteria of the genus Prochloro-coccus and two marine cyanophages. Here we describe arraydesign considerations and microarray experiments carried outto determine the transcriptional program of the T7-likepodovirus P-SSP7 when infecting the cyanobacterial hostProchlorococcus MED4, as well as to assess the transcriptionalresponse of the host to infection. Transcriptional profiles for allphage genes were investigated during the latent period ofinfection and revealed a transcriptional program for thepodovirus P-SSP7 reminiscent of that for T7. The viral genomewas transcribed in three functional clusters from left to right ofthe genome map (Lindell et al. 2007), with the modulethought to be involved in host takeover being transcribed first,then the DNA replication module, and finally the moduleencoding structural and packaging genes (Figs. 4 and 5). UnlikeT7, however, the last three genes of the genome, including thebacterial-like transaldolase gene, were transcribed out of order.These last three genes, together with the cyanobacteria-likephotosynthesis genes and a bacteria-like ribonucleotide reduc-tase gene, were transcribed together with the phage DNA repli-cation module, leading to the hypothesis that the products ofthe bacteria-like genes in the phage genome may be involvedin generating energy and substrates for genome replication.Investigation of the whole-genome response of the cyanobac-terial host to infection revealed that whereas the vast majorityof transcripts were downregulated as infection progressed (75%of the genome), 41 protein coding genes (Lindell et al. 2007)and three ncRNAs (Steglich et al. 2008) were upregulated intwo distinct expression clusters (Fig. 4). The function of theseupregulated genes is still unknown, as is whether they wereupregulated as a host stress response or by the phage for itsown purposes. The questions raised from the results of thesemicroarray results are active areas of current research.

Details of the procedures for this experiment can be foundin the supplemental information of Lindell et al. (2007).Below, we provide a summary of the procedures used in thedesign and implementation of the arrays and the reasonsbehind their use rather than providing the details of the actualexperimental procedures.

Shortly after the sequencing and annotation of twoProchlorococcus strains, MED4 and MIT9313 (Rocap et al.


48

2003), we designed the custom-made MD4-9313 high-densityAffymetrix array to investigate the transcriptional response ofthese cyanobacteria to a variety of environmental stressors,one of which was viral infection. We therefore included therecently sequenced genomes of two cyanophages that infectProchlorococcus MED4, the T7-like podovirus P-SSP7 and theT4-like myovirus P-SSM4 (Sullivan et al. 2005). Importantly,the presence of both host and viral genomes on a single arrayenabled the concurrent investigation of the transcriptionalprogram of the virus during infection, together with the host’stranscriptional response to this infection from the one sampleon a single array. We further decided to design the array withmultiple cyanobacterial and cyanophage genomes to reducethe design price per organism. In addition to the viral infec-tion experiment described here, this array has been used suc-cessfully in numerous studies investigating the transcriptionalresponse of Prochlorococcus to a variety of environmental stres-sors (Martiny et al. 2006, Steglich et al. 2006, Tolonen et al.2006) over a diel cycle (Zinser et al. 2009), as well as in the

investigation of small noncoding RNAs (Steglich et al. 2008).The MD4-9313 array was designed by the researchers (D.

Lindell, M. A. Wright, S. W. Chisholm, and G. M. Church) inconjunction with the Affymetrix design team. We decided ona design arrangement somewhat different from the standardfor Affymetrix expression arrays, which consist of 11 probesper open reading frame (ORF), often biased toward the 3′ end.Probes for the two Prochlorococcus strains were not 3′ biasedbut rather were designed evenly along the annotated ORFs sothat probes were spaced approximately every 80 bases. We fur-ther designed probes for all intergenic regions (longer than 35bp) on both strands at intervals of about 45 bases so thatprobes would be present for both short protein coding genesand noncoding RNA genes not initially annotated. Probe spac-ing was decreased for short ORFs and intergenic regions sothat 11 and four probes, respectively, were designed wherepossible. This same design strategy could not be used for theviral genomes, as they had only just been sequenced and hadyet to be annotated at the time of design. Therefore probeswere designed across the phage genomes at approximate inter-vals of 90 bases on both strands. Seeing as total RNA is used inbacterial arrays, care was taken to ensure that probes had nosimilarity to ribosomal genes to reduce the chance of nonspe-

Fig. 4. Hierarchical cluster analysis of gene expression profiles of the P-SSP7 phage during infection of Prochlorococcus MED4 showing three dis-tinct gene expression clusters. Transcript levels were determined byAffymetrix microarray analysis. The dendrogram appears on the left, heatmap in the middle, and gene names and cluster membership on theright. In the heat map, red indicates an increase and green a decrease inexpression. Time after infection appears above the heat map. Hierarchicalclustering was carried out using Pearson correlation and average linkage.Input data were the average logged expression values of three biologicalreplicates, standardized so that mean expression values for each geneequal 0 and standard deviation equals 1. Reproduced with permissionfrom Lindell et al. (2007).

Fig. 5. Transcriptional profiles of Prochlorococcus MED4 genes with timeafter infection by the podovirus P-SSP7. Transcript levels were determinedby Affymetrix microarray analysis and are presented as log2-fold changein infected cells relative to the paired uninfected cells over the 8-h latentperiod of infection. The results are the average of three biological repli-cates. Only genes whose expression levels were significant at a false dis-covery rate of q < 0.05 are shown. Blue and red indicate two significantlyupregulated gene clusters. Gray indicates genes significantly downregu-lated at 8 h after infection. Reproduced with permission from Lindell et al.(2007).


49

cific binding due to their high transcript levels. Probe pairs ofperfect match (identical to the sequence to be detected) andmismatch (containing a single base change at the center of theprobe) were designed as per the Affymetrix standard design.Probes are designed so that their properties meet criteria deter-mined empirically by Affymetrix. This information is propri-etary and is not available to the researcher.

During analysis, an average of the signal from all the probesassociated with a particular gene or intergenic region (a probeset) is used to determine the relative expression level. Whenusing Affymetrix analysis software (MAS), the mismatch signallevel is used as a measure of nonspecific hybridization. WhenRMA analysis, an alternative to MAS, is used, only the perfectmatch probes are used. Designing perfect match probes onlywould increase the number of genes represented on an arraybut should only be done if the researcher is positive they donot need mismatch probes in their analyses. This is an anti-sense array and is designed for protocols that generate and uselabeled cDNA or cRNA sequences that are antisense to theoriginal RNA. This array contains approximately 200,000probe pairs over an area of 8.1 mm in a Midi array format, andeach feature is 18 µm in size.

Initially, the array was tested by hybridizing labeledgenomic DNA from each of the cyanobacterial strains inde-pendently. The array performed extremely well with all probesets for predicted open reading frames, giving a reproduciblesignal well above background levels (generally 20-fold higherthan background levels), whereas some probes sets for inter-genic regions (known to be of lower quality at the time ofdesign) did not provide a significant signal. It is important tonote that the signal intensity varied up to sixfold across theprobe sets even though equal amounts of DNA were presentfor each gene. This indicates that intrinsic differences inhybridization efficiencies exist between probe sets and high-lights that absolute transcript levels cannot be determinedfrom microarray analysis. However, standardization based ongenomic hybridization signals may be useful in determiningthe relative levels of different transcripts if these are below sat-uration levels on the array. Cross-hybridization between thetwo Prochlorococcus genomes was fairly low (10% of the probesdesigned for one genome gave signals above backgroundwhen hybridized with DNA from the other genome), indicat-ing that it is feasible to include probe sets for multiplegenomes on a single array. Indeed, no detrimental effects ofthe presence of the MIT9313 genome on the array weredetected when running experiments with MED4.

Triplicate cultures of Prochlorococcus were grown under con-tinuous light and concentrated by centrifugation to 108 cellsmL–1. Each of the three cultures was divided into two pairedsubcultures, one of which was infected with the podovirus P-SSP7 at a ratio of three infective viruses per cell, and the otherwas amended with filter-sterilized spent medium and servedas an uninfected control (Lindell et al. 2007). Samples werecollected by centrifugation for RNA extraction (100 mL) every

hour over the 8-h latent period of infection from the infectedcultures as well as their paired uninfected controls. The cellpellet was rapidly resuspended in storage buffer (200 mMsucrose, 10 mM sodium acetate, pH 5.2, 5 mM EDTA), snap-frozen in liquid nitrogen, and stored at –80°C. The Ambionproduct RNAlater has also been successfully used by others forstorage of samples before RNA extraction. The procedure fromcollection of cells until freezing takes approximately 20 min.This time can be significantly reduced if the cells are harvestedby filtration onto Supor-450 (Gelman-Pall) filters that are thenimmersed in the above storage buffer. Very high reproducibil-ity was achieved between these three biological treatments,indicating that this number of biological replicates was suffi-cient and that technical replicates were not necessary whenusing this custom-made Affymetrix array.

We decided to grow Prochlorococcus under continuous lightfor the infection experiment (even though they grow naturallyunder a diel light-dark cycle) to remove the complications asso-ciated with intrinsic diel differences in expression patterns ofthe host (Zinser et al. 2009), although the paired treatmentsand controls would have controlled for such differences. Con-centration by centrifugation of the cells before infectioncaused an unexpected problem—that of differential expressionduring the first 4 h after centrifugation. It was therefore fortu-nate that an experimental design based on paired treatmentsand controls was chosen, enabling us to control for transcrip-tional changes associated with the centrifugation at each timepoint. In this case, a comparison of the transcriptionalresponse at different times after infection relative to t = 0 (lon-gitudinal comparison) would have led to erroneous conclu-sions. A further complication with this particular host–virussystem is that we manage to infect only 50% of the cells dur-ing a first round of infection irrespective of the multiplicity ofinfection used. This did not have any detrimental effects ondetermining the transcriptional program of the virus, as thecells appeared to be synchronously infected. This rather com-plicates elucidation of the host responses to infection, how-ever, as uninfected cells are present in the infected treatmentand could mask moderate level responses by the infected cells.

Care must be taken when extracting RNA samples formicroarray analysis, as with any work using RNA, due to thehigh stability of RNases and the difficulty in their removal. Wetherefore work with nuclease-free molecular biology–gradereagents and plastics, in an RNase-free work space and withpipettes and gel boxes dedicated for the purpose. RNA wasextracted using Ambion’s mirVana RNA isolation kit. Immedi-ately before extraction with this kit, the cells were thawed rap-idly at 25°C and centrifuged for 2 min to exchange the resus-pension buffer with the lysis buffer from the kit. Depending onthe cyanobacterium/bacterium investigated, it may be necessaryto add a lysozyme step to the protocol before cell lysis. After RNAextraction, contaminating DNA was removed by digestion withDNase using Ambion’s Turbo DNA-free kit. The resulting RNAwas then subjected to 3 M sodium acetate-ethanol precipitation


50

and a 70% ethanol wash to both remove the DNase reagents andnucleotides and to concentrate the sample. This yielded approx-imately 10–20 µg RNA after the entire procedure from approxi-mately 1010 starting cells of Prochlorococcus MED4 (i.e., approxi-mately 1–2 fg RNA per cell after all losses).

The mirVana RNA extraction protocol was determined to bethe most suitable for our purposes based on a combination ofconsiderations: (a) the relative ease and speed of the procedure—although it does include a phenol-chloroform step; (b) highyield and quality of resulting RNA; and (c) the retention of RNAsas small as 50 bases, which enables the assessment of expressionpatterns of small noncoding RNAs (Steglich et al. 2008). Otherprotocols tested produced significantly lower yields (Ambion’sRibopure and Qiagen’s RNeasy kit), were more labor intensive(the hot-phenol method, Lindell and Post 2001), or removedRNAs shorter than 200 bases (Ambion’s RNAaqueous and Qia-gen’s RNeasy). However, the mirVana kit is not suitable if cellsare harvested by filtration onto Supor-450 filters. In that case, itis necessary to extract the RNA using the hot phenol method, inwhich the filter is dissolved in the organic phenol phase and allnucleic acids are released from the cells embedded in the filter(Lindell and Post 2001, Steglich et al. 2006).

Once the RNA had been extracted and DNA removed, wedetermined the quantity and purity of the RNA spectrophoto-metrically and assessed RNA integrity on agarose gels. Smear-ing of the rRNA bands on gels indicates that the RNA isdegraded and should not be used. We find that runningagarose gels in Tris-acetate-EDTA (TAE) buffer is sufficient forthis purpose, although denaturing gels are generally used formore sophisticated RNA procedures to prevent secondarystructure from affecting the migration of the RNA in the gel.The amount of single-stranded RNA can be determined bymeasuring absorbance at 260 nm in a 1-cm quartz cuvetteusing the conversion factor of 40 (A260 × 40 = concentration ofRNA in ng µL–1). An absorbance (A) ratio at 260 to 280 nm of1.8–2.1 indicates that the RNA is of high quality with negligentprotein contamination, and a ratio of A230 to A260 of 0.3–0.5indicates little salt or phenol contamination. Agilent bioana-lyzers can also be used for assessing the quantity, purity, andintegrity of your RNA before running a microarray experiment.

RNA labeling and microarray hybridization, staining, andscanning were carried out at the Affymetrix service center sit-uated at the BioPolymers Facility of the Department of Genet-ics, Harvard Medical School, Boston, MA, USA. The proceduresused for bacteria are quite standard, although we found thatreducing the amount of total RNA used per array from the 10µg total RNA suggested by Affymetrix for E. coli to 2 µg forProchlorococcus MED4 provided good results and did not com-promise the quality and sensitivity of the array results. Seeingas we did not test the use of this low amount of RNA with E.coli, we don’t know if this difference is due to an exaggeratedsuggestion for high amounts of RNA by Affymetrix or due tointrinsic differences between the bacteria. One potential dif-ference is related to the high growth rate of E. coli (doubling

every hour) relative to Prochlorococcus MED4 (doubling every1–2 days), which may lead to significantly higher rRNA levelsrelative to mRNA in E. coli and therefore a requirement forgreater amounts of total RNA to gain similar levels of mRNA.Therefore, if material is hard to come by, as is often the case,we suggest that you test a range of RNA concentrations foryour system rather than relying on the Affymetrix suggestion.

Below is an overview of the microarray procedures we used.Details of the labeling, hybridization, staining, and scanningprotocols and data analysis carried out in this study can befound in the supplementary information of Lindell et al.(2007). The standard Affymetrix protocols for bacteria can befound at their Web site: http://www.affymetrix.com/support/technical/manual/expression_manual.affx.

Briefly, 2 µg total RNA was subjected to the labeling proto-col. The RNA was reverse transcribed to produce cDNA whichwas then fragmented to 50–200 nucleotides long, purified,and end-labeled with biotin. Biotin incorporation was verifiedby band shift analysis. The labeled sample was hybridized tothe array in an aqueous solution, and stringency washes wereperformed. Staining was achieved by incubation of the arraywith streptavidin, which has a high affinity for biotin, andfinally with a streptavidin-phycoerythrin conjugate. Scanningof the array for phycoerythrin fluorescence was carried out todetermine the raw signal levels for each probe. Spike-inhybridization controls were added before hybridization. In thefuture, we will also include spike-in RNA labeling controls atconcentrations expected to span the signal intensity range tohelp facilitate array normalization procedures.

Independent verification of the array results were carried outby quantitative reverse-transcriptase polymerase chain reaction(qRT-PCR). Genes were chosen for analysis to include represen-tatives of the different expression profiles detected, genes witha range of transcript intensities and genes of biological interestwhere possible (Lindell et al. 2007). Therefore, we verified thearray results for a subset of the genes from each of the threephage expression clusters as well as from the last gene in thegenome to represent the three genes transcribed out of order onthe genome. Host genes from both upregulated clusters and thedownregulated cluster were included. The qRT-PCR results forthe host genes were also used for determining the appropriatenormalization method. Because the high sensitivity of qRT-PCR, a more rigorous DNase treatment is necessary than thatcarried out for microarray analyses, and controls in the absenceof reverse transcriptase (no RT controls) must be carried out toensure that contaminating genomic DNA has been removed.

Data analyses were carried out in the statistical language Rusing several Bioconductor packages (Gentleman et al. 2004).The array data were normalized and probe set summaries werecalculated from perfect match probe intensities using RMAanalysis (Irizarry et al. 2003) from within the Bioconductor pro-gram. RMA with quantile normalization was chosen based onthe validation with RT-PCR data. Compared to other normal-ization schemes (e.g., based on spike-in controls), it yielded


51

superior performance (Lindell et al. 2007). This was initially sur-prising, since quantile normalization assumes similar overalldistribution of probe intensities in different arrays, whichseemed not to be the case here. However, closer inspectionrevealed that this assumption still holds owing to the largenumber of probe sets that did not show differential expressionfound for intergenic regions, for the P-SSP7 phage genome, andfor the additional Prochlorococcus and phage strain on the array.Statistical significance of differentially expressed genes betweeninfected and control cells at each time point was determinedusing a Bayesian t-test. Hierarchical clustering was carried out todetermine cluster patterns of the phage genes and upregulatedhost genes. The reliability of clustering was determined byrepeated clustering of a random resampling of subsets of genes.For both types of genes, visual inspection indicated the exis-tence of discrete clusters. To determine the numbers of reliableclusters, a resampling strategy was used, where hierarchical clus-ter analysis was performed repeatedly on randomly selectedsubsets of genes. Clusters were considered reliable if theyoccurred persistently for different random subsets of genes. Thisstrategy was used for a range of numbers of clusters and indi-cated that there were three reliable phage gene expression clus-ters and two upregulated host gene clusters (Lindell et al. 2007).

Case study 2: Spotted microarrays and the coccolithoviruses—Emiliania huxleyi, a calcifying marine haptophyte with world-wide distribution, is infected by the coccolithovirus family ofviruses. Whereas the study of Emiliania huxleyi is relativelycommon with international researchers owing to its funda-mental importance to global ecosystem function and model-ing, the study of its viruses has been restricted to a handful ofgroups since their discovery (Wilson et al. 2009). Neverthelessthe rate of discovery for this unique and interesting viral fam-ily has been phenomenal, aided primarily by the developmentof microarray-based tools (Allen and Wilson 2006).

Initially, the first-generation microarray was based aroundthe model strain Emiliania huxleyi virus 86, EhV-86. Designedand fabricated in tandem with an ongoing sequencing project,this microarray was initially used to help annotate a highlyunique genome that contained few genes of known functionand was dominated by coding sequences with no databasehomologs whatsoever (Wilson et al. 2005). Delays in sequenc-ing caused by the highly repetitive nature of the EhV-86genome (three families of repeat elements can be found in thegenome [Allen et al. 2006c]) restricted the design to probes fora mere 425 of the 472 predicted genes annotated on the finalEhV-86 genome. The lack of host genomic information (a dozenor so genes from Emiliania huxleyi were known at the time) andthe relatively small number of probes required, in tandem withthe local availability of an Affymetrix glass slide laser scanner,suggested that a spotted microarray would be the preferred sys-tem. Oligonucleotide probes (75mers) were designed one byone using Oligo 6 (a process that took almost a month to com-plete), synthesized by the commercial company Oligator, andprinted using a classic contact pin printing method at the Scot-

tish Centre for Genomic Technology and Informatics (SCGTi).Each probe was printed in triplicate in a 4 × 4 metagrid (eachsubgrid containing 12 × 13 features) design incorporating 2496features in total and covering a space of approximately 18 × 20mm. Once printed at SCGTi, microarrays were supplied to theresearchers, who then had total control over experimentaldesign, sample labeling, microarray hybridization, and dataanalysis. Despite the incomplete coverage of the EhV-86genome, the first-generation coccolithovirus microarray servesas an excellent example of the versatility of this molecular tool.

Initially, genome annotation of the then-completed EhV-86genome was aided by using the microarray to confirm the pres-ence of transcripts of the predicted genes (Wilson et al. 2005).A severe lack of database matches due to the uniqueness of thegenome created a large amount of uncertainty about whatcould be annotated as an ORF, coding sequence (CDS), or gene.With regard to genome annotation, not all open readingframes are coding sequences, and not all coding sequences aregenes. The results from the direct labeling of total RNA usinganchored oligo dT primers (exploiting the poly adenylated tailof mRNA) from cells infected with EhV-86 were used to aid theannotation of the 472 predicted CDSs, since the presence of atranscript is strong evidence that an ORF is actually a tran-scribed CDS. (For in-depth labeling methodology, see supple-mentary material of Wilson et al. [2005].) Of course, identifica-tion and characterization of the protein products is required toreclassify CDSs as genes (this has recently been done with aproteomic approach and has confirmed 28 genes), but the ini-tial identification of 472 CDSs on a predominantly unknownand unique genome was a huge step forward in the molecularcharacterization of the coccolithoviruses.

The approach of directly labeling total RNA used in this typeof experiment is hugely popular due to its ease and relative sim-plicity. However, it is fairly RNA intensive (at least 25 µg of totalRNA was required), which is not always ideal when studyinghost–virus systems. In particular, during the early stages ofinfection when the ratio of virus to host RNA message is rela-tively low, it is often difficult to detect virus message. This prob-lem is further exacerbated in marine systems which, in general,tend not to provide adequate biomass from manageable culturevolumes. To combat this, we tried a range of protocols whichwould allow us to finely map the transcriptional profile of EhV-86 during the first 4 h of infection. Performing a mRNA purifi-cation from the total RNA increased the sensitivity of detectionfor the handful of host probes on the microarray by approxi-mately 10-fold, yet failed to detect viral transcripts. An indirectlabeling method, whereby an aminoallyl nucleotide was incor-porated into the cDNA which was later cross-linked to a cyaninedye, further increased sensitivity approximately 1.5-fold. Itquickly became apparent that a labeling method capable ofincreasing sensitivity by many more orders of magnitude wouldbe required to detect virus transcripts during the first few hoursof infection. To this end, a derivative of the Eberwine (1996)method was used. The Eberwine method was originally devel-


52

oped to study single-cell systems and has been phenomenallypopular with microarray users. Linear amplification of mRNAmessage is achieved by creating cDNA using a primer contain-ing the T7 promoter site directly adjacent to the anchored oligodT primer. Production of cyanine-labeled cRNA using T7 poly-merase can amplify the original message by up to 1000-fold. Weused a commercially available derivative of this system (RocheApplied Biosciences), which boasts amplification of more than100,000-fold by using an initial PCR-based step. Briefly, first-strand cDNA synthesis is performed using a primer (oligo dT-T7-TAS) containing a random sequence with no significant homol-ogy to any sequence in public databases (Target AmplificationSequence, TAS), a T7 promoter, and oligo dT sequence. Aftersecond-strand synthesis using a random primer coupled to TAS,a double-stranded cDNA product is made that can be amplifiedin a PCR reaction using TAS primers. So long as the PCR reac-tion is kept in the exponential phase, message can be amplifiedin a quantitative manner. The resulting amplified cDNA canthen be transcribed into cRNA by a linear amplification stepusing T7 polymerase. Of course, using a PCR-based amplifica-tion strategy can create a plethora of downstream issues withdata analysis and interpretation, but as long as researchers areaware of this from the beginning, most can easily be avoided. Itis crucial to keep the PCR amplification step in the exponentialphase; that way, transcripts should be evenly represented.

In the early stages of infection, the number of infected cellsis low, but this increases as the experiment progresses; hencethe transcriptional profile becomes “blurred” with infectedcells at different stages of the infection process. This blurringof the transcriptional profile is avoided in well-studied mam-malian systems by the addition of inhibitors of RNA poly-merase, protein synthesis, and DNA polymerase, which allowtranscripts to be assigned into classes referred to as immediateearly, early, mid, and late genes.

We experimented with inhibitors such as phosphonoaceticacid (inhibitor of DNA replication) and cyclohexamide (proteinsynthesis) but found they failed to give reliable inhibition inour seawater-based algal culturing system. Therefore, in the caseof this experimental design, we decided on a simple “when areyou on?” question, which is relatively simple to address usingbioinformatics methods. We did this at 0, 1, 2, and 4 h post-infection (hpi), and this allowed us to group CDSs into sixgeneric groups (expressed 1, 2, 4 hpi; not expressed, not tested,and unconfirmed [for ambiguous results]) (Allen et al. 2006a).From these generic groups, we were able to distinguish betweentwo major transcriptional phases during viral infection: onephase dominated by genes associated with a specific promoterand localized to a specific section of the genome and a secondphase in which the remainder of the genes are transcribed.

The third use of the first-generation array was to study thegenomic content of all the coccolithoviruses in our currentvirus collection (Allen et al. 2007). Direct labeling of genomicDNA was performed using random primers and cyanine-labeled dCTP. Single-channel hybridizations were used, with

each genome hybridized to a single array. Pooling of samples tolabel for the alternative channel was a viable option that wasdiscussed extensively before undertaking the experiment, butsince the array was designed specifically for EhV-86, it was feltthat the usefulness of the additional data provided from theadditional control channel did not justify the added expense oflabeling twice as many samples. In perhaps the simplest of allmicroarray experiments to analyze, and taking the microarrayback to its Southern blot roots, an intensity value cutoff waschosen for each array whereby each spot could be consideredon or off. This initially appeared to be a risky strategy, and wewere worried that some genes would be mislabeled as presentor absent/variable. However, the distribution of spot intensitieswas surprisingly reproducible between strains, and the list ofprobes bordering the cutoff boundary (i.e., the ambiguousspots) was found to be nearly identical on each microarray—many were even found on the control EhV-86 array, suggestingthat poor labeling could be caused by some form of secondarystructure in the surrounding genomic regions. An advantage tousing a small microarray with fewer than 1600 features is thateach can be assessed by eye relatively easily, something that isdifficult to achieve with higher-density arrays. This simpleapproach led to the discovery that at least 70 genes of the 425that we tested are absent or sufficiently variable in one or moreof our dozen or so coccolithovirus strains (Allen et al. 2007).

Therefore, the first-generation coccolithovirus microarrayproved to be a robust, useful, and versatile tool that served uswell and produced and contributed to eight publications overa 3-year period (Allen and Wilson 2006).

The high degree of control we achieved by developing andmodifying our own labeling techniques to answer questionsspecific to our work made the decision easy when it came todeveloping the second-generation microarray. The investment,expertise, and past success in developing and optimizing proto-cols for spotted microarrays suggested we should continue theiruse. The second-generation microarray was developed to pro-vide greater coverage of the EhV-86 genome and to begin thetask of studying the host response to infection both under lab-oratory conditions and in the natural environment. To this end,ESTs from Emiliania huxleyi were sequenced, and the partialsequence (>80%) of a second coccolithovirus, EhV-163, wasobtained (Allen et al. 2006b; Kegel et al. 2007). In parallel to thiswork, the E. huxleyi genome sequencing project (led by BetsyRead) was underway; probes from this project for the ESTs withBLAST scores suggesting reasonable similarity to sequences ofknown function in the GenBank database were also included.In total, more than 4000 oligonucleotides were required to rep-resent every annotated EhV-86 CDS, all unannotated ORFs >100bp in EhV-86, the 2000+ E. huxleyi ESTs, all the genes on the E.huxleyi chloroplast and mitochondrial genomes, and for theadditional and highly variable EhV-163 genes (Allen et al.2006c). Due to the higher number of oligonucleotides requiredfor this array and the substantial investment required in theirsynthesis, we chose to allow the oligonucleotide manufacturer


53

to design them to provide insurance should they not work well.Operon was chosen for this task, as they had both the necessarytrack record in oligo design and, importantly, a close workingrelationship with our chosen microarray printing facility.Whereas pin printing proved to be reliable and robust enoughfor our relatively small 2496-feature first-generation array, sub-stantially more features were needed on this second-generationmicroarray. Fortunately, we have been able to take advantage ofpiezoelectric printing at the Liverpool Microarray Facility nodeof the NERC Molecular Genetics Facilities. The new second-gen-eration microarray is now based on a 7 × 5 metagrid, with eachsubgrid composed of 12 × 52 features, with a total of 21,840 fea-tures. Each probe is printed five times, and the printed area isapproximately 22 × 60 mm. In addition, there has also been sig-nificant investment in a tiling microarray for E. huxleyi; it isunder construction using the Nimblegen system and will beavailable to the community shortly.

Other aquatic virus microarrays: White spot syndrome virus—White spot syndrome virus (WSSV) is a commercially relevantviral pathogen of the cultured shrimp (Penaeus sp.). Since its dis-covery in Japan in 1993, it has spread to shrimp farming regionsthroughout Asia, the Americas, Europe, and Australasia (Dhar etal. 2003). As such, it has been intensively studied over pastdecade by a wide range of geographically distinct and inde-pendent research groups and is one of the best studied aquaticvirology systems. The inevitable consequence of this intensiveyet fractured study is that a variety of WWSV microarrays havebeen developed in tandem and independently.

Dhar et al. (2003; California, USA), developed a glassslide–based microarray consisting of 100 probes primarilyderived from the PCR amplification of EST clones from infectedhost cells. These researchers used a direct labeling method andhybridized fluorescently labeled first-strand cDNA to identifyhow shrimp genes responded to viral infection. Khadijah et al.(2003; Singapore) developed a glass slide microarray consistingof approximately 3000 amplified PCR fragments from a clonelibrary created following the restriction digestion of purifiedWSSV genomic DNA. These researchers estimated that thisallowed complete coverage of the WSSV genome and used thearray to identify latency-related WSSV genes through the T7-based Eberwine amplification method. Liu et al. (2005) and Tsaiet al. (2004) (Taiwan) created a glass slide microarray from PCRproducts (200–600 bp in size) using specific primers represent-ing 532 predicted ORFs (encoding potential proteins >60 aminoacids in size) of WSSV. These researchers used a direct labelingapproach to create fluorescently labeled first-strand cDNA toidentify immediate early genes in cyclohexamide-treatedshrimp and to create a temporal profile for the expression ofWSSV genes during infection. Marks et al. (2005; The Nether-lands) created a glass slide microarray from PCR products(300–1000 bp in size) for 158 of the 184 annotated WSSV CDSs(encompassing two different WSSV strains). For the larger CDSs,additional probes were generated to improve coverage. PCRproducts were generated by using either universal primers with

suitable clones from the library used to sequence the WSSVgenome or specific primers with WSSV genomic DNA as tem-plate for a total of 274 probes. A postlabeling approach (initiallyincorporating an aminoallyl nucleotide) was used to generatefluorescently labeled cDNA. These researchers used theirmicroarray to create a temporal expression pattern for WSSVgenes. Lan et al. (2006; China) created a nylon mem-brane–based microarray using PCR products generated from259 specific primer pairs (400–1000 bp in size) covering 151 ofthe 180 CDSs of WSSV. Using 32P radiolabeled cDNA, theseresearchers generated a temporal transcription profile of WSSVgenes during infection. In addition to these microarrays, whichcontain various numbers of WSSV-specific probes, manyresearch groups have developed “shrimp probe only” microar-rays to study the effects of infection by WSSV. Because theirinterests lie with determining how the host responds and theyare not necessarily interested in what the virus is doing per se,these arrays contain only host (shrimp) probes. This approachhas been successfully used by Robalino et al. (2007; South Car-olina, USA) who developed a glass slide microarray using 2,469PCR-amplified products from a shrimp EST library. PostlabeledcDNA was generated from four types of shrimp tissue to studythe immune response at the transcriptional level.

DiscussionMicroarrays are a very powerful and versatile tool. The plat-

form chosen depends on a variety of reasons, technical, eco-nomic, and sometimes historical. The authors of this chapterare a great example of the diversity in approaches to usingmicroarrays for the study of aquatic viruses: whereas D. Lindellfavors the Affymetrix approach, M. J. Allen favors the spottedoligo approach. Neither author is more right or more wrongthan the other: both systems are fit for their current purposes.Like evolution, when developing a microarray system you canonly work with what you’ve got. Indeed, other, less commonlyused, microarray platforms exist (such as those developed byApplied Biosystems, Eppendorf, GE Healthcare, Illumina, andPhalanx), and these may suit your needs better than the systemsdescribed in this article. Regardless of the final platform choice,most of the issues described here will be of direct relevance toyour experimental requirements. This also applies to experi-mental design and analysis. Many options exist, and theseshould be considered in light of the questions you wish toaddress and the resources, money, and skills available to you.

In this article, we have tried to describe the technology,techniques, and rationale behind microarray experiments,with a focus on the specific issues of virus-associated systems.As the reader can see, microarrays should not be undertaken ona whim, but with proper planning and consideration they canprove to be an excellent weapon in the virologist’s armory.However, it is important to note that there is no genericmethod that is perfect for every system. Thus, in closing, wehave come to the general conclusions that if a researcher isgiven a blank canvas with no prior preconceptions or limita-


54

tions then Affymetrix may be the platform of choice for sys-tems when a large number of experiments will be carried out,Agilent may be the platform of choice when high design flexi-bility is desired and few samples will be investigated, and spot-ted arrays may be the platform of choice for large-volumeexperiments when high design flexibility is required. Yet accessto existing infrastructure or even a researcher’s current knowl-edge and thinking can have a profound impact on the routeand choices taken. The key is to be prepared from the outset.Although they are no easy undertaking, microarrays have thepotential to drive your research in exciting and wonderfuldirections. It is the destination that matters most (i.e., publish-able data of high quality), not the route taken to get there.

ReferencesAllen, M. J., T. Forster, D. C. Schroeder, M. Hall, D. Roy, P.

Ghazal, and W. H. Wilson. 2006a. Locus-specific geneexpression pattern suggests a unique propagation strategyfor a giant algal virus. J. Virol. 80:7699-7705.

———, D. C. Schroeder, A. Donkin, K. J. Crawfurd, and W. H.Wilson. 2006b. Genome comparison of two Coccol-ithoviruses. Virol. J. 3:15.

———, ———, and W. H. Wilson. 2006c. Preliminary charac-terisation of repeat families in the genome of EhV-86, agiant algal virus that infects the marine microalga Emilianiahuxleyi. Arch. Virol. 151:525-535.

———, and W. H. Wilson. 2006. The coccolithovirus microarray:An array of uses. Brief Funct. Genomic Proteomic 5:273-279.

———, J. Martinez-Martinez, D. C. Schroeder, P. J. Somerfield,and W. H. Wilson. 2007. Use of microarrays to assess viraldiversity: From genotype to phenotype. Environ. Micro-biol. 9:971-982.

———, and W. H. Wilson. 2008. Aquatic virus diversityaccessed through omic techniques: A route map to func-tion. Curr. Opin. Microbiol. 11:226-232.

Alwine, J. C., D. J. Kemp, and G. R. Stark. 1977. Method fordetection of specific RNAs in agarose gels by transfer to dia-zobenzyloxymethyl-paper and hybridization with DNAprobes. Proc. Natl. Acad. Sci. U. S. A. 74:5350-5354.

Baldi, P., and A. D. Long. 2001. A Bayesian framework for theanalysis of microarray expression data: Regularized t-testand statistical inferences of gene changes. Bioinformatics17:509-519.

Benjamini, Y., and Y. Hochberg. 1995. Controlling the falsediscovery rate: A practical and powerful approach to multi-ple testing. J. Royal Stat. Soc. Ser. B 57:289-300.

Bier, F. F., M. Von Nickisch-Rosenegk, E. Ehrentreich-Forster, E.Reiss, J. Henkel, R. Strehlow, and D. Andresen. 2008. DNAmicroarrays. Adv. Biochem. Eng. Biotechnol. 109:433-453.

Bittner, M., and others. 2000. Molecular classification of cuta-neous malignant melanoma by gene expression profiling.Nature 406:536-540.

Bolstad, B. M., R. A. Irizarry, M. Astrand, and T. P. Speed. 2003.A comparison of normalization methods for high density

oligonucleotide array data based on variance and bias.Bioinformatics 19:185-193.

Brazma, A., and others. 2001. Minimum information about amicroarray experiment (MIAME)—toward standards formicroarray data. Nat. Genet. 29:365-371.

Choe, S. E., M. Boutros, A. M. Michelson, G. M. Church, andM. S. Halfon. 2005. Preferred analysis methods forAffymetrix GeneChips revealed by a wholly defined controldataset. Genome Biol. 6:R16.

Chou, H. H., A. P. Hsia, D. L. Mooney, and P. S. Schnable. 2004.Picky: Oligo microarray design for large genomes. Bioinfor-matics 20:2893-2902.

Chung, W. H., S. K. Rhee, X. F. Wan, J. W. Bae, Z. X. Quan, andY. H. Park. 2005. Design of long oligonucleotide probes forfunctional gene detection in a microbial community. Bioin-formatics [doi:10.1093/bioinformatics/ bti673].

Cope, L. M., R. A. Irizarry, H. A. Jaffee, Z. J. Wu, and T. P. Speed.2004. A benchmark for affymetrix GeneChip expressionmeasures. Bioinformatics 20:323-331.

Coppee, J. Y. 2008. Do DNA microarrays have their futurebehind them? Microbes Infect. 10:1067-1071.

Dabney, A. R., and J. D. Storey. 2006. A reanalysis of a publishedAffymetrix GeneChip control dataset. Genome Biol. 7:401.

DeRisi, J., and others. 1996. Use of a cDNA microarray toanalyse gene expression patterns in human cancer. Nat.Genet. 14:457-460.

Dhar, A. K., A. Dettori, M. M. Roux, K. R. Klimpel, and B. Read.2003. Identification of differentially expressed genes in shrimp(Penaeus stylirostris) infected with White spot syndrome virusby cDNA microarrays. Arch. Virol. 148:2381-2396.

Dudoit, S., J. Popper Shaffer, and J. C. Boldrick. 2003. Multiplehypothesis testing in microarray experiments. Stat. Sci.18:71-103.

Dufva, M. 2009a. Fabrication of DNA microarray. MethodsMol. Biol. 529:63-79.

———. 2009b. Introduction to microarray technology. Meth-ods Mol. Biol. 529:1-22.

Dunn, J. 1974. Well separated clusters and optimal fuzzy par-titions. J. Cybernetics 4:95-104.

Eberwine, J. 1996. Amplification of mRNA populations usingaRNA generated from immobilized oligo(dT)-T7 primedcDNA. Biotechniques 20:584-591.

Futschik, M., and T. Crompton. 2004. Model selection andefficiency testing for normalization of cDNA microarraydata. Genome Biol. 5:R60.

Gentleman, R. C., and others. 2004. Bioconductor: Open soft-ware development for computational biology and bioinfor-matics. Genome Biol. 5:R80.

Gentry, T. J., G. S. Wickham, C. W. Schadt, Z. He, and J. Zhou.2006. Microarray applications in microbial ecologyresearch. Microb. Ecol. 52:159-175.

Gibbons, F. D., and F. P. Roth. 2002. Judging the quality ofgene expression-based clustering methods using geneannotation. Genome Res. 12:1574-1581.


55

Gresham, D., M. J. Dunham, and D. Botstein. 2008. Compar-ing whole genomes using DNA microarrays. Nat. Rev.Genet. 9:291-302.

Hancock, D., and others. 2005. maxdLoad2 and maxdBrowse:Standards-compliant tools for microarray experimentalannotation, data management and dissemination. BMCBioinformatics 6:264.

He, Z., L. Wu, M. W. Fields, and J. Zhou. 2005. Use of microar-rays with different probe sizes for monitoring gene expres-sion. Appl. Environ. Microbiol. 71:5154-5162.

Herold, K. E., and A. Rasooly. 2003. Oligo Design: A computerprogram for development of probes for oligonucleotidemicroarrays. Biotechniques 35:1216-1221.

Hoffman, R., T. Seidl, and M. Dugas. 2002. Profound effect ofnormalization on detection of differentially expressedgenes in oligonucleotide microarray analysis. Genome Biol.3:research0033.1-research0033.11.

Huber, W., R. A. Irizarry, and R. Gentleman. 2005. Preprocess-ing overview, p. 3-12. In R. Gentleman, V. J. Carey, W.Huber, R. A. Irizarry, and S. Dudoit [eds.], Bioinformaticsand computational biology solutions using R and Biocon-ductor. Springer.

Hughes, A. L., and R. Friedman. 2005. Poxvirus genome evolu-tion by gene gain and loss. Mol. Phylogenet. Evol. 35:186-195.

Irizarry, R. A., B. Hobbs, F. Collin, Y. D. Beazer-Barclay, K. J.Antonellis, U. Scherf, and T. P. Speed. 2003. Exploration,normalization, and summaries of high density oligonu-cleotide array probe level data. Biostatistics 4:249-264.

———, L. M. Cope, and Z. J. Wu. 2006. Feature-level explo-ration of a published Affymetrix GeneChip control dataset.Genome Biol. 7:404.

Jain, A., and R. Dubes. 1988. Algorithms for clustering data.Prentice Hall.

Jain, N., J. Thatte, T. Braciale, K. Ley, M. O'connell, and J. K.Lee. 2003. Local-pooled-error test for identifying differen-tially expressed genes with a small number of replicatedmicroarrays. Bioinformatics 19:1945-1951.

Karaman, M. W., S. Groshen, C. C. Lee, B. L. Pike, and J. G.Hacia. 2005. Comparisons of substitution, insertion anddeletion probes for resequencing and mutational analysisusing oligonucleotide microarrays. Nucleic Acids Res. 33:e33.

Karsten, S. L., and D. H. Geshwind. 2002. Gene expressionanalysis using cDNA microarrays. Curr. Protoc. Neurosci.Chapter 4:Unit 4.28.

Kegel, J., M. J. Allen, K. Metfies, W. H. Wilson, D. Wolf-Gladrow, and K. Valentin. 2007. Pilot study of an ESTapproach of the coccolithophorid Emiliania huxleyi duringa virus infection. Gene 406:209-216.

Kerr, K. F., K. A. Serikawa, C. Wei, M. A. Peters, and R. E. Bum-garner. 2007. What is the best reference RNA? And otherquestions regarding the design and analysis of two-colormicroarray experiments. OMICS 11:152-165.

Kerr, M. K. 2003. Design considerations for efficient and effec-tive microarray studies. Biometrics 59:822-828.

———, and G. A. Churchill. 2001. Experimental design forgene expression microarrays. Biostatistics 2:183-201.

Khadijah, S., S. Y. Neo, M. S. Hossain, L. D. Miller, S. Mathavan,and J. Kwang. 2003. Identification of white spot syndromevirus latency-related genes in specific-pathogen-freeshrimps by use of a microarray. J. Virol. 77:10162-10167.

Lan, Y., X. Xu, F. Yang, and X. Zhang. 2006. Transcriptionalprofile of shrimp white spot syndrome virus (WSSV) geneswith DNA microarray. Arch. Virol. 151:1723-1733.

Letowski, J., R. Brousseau, and L. Masson. 2004. Designing bet-ter probes: Effect of probe size, mismatch position andnumber on hybridization in DNA oligonucleotide microar-rays. J. Microbiol. Methods 57:269-278.

Levine, E., and E. Domany. 2001. Resampling method forunsupervised estimation of cluster validity. Neural Com-put. 13:2573-2593.

Li, W., J. Huang, M. Fan, and S. Wang. 2002. MProbe: Com-puter aided probe design for oligonucleotide microarrays.Appl. Bioinformatics 1:163-166.

Li, X., Z. He, and J. Zhou. 2005. Selection of optimal oligonu-cleotide probes for microarrays using multiple criteria,global alignment and parameter estimation. Nucleic AcidsRes. 33:6114-6123.

Lindell, D., and A. F. Post. 2001. Ecological aspects of ntcAgene expression and its use as an indicator of the nitrogenstatus of marine Synechococcus spp. Appl. Environ. Micro-biol. 67:3340-3349.

———, and others. 2007. Genome-wide expression dynamicsof a marine virus and host reveal features of co-evolution.Nature 449:83-86.

Liu, W. J., Y. S. Chang, C. H. Wang, G. H. Kou, and C. F. Lo.2005. Microarray and RT-PCR screening for white spot syn-drome virus immediate-early genes in cyclohexamide-treated shrimp. Virology 334:327-341.

Liu, X. S. 2007. Getting started in tiling microarray analysis.PLoS Comput. Biol. 3:1842-1844.

Marks, H., O. Vorst, A. M. Van Houwelingen, M. C. Van Hulten,and J. M. Vlak. 2005. Gene-expression profiling of Whitespot syndrome virus in vivo. J. Gen. Virol. 86 Pt 7:2081-2100.

Martiny, A. C., M. L. Coleman, and S. W. Chisholm. 2006.Phosphate acquisition genes in Prochlorococcus ecotypes:Evidence for genome-wide adaptation. Proc. Natl. Acad. Sci.U. S. A. 103:12552-12557.

Millard, A. D., and B. Tiwari. 2009. Oligonucleotide microar-rays for bacteriophage expression studies. Methods Mol.Biol. 502:193-226.

———, K. Zwirglmaier, M. J. Downey, N. H. Mann, and D. J.Scanlan. 2009. Comparative genomics of marine cyanomy-oviruses reveals the widespread occurrence of Synechococ-cus host genes localized to a hyperplastic region: Implica-tions for mechanisms of cyanophage evolution. EnvironMicrobiol. Jun 7 [Epub ahead of print].

Monier, A., A. Pagarete, C. de Vargas, M. J. Allen, B. Read, J. M.Claverie, and H. Ogata. 2009. Horizontal gene transfer of


56

an entire metabolic pathway between a eukaryotic alga andits DNA virus. Genome Res. 19:1441-1449.

Moreira, D., and C. Brochier-Armanet. 2008. Giant viruses,giant chimeras: The multiple evolutionary histories ofMimivirus genes. BMC Evol. Biol. 8:12.

Morrison, N., and others. 2006. Annotation of environmentalOMICS data: Application to the transcriptomics domain.Omics 10:172-178.

Nielsen, H. B., R. Wernersson, and S. Knudsen. 2003. Design ofoligonucleotides for microarrays and perspectives fordesign of multi-transcriptome arrays. Nucleic Acids Res.31:3491-3496.

Nordberg, E. K. 2005. YODA: Selecting signature oligonu-cleotides. Bioinformatics 21:1365-1370.

Page, G. P., S. O. Zakharkin, K. Kim, T. Mehta, L. Chen, and K.Zhang. 2007. Microarray analysis. Methods Mol. Biol.404:409-430.

Pirooznia, M., P. Gong, J. Y. Yang, M. Q. Yang, E. J. Perkins, and Y.Deng. 2008. ILOOP—A web application for two-channelmicroarray interwoven loop design. BMC Genomics (suppl. 2)9:S11.

Reymond, N., H. Charles, L. Duret, F. Calevro, G. Beslon, andJ. M. Fayard. 2004. ROSO: Optimizing oligonucleotideprobes for microarrays. Bioinformatics 20:271-273.

Rich, V. I., K. Konstantinidis, and E. F. Delong. 2008. Designand testing of ‘genome-proxy’ microarrays to profile marinemicrobial communities. Environ Microbiol 10:506-521.

Robalino, J., and others. 2007. Insights into the immune tran-scriptome of the shrimp Litopenaeus vannamei: Tissue-spe-cific expression profiles and transcriptomic responses toimmune challenge. Physiol. Genomics 29:44-56.

Rocap, G., and others. 2003. Genome divergence in twoProchlorococcus ecotypes reflects oceanic niche differenti-ation. Nature 424:1042-1047.

Rouillard, J. M., M. Zuker, and E. Gulari. 2003. OligoArray 2.0:Design of oligonucleotide probes for DNA microarrays using athermodynamic approach. Nucleic Acids Res. 31:3057-3062.

Schretter, C., and M. C. Milinkovitch. 2006. OLIGOFAKTORY:A visual tool for interactive oligonucleotide design. Bioin-formatics 22:115-116.

Simon, R. 2008. Microarray-based expression profiling andinformatics. Curr. Opin. Biotechnol. 19:26-29.

Sipe, C. W., and M. S. Saha. 2007. The use of microarray tech-nology in nonmammalian vertebrate systems. MethodsMol. Biol. 382:1-16.

Smith, B., and others. 2007. The OBO Foundry: Coordinatedevolution of ontologies to support biomedical data integra-tion. Nat. Biotechnol. 25:1251-1255.

Smyth, G. K. 2004. Linear models and empirical Bayes meth-ods for assessing differential expression in microarrayexperiments. Stat. Applic. Genet. Mol. Biol. 3:Article 3.

———, J. Michaud, and H. S. Scott. 2005. Use of within-arrayreplicate spots for assessing differential expression inmicroarray experiments. Bioinformatics 21:2067-2075.

Southern, E. M. 1975. Detection of specific sequences amongDNA fragments separated by gel electrophoresis. J. Mol.Biol. 98:503-517.

Steglich, C., M. Futschik, T. Rector, R. Steen, and S. W. Chisholm.2006. Genome-wide analysis of light sensing in Prochloro-coccus. J. Bacteriol. 188:7796-7806.

———, M. E. Futschik, D. Lindell, B. Voss, S. W. Chisholm, andW. R. Hess. 2008. The challenge of regulation in a minimalphotoautotroph: Non-coding RNAs in Prochlorococcus.PLoS Genet. 4:e1000173.

Stenberg, J., M. Nilsson, and U. Landegren. 2005. ProbeMaker:An extensible framework for design of sets of oligonu-cleotide probes. BMC Bioinformatics 6:229.

Sullivan, M. B., M. L. Coleman, P. Weigele, F. Rohwer, and S.W. Chisholm. 2005. Three Prochlorococcus cyanophagegenomes: Signature features and ecological interpretations.PLoS Biol. 3:790-806.

Tolonen, A. C., and others. 2006. Global gene expression ofProchlorococcus ecotypes in response to changes in nitro-gen availability. Mol. Sys. Biol. 2:53.

Tsai, J. M., H. C. Wang, J. H. Leu, H. H. Hsiao, A. H. Wang, G.H. Kou, and C. F. Lo. 2004. Genomic and proteomic analy-sis of thirty-nine structural proteins of shrimp white spotsyndrome virus. J. Virol. 78:11360-11370.

Tusher, V. G., R. Tibshirani, and G. Chu. 2001. Significanceanalysis of microarrays applied to the ionizing radiationresponse. Proc. Natl. Acad. Sci. U. S. A. 98:5116-5121.

Vera, J. C., C. W. Wheat, H. W. Fescemyer, M. J. Frilander, D. L.Crawford, I. Hanski, and J. H. Marden. 2008. Rapid tran-scriptome characterization for a nonmodel organism using454 pyrosequencing. Microb. Ecol. 17:1636-1647.

Wang, D., L. Coscoy, M. Zylberberg, P. C. Avila, H. A. Boushey,D. Ganem, and J. L. Derisi. 2002. Microarray-based detec-tion and genotyping of viral pathogens. Proc. Natl. Acad.Sci. U. S. A. 99:15687-15692.

———, and others. 2003. Viral discovery and sequence recov-ery using DNA microarrays. PLoS Biol. 1:E2.

Wang, X., and B. Seed. 2003. Selection of oligonucleotide probesfor protein coding sequences. Bioinformatics 19:796-802.

Wilson, W. H., and others. 2005. Complete genome sequenceand lytic phase transcription profile of a Coccolithovirus.Science 309:1090-1092.

———, J. L. Van Etten, and M. J. Allen. 2009. The Phycod-naviridae: The story of how tiny giants rule the world. Curr.Top. Microbiol. Immunol. 328:1-42.

Wit, E., and J. McClure. 2004. Statistics for Microarrays. Wiley.Yang, Y. H., S. Dudoit, P. Luu, D. M. Lin, V. Peng, J. Ngai, and

T. P. Speed. 2002. Normalization for cDNA microarraydata: A robust composite method addressing single andmultiple slide systematic variation. Nucleic Acids Res.30:e15.

Zinser, E. R., and others. 2009. Choreography of the transcrip-tome, photophysiology, and cell cycle of a minimal pho-toautotroph, prochlorococcus. PLoS ONE 4:e5135.

57

Introduction:

Archaea have been shown by molecular techniques to bewidespread in many ecosystems (e.g., Chaban et al. 2006),

but to date only about 50 viruses have been reported thatinfect this large and diverse group of organisms (Prangishviliet al. 2006). Most archaeal viruses have been isolated fromeither extreme thermoacidophiles or extreme halophiles(Prangishvili et al. 2006; Porter et al. 2007). This work waspioneered by the late Wolfram Zillig, but was not systemati-cally addressed until the work of Prangishvili and Dyall-Smith, respectively. Early virus isolates of extreme halophiles(haloviruses) were of the head-and-tail type, the same mor-phology observed in more than 90% of described bacterio-phages; more recent isolates have included representatives ofspindle-shaped and spherical morphotypes (Porter et al.2007; Pietila et al. 2009). In contrast, none of the viruses ofextremely thermophilic crenarchaea are of the head-and-tailtype, but show a fascinating variety of unique morphologiesand genomes, indicating that we have only just begun toappreciate the diversity of archaeal viruses (Prangishvili et al.2006).

Viruses in high temperature acidic environments are sur-prisingly low in abundance, commonly 103 per mL, as deter-mined by either nucleic acid staining techniques (Ortmann etal. 2006; Prangishvili et al. 2006) or direct counting of virus-like particles (VLP) (Ortmann et al. 2006; Prangishvili et al.2006). The reason for this is unknown. Therefore almost all of

The isolation of viruses infecting ArchaeaKenneth M. Stedman1, Kate Porter2, and Mike L. Dyall-Smith3

1Department of Biology, Center for Life in Extreme Environments, Portland State University, P.O. Box 751, Portland, OR 97207,USA2Biota Holdings Limited, 10/585 Blackburn Road, Notting Hill Victoria 3168, Australia3Max Planck Institute of Biochemistry, Department of Membrane Biochemistry, Am Klopferspitz 18, 82152 Martinsried,Germany

AbstractA mere 50 viruses of Archaea have been reported to date; these have been investigated mostly by adapting

methods used to isolate bacteriophages to the unique growth conditions of their archaeal hosts. The most numer-ous are viruses of thermophilic Archaea. These viruses have been discovered by screening enrichment culturesand novel isolates from environmental samples for their ability to form halos of growth inhibition, or by usingelectron microscopy to screen enrichment cultures for virus-like particles. Direct isolation without enrichmenthas not yet been successful for viruses of extreme thermophiles. On the other hand, most viruses of extremehalophiles, the second most numerous archaeal viruses, have been isolated directly from hypersaline environ-ments. Detailed methods for the isolation of viruses of extremely thermoacidophilic Archaea and extremelyhalophilic Archaea are presented in this manuscript. These methods have been extremely effective in isolatingnovel viruses. However, Archaea comprise much more than extreme thermoacidophiles and extreme halophiles.Therefore a vast pool of archaeal viruses remain to be discovered, isolated, and characterized, particularly amongthe methanogens and marine Archaea. Some suggestions for expansion of the described methods are discussed.We hope these suggestions will provide an impetus for future work on these and other Archaeal viruses.


Acknowledgments:Publication costs for the Manual of Aquatic Viral Ecology were pro-


The authors would like to thank D. Prangishvili for his insights andvast knowledge of the isolation of viruses of extremely thermophilicArchaea. KS would like to thank D. Grogan for the suggestion of PEG400 as a bath liquid. The authors would also like to thank an anony-mous reviewer whose suggestions greatly improved the manuscript.Research in the Stedman lab is supported by NSF (MCB: 0702020) andNASA (NNX07AJ26G and NNX07AT63A). MDS is grateful to D.Oesterhelt, and the Department of Membrane Biochemistry, MPI, fortheir continuing support.

ISBN 978-0-9845591-0-7, DOI 10.4319/mave.2010.978-0-9845591-0-7.57Suggested citation format: Stedman, K. M., K. Porter, and M. L. Dyall-Smith. 2010. The isolationof viruses infecting Archaea, p. 57–64. In S. W. Wilhelm, M. G. Weinbauer, and C. A. Suttle[eds.], Manual of Aquatic Viral Ecology. ASLO.


MANUALof


the viruses isolated from thermoacidophilic Archaea comefrom enrichment cultures of environmental samples.

Hypersaline waters are similar to marine ecosystems, withhigh VLP counts, commonly around 108 VLP per mL (Guixa-Boixareu et al. 1996; Oren et al. 1997; Diez et al. 2000; Pedros-Alio et al. 2000a; Danovaro et al. 2005; Bettarel et al. 2006).Despite the high virus levels, low cell growth rates, and fre-quent observations of VLPs inside cells, some studies predictthat haloviruses are not major regulating factors of commu-nity size (Guixa-Boixareu et al. 1996; Pedros-Alio et al. 2000a;Pedros-Alio et al. 2000b). Although the viral role in microbialpopulation control remains unclear, high virus numbers indi-cate that they should be readily isolable directly from watersamples but, to date, only about 21 well-described haloviruseshave been reported in the published literature (Pagaling et al.2007; Porter et al. 2007; Pietila et al. 2009).

Methanogens are the first-identified and probably best- characterized members of the Archaea; however, reports oftheir viruses are surprisingly sparse in the literature, with onlythree different viruses described, one characterized in detail,and ten viruses or proviruses reported. Viruses of methanogenicArchaea have been isolated from anaerobic sludge digesters(Meile et al. 1989; Nolling et al. 1993) and found in super-natants of known cultures (Wood et al. 1989). It is unclearwhether this lack of published viruses is due to the low abun-dance of viruses of Methanogens or insufficient screening.However, a recent bioinformatic analysis of the incompletegenome of Methanococcus voltae strain A3 indicated the pres-ence of at least two different proviruses (Krupovic and Bamford2008) and highlights the need for further study of viruses ofmethanogenic Archaea.

This manuscript gives methods for isolation of viruses ofthe thermoacidophilic archaeon Sulfolobus and close relativesand viruses from hypersaline waters. Methods used for the iso-lation of viruses of other thermophilic Archaea are also dis-cussed. Similar methods have been used for isolation of thefew viruses of methanogenic Archaea but are not discussed indetail here.

Materials and Procedures:Viruses from Sulfolobus and close relatives—The following

methods are basically method “A” of Zillig et al. (1994) andwere described recently in detail by Prangishvili (2006). Thesemethods consist of enrichment cultures followed by host iso-lation and screening for virus production in both these iso-lates and enrichment cultures. These techniques are very sim-ilar to those used for bacteriophages, with the major exceptionbeing the extreme growth conditions (80˚C, pH 3).

Preparation of anaerobic tubes for sample transport—For eachsample to be collected, one anaerobic collection vessel (Fig. 1Ainset) is prepared. A small amount (ca. 50–220 mg) of elemen-tal sulfur (e.g., Riedel-deHaën) is placed in an anaerobic tube,and 0.1 mL of a 2% resazurin solution and 0.1 mL of water sat-urated with H2S are added (a fresh Na2S solution can also be

used for reduction, but with less success; Stedman, unpub-lished). The air in the tube is displaced with CO2 and N2 by theHungate technique and the tube is stoppered (Hungate et al.1966). A cap is placed on the tube and the assemblage auto-claved. A gas phase of 160 kPa of CO2 and 1 kPa of H2S has alsobeen used successfully (Prangishvili 2006).

Sample collection and transportation for thermoacidophilicArchaea—Liquid and wet sediment samples are collected fromturbid terrestrial hot springs with high temperature >70˚C andlow pH <4. The pH is often tested with pH paper because it isless susceptible to temperature changes than most pH elec-trodes. Samples are collected in sterile 50-mL conical flasks atthe end of an extendible pole with a clamp (see Fig. 1A). Aftermost of the sediment is allowed to settle, the pH of the liquidis carefully adjusted to ca. 5.5 with solid CaCO3 by slow addi-tion and stirring. Once the pH is adjusted the sample is trans-ferred to a pre-prepared anaerobic tube using a syringe (seeabove and Fig. 1A inset). If the resazurin indicator changes topink, drops of H2S-saturated water are added until the sampleclears. Samples can be maintained for up to 2 weeks at roomtemperature before enrichment.

Alternative sample collection—If the laboratory is relativelyclose to the sampling location, water and sediment samplesare collected as above, but instead of an anaerobic tube, a ster-ile screw-cap vial or centrifuge bottle is completely filled sothat very little air is present. Samples can then be transportedat ambient temperature and should be enriched within 8–10 hof collection (Rice et al. 2001).

Enrichment culture for host and virus isolation—Samples col-lected either in anaerobic tubes or filled centrifuge tubes arediluted 1:50 or 1:100 in Sulfolobus growth medium (Zillig et al.1994) containing either yeast extract (0.1% w/v) and sucrose(0.2% w/v) as carbon sources or Tryptone (0.2% w/v) in long-necked Erlenmeyer flasks (see Fig. 1B inset), and incubated at80˚C with shaking (150 rpm) for up to 2 weeks. The salts inSulfolobus growth medium are, per liter: 3 g (NH4)2SO4, 0.5 gK2HPO4 × 3 H2O, 0.1 g KCl, 0.5 g MgSO4 × 7 H2O, 0.01 gCa(NO3)2 × 4 H2O, 1.8 mg MnCl2 × 4 H2O, 4.5 mg Na2B4O7 ×10 H2O, 0.22 mg ZnSO4 × 7 H2O, 0.05 mg CuCl2 × 2 H2O, 0.03mg Na2MoO4 × 2 H2O, 0.03 mg VOSO4 × 5 H2O, 0.01 mgCoSO4 × 7 H2O. The medium was buffered with 0.7 g glycineper liter and the pH was adjusted to pH 3–3.5 with 1:2 dilutedsulfuric acid. For long-term 80˚C growth, our favorite bath liq-uid is PEG 400, which is a noncorrosive, nontoxic, water sol-uble compound that does not evaporate (see Fig. 1B); mineraloil and water can be used as bath liquid but are suboptimaldue to cleanup and evaporation, respectively. When growth isdetected by either an increase in turbidity or production of acharacteristic “damp sock” odor (W. Zillig pers. comm.), sam-ples are plated on Gelrite® plates (see below and Fig. lC), redi-luted 1:50, and screened for VLP production by a spot-on-lawn assay (see below and Fig. 1D) or electron microscopy (seebelow and Fig. 2). The second round of enrichment culture isalso plated and screened for virus production.

Stedman et al. Archaeal virus isolation

58

Plating on Gelrite plates for host and virus isolation—Plates aremade by slowly adding 6–10 grams/L Gelrite (Kelco) to Sul-folobus media (see above) and boiling until dissolved. Gelrite, axanthan gum, is used instead of agar because set Gelrite plates

remain solid up to 90˚C. Alternatively, a 2 × Gelrite concen-trate (12–20 g/L) is made in water, melted by boiling, andadded to an equal volume of 2 × concentrated Sulfolobusmedium (Grogan 1989). Calcium (Ca(NO3)2) and magnesium(MgCl2) are added to a final concentration of 1.5 and 5 mM,respectively, to stabilize the gel. Before the gel solidifies, ca.25 mL is poured into standard (90 mm) Petri plates with cams.After the Gelrite solidifies, plates can be stored at 4˚C indefi-nitely. Approximately 0.1 mL, from undiluted to 10–3, ofenrichment cultures are spread on Gelrite plates in the pres-ence or absence of 0.5 mL 0.2% Gelrite dissolved in Sulfolobusmedium. Plates are incubated inverted in airtight moist con-tainers at 75–80˚C for approximately 1 week before coloniesappear (Fig. 1C). Multiple wet paper towels and a 90-mm Petridish filled with water at the bottom of a sealable container(e.g., Tupperware®) is sufficient.

Spot-on-lawn (halo) assay for screening enrichment cultures andisolates for viruses—This protocol is based on Schleper et al.(1992) as modified by Stedman et al. (2003). Gelrite plates arepreincubated ca. 10 min at 80˚C to dry, then 10 mL of Sul-folobus medium with ca. 0.2 % (w/v) Gelrite is boiled to dis-solve the Gelrite. This “softlayer” is allowed to cool slightly (toca. 80˚C). Approximately 3 mL of softlayer are added to ca. 0.2mL of exponentially growing host cells, generally Sulfolobus sol-fataricus, and spread on a plate by swirling. After the Gelritesolidifies, 1–2 µL of culture or supernatant to be screened isspotted on the plate. For a positive control, 1 µL of a 0.01%(v/v) Triton X-100 solution is spotted. Plates are incubated asabove for 2–3 d and plates examined for clearing around spots(Fig. 1D).

Electron microscopy for virus identification and virus assemblagecharacterization—Generally, 5µL of an enrichment culture, or


59

Fig. 1: Pictorial overview of isolation of Sulfolobus viruses. (A) Wolfram Zillig sampling at a typical Sulfolobus-containing pool in Yellowstone NationalPark, USA, September 2000 (inset shows anaerobic tubes with samples). (B) 80˚C incubator with long-necked growth flasks (detail in inset). (C) Single-colony isolates of Sulfolobus solfataricus on a Gelrite® plate. This plate contains a mixture of S. solfataricus containing (blue colonies) and lacking (brown)a vector expressing the lacS gene from S. solfataricus and was sprayed with 5-bromo-4-chloro-3-indolyl-b-D-galactopyranoside (X-gal) (see Jonuscheit etal. 2003). (D) Lawn of S.solfataricus strain P1 with halos of growth inhibition due to virus production by 2-µl spots of virus-infected strains. Spots labeledSV2P1 and SV2P2 are from S. solfataricus strains P1 and P2 infected with SSV-I2 respectively (Stedman et al. 2003). Spot labeled C is a detergent-posi-tive control. Spot labeled P2- is uninfected S. solfataricus strain P2 as a negative control.

Fig. 2: TEM of Sulfolobus viruses and VLPs. (A) Sulfolobus spindle-shapedvirus SSV-I2 particles. (B) Sulfolobus turreted icosahedral virus (STIV). (C)Three different VLPs from an enrichment culture from AmphitheaterSprings, Yellowstone National Park, USA. Note end of a Sulfolobus islandi-cus rod-shaped virus (SIRV)-like particle in upper right of image). (D)Virus-like particles from Amphitheater Springs. All scale bars 200 nm.Negative stain with uranyl acetate.

0.2 µm filtered and centrifuged (10 min at 3000g) cell-freesupernatant, is spotted onto carbon/formvar-coated electronmicroscope grids (Ted Pella or EM Sciences), allowed to absorbfor 2 min, and then stained with 0.2% (w/v) uranyl-acetate for15–30 s. Samples are examined by transmission electronmicroscopy (TEM), e.g., JEOL 100 cx, operated at 100 keV.VLPs can generally be discerned at ×16,000–20,000 magnifica-tion (Fig. 2). Generally this method is successful only if thereis an indication for the presence of virus, for instance a haloon a lawn. Even when halos are formed, finding viruses byTEM can be challenging; often supernatants are concentrated 10- through 1000-fold by ultrafiltration or ultracentrifugation(Rice et al. 2001).

Viruses from hypersaline waters—Artificial salt water andmedium MGM: Artificial salt water solutions are designed tomimic the natural concentrated brines where haloarchaeaare found. The formulation used by M. Dyall-Smith(described in the online handbook, the Halohandbook,http://www.haloarchaea.com/resources/halohandbook/) isbased on that described by Rodriguez-Valera et al.(Rodriguez-Valera et al. 1980; Torreblanca et al. 1986). Perliter, it contains 4 M NaCl, 150 mM MgCl2, 150 mM MgSO4,90 mM KCl, 3.5 mM CaCl2, adjusted to pH 7.5 using ca. 2 mL1 M Tris-HCl (pH 7.5). At 30% w/v, the total salts are presentin a much higher concentration than in seawater, but inapproximately the same proportions. Adjustments of Mg2+,pH, or other conditions may be necessary for specific haloar-chaeal groups.

Modified growth medium (MGM) contains 5 g peptone and1 g yeast extract per liter of salt water. The salt concentrationis varied according to the host strain (see below), and isdetailed in the Halohandbook.

Isolation of haloviruses from natural waters—Salt lake sam-ples are collected from hypersaline waters, which typicallyrange from 15% w/v total salt, up to saturation (ca. 35%).Samples are collected in sterile 5–10 mL vessels and may bestored for several weeks at room temperature. In the labora-tory, cells and cellular debris are removed by centrifugation(5,000g, 10 min, room temperature). The supernatant is thenscreened directly for viruses by plaque assay. The use of chlo-roform is avoided, because it is known to have a detrimentalaffect on both phage-like and lipid-containing haloviruses.The choice of host strains depends on the experimentalobjectives, and includes well-characterized members of theHalobacteriaceae, such as Hbt. salinarum (host for ΦH andseveral others) or Har. hispanica (host for SH1, His1, His2, andothers), or natural isolates from the same source, such as Hrr.coriense (host for HF2). To maximize isolation success, severalhosts should be used for the same sample. The advantage ofthe use of characterized hosts is that methods for geneticmanipulation are often established and their genomesequences have been determined.

Base and overlay plates (90 mm) are made with MGM (seeabove), solidified using 1.5% w/v agar. A range of salt water

concentrations should be examined, because salt concentra-tion seems to greatly affect the size and clarity of plaques.Using salt water concentrations that are 2% to 5% lower thanthe optimum for host growth commonly gives better plaques.Incubation temperature is also important, because somehaloviruses plaque poorly or not at all at 37–42°C, whereasthey form clear plaques at 30°C. Plates can be stored indefi-nitely at 4°C (wrapped in plastic to prevent dessication), butshould be warmed to room temperature or warmer for use. Forvirus isolation, 100–500 µl of the cleared water sample is mixedwith 150 µl of exponentially growing host cells. These may becharacterized strains of haloarchaea, or natural isolates, forexample, isolates from the natural water sample. Then, 3–4 mLof molten (50°C) top-layer MGM (with 0.7% w/v agar) isadded, and the solution mixed gently and poured evenly overthe plate. After setting on a level surface for 5–10 min, platesare incubated aerobically, inverted in airtight containers at30°C and 37°C for 1–4 d, and checked every day for plaques.

Any visible plaques are picked using sterile glass Pasteurpipettes, or sterile plastic micropipette tips. These agar plugsare then transferred to tubes containing 500 µl of halovirusdiluent (2.47 M NaCl, 90 mM MgCl2, 90 mM MgSO4, 60 mMKCl, 3 mM CaCl2, 10 mM Tris-HCl pH 7.5), and vortexed tohomogenize the sample. These suspensions are thenreplaqued on overlay plates to purify the isolates and to elim-inate “false plaques” caused by artifacts in the agar overlay orcontaminants in the water sample.

Isolation from lysogens—Several haloviruses have been iso-lated from laboratory strains of haloarchaea. Most were inad-vertent discoveries, based on the spontaneous lysis of the hostculture (e.g., ΦH, ΦCh1), or the detection of virus particles inpurified preparations of flagella (Hs1). A more systematicapproach would be to use induction by mitomycin C, andthen to plaque cell supernatants on related (nonlysogenic)host strains. Indeed, this has been recently used to isolate anew halovirus, SNJ1 (Mei et al. 2007), from a strain ofNatrinema, and could be used more widely.

Electron microscopy—Standard negative stain EM works beston samples with low salt concentrations, but manyhaloviruses are stable only at high salt concentrations. If oneuses high salt preparations, the salts can crystallize on thegrid, occluding the particles and heating up the specimen.One way to overcome the problem is to first fix the sampleusing gluteraldehyde. Another problem is poor adsorption toplastic-coated grids (e.g., formvar). Pretreatment of the gridswith poly-L-lysine can alleviate this issue. The followingmethod for examining haloviruses was adapted from thatdescribed by Tarasov et al. (2000). A sample of virus is placedon a sterile surface and the grid, plastic-coated side down,placed on the droplet for 1.5–2 min. The grid is then placed,for 1–1.5 min, on a drop of freshly filtered 2% w/v uranylacetate and excess stain removed with filter paper. After airdrying, grids may be examined by transmission electronmicroscopy, as described above (Fig. 3A, B, and C).


60

Assessment

Isolation of viruses of thermophilic Archaea—Use of the meth-ods described above has been highly successful with theunprecedented discovery of three new virus families(Fuselloviridae, Rudiviridae, and Lipothrixviridae), one floatinggenus (Guttaviridae) and one proposed family (Turriviridae) ofviruses just from Sulfolobus hosts (reviewed in Stedman et al.2005). Approximately 10% of samples collected from Icelandicand other hot springs yielded viruses or other extrachromoso-mal elements on enrichment (Stedman et al. 2005). Anaerobicenrichment by Prangishvili and coworkers using otherwisesimilar procedures has allowed the isolation of three addi-tional virus families (Prangishvili et al. 2006). A plethora ofFuselloviruses have also been isolated (Martin et al. 1984;Schleper et al. 1992; Arnold et al. 1999; Stedman et al. 2003;Wiedenheft et al. 2004; Stedman et al. 2006; Peng 2008) (Fig.2A). Despite this success, eight new virus families each with adifferent morphotype, and on the order of 30 unique viruses,this is undoubtedly an undersampling of the diversity andprevalence of viruses in acidic hot springs, let alone in otherenvironments. It is highly likely that these techniques andmodifications thereof will allow the isolation of more anddiverse viruses. The current limitation seems to be more lack

of manpower than technique. Beyond manpower, moreprogress in host isolation and cultivation is likely to be themost critical step in allowing the discovery of more viruses.

Isolation of haloviruses—In the early days of halovirusresearch (1974–1993), deliberate virus isolation from naturalhypersaline waters was uncommon. Major exceptions to thiswere the superb ecological studies of haloviruses reported byDaniels and Wais, who isolated Halobacterium species andtheir viruses from Jamaican salt lakes and noted the signifi-cant effect of salt concentration on virulence (Wais et al. 1975;Wais and Daniels 1985; Daniels and Wais 1990). Currently,there are only around 21 described haloviruses, of which 17belong to the Caudovirales, 2 are members of the Salterprovirusgroup, and two are as yet unclassified (SH1, HRPV-1). Themore recent isolates are morphologically and genetically morediverse (spindle and round morphotypes), and most were iso-lated directly from hypersaline water sources using methodsdescribed above (Porter et al. 2007). Currently, about 10haloviruses are under active study (ΦCh1, BJ1, HF1, HF2, His1,His2, HRPV-1, SNJ1, and SH1), and these examples encompassthe three known dominant morphotypes—head-and-tail,spindle-shaped and round—of haloviruses so far observed bydirect EM of natural waters (Guixa-Boixareu et al. 1996; Orenet al. 1997; Diez et al. 2000; Santos et al. 2007). By negativestain TEM, HRPV-1 particles are reported to be pleomorphic(Pietila et al. 2009), but because this method often distorts par-ticles (because of the low salt), it will be important to confirmthis by cryo-EM. Filamentous VLPs, observed by F. Santos andcolleagues (Santos et al. 2007) must be isolated to prove thatthey are not dissociated tail fragments from the head-and-tailVLPs. Nevertheless, because the cultivation barrier of haloar-chaea has recently been overcome (Bolhuis et al. 2004; Burnset al. 2004a; Burns et al. 2004b), a better representation of thetrue viral population of salt lakes is now possible, and progressin field should improve dramatically.

DiscussionIsolation of viruses of thermophilic Archaea—Sulfolobus are

often not the dominant organisms in hot springs with tem-perature >70˚C and pH <4 (Snyder et al. 2004). Furthermore,Sulfolobus virus sequence diversity decreases with enrichment(Snyder et al. 2004). Therefore it is likely that the virusesreported to date are considerably fewer and less diverse thanthe viruses present in situ. Direct TEM imaging of concen-trated samples from both acidic and neutral hot springs indi-cate that a number of viruses with novel morphology remainto be isolated. (Rice et al. 2001; Rachel et al. 2002).

Comparison of methods for isolation of haloviruses and virusesof thermophilic Archaea—The main difference between the twomain techniques described here is due to the relative abun-dance of viruses and VLPs in the environments of the hosts.There are many more viruses in hypersaline environmentsthan in thermoacidophilic ones. Therefore direct isolationhas been successful for haloviruses, but not for viruses of ther-


61

Fig. 3: Electron micrographs of haloviruses and VLPs. (A) Spindle-shapedparticles of His1 virus (host is Har. hispanica). (B) Spherical particles of SH1virus. Also seen is a flagellar filament from the host (Har. hispanica). (C)Head-tail VLPs, and other structures, seen in a natural hypersaline watersample (Serpentine lake, Rottnest Island, Western Australia). All scale bars200 nm. Negative stain with uranyl acetate.

moacidophiles. The conditions for host growth are also verydifferent. Halovirus hosts grow at moderate temperatures, butat saturating salt conditions, whereas the thermoacidophilesgrow only at temperatures greater than 70˚C, requiring theuse of Gelrite for plates and lawns and the use of long-neckedflasks and PEG400 bath fluid for liquid culture.

Isolation of viruses from other Archaea—The Sulfolobales arerelatively well studied, but are only one relatively small groupof Archaea (Huber and Stetter 2001). Of the other Archaea,only the viruses of extreme halophiles viruses have been stud-ied in any depth (Porter et al. 2007). A few VLPs have beenobserved and one genome has been sequenced from enrich-ment cultures from deep-sea hydrothermal vent samples atvery high temperatures (Geslin et al. 2003a; Geslin et al.2003b; Geslin et al. 2007). Very little work has been done withmethanogen viruses. Those that have been characterizedappear to be more like bacterial Caudoviruses than the char-acterized viruses of thermoacidophilic Archaea. There are twoexceptions, the VLP reported by Wood et al. (1989), and twopossible proviruses in the Methanococcus voltae A3 draftgenome sequence (Krupovic and Bamford, 2008). Nothing isknown of viruses of the extremely abundant mesophilicArchaea that are present in soils and the oceans (reviewed inChaban et al. 2006). The long-awaited isolation of one of thelatter, Nitrosopumilus maritimus by Stahl and coworkers,should allow screening to take place (Konneke et al. 2005).

The genome sequences of many uncultured Archaea mayprovide clues from possible proviruses in their sequences thatwill allow the molecular screening of virus-sized samples fromthe oceans or soils without the need for cultivation of hosts,which is the critical bottleneck in the study of archaealviruses. Additionally, metagenomic projects allow the identifi-cation of further viruses. For example, a halovirus-likesequence, EHP-1, has been recovered directly from a crystal-lizer pond, although the virus itself has yet to be isolated (San-tos et al. 2007). Recently the genome of an apparentlyarchaeal virus from an extremely acidic acid mine drainagesite was cleverly determined by analysis of CRISPR sequencesin a metagenome databank (Andersson and Banfield, 2008).The human gut metagenome project may also provide someclues to the presence of currently undetected viruses ofArchaea. Methanogens are associated with gum disease (Leppet al. 2004) and have long been known to be in human gutsamples (e.g., Nottingham and Hungate 1968). Therefore, it ishighly likely that their viruses are also present.

Clearly, there is a great deal remaining to be discovered inviruses of Archaea. Implementation and expansion of themethods described and proposed herein should greatly stimu-late progress in the discovery, characterization and under-standing of this understudied group of viruses.

ReferencesAndersson, A.F., and J. F. Banfield. 2008. Virus population

dynamics and acquired virus resistance in natural microbial

communities. Science 320:1047-1050.Arnold, H. P., and others. 1999. The genetic element pSSVx of

the extremely thermophilic crenarchaeon Sulfolobus is ahybrid between a plasmid and a virus. Mol. Microbiol.34:217-226.

Bettarel, Y., M. Bouvy, C. Dumont, and T. Sime-Ngando. 2006.Virus-bacterium interactions in water and sediment of WestAfrican inland aquatic systems. Appl. Environ. Microbiol.72:5274-5282.

Bolhuis, H., E. M. T. Poele, and F. Rodriguez-Valera. 2004. Iso-lation and cultivation of Walsby’s square archaeon. Envi-ron. Microbiol. 6:1287-1291.

Burns, D. G., H. M. Camakaris, P. H. Janssen, and M. L. Dyall-Smith. 2004a. Combined use of cultivation-dependent andcultivation-independent methods indicates that membersof most haloarchaeal groups in an Australian crystallizerpond are cultivable. Appl. Environ. Microbiol.70:5258-5265.

———, ———, ———, and ———. 2004b. Cultivation ofWalsby’s square haloarchaeon. FEMS Microbiol. Lett.238:469-473.

Chaban, B., S. Y. Ng, and K. F. Jarrell. 2006. Archaeal habitats—from the extreme to the ordinary. Canad. J. Microbiol.52:73-116.

Daniels, L. L., and A. C. Wais. 1990. Ecophysiology of bacte-riophage S5100 infecting Halobacterium cutirubrum. Appl.Environ. Microbiol. 56:3605-3608.

Danovaro, R., C. Corinaldesi, A. Dell’anno, M. Fabiano, and C.Corselli. 2005. Viruses, prokaryotes and DNA in the sedi-ments of a deep-hypersaline anoxic basin (DHAB) of theMediterranean Sea. Environ. Microbiol. 7:586-592.

Diez, B., J. Anton, N. Guixa-Boixereu, C. Pedros-Alio, and F.Rodriguez-Valera. 2000. Pulsed-field gel electrophoresisanalysis of virus assemblages present in a hypersaline envi-ronment. Int. Microbiol. 3:159-164.

Geslin, C., M. Le Romancer, G. Erauso, M. Gaillard, G. Perrot,and D. Prieur. 2003a. PAV1, the first virus-like particle iso-lated from a hyperthermophilic euryarchaeote, “Pyrococcusabyssi.” J. Bacteriol. 185:3888-3894.

———, ———, M. Gaillard, G. Erauso, and D. Prieur. 2003b.Observation of virus-like particles in high temperatureenrichment cultures from deep-sea hydrothermal vents.Res. Microbiol. 154:303-307.

———, and others. 2007. Analysis of the first genome of ahyperthermophilic marine virus-like particle, PAV1, iso-lated from Pyrococcus abyssi. J. Bacteriol. 189:4510-4519.

Grogan, D. W. 1989. Phenotypic characterization of thearchaebacterial genus Sulfolobus: comparison of five wild-type strains. J. Bacteriol. 171:6710-6719.

Guixa-Boixareu, N., J. I. Calderón-Paz, M. Heldal, G. Bratbak,and C. Pedrós-Alió. 1996. Viral lysis and bacterivory asprokaryotic loss factors along a salinity gradient. Aquat.Microb. Ecol. 11:215-227.

Huber, H., and K. O. Stetter. 2001. Family Sulfolobaceae, p. 198.


62

In D. Boone, R. Castenholz and G. Garrity [eds.], Bergey’smanual of systematic bacteriology. Springer.

Hungate RE, W. Smith W., and R.T. Clarke. 1966 Suitability ofbutyl rubber stoppers for closing anaerobic roll culturetubes. J. Bacteriol. 91(2):908-909.

Jonuscheit, M., E. Martusewitsch, K. M. Stedman, and C.Schleper. 2003. A reporter gene system for the hyperther-mophilic archaeon Sulfolobus solfataricus based on a selec-table and integrative shuttle vector. Mol. Microbiol.48:1241-1252.

Konneke, M., A. E. Bernhard, J. R. De La Torre, C. B. Walker, J.B. Waterbury, and D. A. Stahl. 2005. Isolation of anautotrophic ammonia-oxidizing marine archaeon. Nature437:543-546.

Krupovic, M., and D. H. Bamford. 2008. Archaeal provirusesTKV4 and MVV extend the PRD1-adenovirus lineage to thephylum Euryarchaeota. Virology 375:292-300.

Lepp, P. W., M. M. Brinig, C. C. Ouverney, K. Palm, G. C.Armitage, and D. A. Relman. 2004. Methanogenic Archaeaand human periodontal disease. Proc. Natl. Acad. Sci. USA101:6176-6181.

Martin, A., S. Yeats, D. Janekovic, W. D. Reiter, W. Aicher, andW. Zillig. 1984. SAV 1, a temperate U.V.-inducible DNAvirus-like particle from the archaebacterium Sulfolobus aci-docaldarius isolate B12. The EMBO J. 3:2165-2168.

Mei, Y., J. Chen, D. Sun, D. Chen, Y. Yang, P. Shen, and X.Chen. 2007. Induction and preliminary characterization ofa novel halophage SNJ1 from lysogenic Natrinema sp. F5.Can. J. Microbiol. 53:1106-1110.

Meile, L., U. Jenal, D. Studer, M. Jordan, and T. Leisinger. 1989.Characterization of PsiM1, a virulent phage of Methanobac-terium thermoautotrophicum Marburg. Arch. Microbiol.152:105-110.

Nolling, J., A. Groffen, and W. M. De Vos. 1993. PHI-F1 andPHI-F3, two novel virulent, archaeal phages infecting dif-ferent thermophilic strains of the genus Methanobacterium.J. Gen. Microbiol. 139:2511-2516.

Nottingham, P. M., and R. E. Hungate. 1968. Isolation ofmethanogenic bacteria from feces of man. J. Bacteriol.96:2178-2179.

Oren, A., G. Bratbak, and M. Heldal. 1997. Occurrence ofvirus-like particles in the Dead Sea. Extremophiles 1:143-149.

Ortmann, A. C., B. Wiedenheft, T. Douglas, and M. Young.2006. Hot crenarchaeal viruses reveal deep evolutionaryconnections. Nature Rev. Microbiol. 4:520-528.

Pagaling, E., R. D. Haigh, W. D. Grant, D. A. Cowan, B. E.Jones, Y. Ma, A. Ventosa, and S. Heaphy. 2007. Sequenceanalysis of an archaeal virus isolated from a hypersalinelake in Inner Mongolia, China. BMC Genomics 8:410.

Pedros-Alio, C., and others 2000a. The microbial food webalong salinity gradients. FEMS Microbial. Ecol. 32:143-155.

———, J. I. Calderon-Paz, and J. M. Gasol. 2000b. Compara-tive analysis shows that bacterivory, not viral lysis, controls

the abundance of heterotrophic prokaryotic plankton.FEMS Microbial. Ecol. 32:157-165.

Peng, X. 2008. Evidence for the horizontal transfer of an inte-grase gene from a fusellovirus to a pRN-like plasmid withina single strain of Sulfolobus and the implications for plas-mid survival. Microbiol. 154:383-391.

Pietila, M. K., E. Roine, L. Paulin, N. Kalkkinen, and D. H.Bamford. 2009. An ssDNA virus infecting archaea: a newlineage of viruses with a membrane envelope. Mol. Micro-biol. 72:307-19.

Porter, K., B. E. Russ, and M. L. Dyall-Smith. 2007. Virus-hostinteractions in salt lakes. Curr. Opin. Microbiol. 10:418-424.

Prangishvili, D. 2006. Hyperthermophilic virus–host systems:Detection and isolation, p. 331-347 In F. Rainey and A.Ohren [eds.], Extremophiles. Academic Press.

Prangishvili, D., P. Forterre, and R. A. Garrett. 2006. Viruses ofthe Archaea: a unifying view. Nature Rev. Microbiol. 4:837-848.

Rachel, R., and others 2002. Remarkable morphological diver-sity of viruses and virus-like particles in hot terrestrial envi-ronments. Arch. Virology 147:2419-2429.

Rice, G., and others 2001. Viruses from extreme thermal envi-ronments. Proc. Natl. Acad. Sci. USA 98:13341-13345.

Rodriguez-Valera, F., F. Ruiz-Berraquero, and A. Ramos-Cor-menzana. 1980. Behaviour of mixed populations ofhalophilic bacteria in continuous cultures. Canad. J. Micro-biol. 26:1259-1263.

Santos, F., A. Meyerdierks, A. Pena, R. Rossello-Mora, R.Amann, and J. Anton. 2007. Metagenomic approach to thestudy of halophages: The environmental halophage 1. Env-iron. Microbiol. 9:1711-1723.

Schleper, C., K. Kubo, and W. Zillig. 1992. The particle SSV1from the extremely thermophilic archaeon Sulfolobus is avirus: demonstration of infectivity and of transfection withviral DNA. Proc. Natl. Acad. Sci. USA 89:7645-7649.

Snyder, J. C., J. Spuhler, B. Wiedenheft, F. F. Roberto, T. Dou-glas, and M. J. Young. 2004. Effects of culturing on the pop-ulation structure of a hyperthermophilic virus. Microb.Ecol. 48:561-566.

Stedman, K. M., and others. 2003. Relationships betweenfuselloviruses infecting the extremely thermophilicarchaeon Sulfolobus: SSV1 and SSV2. Res. Microbiol.154:295-302.

———, D. Prangishvili, and W. Zillig. 2005. Viruses of Archaea,p. 499-516. In R. Calendar [ed.], The bacteriophages, 2nded. S. T. Abedon, editorial assistant. Oxford Univ. Press.

———, A. J. Clore, and Y. Combet-Blanc. 2006. Biogeographi-cal diversity of archaeal viruses, p. 131-144. In N. A. Logan,H. M. Pappin-Scott, and P. C. F. Oynston [eds.], Prokaryoticdiversity: Mechanisms and significance. Cambridge Univ.Press.

Tarasov, V. Y., M. G. Pyatibratov, S. L. Tang, M. Dyall-Smith,and O. V. Fedorov. 2000. Role of flagellins from A and B loci


63

in flagella formation of Halobacterium salinarum. Mol.Microbiol. 35:69-78.

Torreblanca, M., F. Rodriguez-Valera, G. Juez, A. Ventosa, M.Kamekura, and M. Kates. 1986. Classification of non- alkaliphilic halobacteria based on numerical taxonomyand polar lipid-composition, and description of Haloarculagen. nov. and Haloferax gen. nov. System. Appl. Microbiol.8:89-99.

Wais, A. C., M. Kon, R. E. MacDonald, and B. D. Stollar. 1975.Salt-dependent bacteriophage infecting Halobacteriumcutirubrum and H. halobium. Nature 256:314-315.

———, and L. L. Daniels. 1985. Populations of bacteriophage

infecting Halobacterium in a transient brine pool. FEMSMicrobial. Ecol. 31:323-326.

Wiedenheft, B., and others. 2004. Comparative genomicanalysis of hyperthermophilic archaeal Fuselloviridaeviruses. J. Virology 78:1954-1961.

Wood, A. G., W. B. Whitman, and J. Konisky. 1989. Isolationand characterization of an archaebacterial viruslike particlefrom Methanococcus voltae A3. J. Bacteriol. 171:93-98.

Zillig, W., and others. 1994. Screening for Sulfolobales, theirplasmids and their viruses in Icelandic solfataras. System.Appl. Microbiol. 16:609-628.


64

65

Introduction

Viruses are known to infect and lyse a wide range ofautotrophs (see Brussaard 2004 for review). Because viral lysiscan affect the carbon fixation rates of eukaryotic and prokary-otic primary producers, this suggests a significant effect ofviruses on primary production-mediated carbon cycling (Suttle2007). Modeling exercises have estimated that between 6% and26% (Fuhrman 1999; Wilhelm and Suttle 1999; Ruardij et al.2005) of primary production may bypass higher trophic levelsbecause of viral-induced transformation of phytoplankton cells

to dissolved organic matter (viral shunt). However, althoughviruses are predicted to have a significant impact on primaryproduction, viral-induced mortality rates from natural watersare scarce.

Various methods have been developed to determine theimpact of viruses on microbial populations; however, mostwere developed for bacterioplankton (Proctor and Fuhrman1990; Heldal and Bratbak 1991; Steward et al. 1992; Weinbaueret al. 1993; Noble and Fuhrman 2000; Wilhelm et al. 2002;Parada et al. 2008) and may not be applicable for autotrophs.One of the most straightforward involves quantifying rates ofcell lysis based on net increases in viral abundance over time(Bratbak et al. 1990). However, this method is limited becauseit relies on the virus of interest being distinguishable fromother viruses in the sample, is only appropriate when viralnumbers are increasing, and requires the use of assumed burstsizes to calculate lysis rates. Also, unless loss through viraldecay is also accounted for, this approach provides only a min-imum estimate of viral productivity (Noble and Steward 2001),and the resulting productivity estimates are highly influencedby the frequency and timing of sampling (Bratbak et al. 1990).

The approach based on the frequency of visibly infectedcells (Proctor and Fuhrman 1990; Brussaard et al. 1996) isanother relatively straightforward method with the advantage

Estimation of viral-induced phytoplankton mortality using themodified dilution methodSusan A. Kimmance1* and Corina P. D. Brussaard2

1Plymouth Marine Laboratory, Prospect Place, The Hoe, Plymouth, Devon, PL1 3DH, United Kingdom2Department of Biological Oceanography, Royal Netherlands Institute for Sea Research (NIOZ), PO Box 59, 1790 AB Den Burg,Texel, the Netherlands

AbstractThe modified dilution assay aims to partition phytoplankton mortality into virus- versus grazing-induced

fractions and has previously been applied to several different environments to determine viral lysis rates of nat-ural phytoplankton. The method involves creating a gradient of both grazing and viral lysis by dilution withdifferent proportions of grazer- and virus-free filtrate, and assessing the subsequent impact on phytoplanktongrowth rates. We have conducted a critical evaluation of this method, and reviewed published data setsobtained using this approach to examine the utility of the modified dilution assay for estimating viral mortal-ity rates. We provide modifications and improvements that have been incorporated into the method since itwas first developed, and suggest recommendations for improving experimental success in less productive olig-otrophic environments. Published data show that viral lysis rates vary between different algal groups and inresponse to environmental conditions. Results also suggest that this method has the potential to be a useful toolfor estimating the impact of viruses on phytoplankton populations, but that the measurement of natural, lowviral lysis rates (<0.1 d–1) can challenge the application of this approach. Ultimately, however, the limitation ofthis method is associated with dilution of specific phytoplankton populations at low abundance.




ISBN 978-0-9845591-0-7, DOI 10.4319/mave.2010.978-0-9845591-0-7.65Suggested citation format: Kimmance, S. A., and C. P. D. Brussaard. 2010. Estimation of viral-induced phytoplankton mortality using the modified dilution method, p. 65–73. In S. W.Wilhelm, M. G. Weinbauer, and C. A. Suttle [eds.], Manual of Aquatic Viral Ecology. ASLO.


MANUALof


that incubations or culturing are not necessary. The methodinvolves determining the frequency of virally infected cellsusing transmission electron microscopy to derive the fractionof the host population that is infected by viruses, and thus thepotential number of cells that will be lost as a result of virallysis. However, this method is reliant on correct identificationof the host species from thin sections, which if not accuratelydetermined may lead to over- or underestimation of theimpact of viral infection. A second limitation of this methodis the requirement for estimates of the proportion of the lyticcycle during which viral particles are visible and knowledge ofthe virus latent period; presently there are few data for virus-phytoplankton systems (Proctor and Fuhrman 1990; Brus-saard et al. 1996). To obtain potential rates of infection, a thirdapproach is to use the contact rate model of Murray and Jack-son (1992); for example, see Suttle and Chan (1994). Assump-tions must be made, however, that host abundance estimatesare accurate and that all viruses have the potential to infect allhosts (Suttle 2000). In natural systems, specificity and range ofviral-strain hosts may lead to deviations from the model,which may change the proportion of host cells that areinfected (Suttle 2000).

Viral production rates have also been estimated from decayrates of virus communities (Heldal and Bratbak 1991). Theprincipal of this approach is that host mortality rates are deter-mined indirectly as the product of the decay rate of virusesdivided by the burst size, and the assumption is that viral pro-duction and removal rates are balanced. By the use of fluores-cently labeled viruses as tracers (Noble and Fuhrman 2000)rates of both virus production and removal can be simultane-ously determined as the change in ratio of labeled versus unla-belled viruses over time. However, both of these methods relyon the caveat that burst size (the number of viruses producedupon host cell lysis) is known. Most studies apply burst sizevalues reported in the literature to derive mortality rates; thesereported values may not be appropriate because of the highlyvariable nature of the burst size parameter to environmentalconditions (Wilson et al. 1996; Bratbak et al. 1998). In addi-tion, methods using viral tracers and decay of infectivity (Sut-tle and Chen 1992) assume that viral tracers are representativeof the natural virus communities and rarely take into accountthe effect of sunlight on viral DNA, which destroys infectivity,but not viral particles (Wilhelm et al. 1998). The latter tech-nique also incurs an additional disadvantage because it is lim-ited to dark incubations to protect the fluorescent stain fromlight degradation, which negates its use for autotrophic virus-host systems (Noble and Steward 2001). Extrapolation fromviral DNA synthesis rates determined using radiolabelled inor-ganic phosphate can provide phytoplankton virus productionrates (Steward et al. 1992). However, this method requiresusing isotopes with short half-lives, which are challenging tohandle, and the use of poorly constrained conversion factorsto convert radioactivity into viral production (Noble andSteward 2001).

All of the above methods attempt to derive viral-inducedmicrobial mortality rates indirectly by assessing changes to thevirus, i.e., changes in viral abundance, production, or decay.Measurements of viral replication rates must be inextricablylinked to host mortality rates, so this seems to be a reasonableassumption; however, it does not provide us with the param-eter that is most often lacking, i.e., a direct rate of host cell lossvia viral lysis. Viral-induced mortality rates should ideally bemeasured as directly as possible without the use of inferredassumptions and conversion factors. The “modified dilution”approach was introduced to provide this direct measurementfor specific phytoplankton. This method is an adaptation ofthe original dilution technique developed by Landry and Has-sett (1982), which has been used extensively to provide esti-mates of phytoplankton growth and microzooplankton graz-ing (e.g., Gallegos 1989; Landry et al. 1995; Worden andBinder 2003). The original dilution approach is based on thetheory that by combining whole seawater with grazer-free fil-trate in different proportions, grazing impact will be progres-sively decreased with increasing dilution. Linear regressionanalysis (Fig. 1) of observed phytoplankton net growth rateover the 24-h incubation period versus experimental dilution(D) provides estimates of growth rate in the absence of grazing(y-intercept), and grazing mortality rate (slope). The grazer-free water still allows most free viruses to pass through the fil-ter. The modified dilution approach includes an additionaldilution step created by combining whole seawater with virus-free filtrate in different proportions, thus allowing the directmeasurement of grazing versus grazer and viral-induced mor-tality. Viral mortality of the phytoplankton can be obtainedfrom the difference in the two relationships (Fig. 2).

To date there are few published studies that have used themodified dilution approach to determine phytoplankton mor-tality (Evans et al. 2003; Baudoux et al. 2006; 2007; Kimmanceet al. 2007; Baudoux et al. 2008; Brussaard et al. 2008), partlybecause these experiments can be difficult and time-consum-ing to set up and analyze (e.g., counting live phytoplankton).The labor-intensive nature of these experiments limits theiruse to perhaps one or two depths per day during field studies.However, at present this method is the only one that attemptsto directly measure viral lysis rates of phytoplankton, there-fore more studies need to be undertaken to fully validate anddevelop its utility. Note that this method is distinct from the“viral reduction” approach, which was developed to estimateviral-mediated bacterioplankton mortality (e.g., Parada et al.2008). This method is often referred to as a dilution approach;however, although free virus abundance is diluted duringexperimental setup, the prokaryote hosts are concentrated.Viral-induced mortality rates are then obtained indirectlythrough changes in viral abundance over time. This procedureis distinctly different to the modified dilution method, inwhich both viruses and host are diluted to create a gradient oflysis pressure and viral lysis rates are derived directly from lossof phytoplankton cells.

Kimmance and Brussaard Estimating viral lysis of phytoplankton

66


Experiments are set up based on the original protocol of themodified dilution method of Evans et al. (2003), and the fol-lowing method is appropriate for both natural and culturedphytoplankton–virus systems (Fig. 3). The approach is basedaround combining mesoplankton-free seawater (whole water)with either grazer-free or virus-free diluents. To prevent con-tamination of seawater samples by handling, vinyl gloves areworn throughout the water collection and experimentalsetup. To create mesoplankton-free whole water for experi-ments, seawater is gently siphoned through 200-µm mesh (orreverse sieved, Baudoux et al. 2006) using silicone tubing intoclean, polypropylene carboys. Additional (reverse) prefiltra-tion may be required, e.g., during sampling of colonial phyto-plankton (Baudoux et al. 2006) or in productive, coastal, orsediment-filled waters, to prevent clogging during the laterfine-scale filtration. Gentle filtration and avoidance of air bub-bling are crucial to prevent destruction of the grazer, virus,and phytoplankton populations. To protect the phytoplank-ton communities within the seawater sample from excessivelight (light-shock), experiments using natural phytoplanktonpopulations are best set up predawn in dimmed-light condi-tions using light-proofed carboys. It is also recommended thatexperiments be performed at the same time of day because ofthe synchronicity of phytoplankton cell division and poten-tial diel effects on viral infection processes. Filtration and han-dling should be conducted at in situ temperature. Handlingtime once the seawater has been collected should be kept to aminimum before the start of the experimental incubation.

Ideally experimental setup should be within an hour of watercollection if possible.

To create the grazer-free diluent whole seawater is gravityfiltered through acid-washed (see below) 0.2–0.45 µm filters(e.g., PALL Acropak™ Supor® membrane capsules) into a cleancarboy; a new filter should be used for each dilution experi-ment (Fig. 3). Prior to each experiment 0.2–0.45 µm filters, sil-icon tubing, carboys, and bottles should be acid-washed in 5%HCL and rinsed thoroughly with Milli-Q water and then twicewith the sample. The 0.2–0.45-µm and kDa filters should beflushed prior to use with freshly prepared Milli-Q and the firstfew liters of both filtrates discarded. Ensure that all air bubblesare removed from the filter capsule and tubing during filtra-tion by opening the filter valves. To speed up this process youmay use two filters in parallel. After gravity filtration half ofthe grazer-free diluent is passed through a tangential flow fil-tration system with kilodalton (kDa) pore size to create thevirus-free diluent. The designated pore size is variable but typ-ically between 10–100 kDa (Evans et al. 2003; Baudoux et al.2006; Kimmance et al. 2007). This pore size range is sufficientfor removal of both bacteriophage and larger algal virusessuch as Emiliania huxleyi viruses. The effectiveness of the mod-ified dilution technique is dependent on the efficiency of the0.2–0.45-µm filtration step to create a gradient of grazing pres-sure and the kDa filtration step to remove viruses from the0.2–0.45-µm filtrate, thus creating an additional viral gradi-ent. Assessing the differences in virus abundance between thediluents in comparison with natural samples determines howeffective the additional filtration step was at removing virusesand thus producing a gradient of viral pressure.


67

Fig. 1. Theoretical dilution plot of apparent phytoplankton growth rateversus fraction of whole water (D). Dilution of whole water with <0.45-µmfiltrate creates a gradient of grazing, and thus mg = mortality rate due tograzing and y-intercept = gross µ; phytoplankton growth rate in theabsence of grazing pressure (µ, d–1). Net µ = actual observed phyto-plankton growth rate (µ, d–1).

Fig. 2. Idealized dilution plots and analysis when whole water is dilutedwith either <0.45 µm filtered water (� = grazer-free) or kDa filtrate (� =grazer- and virus-free) to make up the dilution series for incubation.Reducing viral mortality in the kDa dilution series (�) is expected toincrease phytoplankton growth rates. Viral mortality rate of the phyto-plankton is obtained from the difference in the relationship between µand D in the parallel dilution series (µ = instantaneous growth rate of phy-toplankton; mv = mortality due to viral lysis; mg = mortality due to micro-zooplankton grazing; D = fraction of whole water).

The grazer-free and virus-free diluents are added to 10-Lpolycarbonate bottles in the correct proportions to create theparallel t0 dilution series, e.g., 20%, 40%, 70%, and 100%whole water. The mesoplankton-free whole water is then gen-tly added by siphoning. From each of these t0 bottles, tripli-cate 1-L polycarbonate bottles were rinsed twice and then gen-tly filled by siphoning (to minimize physical damage to thegrazers, viruses, and phytoplankton populations), ensuringthat bottles are filled completely to avoid trapping air bubblesinside upon closure. After filling, triplicate experimental bot-tles are placed randomly into experimental conditions. Fornatural samples, if in situ incubation is not feasible, then theexperimental environment should match the in situ tempera-ture and light conditions (including light–dark period) as

closely as possible. Therefore outdoor incubators should betemperature controlled and covered with neutral densityscreening to match natural light intensity at the sampledepth. Laboratory assays should be set up as described aboveexcept that viral lysate capable of infecting the chosen hostshould be added (multiplicity of infection >1) to the culturedhost before dilution (Baudoux et al. 2006) and experimentalbottles incubated under the appropriate conditions for thechosen cultured virus–host system.

For determination of phytoplankton composition and abun-dance and virus abundance, triplicate subsamples (5 mL) aretaken from the 10 L t0 dilution bottles and the 0.2–0.45-µm andkDa diluents. However, an alternative method is to take samplesout of each and every 1-L experimental bottle directly at t0,


68

Fig. 3. Modified dilution assay experimental design. Mesoplankton-free whole water (1) is combined with either <0.45-µm filtrate (2) or kDa filtrate (3)in the correct proportions to create the parallel t0 dilution series: <0.45µm series, with reduced grazing mortality (4) kDa series, with reduced grazingand viral mortality (5). Replicate sample bottles from the <0.45-µm and kDa dilution series are then created (6 and 7) and incubated under experimen-tal conditions.

making sure that there are no air bubbles in the bottles uponclosing. Final time-point samples are also taken from everyexperimental bottle at the end of the experimental period, 24 h.Initial (t0) and final (t24) phytoplankton composition and abun-dance estimates are typically determined by analysis of samplesusing flow cytometry (Evans et al. 2003; Baudoux et al. 2006;Kimmance et al. 2007). If flow cytometry is not available othercell-counting methods could be used (e.g., microscopy orCOULTER COUNTER®). In the case of monoalgal blooms,chlorophyll measurements are a possible alternative (Evans etal. 2003). However, this is not recommended because of thehighly specific nature of viral infection. It is more desirable todetect changes within specific phytoplankton groups, and withflow cytometry distinct groups can easily be discriminated bydifferences in fluorescence characteristics.

Apparent phytoplankton growth rates (µ, d–1) are calculatedfrom each experimental bottle as the changes in abundanceduring the incubation using the equation:

µ = ln (Pt/P0)/t (1)

where Pt and P0 are the final and initial measured phyto-plankton abundance, respectively, and t is the duration of theexperiment. The actual dilution rate is calculated by dividingthe t0 phytoplankton abundance in each bottle by the aver-aged abundance of the replicate t0 100% counts (3–6 bottles).Model 1 linear regression analysis of apparent growth ratesagainst fraction of whole water is applied to each of the dilu-tion experimental series (0.2–0.45 µm and kDa) to estimateinstantaneous growth and mortality due to grazing and/orviral lysis. As explained above, the regression coefficient ofapparent growth rate versus fraction of whole water for the<0.45-µm dilution series represents the microzooplanktongrazing rate [mg = (µ in the absence of grazing – observednet µ) = slope], whereas the regression coefficient from thekDa dilution series represents the combined mortality impactof both grazers and viruses (mg + mv), where mv = viral mortal-ity (Fig. 2). The viral-induced phytoplankton mortality ratetherefore is the difference between these two regression slopes,i.e., mv = [(mv + mg) – mg]. However, a significant rate can bederived only when there is a significant difference betweenthe two mortality slopes. To determine this, experiments thatproduce significant regression slopes for both the 0.2–0.45-µmand kDa dilution series are further analyzed, e.g., using anF-test. Only then can the difference between the two slopes beconsidered a significant viral lysis rate.

AssessmentThe modified dilution approach is promising in that it is

the only method that can potentially partition phytoplanktonmortality into grazer- versus viral-induced fractions and doesnot require the use of conversion factors, which can introduceerror. However, there are limitations associated with thismethod. As described above, one of the fundamental assump-tions of the original Landry and Hassett dilution method

(1982) is that phytoplankton mortality during the incubationperiod is a result of an encounter between grazer and prey, andthe grazing impact will vary with dilution. The concept ofprey (host) encounter is also fundamental to the modifieddilution approach. However, phytoplankton cells may alreadybe infected at the beginning of the experiment, and thus thiscomponent of cell lysis will not vary with dilution. Only newinfection is detected during the experiment. The method isbased on changes in phytoplankton cell abundance and thusthe period to cell lysis of the algal host will dictate whethermeasurable cell lysis occurs during the incubation (Evans et al.2003; Baudoux et al. 2006) and not the latent period of thevirus. As long as the time to lysis of cells is longer than 12 hand shorter than 24 h we can determine the viral lysis rate. Itdoes not matter whether reinfection is happening during thisexperimental time period because the cells that are infected bythe newly released viruses will not undergo lysis before theend of the experiment. An incubation period of 24 h is opti-mum because a shorter time is not suitable for investigatingphased algal growth, and a longer time period induces bottleeffects due to containment. During long experimental incuba-tions containment will influence the apparent growth rateestimates and affect the outcome of experiments.

Although nutrient adjustment has previously been sug-gested to ensure that phytoplankton growth rates are inde-pendent of the dilution effect (Landry et al. 1995), the addi-tion of nutrients is not recommended here as it may produceunnatural growth rates. Care should be taken with the inter-pretation of results, especially under incubation conditions inwhich, e.g., nutrients may have become limiting. If nutrientdepletion occurs to the same degree in all dilutions thengrowth rate may be underestimated but the grazing/viralmortality rate should not be affected. However, depletion ofa recycled limiting nutrient would occur to a greater degree inthe highest dilution with the fewest grazers/viruses; thiswould lower the slope of the regression and underestimateboth grazing and growth rates (Dolan and McKeon 2004).Typically, dilution effects on phytoplankton growth rate aredetectable when plotting net growth against dilution factor,particularly if combined with physiological assessment dur-ing the experiments. Therefore if nutrient limitation effectsare visible then it may be acceptable to remove the datapoints at the highest level of dilution when determining theregression slopes.

The original dilution protocol was modified to include anindependent measure of “relative grazing activity” to improvethe accuracy of grazing rate measurements (Landry et al. 1995).This internal measure (uptake of fluorescently labeled prey)was introduced because of the potential violation of theassumption that grazing impact varies in direct proportion tothe dilution of grazer population density. In theory the mech-anism of viral infection is simpler than grazing in that a singlevirus infects only a single host. Therefore, the premise thatgrazer–prey interactions should follow a linear rather than


69

nonlinear response with dilution should also apply tovirus–host encounter and infection. However, like grazers,viruses and their hosts are not just inert particles and so in real-ity may not follow simple encounter rate theory. Whether thishypothesis remains true throughout the experimental periodhas not yet been tested for natural virus–host communities toour knowledge; however, viral lysis has been shown to remainlinear to dilution factor in a model Phaeocystis globosa system(Baudoux et al. 2006). This is an important assumption of themodified dilution technique that needs further clarificationusing natural populations. Other factors such as threshold hostconcentrations and host defense mechanisms may also cause adeviation from the linear viral lysis model. Gallegos (1989)showed that dilution regressions can become nonlinear as aresult of prey density thresholds for microzooplankton grazing.The possibility of threshold concentrations also needs to beconsidered during modified dilution experiments, and thelevel of abundance of both phytoplankton and grazers/virusesmust be adjusted to allow a high enough encounter rate evenat the lowest level of dilution. The effect of reduced viral mor-tality because of a host threshold in the dilute treatments couldproduce a nonlinear response and thus an underestimation ofviral lysis rate. Another factor to consider is the possibility thatgrazers may preferentially graze viral-infected cells. As yet thistheory remains to be tested for other phytoplankton speciesand in natural populations; however, Evans and Wilson (2008)suggest that under culture conditions grazers may activelyselect for infected versus noninfected phytoplankton prey. Ifthis does occur typically during the dilution experiments, itwould cause an underestimation of mortality rates and haveserious implications for the use of the modified dilutionapproach to measure viral lysis rates.

Ultimately, the success of this approach is reliant on the pro-duction of two significant mortality rates (mv = [(mv + mg) – mg])and thus it is vital to have an appropriate sample number inboth dilution series to be able to detect a significant differencebetween the two regression slopes. The accuracy of the mortal-ity data are dependent on the precision of the apparentgrowth-rate measurements, which ultimately depends on thephytoplankton cell counts. The ability to quantify initial t0 andt24 phytoplankton abundance within acceptable levels of preci-sion (i.e., <5 % standard error) is fundamental for the successof this method. Thus it is critical that analyses of t0 and t24 phy-toplankton cell counts conducted using fresh samples are ana-lyzed as quickly as possible. Relatively precise estimates of ini-tial and final abundances are needed to minimize the errorgenerated through calculations of population rates of change.Sources of error can arise from imprecise subsampling and vari-ability between replicates at the same dilution. Furthermore,attempting to enumerate distinct subpopulations can be par-ticularly challenging in areas of low productivity, becausedetecting mortality rates based on small changes in cell con-centrations is not easy. Phytoplankton abundance is rela-tively low in oligotrophic waters, which creates a problem

because of the high levels of dilution required for this assay.As mentioned above, cell counts are typically analyzed usingflow cytometry and live samples. The use of flow cytometryinstead of chlorophyll a is recommended because moredetailed information can be obtained for viral-induced mor-tality of specific algal groups. Because viruses are specific tophytoplankton species and at times even strains, it is impor-tant not to use bulk measurements such as chlorophyll a.However, in some environments sample analysis can take toolong because of the difficulty of gaining a statistically viablecell count at the most dilute level of dilution. This difficultyhas serious implications because it extends the analysis timefor the whole dilution series (21 samples if experiment hastriplicate bottles at four dilution levels). During this longanalysis time there may be changes in phytoplankton hostphysiology or loss of cells, which may affect the outcome ofthe experiments. Therefore fixation is occasionally necessary.However, because it is well established that fixation in itselfcauses some degree of cell loss (e.g., Vaulot et al. 1989), thiseffect of fixation should be tested to ensure that the percent-age of cell loss is the same regardless of dilution factor.

As described above, the ability to measure a significant virallysis rate during modified dilution experiments is dependenton the presence of a detectable difference between the tworegression slopes, and when viral mortality rates are low thedetection of this difference becomes difficult. Therefore toincrease the chance of detecting low rates it is critical that theexperimental design is appropriate for the phytoplanktoncommunities studied. To detect significant differencesbetween the two slopes, experiments need to have a largeenough n number (number of samples in each parallel dilu-tion series). Slight slopes or low grazing rates are difficult todetect with regression analysis using the small n values com-monly employed during dilution experiments (Dolan andMcKeon 2004). This can be tested, however, and to alleviatethis problem Kimmance et al. (2007) suggested adding a sen-sitivity test to the modified dilution assay. This additionalanalysis may improve the efficiency of dilution experimentsby demonstrating the level of modification that needs to bemade to the experimental design, i.e., determining the appro-priate n value for production of significant results (Kimmanceet al. 2007). However, even if the analysis is not applied beforeexperiments are conducted it can be useful for determiningretrospectively the level at which the regression analyses andsubsequent F-tests would have had the power to detect differ-ences between the mortality slopes. From the literature stud-ies conducted so far, testing the sensitivity of the experimen-tal approach indicates that viral lysis rates <0.1 d–1 were noteffectively estimated using the modified dilution method(Kimmance et al. 2007, Baudoux et al. 2008).

DiscussionSince its development in 2003 the modified dilution assay

has been applied to several different environments, including


70

seminatural mesocosm (Evans et al. 2003), coastal (Baudoux etal. 2006; Kimmance et al. 2007), open ocean (Brussaard et al.2008), and oligotrophic (Baudoux et al. 2007; 2008), to test itsability to determine viral lysis rates of natural phytoplanktoncommunities. Results from these published studies clearlyshow that rates vary in response to environmental conditionsand between different algal groups. The modified dilutionmethod was first tested during such conditions in a nutrient-enriched mesocosm experiment. Using this approach Evans etal. (2003) successfully determined viral-induced mortalityrates of Micromonas spp. of 0.10–0.29 d–1 and showed that upto 34 % of Micromonas spp. production could be lysed daily byviruses. More recently Baudoux et al. (2006) demonstratedthat this approach could also be applied successfully during abloom of P. globosa. In this more productive, coastal systemthe viral-induced mortality was higher, accounting for up to66% of the total P. globosa mortality, with maximum lysis ratesof 0.35 d–1 measured. However, in both of these studies therewas a dominant phytoplankton species and viruses were themajor mortality agents.

Other published data sets suggest that in more diverse com-munities significant viral lysis rates are not always as readilyattainable with this method (Kimmance et al. 2007; Baudouxet al. 2008). In microbial systems in which low biomass andhigh diversity predominate, the chances of successfully meas-uring lysis events using the modified dilution approach maybe lower. Recently Baudoux et al. (2007) and Brussaard et al.(2008) tested this method in less productive, oligotrophicwaters. In the subtropical northeastern Atlantic, Baudoux etal. (2007) found that significant viral lysis rates could beobtained at five of six stations, and lysis of 1 group ofpicoeukaryotes was responsible for 50–100% of the total cellmortality with rates of 0.1–0.8 d–1 (Baudoux et al. 2007). How-ever, these experiments were restricted to phytoplankton pop-ulations from the deep chlorophyll maximum where cellnumbers were sufficiently high to avoid the problems associ-ated with a three- to four-fold dilution, which is essential forthis methodology (Baudoux et al. 2007). Substantial viral lysisrates (0.16–0.23 d–1) were also obtained for picoeukaryotesduring true oligotrophic conditions in the North Sea, butlower viral lysis rates were not significantly different from zero(Baudoux et al. 2008). Kimmance et al. (2007) were unable tomeasure any significant viral-induced mortality rates at acoastal site in the western English Channel where picophyto-plankton were the dominant phytoplankton group. Low, non-significant lysis rates (0.09–0.18 d–1) were also obtained forphytoplankton populations in the Southern Ocean (Brussaardet al. 2008). Such low rates of viral lysis may be common innatural nonbloom conditions (which may present a challengefor the use of this method in some environments). Thereforemore studies need to be conducted to determine the ability ofthe modified dilution assay to accurately detect such low virallysis rates. However, we should also investigate virus dynam-ics at an array of host densities using laboratory cultured sys-

tems. By modeling virus–host encounter rates and infectiondynamics from a variety of cultured phytoplankton groups wecan determine the sensitivity range, i.e., the highest and low-est viral lysis rates that can be measured using this method.Thus, we can assess the suitability of this approach for esti-mating viral lysis rates in both low and high productivityenvironments.

Unfortunately there are limited data available to comparethe rates obtained using the modified dilution assay to thoseobtained with an alternative method. Of the viral lysis studiespublished so far most do not include a comparative assay forestimating the viral impact on phytoplankton. However, Bau-doux et al. (2006) demonstrated the utility and validity of theviral dilution method using both a cultured system and natu-ral populations. Comparison of total cell lysis rates (obtainedusing the dissolved esterase activity assay) and growth-cyclelysis experiments to viral-induced mortality rates (derivedfrom the viral dilution assay) showed good agreement. Thus,the consistency of the viral lysis rates obtained by this methodwith other means for assessing cell lysis provides confidencein the suitability of this method to infer the impact of viruseson phytoplankton mortality (Baudoux et al. 2006).

Comments and recommendationsAlthough there are assumptions associated with the mod-

ified dilution approach that have not yet been tested andfurther application of this technique needs to be conductedwith natural populations of phytoplankton, at present it isthe only method that can potentially directly derive virallysis rates from phytoplankton mortality. It is also the onlysingle method that can in theory provide concurrent esti-mates of grazing and viral-induced mortality. However, low(natural) levels of viral mortality will challenge this methodunless it can be modified to include more replicates and bet-ter precision. By improving experimental design throughincreased number of replicates, diluting biomass to anappropriate level to allow virus–host encounters even at thelowest level of dilution, prolonging counting time for themore diluted samples (provided the in situ temperature canbe maintained), assessing taxon-specific mortality rates, andcombining this approach with methods that assess host/viraldiversity and host specificity, this method may be able toprovide realistic estimates of viral mortality in naturalwaters. Other practical issues to consider include: ensuringprecision in measurements of variables used to determinephytoplankton apparent growth rates, considering the use ofnonstandard levels of dilution appropriate to abundance lev-els of natural host/virus populations, assessing the physio-logical state of host populations during incubation, anddetermining the sensitivity of the regression analyses byassessing the power of the test to provide an indication ofthe accuracy and error associated with the analyses, and thusprovide a more powerful indication of “real” versus experi-mentally induced variation.


71

This report highlights the importance of monitoring vari-ous parameters during experimental incubations, particularlywhen trying to estimate phytoplankton growth and mortalityrates from natural waters. Certain aquatic environments pres-ent more challenges than others and it may be that the mod-ified dilution assay is best suited to productive waters. Experi-ments conducted in oligotrophic conditions may requiremodifications from the typical methodological approach,such as fixed versus live cell counts (see “Assessment” above).Although the modified dilution approach requires adaptationsto suit the environmental conditions being investigated andneeds further testing in natural waters, it has the potential tobe a useful tool for estimating the impact of viruses on phyto-plankton populations. Furthermore, because of the removal ofgrazers and viruses, these studies have the potential to provideestimates of other variables of interest to marine microbiolo-gists; factors such as nutrient recycling and dissolved organicmatter production can be assessed in the presence and absenceof key food web components. What must be maintained, how-ever, is the statistical rigor that is fundamental to this method.Only then will the rates obtained from this approach improveour understanding of the role of viruses on primary produc-tion-mediated carbon cycling.

ReferencesBaudoux, A.-C., A. A. M. Noordeloos, M. J. W. Veldhuis, and

C. P. D. Brussaard. 2006. Virally induced mortality of Phaeo-cystis globosa during two spring blooms in temperate coastalwaters. Aquat. Microb. Ecol. 44:207-217.

———, M. J. W. Veldhuis, H. J. Witte, and C. P. D. Brussaard.2007. Viruses as mortality agents of picophytoplankton inthe deep chlorophyll maximum layer during IRONAGESIII. Limnol. Oceanogr. 52:2519-2529.

———, ———, A. A. M. Noordeloos, G. van Noort, and C. P. D.Brussaard. 2008. Estimates of virus- vs. grazing induced mor-tality of picophytoplankton in the North Sea during sum-mer. Aquat. Microb. Ecol. 52:69-82.

Bratbak, G., M. Heldal, S. Norland, and T. F. Thingstad. 1990.Viruses as partners in spring bloom microbial trophody-namics. Appl. Environ. Microbiol. 56:1400-1405.

———, A. Jacobsen, M. Heldal, K. Nagasaki, and F. Thingstad.1998. Virus production in Phaeocystis pouchetii and its rela-tion to host cell growth and nutrition. Aquat. Microb. Ecol.16:1-9.

Brussaard, C. P. D. 2004. Viral control of phytoplankton pop-ulations: A review. J. Euk. Microbiol. 51:125-138.

———, R. S. Kempers, A. J. Kop, R. Riegman, and M. Heldal.1996. Virus-like particles in a summer bloom of Emilianiahuxleyi in the North Sea. Aquat. Microb. Ecol. 10:105-113.

———, K. R. Timmermans, J. Uitz, and M. J. W. Veldhuis.2008. Virioplankton dynamics and virally induced phyto-plankton lysis versus microzooplankton grazing southeastof the Kerguelen (Southern Ocean). Deep-Sea Res. II 55:752-765.

Dolan, J. R., and K. McKeon. 2004. The reliability of grazingrate estimates from dilution experiments: Have we over-estimated rates of organic carbon consumption? Ocean Sci.Disc. 1:21–36.

Evans, C., S. D. Archer, S. Jacquet, and W. H. Wilson. 2003.Direct estimates of the contribution of viral lysis and micro-zooplankton grazing to the decline of a Micromonas spp.population. Aquat. Microb. Ecol. 30:207-219.

———, and W.H. Wilson. 2008. Preferential grazing ofOxyrrhis marina on virus infected Emiliania huxleyi. Limnol.Oceanogr. 53:2035-2040.

Fuhrman, J. A. 1999. Marine viruses and their biogeochemicaland ecological effects. Nature 399:541-548.

Gallegos, C. L. 1989. Microzooplankton grazing on phyto-plankton in the Rhode River, Maryland: Nonlinear feedingkinetics. Mar. Ecol. Prog. Ser. 57:23-33.

Heldal, M., and G. Bratbak. 1991. Production and decay ofviruses in aquatic environments Mar. Ecol. Prog. Ser.72:205-212.

Kimmance, S. A., W. H. Wilson, and S. D. Archer. 2007. Use ofthe modified dilution technique to estimate viral versusgrazing mortality of phytoplankton: limitations associatedwith method sensitivity in natural waters. Aquat. Microb.Ecol. 49:207-222.

Landry, M. R., and R. P. Hassett. 1982. Estimating the grazingimpact of marine microzooplankton. Mar. Biol. 67:283-288.

———, J. Kirshtein, and J. Constantinou. 1995. A refined dilu-tion technique for measuring the community grazingimpact of microzooplankton, with experimental tests in thecentral equatorial Pacific. Mar. Ecol. Prog. Ser. 120:53-63.

Murray, A. G., and G. A. Jackson. 1992. Viral dynamics: Amodel of the effects of size, shape, motion and abundanceof single-celled planktonic organisms and other particles.Mar. Ecol. Prog. Ser. 89:103-116.

Noble, R. T., and J. A. Fuhrman. 2000. Rapid virus productionand removal as measured with fluorescently labeled virusesas tracers. Appl. Environ. Microbiol. 66:3790-3797.

———, and G. F. Steward. 2001. Estimating viral proliferationin aquatic samples, p. 67-84. In J. H. Paul [ed.], Marinemicrobiology. Academic Press.

Parada, V., A.-C. Baudoux, and E. Sintes. 2008. Dynamics anddiversity of newly produced virioplankton in the NorthSea. ISME J. 2:924–936.


Ruardij, P., M. J. W. Veldhuis, and C. P. D. Brussaard. 2005.Modeling the bloom dynamics of the polymorphic phyto-plankter Phaeocystis globosa: Impact of grazers and viruses.Harmful Algae 4:941-963.

Steward, G. F., J. Wikner, D. C. Smith, W. P. Cochlan, and F.Azam. 1992. Estimation of virus production in the sea, I:Method development. Mar. Microb. Food Webs 6:57-78.

Suttle, C. A. 2000. Ecological, evolutionary, and geochemicalconsequences of viral infection of cyanobacteria and


72

eukaryotic algae. p 247-296. In: C. J. Hurst. 2000. Viral ecol-ogy. Academic Press, London.

———. 2007. Marine viruses-major players in the globalecosystem. Nature Rev. Microbiol. 5:801-812.

———, and A. M. Chan. 1994. Dynamics and distribution ofcyanophages and their effect on marine Synechococcus spp.Appl. Environ. Microbiol. 60:3167-3174.


Vaulot, D., C. Courties, and F. Partensky. 1989. A simplemethod to preserve oceanic phytoplankton for flow cyto-metric analyses. Cytometry 5:629-635.

Weinbauer, M. G., Fuks, D. and Peduzzi, P. 1993. Distributionof viruses and dissolved DNA along a coastal trophic gradi-ent in the Northern Adriatic Sea. Appl. Environ. Microbiol.59:4074-4082.

Wilhelm, S. W., M. G. Weinbauer, C. A. Suttle, and W. H. Jef-

frey. 1998. The role of sunlight in the removal and repair ofviruses in the sea. Limnol. Oceanogr. 43:586-592.

———., and C. A. Suttle. 1999. Viruses and nutrient cycles inthe sea: Viruses play critical roles in the structure and func-tion of aquatic food webs. Bioscience 49:781-788.

———, S. M. Brigden, C. A. Suttle. 2002. A dilution techniquefor the direct measurement of viral production: a compari-son in stratified and tidally mixed coastal waters. Microb.Ecol. 43:168-173

Wilson, W. H., Carr, N. G., and Mann, N. H. 1996. The effectof phosphate status on the kinetics of cyanophage infec-tion in the oceanic cyanobacterium Synechococcus sp.WH7803. J. Phycol. 32:506-516.

Worden, A. Z., and B. J. Binder. 2003. Application of dilutionexperiments for measuring growth and mortality ratesamong Prochlorococcus and Synechococcus populations inoligotrophic environments. Aquat. Microb. Ecol. 30:159-174.


73

74

Introduction

Viruses are the most abundant and diverse biological com-ponent of aquatic environments (Bergh et al. 1989; Proctor

and Fuhrman 1990). The estimated overall abundance in theworld’s oceans is on the order of 1030 (Suttle 2005), a valuethat exceeds the abundance of prokaryotes by one order ofmagnitude (Whitman et al. 1998; Suttle 2005). Moreover, ithas become increasingly evident that viruses play critical rolesin shaping aquatic communities and determining ecosystemdynamics (Fuhrman 1999; Danovaro et al. 2008). Morerecently, the interest for the determination of viral enumera-tion has extended to the benthic compartment, as it hasbecome increasingly evident that viral abundances in surfacesediments at all depths down to abyssal sediments exceedthose in the water column by orders of magnitude (typicallyreaching values of 108–109 viral particles mL–1; e.g., Danovaroand Serresi 2000; Danovaro et al. 2002; Corinaldesi andDanovaro 2003; Maranger and Bird 1996; Middelboe et al.2006; Middelboe and Glud 2006; Siem- Jørgensen et al. 2008).Moreover, high viral abundances have been reported also insubsurface sediment (Bird et al. 2001.) There is a clear need todevelop accurate protocols for viral enumeration in marinesediments for detecting potential changes in viral abundance

Separation of free virus particles from sediments in aquatic systemsRoberto Danovaro1* and Mathias Middelboe2

1Department of Marine Science, Polytechnic University of Marche, Via Brecce Bianche, 60131 Ancona, Italy2Marine Biological Laboratory, University of Copenhagen, Helsingør, Denmark

AbstractThe number of benthic viruses per unit of volume, at all depths (from shallow down to abyssal sediments), exceeds

water column abundances by orders of magnitude. The need of methods for the determination of viral counting inaquatic sediment is becoming increasingly urgent along with the increasing evidence of the relevance of viruses inthe benthic domain. The procedures used for determining viral abundances in sediments require specific modifica-tions to release the viruses from sediment particles and to minimize the physical and chemical interferences of sedi-mentary matrix with the analysis. Dislodging viruses from sediment samples is the first crucial step for the analysesof viral abundance in benthic samples. Here we present the results of several tests aimed at optimizing the protocolfor viral counts based on (i) the chemical treatment (type and modality of the use of surfactants), (ii) mechanical treat-ment (ultrasounds), (iii) cleaning of the samples (by enzymatic digestion of the extracellular DNA by means ofDNases), and (iv) the limitations associated with viral recovery from the sediment (by serial washing steps). Sedimenttexture and composition vary considerably along horizontal and vertical gradients, and here we compare shallowsandy sediments with more silty deep-sea sediments. We found that the use of the surfactant tetrasodium pyrophos-phate (final concentration 5 mM for 15 min), followed by ultrasound treatments (3 times for 1 min with 30 s inter-vals) and by the addition of an enzymatic cocktail composed of DNase I, nuclease P1, nuclease S1, and esonuclease3, increased the detectability by staining with fluorochrome, thus resulting in significantly higher and more accurateviral counting, determined by epifluorescence microscopy. Our results also indicate that sediment samples processedusing this optimized protocols displayed a significantly lower coefficient of variation, thus making sufficient thecounting of a lower number of optical fields. Centrifugation of sediment samples after extraction procedures couldunderestimate viral counting, and we recommend here an accurate check of the potential loss or an alternative pro-cedure based on sediment dilution prior to quantification by epifluorescence microscopy.

*Corresponding author: E-mail: [email protected]; Tel: +39 71 220 4654; Fax: +39 71 220 4650



This work was carried out within the frame of the program BIOFUN:“Biodiversity and ecosystem functioning in contrasting southernEuropean deep-sea environment” supported by European ScienceFoundation and by the Danish Natural Sciences Research Council.

ISBN 978-0-9845591-0-7, DOI 10.4319/mave.2010.978-0-9845591-0-7.74Suggested citation format: Danovaro, R., and M. Middelboe. 2010. Separation of free virus parti-cles from sediments in aquatic systems, p. 74–81. In S. W. Wilhelm, M. G. Weinbauer, and C. A.Suttle [eds.], Manual of Aquatic Viral Ecology. ASLO.


MANUALof


Danovaro and Middelboe Viral extraction from aquatic sediments

75

and because of viral production’s being typically assessed byuse of total counts (Danovaro et al. 2008).

Approaches and procedures used for determining viralabundances in sediments are derived from those applied towater samples. However, due to the specific characteristics ofthe sediments and of the benthic environment, significantmodifications of the protocols are needed to minimize thephysical and chemical interferences of sedimentary matrixwith the analysis. Efficient dislodging of viruses from sedi-ment samples is the first crucial step for the analyses of viralabundance in benthic samples. This step is currently requiredfor all of the available techniques of viral enumeration in sed-iments, including the technique based on the largely used epi-fluorescence microscopy (EFM; Danovaro et al. 2001; Patel etal. 2007), transmission electron microscopy (TEM; Middelboeet al. 2003), or the use of flow cytometry (FCM; Duhamel andJacquet 2006).

Quantifying their abundance is a fundamental step in anyattempt to understand the role of viruses in sediments, andtheir spatial and temporal dynamics. However, quantificationof viral abundance in benthic environments is a much morecomplex task than is the case in the water column becausethe benthic viral particles have to be dislodged from thematrix in which they are embedded prior to quantification.Numerous approaches for extracting viruses from sedimentshave been proposed with quite variable results, and there is,therefore, a strong need for an evaluation of previous experi-ences and, subsequently, to move toward a consensus aboutwhich general extraction principle that provides the mostaccurate determination of viral abundance in sediments. Theseparation of viruses from sediment particles requires thebreaking of the links between viruses and sediment particles.To do this, one of the most common approaches is based ona chemical treatment with surfactants, which createhydrophylic links among particles, thus enhancing theirinterdispersion. Among the available surfactants, tetrasodiumpyrophosphate and polyoxyethylene-sorbitanmonooleate arethe most widely used in the last decade (Maranger and Bird1996; Hewson and Fuhrman 2003; Corinaldesi et al. 2007).The dislodgement of viruses from the sediment particles istypically accompanied by a mechanical shaking (by means ofmanual shaking, ultrasonication, and/or vortexing; Danovaroet al. 2001).

Here we present a detailed description of what we find isthe most efficient protocol for dislodging viruses from sedi-ment particles for subsequent analysis by EFM, flow cytome-try, or TEM. The treatments compared in this study were per-formed with different types of sediments covering a widerange of depths (from shallow sandy sediments to silty deep-sea systems). To test the extraction efficiency, we focused ourattention on EFM counting using SYBR Green I (or SYBR Gold)as a stain. The following methodological aspects were investi-gated: (i) virus dislodgment from sediment particles (usingsurfactant and ultrasound treatments), (ii) efficiency of virus

extraction from bulk sediment (by the number of postsonica-tion washings), (iii) stain-counting accuracy and efficiency (byremoving possible interferences due to extracellular DNA invirus counting and by comparison with TEM counts), and (iv)effects of centrifugation versus dilution of the sediment priorto counting viral particles.

Materials and proceduresExtraction of viruses for the analysis on porewater samples—

Studies dealing with the determination of viral abundance orviral production in sediments have been carried out so far onboth viruses dispersed in the porewater (nonattached to sedi-ment particles) and on viruses attached to sediment grains.The counting of viruses in the porewater does not require aspecific modification of the protocol used for aquatic samples,except the squeezing of the sediment for the extraction of theporewater (e.g., by centrifugation). Alternatively, wet sedi-ment samples are filtered on glass fiber (GF/F) filters and theporewater recovered in a sterile flask for subsequent analysis(Hewson and Fuhrman 2003). Conversely, the analysis of viralabundance in sediments (i.e., including viruses attached tosedimentary matrix) requires a specific treatment.

Extraction of viruses for the analysis on sediment samples—Extracting viruses from sediments is a relatively recent andunexplored discipline in aquatic viral methodology, and thereexist no established consensus on how this is done most effi-ciently. We present here studies for examining the efficiencyof various treatments in extracting viruses for subsequentcounting by epifluorescence microscopy.

Sediment samples can be collected by means of manualcorers, multiple-corer, or other sampling devices allowing torecover a perfectly undisturbed sediment surface. Here wecompared two sediment types: (a) sandy coastal sediments,collected in the Adriatic Sea (Mediterranean) and in Øresund,Denmark, and (b) muddy deep-sea sediments collected in theNortheastern Atlantic Ocean (at 4800-m depth).

Immediately after the recovery of samples, sediment slurrieswere made with about 0.5 mL of the top 1 cm of both sedimenttypes (taken from independent cores and independent deploy-ments), and 4.5 mL sterile and 0.02 μm prefiltered seawater, toavoid the burst of potentially infected cells (and the conse-quential release of viruses and DNA) due to changes of osmoticpressure. This protocol is also applicable for the estimate ofviral abundance in freshwater sediments; in this case, the dilu-tion steps should be performed with freshwater. Samples wereimmediately brought in the laboratory for their analysis with-out the use of preservatives, to avoid virus loss due to formalinfixation (Danovaro et al. 2001). Alternatively, the slurry isimmediately frozen at –20°C or frozen in liquid nitrogen andstored at –80°C until further processing (see below). In ourexperience, no changes in viral counts can be detected within6 months of storage. If sampling of the sediment is not possi-ble at sea, the intact cores can be closed with rubber stoppersand transported to the laboratory in a water bath at in situ tem-


76

perature, where they are then processed. The spatial distribu-tion of viruses in sediments is very heterogeneous even onmicrometers to centimeter scales (Middelboe et al. 2006). It istherefore important to consider whether the given sedimentshould be homogenized before sampling to obtain an averageviral density in a given sediment layer or whether one is inter-ested in including the small scale variability in the analysis.

Materials used for the assessment and detailed informationon buffers and solutions for the optimized protocol are givenbelow. The optimized protocol for the separation of free virusparticles from sediments is shown in Fig. 1.

Chemical treatment—The extraction of the porewater can beaccomplished by centrifugation or filtration. Conversely, thefirst step for the detachment of the viruses from the sedimentis a chemical treatment. Among the surfactants currently usedby different authors to detach the viruses from sediment par-ticles, the most widely used are (i) tetrasodium pyrophosphate(SIGMA; Danovaro et al. 2001), (ii) a mix of 10 mM pyrophos-phate and 5 mM EDTA (Hewson and Fuhrman 2003), (iii) PBS(1:2 vol: vol, sediment and PBS, respectively; Hewson et al.2001), (iv) polyoxyethylene-sorbitan monooleate (Tween 80,SIGMA). All of these surfactants are suitable for viral countingunder epifluorescence microscopy, with the exception ofTween 80 that displays lower performance for interferencewith the visualization of the viruses on EFM but has beenreported to be highly efficient for analysis by flow cytometry(Duhamel and Jacquet 2006).

In this study, we optimized the protocol for the counting ofbenthic viruses under epifluorescence microscopy using tetra-sodium pyrophosphate (5 and 10 mM; i.e., 250 or 500 µL of a100 mM solution in a 5 mL slurry). Samples were then incu-bated for 15 min in the dark at room temperature. Additionalsediment samples (n = 3; 0.5 mL) without pyrophosphate wereadded to 4.5 mL of MilliQ water and served as controls. Afterincubation, all samples were shaken manually for 1 min andthen centrifuged (800g; 1 min) to reduce interference due tosuspended particles.

Aliquots of the supernatant were diluted 500 to 1000 timesbefore filtration onto 0.02-µm-pore-size Anodisc 25 mem-brane filters (Whatman) under void-pressure <100 mmHg. Thefinal dilution of sediments depends on the actual viral abun-dance on the filter. Typically, the final dilution of the sedi-ments is 100 times for sandy sediments, 500 times for sandy-muddy sediments, whereas in virus-rich sediment samples theoptimal final dilution could be up to 1000 times. The choiceof the correct dilution, which must be checked for all sedi-ment types, is of vital importance for the final result: an exces-sive or an insufficient number of viruses on the filter couldinvalidate the final calculations. A low filtration pressure isneeded to avoid the damages or burst of prokaryotic cells, andthe consequential release of nucleic acids that, binding thefluorochrome, could interfere with virus counting.

For sediment samples, the filters are typically stained with20 µL of SYBR Green I (Lot no. 4967-30; diluted to 500× in

sterile MilliQ water; optical density at 495 nm = 1.357) andincubated for 15 min in the dark, rinsed twice with 1 mLMilliQ water (to eliminate fluorescence background noise),and analyzed by EFM using a Zeiss Axioplan microscopeequipped with a 100-W lamp. Ten to 50 fields were viewed at1000× magnification, and a minimum of 400 viruses wascounted. Virus-like-particles (VLP) were discriminated frombacteria (0.2- to 2-mm diameter) by their dimensions (0.015-to 0.2-mm diameter; Noble and Fuhrman 1998). As sodiumpyrophosphate–enhanced virus extraction efficiency, all sub-sequent steps were carried out using this surfactant. Detailson the procedure for staining the samples before countingunder epifluorescence microscopy are reported in Patel et al.(2007) and can be found also in Suttle and Fuhrman (2010,this volume).

Physical treatment—To test the combined effects ofpyrophosphate and ultrasound treatments on virus extraction,muddy and sandy sediments (n = 3; 0.5 mL for both sedimenttypes) were added to pyrophosphate (5 mM final concentra-tion) and treated by ultrasounds (Branson 2200 sonifier; 100W; 47 kHz) for 0, 1, 3, 8, and 15 min. To prevent overheating,

Fig. 1. Protocol illustrating the steps required for the separation ofviruses from the sediment particle and subsequent counting.


77

ultrasonication was performed in ice bath. If the same sampleis used for counting prokaryotes, it must be noted that theaddition of ice to the sonication bath could result in a signifi-cant decrease of free prokaryotes in the sample (Duhamel andJacquet 2006). The ultrasonication (with 30 s intervals everymin) interval was alternated with gentle shaking. Centrifuga-tion have been shown to remove a fraction of the viruses andprokaryotes in the sample, which are still associated with par-ticles (see below), thus potentially underestimating their totalabundance. In this case, however, it was used to analyze theefficiency of sonication treatment for releasing viruses fromparticles. After centrifugation, aliquots of the supernatantwere processed as described above.

Postsonication extraction efficiency—Many of the initial stud-ies of benthic viruses applied a centrifugation step and a seriesof washing steps following the sonication of the sedimentslurry (e.g., Fischer et al. 2005; Danovaro et al. 2001; Middel-boe et al. 2003). The purpose of centrifugation is to reduce thenumber of particles, which interfere with the subsequentanalysis of viral particles by epifluorescence microscopy (e.g.,by covering the viral particles and by autofluorescence of col-loidal sized sediment particles). In the present study, the effi-ciency of virus detachment from sediment particles waschecked by estimating the ratio of virus abundance after thefirst extraction with ultrasound and pyrophosphate treatmentversus the cumulative virus abundance obtained by this pro-cedure plus three further washing steps. The added steps were(i) an aliquot of supernatant obtained from deep-sea sedimentsamples (0.5 mL sediment plus 4.5 mL MilliQ water andsodium pyrophosphate) after sonication (3 min) and centrifu-gation was withdrawn and treated for counting as describedabove; (ii) the remaining supernatant was carefully dis-charged, the pellet was resuspended with 5 mL MilliQ water,shaken for 1 min, and centrifuged again, an aliquot of thesupernatant was withdrawn, and viruses were counted asdescribe d above; and (iii) this procedure was repeated threetimes (since after the third washing, less than 5% of the totalvirus abundance was encountered).

Alternatively, instead of washing and centrifugation, thesample can be diluted to a point where the sediment particlesdo not interfere significantly with the analysis of viruses in theepifluorescence microscope. In that case the last sonicationstep should be followed by dilution of the sample with 0.02µm filtered bottom water up to 40 mL (instead of centrifuga-tion). It is convenient to perform the whole extraction proce-dure (slurry, incubation with surfactant, sonication, and dilu-tion) in a 50 mL centrifuge tube. The diluted sample is thengently mixed, and one can either prepare a slide with 10-200µL sample and/or take a 1 mL subsample, snap freeze it in liq-uid nitrogen, and store at –80°C.for later slide preparation.

One of the implications of this is that it is slightly more dif-ficult and time consuming to count viruses in the microscope,especially with samples collected deeper sediment layers (e.g.,>10 cm) as the ratio between background noise (from sediment

particles) to actual viral abundance increases with sedimentdepth. The advantage, on the other hand is that the extractionprocedure is faster when compared with a series of centrifuga-tion and washing steps, hence also reducing the handling timeof viruses and therefore the decay of viruses that goes on in theperiod between sampling and slide preparation.

Interference with virus enumeration due to extracellular DNA—Toeliminate uncertainties in virus counting that we found some-times associated with the presence of a matrix of extracellularDNA (Danovaro unpubl. data), we tested the effect of nucleasetreatment on sediment samples. Twenty-five microliters ofDNase I from bovine pancreas (1.9 U mL mL–1), 10 mL nucleaseP1 from Penicillium citrinum (4 U mL mL–1), 10 mL nuclease S1from Aspergillus orizae (2.3 U mL–1), and 10 mL esonuclease 3from Escherichia coli (1.9 U mL mL–1) were added to 1.0 mLaliquots of the supernatant obtained from fresh sandy sedimentsand incubated for 15 min at room temperature. Additionalaliquots of the supernatant (1.0 mL) without enzymes were incu-bated under the same conditions and served as controls.

AssessmentThe extraction optimization was divided into 3 phases,

which were optimized to obtain the highest virus recoveryfrom sediments as described here below.

Chemical treatment—The pyrophosphate extraction effi-ciency was tested comparing treated sediments (final concen-trations 5 and 10 mM) with untreated sediments (Fig. 2).Results presented here indicate that the use of pyrophosphateis highly effective for the dislodgement of viruses from sedi-ment particles, but the choice of an optimal concentration iscrucial for the subsequent viral counting. Deep-sea sedimentsamples incubated with sodium pyrophosphate (5 mM finalconcentration) displayed higher viral counts than untreatedsamples (14.6 ± 2.79 × 109 versus 9.94 ± 6.12 × 109 viruses g–1,respectively). Pyrophosphate concentrations >5 mM in sedi-ment samples did not increase significantly the extraction effi-ciency. On the contrary, too high pyrophosphate concentra-tions were found to decrease the stain contrast thusinterfering with viral enumeration. Shorter incubation times(3-10 min) were less effective in detaching viral particles fromthe sediments, whereas longer incubation times did notincrease the recovery efficiency (data not shown).

Physical treatment—The effects of sonication on viruscounts was tested for 0, 1, 3, 8, and 15 min on coastal anddeep-sea sediments. Results are reported in Fig. 3. The highestvirus counts for both sediment types (1.1 and 11.1 × 109

viruses g–1 for coastal and deep-sea sediments, respectively)were obtained after 3 min of sonication and were significantlyhigher (two- to fourfold; t-test; P < 0.01) than values obtainedwithout sonication (i.e., with simple shaking). Treatmentslonger than 3 min resulted in a progressively lower viralcount, and a sonication lasting 15 min reduced virus countsby about 1 order of magnitude (t-test; P < 0.01 for both sedi-ment types). It must be underlined that our results were


78

obtained in both coastal and deep-sea samples. However, themaximum extraction efficiency at the third minute of sonica-tion varied significantly among the two sediment types, rang-ing from 55% in deep sea muds to 70% in coastal sands.

When we used the pyrophosphate-ultrasound treatment todislodge viruses from sediment particles, we paid special atten-tion to avoid the disruption of virus-infected prokaryotic cells,which might release virus particles, and thus falsifying thenumber of free viruses originally present in the sample. Thisrisk was minimized by performing the ultrasound treatmentin an ice bath and interrupting the treatment every minute for30 s, to prevent overheating and further alterations of samplesbefore counting. Finally, it should be taken into account that,as observed for prokaryotes, the optimal sonication time maystrongly depend on the sonicator model and settings (Epsteinand Rossel 1995), and may therefore vary considerably amonglaboratories.

Postsonication extraction efficiency—The efficiency of virusextraction by pyrophosphate-ultrasound treatment can varysignificantly among different sediment types. Typically ca.60% total extractable viruses are obtained after the first step,but the abundances of viruses extracted by this procedurewere significantly lower than the total cumulative virus abun-dance (t-test; P < 0.01). The subsequent first and second wash-ings recovered 27.5% and 9.0% of the total virus abundance,and after the third wash step, <5% was recovered (Fig. 4). Aquick centrifugation (at 800g per 1 min) can be conductedafter the chemical and physical treatments to reduce the pres-ence of sediment particles and pyrophosphate in theprocessed supernatant, but it has to be checked carefully thepotential loss of viruses associated with this step. In muddydeep-sea sediment samples, a reduction of sediment particlesby centrifugation can be required to count viruses on micro-scope slides without severe interference from nonviral parti-cles. However, a recent study (Siem-Jørgensen et al. 2008)examining extraction of viruses and bacteria from a silty, shal-

low (35 m water depth) sediment, found that the number ofviruses and bacteria kept increasing for 9 subsequent wash-ings. In that case, the extracted fraction after three washesonly represented ~60% and ~45% of viruses and bacteria,respectively, of the numbers that would have been extractedafter 8 washes (Fig. 4b). Tests have been performed to comparethe centrifugation/washing procedure with the more simpledilution procedure, where the sonicated sediment is dilutedrather than going through repeated steps of centrifugationand washing. The tests showed that even very low centrifuga-tion removes bacteria and viruses entrapped by settling parti-cles, and underestimated viral and bacterial abundance by anaverage factor of 2.2 ± 0.17 and 7.7 ± 0.27 (n = 15), respec-tively, relative to the dilution method (Fig. 4c). Centrifugationand washing of sediment samples during extraction can thusrepresent a significant source of error in the quantification ofviruses in sediment compared with the more simple dilutionprocedure, which probably vary between different types ofsediment, depending on the efficiency of viral attachment tothe particles. Furthermore, the washing and centrifugation

Fig. 2. Virus abundance in deep-sea sediments treated and untreatedwith tetrasodium pyrophosphate 5 mM final concentration. Standarddeviations (n = 10) are shown.

Fig. 3. Effect of sonication on virus abundance in surface (a) and deep-sea (b) sediments. Standard deviations (n = 3) are shown.


79

procedures are quite time consuming, especially if such testsneed to be performed for each type of sediment, or perhapseven each sediment depth.

Efficiency of DNase treatment—Viral abundances obtained afternuclease treatment were significantly higher than those observedin untreated samples (5.11 ± 0.15 × 108 and 4.62 ± 0.19 × 108

viruses g–1 of sediment dry weight, respectively; t-test; P < 0.05).

DiscussionIn the present work, we have optimized the original proto-

col for viral extraction from aquatic sediments using as a modelthe extraction process of prokaryotes from sediment samplesoptimized by Ellery and Scheyer (1984). All of the extractionsteps of the protocol were tested for two common sedimenttypes of aquatic environments: shallow sands and deep-seamuds. The extraction efficiency of each step was determined asviral counts determined by epifluorescence microscopy.

The use of surfactants as first step of chemical treatment iswidely used in the study of benthic microbial ecology, virusesincluded (Danovaro et al. 2001; Middelboe et al. 2003; Fischeret al. 2005). Surfactants are able to weaken hydrophilic linksamong sedimentary and biological particles thus allowing thesubsequent separation between viruses and sediment grains.Among the different surfactants commercially available, themost widely used in studies of benthic viral abundance istetrasodium pyrophosphate. The literature provides contrast-ing results about the use of this surfactant. Middelboe et al.(2003) reported that the use of tetrasodium pyrophosphatesignificantly increases the extraction efficiency of viruses fromestuarine sands, in a percentage ranging from ~60 to ~70%,whereas Duhamel and Jacquet (2006) found that the use ofthe only pyrophosphate before FCM did not increase signifi-cantly viral counts in freshwater sands, and suggested to use amix of pyrophosphate and Tween 80 (a nonionic detergentand an emulsifier) to obtain extraction efficiencies rangingfrom 25 to 40%. The analyses conducted on deep-sea mudsrevealed that the simple addition of pyrophosphate, althoughaugmenting the average counts (Fig. 2), did not increase theextraction efficiency significantly (t-test; P = 0.296). The dis-crepancy observed comparing different sediment types wasnot due to differences in the final concentration of pyrophos-phate solution (5 mM in all cases) or in the incubation time(15 min room temperature in all cases), and is likely depend-ent on the different mineralogical compositions of the sub-strates. Deep-sea muds, when compared with surface sands,displayed a finer sediment texture, and it is known that thesilt fraction is able to create stronger electrostatic links withbiological particles, including both benthic prokaryotes andviruses. This result suggests a lower extractability of viruses insilty sediments and requires the ultrasound treatment toenable the extraction of viruses (Fig. 3). Moreover, since viralsorption to sediment particles usually increases with increas-ing cation concentration in solution, particularly in the pres-ence of divalent cations (Schijven and Hassanizadeh 2000),the observed differences in the extractability of virusesbetween marine and freshwater sediments might be due to dif-ferences in cation concentrations. Finally, because viral counts

Fig. 4. Postsonication extraction efficiency of virus recovery from (a)deep-sea sediments and (b) from shallow coastal silty sediments (panel bredrawn from a supplemental image from Siem-Jørgensen et al. 2008 andused with permission). (c) Effects of centrifugation and dilution, respec-tively, of sediment slurries prior to preparation of EFM slides for the recov-ery of viruses and bacteria in samples from a shallow coastal sediment.


80

are strongly influenced by the kind of analytical approachused, results obtained by epifluorescence microscopy (Middel-boe et al. 2003) are difficult to compare with results obtainedby flow cytometry (Duhamel and Jacquet 2006), and protocolsdeveloped for one approach could give different results ifapplied to the other.

Our results also indicate that pyrophosphate-treated sam-ples were characterized by significantly lower CVs thanuntreated samples (19.1 versus 61.6%; t-test; P < 0.05). Similarresults were reported for benthic prokaryotes (Epstein andRossel 1995), suggesting that the use of pyrophosphateincreases counting accuracy, making sufficient the counting ofa lower number of optical fields. Minimizing the variabilityamong replicates (i.e., obtaining low CVs) is also particularlyimportant in viral counting, especially when differencesamong samples collected in different stations or in differentincubation times are very low.

The extraction efficiency of viruses from sediment samplesis significantly increased in all our samples with a furtherphysical detachment by ultrasonication. In this work, wetested the optimal sonication time comparing virus abun-dance obtained by 6 different treatments, from 0 to 15 min(Fig. 3). By our results, the highest virus recovery is obtainedafter 3 min of sonication in both coastal sands and deep-seamuds. We obtained a higher extraction efficiency in sands(70%), whereas in deep-sea sediments viral abundanceincreased by 55% after 3 min of sonication. These results sug-gest that the optimal sonication time could vary amongmarine and freshwater sediments and could depend on thesediment grain size. Fischer et al. (2005) reported an optimalsonication time of 1 min in silty freshwater sediments, withan increase in virus counts of 11% to 27% compared with not-sonicated samples, whereas Middelboe et al. (2003) found thata 3-min sonication increased the extraction efficiency of 65%to 78% in estuarine sands.

The presented protocol allows extraction of most extractableviruses without postsonication washing. The postsonicationextraction efficiency, tested in the deep-sea mud, revealed thatextraction alone only released 60% of the extractable viruses.In these sediments, however, 90% of total extractable viruseswere dislodged after a single subsequent washing. The effect ofthe washing procedure can vary among samples, but generallypostsonication washings increase virus counts significantlyfrom ca. 2% in coastal sands (Hewson and Fuhrman 2003) andca. 11% to 40% in estuarine sediments (Middelboe et al. 2003;Siem-Jørgensen et al. 2008). Tests on the efficiency of theapproach based on repeated washings versus the test based onsediment dilution have provided conflicting results. The firstapproach has the advantage of making clear slides easily readunder epifluorescence microscopy, but requires a carefulexamination a priori to determine the extraction efficiency,the potential loss of viruses during centrifugation, and thesubsequent coefficient for the calculation of the total viralabundance of the sediments. The approach based on sediment

dilution has the advantage of avoiding the problems createdby the centrifugation and the need for determining the coeffi-cient of extraction efficiency, but depending on the sedimenttype, is possibly complicated by the difficulty of countingviruses in an optical field rich in sediment particles and thepotential masking effects.

Comments and recommendationsOur results also indicate that the use of DNase treatment is

extremely useful, especially in muddy sediments (Fig. 5). It isconceivable that dissolved DNA within sediment samplescould bind to fluorochromes and thereby inflate viral esti-mates. To eliminate uncertainties in viral counting due toextracellular DNA interference, we tested the effect of DNasetreatment. Free, extracellular DNA could be extremely abun-dant in sediments (Danovaro et al. 1999; Dell’Anno et al.1998; Dell’Anno et al. 1999; Dell’Anno and Danovaro 2005)and a nuclease treatment significantly reduce the fluorescencenoise (due to the SYBR staining of extracellular DNA), whichcan have an important masking effect and determine theunderestimation of the actual viral abundance in the sedi-ment. We found that the sediment samples treated DNasesdisplayed viral counts significantly higher than untreatedsamples (on average ~10% higher values). At the same time, ithas been reported that in seawater samples DNases are able todegrade viral particles, thus influencing viral counts(Maruyama et al. 1993). To what extent this applies to the sed-iment is unknown yet, but data accumulated so far indicatethat a final concentration of 1 U mL–1 increases viral countsthrough the effect of increased visibility of the optical fieldsduring counting.

Accurate quantification of benthic viral abundance is fun-damental prerequisite for understanding spatial and temporaldynamics of benthic viruses. However, the study of benthicviruses is still a relatively new discipline in aquatic viral ecol-ogy and obviously more work is needed to optimize theextraction of viruses from sediments and clarify the efficiencyof various procedures in dislodging viruses from particles indifferent types of sediments.

Fig. 5. Effect of DNase treatment extraction efficiency on muddy sedi-ments. The standard deviations (n = 3) are shown.


81

References

Berg, Ø., K. Y. Børsheim, G. Bratbak, and M. Heldal. 1989.High abundance of viruses found in aquatic environments.Nature 340:467-468.

Bird, D. F., S. K. Juniper, M. Ricciardi-Rigault, P. Martineu, Y. T.Prairi.e., and S. E. Calvert. 2001. Subsurface viruses and bac-teria in Holocene/Late Pleistocene sediments of SaanichInlet, BC: ODP Holes 1033B and 1034B, Leg 169S. Mar.Geol. 174:227-239.

Corinaldesi, C., and R. Danovaro. 2003. Ecology of viruses inaquatic sediments. Rec. Res. Devel. Microbiol. 7:119-134

———, A. Dell’Anno, and R. Danovaro. 2007. Viral infectionplays a key role in extracellular DNA dynamics in marineanoxic systems. Limnol. Oceanogr. 52:508-516.

Danovaro, R., A. Dell’Anno, A. Pusceddu, and M. Fabiano.1999. Nucleic acid concentrations (DNA, RNA) in the conti-nental and deep-sea sediments of the Eastern Mediter-ranean: relationships with seasonally varying organic inputsand bacterial dynamics. Deep Sea Res. I 46:1077-1094.

——— and M. Serresi. 2000. Viral abundance and virus-tobacterium ratio in deep-sea sediments of the EasternMediterranean. Appl. Environ. Microbiol. 66:1857-1861.

———, A. Dell’Anno, A. Trucco, and S. Vannucci. 2001. Deter-mination of virus abundance in marine sediments. Appl.Environ. Microbiol. 67:1384-1387.

———, E. Manini, and A. Dell’Anno. 2002. Higher abundanceof bacteria than viruses in deep Mediterranean sedimentsAppl. Environ. Microbiol. 68:1468-1472.

———, A. Dell’Anno, C. Corinaldesi, M. Magagnini, R. Noble,C. Tamburini, and M. Weinbauer. 2008. Major viral impacton the functioning of benthic deep-sea ecosystems. Nature454:1084-1087.

Dell’Anno, A., M. Fabiano, G. C. A. Duineveld, A. Kok, andR. Danovaro. 1998. Nucleic acid (DNA, RNA) quantificationand RNA/DNA ratio determination in marine sediments:comparison of spectrophotometric, fluorometric, and high-performance liquid chromatography methods and estimationof detrital DNA. Appl. Environ. Microbiol. 64:3238-3245.

———, ———, M. L. Mei, and R. Danovaro. 1999. Pelagic- benthic coupling of nucleic acids in an abyssal location ofthe northeastern Atlantic Ocean. Appl. Environ. Microbiol.65:4451-4457.

———, and R. Danovaro. 2005. Extracellular DNA plays a keyrole in deep-sea ecosystem functioning. Science 309:2179.

Duhamel, S., and S. Jacquet. 2006. Flow cytometric analysis ofbacteria and virus-like particles in lake sediments. J. Micro-biol. Met. 64:316-322.

Ellery, W. N., and M. H. Scheyer. 1984. Comparison of homoge-nization and ultrasonication as techniques in extractingattached sedimentary bacteria. Mar. Ecol. Prog. Ser. 15:247-250.

Epstein, S. S., and J. Rossel. 1995. Enumeration of sandy sedi-ment bacteria: search for optimal protocol. Mar. Ecol. Prog.Ser. 117:289-298.

Fischer, U. R., A. K. T. Kirschner, and B. Velimirov. 2005. Opti-mization of extraction and estimation of viruses in siltyfreshwater sediments. Aquat. Microb. Ecol. 40:207-216.

Fuhrman, J. A. 1999. Marine viruses and their biogeochemicaland ecological effects. Nature 399:541-548.

Hewson, I., J. O’Neil, C. Heil, G. Bratbak, and W. Dennison.2001. Effects of concentrated viral communities on photo-synthesis and community composition of co-occurringbenthic microalgae and phytoplankton. Aquat. Microb.Ecol. 25:1-10.

———, and J. A. Fuhrman. 2003. Viriobenthos production andvirioplankton sorptive scavenging by suspended sedimentparticles in coastal and pelagic waters. Microb. Ecol.46:337-347.

Maranger, R., and D. F. Bird. 1996. High concentrations ofviruses in the sediments of Lac Gilbert, Quebec. Microb.Ecol. 31:141-151.

Maruyama, A., M. Oda, and T. Higashihara. 1993. Abundanceof virus-sized non-DNase-digestible DNA (coated DNA) ineutrophic seawater. Appl. Environ. Microbiol. 59:712-717.

Middelboe, M., R. N. Glud, and K. Finster. 2003. Distributionof viruses and bacteria in relation to diagenetic activity inan estuarine sediment. Limnol. Oceanogr. 48:1447-1456.

———, and ———. 2006. Viral activity along a trophic gradi-ent in continental margin sediments off central Chile. Mar.Biol. Res. 2:41-51.

———, ———, F. Wenzhöfer, K. Oguri, and H. Kitazato. 2006.Spatial distribution and activity of viruses in the deep-seasediments of Sagami Bay, Japan. Deep-Sea Res. 53:1-13.


Patel, A., R. T. Noble, J. A. Steele, M. S. Schwalbach, I. Hewson,and J. A. Fuhrman. 2007. Virus and prokaryote enumera-tion from planktonic aquatic environments by epifluores-cence microscopy with SYBR Green I. Nature Protocols2:269-276.


Schijven, J. F., and S. M. Hassanizadeh. 2000. Removal ofviruses by soil passage: overview of modeling, processesand parameters. Crit. Rev. Environ. Sci. Technol. 30:49-127.

Siem-Jørgensen, M., R. N. Glud, and M. Middelboe. 2008. Viraldynamics in a coastal sediment: Seasonal pattern, control-ling factors and relations to the pelagic-benthic coupling.Mar. Biol. Res. 4:165-179; supplemental material.

Suttle, C. A. 2005. Viruses in the sea. Nature 437:356-361.———, and J. A. Fuhrman. 2010. Enumeration of virus particles

in aquatic or sediment samples by epifluorescencemicroscopy, p. 145-153. In S. W. Wilhelm, M. G. Weinbauer,and C. A. Suttle [eds.], Manual of Aquatic Viral Ecology. ASLO.

Whitman, W. B., D. C. Coleman, and W. J. Wiebe. 1998.Prokaryotics: the unseen majority. Proc. Nat. Acad. Sci.U.S.A. 95:6578-6583.

82

Introduction

Given the inherent difficulties associated with the isolationof purified viruses from aquatic environments, manyresearchers have chosen to explore the diversity and distribu-tion of viruses using culture-independent molecular tech-niques. Due to their nature as obligate intracellular parasites,examination of viruses in the lab setting requires the con-

comitant maintenance and growth of the host organisms thatthey infect. Culture-independent approaches circumvent thisconstraint, allowing the researcher to characterize complexviral consortia directly. To achieve this however, one firstrequires sufficient knowledge of the genetic composition ofthe virus population in question. Within this chapter, we willdescribe how to interrogate the ecology of specific viruses innatural systems based on the limited amount of genetic infor-mation available from characterized viral isolates.

The characterization of viruses by these methods can brieflybe described within a flow diagram that outlines the major stepsin the construction and analysis of marker gene libraries (Fig. 1).Successful execution of this process however, requires carefulapplication of appropriate controls and independent valida-tions of individual steps. Within this chapter, we endeavor tohighlight major components of these methods, discussingoptions and considerations in the specific step-by-step details.By building on the previous experience of numerous labs, thischapter should not only be useful to the new virus ecologist, butalso serve as a valuable resource to established research groups.

Marker genes for viruses are typically amplified from aquaticsamples for one of three purposes: 1) determining the presenceof specific viruses, 2) determining the diversity of a group ofrelated viruses, or 3) determining the abundance of a specificvirus population based on the abundance of a marker gene.Within the context of this chapter, we will focus on the meth-ods associated with 1 and 2 above, constraining our foci toviruses infecting algae, bacteria, and heterotrophic flagellates.

The construction and analysis of marker gene librariesSteven M. Short1*, Feng Chen2, and Steven W. Wilhelm3

1Department of Biology, University of Toronto Mississauga, Mississauga, ON L5L 1C6, Canada2Center of Marine Biotechnology, The University of Maryland Biotechnical Institute, Baltimore, MD 21202, USA3Department of Microbiology, The University of Tennessee, Knoxville, TN 37922, USA

AbstractMarker genes for viruses are typically amplified from aquatic samples to determine whether specific viruses

are present in the sample, or to examine the diversity of a group of related viruses. In this chapter, we will pro-vide an overview of common methods used to amplify, clone, sequence, and analyze virus marker genes, andwill focus our discussion on viruses infecting algae, bacteria, and heterotrophic flagellates. Within this chapter,we endeavor to highlight critical aspects and components of these methods. To this end, instead of providinga detailed experimental protocol for each of the steps involved in examining virus marker gene libraries, wehave provided a few key considerations, recommendations, and options for each step. We conclude this chap-ter with a brief discussion of research on a major capsid protein (g20) of cyanomyoviruses using this work as acase study for polymerase chain reaction primer design and development. By building on the experience ofnumerous labs, this chapter should not only be useful to the new virus ecologist, but also serve as a valuableresource to established research groups.




The authors would like to thank colleagues and particularly previousstudents whom have helped work out the bugs of the described tech-niques. The authors also acknowledge the support of the NaturalScience and Engineering Research Foundation Canada (SMS) and theNational Science Foundation (NSF-OCE 0452409 SWW) for the supportof their research programs that lead to the development of these ideas,and the Scientific Committee for Oceanographic Research for support-ing working group 126 (marine virus ecology).

ISBN 978-0-9845591-0-7, DOI 10.4319/mave.2010.978-0-9845591-0-7.82Suggested citation format: Short, S. M., F. Chen, and S. W. Wilhelm. 2010. The construction andanalysis of marker gene libraries, p. 82–91. In S. W. Wilhelm, M. G. Weinbauer, and C. A. Suttle[eds.], Manual of Aquatic Viral Ecology. ASLO.


MANUALof


Short et al. Viral marker genes

83


Viral gene markers—Viruses are probably the most diverse bio-logical entity in the biosphere. Despite the fact that no universalgene marker (like the 16S and 18S ribosomal ribonucleic acidgenes from prokaryotes and eukaryotes, respectively) is availablefor all viruses, many studies have demonstrated that certaingenes are conserved among certain groups of viruses that infectclosely related hosts. By designing oligonucleotide primers thathybridize to conserved regions of these marker genes, manyresearchers have used PCR to amplify virus marker genes fromenvironmental samples to investigate the genetic diversity ofspecific groups of viruses in variety of aquatic environments (seeTable 1). Currently, viral capsid related genes and virus-encodeddeoxyribonucleic acid (DNA)/ribonucleic acid (RNA) polymerasegene are the most widely used genetic markers for aquaticviruses, and various PCR primer sets have been designed to tar-get these genetic markers (Table 1). Studies of these virus markergenes have demonstrated that viruses in the marine environ-ments are much more diverse than might be expected based onthe limited numbers of cultivated viruses. With the recent rapidincrease in the number of microbial genes and genomes avail-able in public sequence databases, many viral signature genes(e.g., genes involved in photosynthesis or DNA replication) havebeen identified. By taking advantage of the plethora of informa-tion now available in sequence databases (e.g., NCBI’s GenBankdatabase at http://www.ncbi.nlm.nih.gov/Genbank/), poly-merase chain reaction (PCR)-based methods can provide arapid, sensitive, and economical approach to explore thediversity of viral genes or viral groups in nature and addressimportant questions about the distribution, diversity, andeven activity of virus in aquatic ecosystems.

Sample collection and preparation—The history and details ofproper sample collection and processing before PCR amplifica-tion of virus genes are numerous. In some cases, virus markerscan readily be amplified directly from unaltered whole watersamples. In other cases, preconcentration of virus particles maybe required; this process is thoroughly explained in anotherchapter (Wommack et al. 2010, this volume). For qualitativepurposes, PCR amplification is often most successful from con-centrated virus communities. However, the variety of stepsinvolved in either ultrafiltration or ultracentrifugationincreases the potential for particle loss, which can complicatequantitative analyses. Ultimately, the ambient abundance ofviruses and the sensitivity of the particular assay will dictatethe approach taken in preparing samples for analysis.

Similarly, a debate continues as to whether nucleic acidsneed to be extracted from virus samples prior to PCR amplifica-tion, or whether viral genetic material can be directly amplified.Many of the early studies on virus diversity in aquatic systemsemployed virus concentrates (see Wommack et al. 2010, thisvolume) as starting material. More recently, researchers havedirectly amplified marker elements from unextracted virus- bearing samples (Short and Short 2008; Wilhelm and Matteson2008). Moreover, when comparing PCR amplification of unex-tracted virus concentrates and polyethylene glycol (PEG) pre-cipitated virus concentrates to extracted viral DNA, theextracted DNA often produced poor PCR amplification yields(Chen et al. unpubl. results). While the approach of using unex-tracted virus DNA is often quite successful and requires onlyslight changes to the PCR protocol, its efficacy may depend onthe capsid/membrane composition of the virus in question.Nonetheless, a simple freeze/heat treatment consisting of 3 rep-etitions of freezing virus samples until solid followed heating to95°C for 2 min has been used to generate PCR-amplifiable virusDNA from a variety of aquatic samples (Chen et al. 1996; Shortand Short 2008; Short and Suttle 2002).

Primer design—In targeting a specific population or group ofmicroorganisms in aquatic environments using PCR-basedmethods, primer design is often the most critical and challeng-ing step. Thankfully, because PCR is a well established technique,many excellent volumes have been written on the optimizationand application of PCR, and most include some discussion of thecritical considerations for primer design (e.g., Altshuler 2006;Atlas 1993; Innis et al. 1990; Mcpherson and Moller 2006), andsome focus entirely on primer design (Yuryev 2007). In addition,freely available software can be found on the Internet that canaid in primer design. For example, the program OligoAnalyzer3.1 is available at the Integrated DNA Technologies Web site(http://www.idtdna.com/analyzer/Applications/OligoAnalyzer/).This particular software allows the user to enter oligonucleotide(primer) sequences and provides general analytical informationsuch as predicted melting temperatures for each primer, as wellas more complicated but useful information such as the primer’spotential for hairpin formation, self-dimer formation, and hetero-dimer formation. This software also allows the user to

Fig. 1. Flowchart of steps from harvesting nucleic acids from virus sam-ples to data analysis.


84

Tab

le 1

.C

onse

rved

gen

es a

nd p

rimer

s us

ed in

the

con

stru

ctio

n an

d an

alys

is o

f m

arke

r ge

ne li

brar

ies

Taxo

no

mic

gro

upG

ene

Gen

e p

rod

uct

Prim

er s

eque

nce

sR

efer

ence

s

Cya

nop

hage

psbA

pho

tosy

nthe

tic g

ene

D1

58-V

DID

GIR

EP-6

6:G

TN G

AY A

TH G

AY G

GN

ATH

MG

N G

AR

CC

(Zei

dner

et

al.

2003

)33

1-M

HER

NA

HN

FP-3

40:

GG

R A

AR

TTR

TGN

GC

R TT

N C

KY T

CR

TGC

AT

Cya

nop

hage

psbA

& p

sbD

pho

tosy

nthe

tic g

ene

psb

AF:

GTN

GAY

ATH

GAY

GG

N A

TH M

GN

GA

R C

C(M

illar

d et

al.

2004

)D

1 &

D2

psb

AR

:G

GR

AA

R TT

R TG

N G

CC

yano

phag

e p

sbA

Phot

osyn

ethe

tic g

ene

D1

psb

A-9

3F:

TAY

CC

N A

TY T

GG

GA

A G

C(W

ang

and

Che

n 20

08)

& C

yano

bact

eria

psb

A-3

41R

:G

TR T

TR A

AG

GG

D G

AR

CT

Myo

virid

aeg2

0vi

ral c

apsi

d as

sem

bly

pro

tein

CPS

1:G

TA G

WA

TTT

TC

T A

CA

TTG

AYG

TTG

G(F

ulle

r et

al.

1998

)C

PS2:

GG

T A

RC C

AG

AA

A T

CY

TCM

AG

C A

TM

yovi

ridae

g20

vira

l cap

sid

asse

mbl

y p

rote

inC

PS3:

TGG

TAY

GTY

GAT

GG

M A

GA

(Zho

ng e

t al

. 20

02)

CPS

4:C

AT W

TC W

TC C

CA

HTC

TTC

CPS

8:A

AA

TAY

TTD

CC

A A

CA

WAT

GG

AM

yovi

ridae

g23

maj

or c

apsi

d p

rote

inM

ZIA

1bis

:G

AT A

TT T

GI G

GI G

TT C

AG

CC

I ATG

A(F

ilee

et a

l. 20

05)

MZ

IA6:

CG

C G

GT

TGA

TTT

CC

A G

CA

TG

A T

TT C

Myo

virid

aeM

CP

maj

or c

apsi

d p

rote

inA

N15

MC

PF5:

GTT

CC

T G

GC

AC

A C

CT

GA

A G

CG

(Bak

er e

t al

. 20

06)

(cya

nop

hage

tha

t A

N15

MC

PR5:

CTT

AC

C A

TC G

CT

TGT

GTC

GG

C A

TCin

fect

Ana

baen

a&

Nos

toc)

Nod

ular

ia-s

pec

ific

g23

maj

or c

apsi

d p

rote

inC

AP1

:AT

T TG

Y G

GY

GTT

CA

G C

CK

ATG

A(J

enki

ns a

nd H

ayes

200

6)cy

anop

hage

C

AP2

:A

AC

RAY

TTC

RC

G G

TT G

AT T

TC C

A(m

yovi

rus

& s

ipho

viru

s)T7

-like

Pod

opha

geT7

-DN

A

DN

A p

olym

eras

eT7

DPo

l230

F:A

RG A

RM R

IA A

YG G

IT(B

reitb

art

et a

l. 20

04)

pol

ymer

ase

T7D

Pol5

10R

:G

TR T

GD

ATR

TC

I CC

HEC

TOR

Pol2

9F:

GC

A A

GC

AA

C T

TT A

CT

GTG

GH

ECTO

RPo

l711

R:

CG

A G

AG

ATA

CA

C C

AA

CG

A A

PAR

ISPo

l25F

:AT

A C

TA C

AC

GC

T A

CT

CTG

GPA

RIS

Pol7

01R

:G

AG

TG

G C

AA

GA

G G

AG

TTA

TC

yano

pha

ge

Shea

th R

TF:

AC

A T

CA

GC

G T

TC G

TT T

CG

G(T

akas

him

a et

al.

2007

)of

Mic

rocy

stis

Shea

th R

TR:

CA

A T

CT

GG

T TA

G G

TA G

GT

CG

EhV

MC

Pm

ajor

cap

sid

pro

tein

MC

P-F2

:TT

C G

CG

CTC

GA

G T

CG

ATC

1(S

chro

eder

et

al.

2003

)M

CP-

R2:

GA

C C

TT T

AG

GC

C A

GG

GA

GPh

ycod

navi

ridae

MC

Pm

ajor

cap

sid

pro

tein

MC

P Fw

d:

GG

Y G

GY

CA

R C

GY

ATT

GA

(Lar

sen

et a

l. 20

08)

MC

P R

ev:

TGI A

RY T

GY

TCR

AYI A

GG

TA

Phyc

odna

virid

aePo

lD

NA

pol

ymer

ase

AV

S1:

GA

RGG

IGC

IAC

IGTI

YTIG

AYG

C(C

hen

and

Sutt

le 1

995)

AV

S2:

GC

IGC

RTA

ICKY

TTYT

TISW

RTA

POL:

SWRT

CIG

TRTC

ICC

RTA

HaV

PKN

61A

:G

AT C

TG A

CT

CAT

G A

CC

CA

A C

G(N

agas

aki e

t al

. 20

01)

PKN

61B

:C

CA

CC

A T

CA

GA

A T

CA

TC

A C

CC

AG

TPK

N62

A:

AG

A T

GA

AG

A C

GA

TG

A T

GA

CG

A C

GA

TPK

N62

B:

GTG

AA

G A

TG A

AA

AG

G A

AA

GC

A A

GG

APA

GB

01A

:G

TC A

AG

AA

C A

AT C

GA

AC

C G

TA A

TPA

GB

01B

:TT

T AT

T C

TG A

TT T

CT

GTC

CC

GT7

-like

Pod

opha

geT7

-DN

A

DN

A p

olym

eras

eT7

DPo

l230

F:A

RG A

RM R

IA A

YG G

IT(C

ulle

y an

d St

ewar

d 20

07)

pol

ymer

ase

T7D

Pol5

10R

:G

TR T

GD

ATR

TC

I CC

HEC

TOR

Pol2

9F:

GC

A A

GC

AA

C T

TT A

CT

GTG

GH

ECTO

RPo

l711

R:

CG

A G

AG

ATA

CA

C C

AA

CG

A A

PAR

ISPo

l25F

:AT

A C

TA C

AC

GC

T A

CT

CTG

GPA

RIS

Pol7

01R

:G

AG

TG

G C

AA

GA

G G

AG

TTA

T

1. T

he 4

0 bp

GC

-cla

mp

seq

uenc

e as

soci

ated

with

thi

s p

rimer

was

om

itted

for

bre

vity

.2.

The

cod

es f

or m

ixed

bas

es a

re:

R =

A,

G;

Y =

C,

T; M

= A

, C

; K

= G

, T;

S =

C,

G;

W =

A,

T; H

= A

, C

, T;

B =

C,

G,

T; V

= A

, C

, G

; D

= A

, G

, T;

n =

A,

G,

C,

T; I

= in

osin

e.


85

directly compare their primer sequences to sequences archivedin the GenBank database.

As a general guideline, several criteria should be consideredwhen designing PCR primers for the analysis of aquaticviruses:1) the target gene should be evolutionarily conserved among

the viruses of interest;2) at least one region with a minimum of 6 consecutive

amino acids (or >16 nucleotides) that is conserved onlyamong the target organisms can be identified in multiplesequence alignments;

3) when multiple regions are available for primers, regionswith the least degeneracy should be considered;

4) at sites of 4-fold degeneracy where G, A, T, and C shouldall be considered, the practical degeneracy of the primercan be reduced by using an inosine residue;

5) the desired size of PCR products may vary for differentapplications (e.g., shorter PCR amplicons ranging from150–400 bp are ideal for DGGE applications and quantita-tive PCR (qPCR), whereas longer amplicons ranging from500–800 bp are desirable for the phylogenetic analyses ofclone libraries;

6) more than one set of primers should be designed andtested when multiple target regions are available;

7) because the design of specific PCR primers relies on thenumber of known target sequences, it is important toinclude as many related sequences as possible when creat-ing sequence alignments for primer design;

8) PCR primers should be modified (redesigned) as moresequences belonging to the target organisms become available.

In some cases, PCR primers (e.g., primers that target the g20gene of cyanomyoviruses) were originally designed based on alimited number of gene sequences. This can result in poorlyconstrained sequence information since the specificity ofprimers was not well defined in the first place. Although it caneasily be argued that poorly constrained sequence data aremore valuable than no data at all, it is nonetheless importantto use as much sequence data from representative groups ofviruses when designing and redesigning PCR primers (Fig. 1).For example, using newly available sequence data, g20 primersspecific for cyanomyoviruses have been modified, and muchhigher PCR specificity has been achieved (Chen et al. unpubl.data; more details are described below; Fig. 2).

PCR amplification—PCR is a widely used in vitro techniquethat generates millions or even billions of copies of specificgene fragments. There are numerous general and field-specificprocedural references for PCR, and almost all of the major sci-entific vendors distribute PCR reagents and equipment. There-fore, this section will only provide a simple guide to help neo-phyte molecular biologists get started; obviously there are fartoo many options that could be considered for a particularPCR application to discuss them here. Whenever possible, theprocedure outlined in published literature describing the use

of a particular set of primers should be followed. However,researchers should not be surprised when they need to trou-bleshoot previously described conditions for a particular reac-tion. In our experience, different Taq DNA polymerases, ther-mal cyclers and reagents, and even different workers, can havea dramatic influence on PCR results.

One of the most important considerations for PCR is labhygiene. Because of its sensitivity, PCR reactions can easily becontaminated with amplifiable DNA. It is much easier to takeproactive measures to prevent contamination that to have totrack down the source of contamination after it has beendetected. All reagents should be dispensed into small portionsor working stocks before their use. This practice has the doublebenefit of preventing the loss of large stocks of reagents in theevent that they become contaminated, and it also minimizesthe number of freeze-thaw cycles that a reagent endures. Theuse of aerosol barrier tips for automatic pipettors, frequent san-itization of lab benches, and dedicated lab spaces or sterilehoods for setting up PCRs are also highly recommended.Although lab coats are generally recommended as essential per-sonal protective equipment, they must be washed frequently ifworkers are to wear them when setting up PCR reactions; adirty sleeve can be a major reservoir for contaminating nucleicacids! As a final comment, although it may seem obvious, itcannot be stressed enough that positive and negative controlsmust be included in every single PCR experiment.

PCR reactions are set up via the creation of a master mixthat includes all reagents except the template nucleic acid.Generally, it is wise to prepare a slightly larger volume mastermix that is absolutely necessary because the wasted reagentsrepresent a trivial expense, and minor pipettor inaccuraciescan lead to a short fall when dispensing the master mix intoindividual reaction tubes. The following reagents and concen-trations are typical for many PCR reactions:

• PCR buffers are usually supplied at a 10× or 2× concen-tration with the polymerase enzyme. The buffers compo-nents are somewhat variable and are optimized by themanufacturer for use with a particular thermally-stableDNA polymerase enzyme.

• MgCl2 is usually supplied in a 50 or 25 mM stock. The

working concentration can vary between 1.5 to 4.0 mMdepending on the primer sequences. For any particularPCR protocol, the optimal working concentration shouldbe empirically determined as it can have a dramatic effecton the yield of PCR products and the stringency of thereaction.

• dNTPs can be purchased individually, or in mixtures of allfour nucleotides. Generally, dNTPs are mixed and storedas stock solutions with each dNTP at a concentration of10 mM, or 40 mM total for all dNTPs. For most PCR pro-tocols, final concentrations of 0.2 mM of each dNTP issufficient and provides ample product yield without neg-atively affecting the PCR specificity or fidelity.

• oligonucleotide primers can be ordered as lyophilized


86

stocks and can be reuspended in sterile, pure water or TEbuffer (10 mM Tris, 0.1 mM EDTA, pH 7.5) for long-termstorage at a concentration of 100 µM. Generally, aliquots ofworking stocks are made at 10 µM and the final concen-tration of each primer in a PCR reaction can range from 0.1to 1.0 µM (i.e., a total of 10 to 100 pmol of primer in a finalreaction volume of 50 µL) depending on the primer. Gen-erally, PCR with degenerate primers require slightly higher

primer concentrations, but the optimal primer concentra-tion should be determined empirically.

• thermally stable DNA polymerases are the key ingredi-ent in PCR as these enzymes withstand the extreme tem-perature fluctuations of thermal cycling. For manyyears, Taq DNA polymerase was the standard enzymeusing for PCR. However, many vendors now producevarious enzymes or enzyme mixtures that are optimized

Fig. 2. Left: Phylogenetic analysis of cyanomyovirus g20 gene sequences (ca. 390 bp) from excised DGGE bands. The bootstrap values (>50) wereshown on the major nodes. Right: DGGE profile of PCR-amplified g20 gene fragments at Sta. 804 in the Chesapeake Bay from September 2002 to May2004.


87

for long PCR (amplification of fragments >10 Kb), orhigh fidelity amplification. Additionally, most manufac-turers now produce reasonably priced hot-start enzymesthat are not active until after the initial denaturationstep. These hot-start enzymes are very useful as they pre-vent amplification artifacts produced by nonspecificprimer annealing during the initial ramping up to thedenaturation temperature.

• H2O is added in sufficient volume to bring the total vol-ume up to that desired for each reaction (the total volumefor individual reactions is typically 25 or 50 µL dependingon the desired yield). Although it is often overlooked as apotential source of amplification difficulties, H2O qualityis critically important. When possible, certified nuclease-free water should be used, but good results can beobtained with pure water that has been ultrafiltered andis ion free (i.e., 18.3 MΩ-cm resistivity).

Cloning and sequencing—By design, PCR methods for ampli-fying nucleic acids from aquatic viruses use universal primersthat target related but different gene sequences. Because Sanger(dideoxy-based) sequencing reactions are confounded whenmore than one template is present, gene fragments from natu-ral populations must be separated before sequencing reactionscan be conducted. The most common approach to separateindividual amplified gene fragments is to clone the PCR prod-ucts into a plasmid vector, transform bacterial cells with therecombinant plasmids, and purify plasmids from individual iso-lated bacterial colonies; generally, each colony will contain onlyone type of recombinant plasmid. Purified plasmids can then beused as templates for sequencing reactions. Other methods likedenaturing gradient gel electrophoresis (DGGE) can also beused to separate individual gene fragments from complex mix-tures of PCR products. For these methods, individual bands thattheoretically represent only a single DNA fragment are excisedfrom the gel and are re-amplified with second round of PCR.After these second round PCR products are purified, they canthen be used as templates for sequencing reactions.

Cloning PCR products has become relatively routine, andmany manufacturers produce kits that can be used. Althoughthe cost of the kits may exceed the cost of reagents prepared in-house, the time savings and efficacy of the kits far exceeds therelatively minor increased cost of cloning. The same statementcan be made for most kit-based molecular methods, and there-fore we have included below, lists of some common kits thatcan be used for many of the steps involved in the creation ofmarker gene libraries. The list of kits that we have provided isnot meant to indicate any preferences or be all inclusive. Rather,the lists that follow are included to simply suggest a few reliablesources for these kits; many other manufacturers produce simi-lar kits that may be equally cost effective and efficient. Mostcloning or DNA purification kits include detailed instructionsand trouble-shooting guides, and generally the manufacturer’srecommendations and protocols should be followed. The com-petent cells used for bacterial plasmid transformation are often

included in cloning kits, or they can be purchased separately.Although competent cells prepared by individual labs are con-siderably less expensive than commercially prepared cells, theeffort to produce them may not be worth the cost savings unlessthey will be used routinely. Most general molecular biologymanuals provide a protocol for the preparation of competentcells (Ausubel et al. 2002; Sambrook et al. 1989). Two types ofkits are available for cloning PCR products. Some are based ona TA-cloning method that takes advantage of the single deox-adenosine overhang left by Taq DNA polymerase and othernon-proofreading polymerase enzymes, while others aredesigned to clone blunt-ended PCR products. In either case, thenumber of colonies that contain recombinant plasmids withthe desired PCR fragment can be greatly enhanced by loadingall of the PCR reaction in an agarose gel, excising the fragmentof the appropriate size, and purifying the fragment using a com-mercial gel extraction kit. In our experience, this step greatlyreduces the possibility of ligating primer-dimers or other PCRartifacts into the plasmid vector, thereby enhancing the recov-ery of clones containing the gene fragment of interest.

Like PCR product cloning, DNA sequencing has become rou-tine despite the high cost of the instruments used for auto-mated sequence analysis. Generally, because high throughputor multi-user sequencing facilities offer sequencing services atsignificantly reduced cost compared with sequencing withinindividual labs, they have become the most common option fornucleotide sequencing. Many academic institutions and privatecompanies provide sequencing services at a reasonable cost, anda brief web search should reveal many options for sequencingservices. Sequencing reagents are produced by several manufac-turers and vary depending on the automated sequencing instru-ment used. Most, if not all, sequencing facilities will recom-mend specific reagent kits and protocols for their users. Themost important consideration for obtaining good sequencingresults is the purity of the sequencing template as a poor qual-ity template DNA is the most common cause for failed sequenc-ing reactions. Therefore, no matter if sequencing templates arepurified plasmids or PCR products, we highly recommend theuse of commercial DNA purification kits because of their ease ofuse and the consistent DNA purity that they provide.Common UA- or TA-based PCR cloning kits:

• Fermentas InsTAclone™ PCR Cloning Kit(http://www.fermentas.com/)

• Invitrogen TOPO TA Cloning® Kit(http://www.invitrogen.com/)

• Promega pGEM-T and pGEM-T Easy Vector Systems(http://www.promega.com/)

• Stratagene StrataClone™ PCR Cloning Kit(http://www.stratagene.com/)

Common blunt-end PCR cloning kits:• Clontech In-Fusion™ PCR Cloning Kits

(http://www.clontech.com/). Note: although this kit does notrequire deoxyadenosine (“A”) overhangs on PCR fragmentsto be cloned; blunt-end polishing is also not required.


88

• Fermentas CloneJET™ PCR Cloning Kit(http://www.fermentas.com/)

• Invitrogen Zero Blunt® TOPO® PCR Cloning Kit(http://www.invitrogen.com/)

• Stratagene StrataClone™ Blunt PCR Cloning Kit(http://www.stratagene.com/)

Common gel extraction kits:• Fermentas DNA gel extraction kit

(http://www.fermentas.com/)• Invitrogen PureLink™ Gel Extraction Kit

(http://www.invitrogen.com/)• Promega Wizard®” DNA Clean up system

(http://www.promega.com/)• Qiagen QIAquick Gel Extraction kit

(http://www.qiagen.com/)• Stratagene StrataPrep® DNA Gel Extraction Kit

(http://www.stratagene.com/)Common plasmid miniprep kits:

• Fermentas GeneJET™ Plasmid Miniprep Kit(http://www.fermentas.com/)

• Invitrogen ChargeSwitch® NoSpin Plasmid Micro Kit(http://www.invitrogen.com/)

• Promega Wizard® Plus Minipreps DNA purification system(http://www.promega.com/)

• Qiagen QIAprep Spin Miniprep Kit(http://www1.qiagen.com/)

• Stratagene StrataPrep® Plasmid Miniprep Kit(http://www.stratagene.com/)

Common PCR cleanup kits:• Applied Biosystems DNAclear™ kit

(http://www.appliedbiosystems.com/)• Fermentas DNA gel extraction kit

(http://www.fermentas.com/)• Invitrogen ChargeSwitch® PCR Clean-Up Kit

(http://www.invitrogen.com/)• Promega Wizard® DNA Clean up system

(http://www.promega.com/)• Qiagen QIAquick PCR purification kit

(http://www1.qiagen.com/)• Stratagene StrataPrep® PCR Purification Kit

(http://www.stratagene.com/)Bioinformatic analysis—Once sequences have been

obtained from a marker gene clone library, the stepsinvolved in sequence analysis include 1) sequence editing,2) sequence alignment, 3) phylogenetic inference, 4) draw-ing phylograms, and 5) calculating diversity indices (Fig. 1).Although the analysis of clone library sequences can seemdaunting to the uninitiated, references such as Hall’s bookPhylogenetic Trees Made Easy (2008) offer excellent adviceand background information that will walk beginnersthrough the essential elements of sequence analysis; morein-depth discussions of phylogenetic inference can be foundin advanced texts (Felsenstein 2004; Graur and Li 2000;Hillis et al. 1996).

By its very nature, bioinformatic analysis is computationallyintensive and is conducted using a variety of software. In recentyears, computer software and hardware has changed dramati-cally, and most of these changes have resulted in easy to use andwidely available bioinformatic software. For example, Macintoshcomputers now use an Intel chip that allows them to use theWindows operating system, and there are Windows emulatorsavailable for both Linux and Unix operating systems. Therefore,to ensure that this discussion is useful to the broadest possibleaudience, we have focused on the use of Windows-based soft-ware that is freely available on the World Wide Web (most of thesoftware listed in the following paragraphs is also available inversions compatible with Unix or Macintosh operating systems).For the sake of brevity, we will not discuss the parameters thatmust be considered when analyzing genetic libraries. Instead, wewill simply point readers to the excellent texts mentioned in thepreceding paragraph, and provide a brief list of some of the avail-able free software, noting their major functions and the Web sitefrom which they can be downloaded:

• BioEdit (Hall 1999). This software can be used for sequenceediting and much more. It is available free of charge athttp://www.mbio.ncsu.edu/BioEdit/BioEdit.html. Thissoftware package can be used to view the chromatogramsproduced by several different types of automatedsequencers, and it can also be used to analyze the physi-cal properties of nucleic acid or amino acid sequences.Further, it can be used to translate DNA sequences, searchsequences for defined motifs, conduct BLAST searcheslocally or to the GenBank database, align sequences usingClustalW, and it produces publication quality prints ofsequence alignments. This is an extremely useful programthat has far too many functions to list here.

• ClustalX (Thompson et al. 1997). This software is the mostwidely used sequence alignment software available. It canbe used generate pairwise and multiple alignments ofnucleotide and amino acid sequences, and a variety ofparameters such as gap penalties and the substitutionmatrix can be set by the user. It is downloadable for freefrom http://bips.u-strasbg.fr/fr/Documentation/ClustalX/.

• Mega 4 (Tamura et al. 2007). This software can be used toalign nucleic acid or amino acid sequences, estimate evo-lutionary distances using a variety of models, build phylo-genetic tress via neighbor joining or maximum parsimonymethods, and test phylogenetic tree reliability via interiorbranch tests or bootstrap analysis. In addition, Mega 4 hasextensive tree viewing, manipulation, and editing toolsthat can be used to create publication quality trees in avariety of file formats. This software is free and can bedownloaded from http://www.megasoftware.net/.

• MrBayes (Ronquist and Huelsenbeck 2003). This softwareis used for Bayesian phylogenetic inference. Bayesianinference of phylogeny has become very popular amongmolecular systematists and is based on the posterior prob-ability distribution of trees using a Markov chain Monte


89

Carlo simulation technique that approximates these pos-terior probabilities. Although this software is operatedthrough command lines and is not as easy user friendly asother graphical interface programs, excellent documenta-tion is provided with the software, and Hall (2008) pro-vides a good tutorial to help beginning users get started.MrBayes is available for free download from http://mrbayes.csit.fsu.edu/.

• Phylogeny.fr: robust phylogenetic analysis for the non-specialist (Dereeper et al. 2008). This free web serviceincorporates several alignment and phylogenetic toolsinto a user friendly website that can be used to reconstructand analyze phylogenetic relationships between molecu-lar sequences in a single-step or, for more experiencedusers, an “A la carte” menu can be used to tailor variousaspects of the phylogenetic workflow. This site alsoincludes extensive documentation. The site can beaccessed at http://www.phylogeny.fr/.

• EstimateS: Statistical estimation of species richness andshared species from samples. Version 8.0.0, R. K. Colwell.2006. This software can be used to calculate a variety ofbiodiversity functions, estimators, and indexes based on arange of biological data. For example, EstimateS can beused to compute rarefaction and species accumulationcurves, as well as a variety of different species richnessestimators for data from marker gene libraries. EstimateSis a free software application that can be downloadedfrom http://viceroy.eeb.uconn.edu/estimates. Excellentsupporting documentation for the software is also avail-able at the same Web site.

• Rarefaction Calculator (http://www2.biology.ualberta. ca/jbrzusto/rarefact.php), Analytic Rarefaction (http://www.uga.edu/strata/software/index.html), and DOTUR(http://schloss.micro.umass.edu/software/) (Schloss andHandelsman 2005) are other free software applicationsthat can be used to estimate rarefaction curves for datafrom marker gene libraries. We have included thembecause of their simplicity and ease of use.

AssessmentAs mentioned above, the genetic diversity of cyanomy-

oviruses in various aquatic environments has been investi-gated extensively. However, a large proportion of environ-mental g20 sequences do not appear to be from myovirusesthat infect Synechococcus and Prochlorococcus since they clusteroutside clades containing sequences from laboratory isolates(Marston and Sallee 2003; Short and Suttle 2005; Wang andChen 2008; Wilhelm et al. 2006; Zhong et al. 2002). For exam-ple, among 207 clones retrieved from diverse marine environ-ments, about 80% did not cluster with known cyanomy-oviruses (Zhong et al. 2002). More than 60% of DGGE bandsequences recovered from both marine and freshwater envi-ronments were outside the cyanomyovirus cluster (Short andSuttle 2005). This problem occurs because the PCR primers

were designed based on the limited cyanomyovirus g20 genesequences. More specific PCR amplification can be achievedwhen more gene sequences become available. By redesigningthe g20 gene primers based on the newly available cyanomy-ovirus g20 sequences, a high proportion of environmentalclones fell into the Cyanomyovirus cluster (Fig. 2, left panel).The modified g20 primer set SMP-1F and SMP-2R (Wang andChen unpubl. data) were designed based on nearly 30cyanomyovirus g20 sequences, and could be useful for specif-ically monitoring the population dynamics of cyanomy-oviruses in the natural environment. With modified g20primers, 75% of DGGE band sequences fell within theCyanomyovirus cluster. The seasonal shift on cyanomyoviruspopulations in the Chesapeake Bay can be seen from theDGGE analysis of g20 amplicon (Fig. 2, right panel).

The g20 gene study is just one of many examples showingthe difficulty or limitation of using molecular tools to explorethe diversity of microbes in nature. Many steps related to PCRamplification (i.e., Taq enzymes, number of PCR cycles, etc.)could also cause the biased results. Therefore, it is importantto optimize the PCR conditions before a large quantity of sam-ples are analyzed. Finally, this study is also limited by theavailability of sequences in publicly available databases. Whilemany g20 amplicons fall outside the clusters associated withknown cyanophage isolates, the highest identity remains thatof cyanophage g20 genes. As such, the investigator must ulti-mately understand that the interpretation of molecular datafrom culture independent studies is at the mercy of the avail-able data in molecular repositories. While this will no doubtimprove over time, in the case of some understudied virusgroups, data reanalysis in subsequent years may result in dif-ferent interpretations.

DiscussionThe application of molecular tools to questions concerning

the ecology of viruses is a rapidly changing area. Already inthe last several years, advances in DNA sequencing technolo-gies have exponentially expanded the available database ofgenetic information from viruses (Zeidner et al. 2003). Giventhe rate of advancement in both the theory and technologyassociated with this area of research, it is perhaps most impor-tant to caution researchers to be sure that they have fullyexamined the most recent literature prior to establishing anew program of research. Ultimately though, different labora-tories use different tools, and researchers are encouraged toadapt their own available tool sets and materials whenaddressing questions of marine virus diversity.

With respect to choices regarding the use of establishedprimer sets, it is important that investigators carefully followrecommended protocols when adapting techniques developedin another lab (and as such that these protocols are well doc-umented for publication). Sometimes even slight changes ininstrumentation (e.g., the type of thermal cycler) or basicsources of reagents (e.g., similar polymerases from different


90

vendors) can markedly influence the success of a molecularbiological exercise. As with so many other biological systems,much of the molecular biology of aquatic viruses comes downto proper validation, optimization, and the use of both posi-tive and negative controls to get the best possible data.

Comments and recommendationThe molecular examination of viruses in aquatic communi-

ties is just one of the many areas of virus ecology whereresearchers are making tremendous and rapid strides forward.As PCR-based molecular techniques have improved our quali-tative understanding of microbial diversity, quantitativemolecular approaches for studying virus communities,although in their infancy, will allow us to better understandprocesses associated with either the entire virus community orspecific virus populations. While many challenges remain inthe adaptation of lab techniques (e.g., quantitative PCR) tofield studies, these challenges and others associated with PCR-based approaches will undoubtedly be solved in the nearfuture. As such, perhaps the most important recommendationto both the neophyte and the experienced researcher is to com-plete a thorough examination of the peer-reviewed literatureprior to taking on any project. While we provide what we feelare sound recommendations in the current review, the trajec-tory of this field of research is steep and demands that studentsof this field need to be up-to-date on the most recent advances.

ReferencesAltshuler, M. L. 2006. PCR troubleshooting: The essential guide.

Caister Academic Press.Atlas, R. M. 1993. Detecting gene sequences using the poly-

merase chain reaction, p. 267-270. In P. F. Kemp, B. F. Sherr,E. B. Sherr, and J. J. Cole [eds.], Handbook of methods inaquatic microbial ecology. Lewis Publishers.

Ausubel, F. M., and others [eds.]. 2002. Short protocols inmolecular biology: a compendium of methods from currentprotocols in molecular biology, 5th ed. Wiley.

Baker, A. C., V. J. Goddard, J. Davy, D. C. Schroeder, D. G.Adams, and W. H. Wilson. 2006. Identification of a diagnos-tic marker to detect freshwater cyanophages of filamentouscyanobacteria. Appl. Environ. Microbiol. 72:5713-5719.

Breitbart, M., J. H. Miyake, and F. Rohwer. 2004. Global distri-bution of nearly identical phage-encoded DNA sequences.FEMS Microbiol. Lett. 236:249-256.

Chen, F., and C. A. Suttle. 1995. Amplification of DNA-poly-merase gene fragments from viruses infecting microalgae.Appl. Environ. Microbiol. 61:1274-1278.

———, ———, and S. M. Short. 1996. Genetic diversity inmarine algal virus communities as revealed by sequenceanalysis of DNA polymerase genes. Appl. Environ. Micro-biol. 62:2869-2874.

Culley, A. I., and G. F. Steward. 2007. New genera of RNAviruses in subtropical seawater, inferred from polymerasegene sequences. Appl. Environ. Microbiol. 73:5937-5944.

Dereeper, A., and others. 2008. Phylogeny.fr: robust phyloge-netic analysis for the non-specialist. Nucleic Acids Res.36:W465-W469.

Felsenstein, J. 2004. Inferring phylogenies. Sinauer Associates.Filee, J., F. Tetart, C. A. Suttle, and H. M. Krisch. 2005. Marine

T4-type bacteriophages, a ubiquitous component of thedark matter of the biosphere. Proc. Natl. Acad. Sci. USA102:12471-12476.

Fuller, N. J., W. H. Wilson, I. R. Joint, and N. H. Mann. 1998.Occurrence of a sequence in marine cyanophages similar tothat of T4 g20 and its application to PCR-based detectionand quantification techniques. Appl. Environ. Microbiol.64:2051-2060.

Graur, D., and W. Li. 2000. Fundamentals of molecular evolu-tion, 2nd ed. Sinauer Associates.

Hall, B. G. 2008. Phylogenetic trees made easy: a how-to man-ual, 3rd ed. Sinauer Associates.

Hall, T. A. 1999. BioEdit: a user-friendly biological sequencealignment editor and analysis program for Windows95/98/NT. Nucl. Acids. Symp. Ser. 41:95-98.

Hillis, D. M., C. Moritz, and B. K. Mable [eds.]. 1996. Molecu-lar systematics, 2nd ed. Sinauer Associates.

Innis, M. A., D. H. Gefland, J. J. Sninsky, and T. J. White. 1990.PCR protocols: A guide to methods and applications. Aca-demic Press.

Jenkins, C. A., and P. K. Hayes. 2006. Diversity of cyanophagesinfecting the heterocystous filamentous cyanobacteriumNodularia isolated from the brackish Baltic Sea. J. Mar. Biol.Assoc. U.K. 86:529-536.

Larsen, J. B., A. Larsen, G. Bratbak, and R. A. Sandaa. 2008.Phylogenetic analysis of members of the Phycodnaviridaevirus family, using amplified fragments of the major capsidprotein gene. Appl. Environ. Microbiol. 74:3048-3057.

Marston, M. F., and J. L. Sallee. 2003. Genetic diversity andtemporal variation in the cyanophage community infect-ing marine Synechococcus species in Rhode Island’s coastalwaters. Appl. Environ. Microbiol. 69:4639-4647.

McPherson, M. J., and S. G. Møller. 2006. PCR, 2nd ed. Taylor& Francis.

Millard, A., M. R. J. Cloki.e., D. A. Shub, and N. H. Mann.2004. Genetic organization of the psbAD region in phagesinfecting marine Synechococcus strains. Proc. Natl. Acad.Sci. USA 101:11007-11012.

Nagasaki, K., K. Tarutani, M. Hamaguchi, and M. Yamaguchi.2001. Preliminary analysis of Heterosigma akashiwo virusDNA. Microbes Environ. 16:147-154.

Ronquist, F., and J. P. Huelsenbeck. 2003. MrBayes 3: Bayesianphylogenetic inference under mixed models. Bioinformat-ics 19:1572-1574.

Sambrook, J., E. F. Fritsch, and T. Maniatis. 1989. Molecularcloning: a laboratory manual, 2nd ed. Cold Spring HarborLaboratory Press.

Schloss, P. D., and J. Handelsman. 2005. Introducing DOTUR,a computer program for defining operational taxonomic


91

units and estimating species richness. Appl. Environ.Microbiol. 71:1501-1506.

Schroeder, D. C., and C. A. Suttle. 2002. Sequence analysis ofmarine virus communities reveals that groups of relatedalgal viruses are widely distributed in nature. Appl. Envi-ron. Microbiol. 68:1290-1296.

———, J. Oke, M. Hall, G. Malin, and W. H. Wilson. 2003.Virus succession observed during an Emiliania huxleyibloom. Appl. Environ. Microbiol. 69:2484-2490.

Short, C. M., and C. A. Suttle. 2005. Nearly identical bacterio-phage structural gene sequences are widely distributed inboth marine and freshwater environments. Appl. Environ.Microbiol. 71:480-486.

Short, S. M., and C. M. Short. 2008. Diversity of algal virusesin various North American freshwater environments.Aquat. Microb. Ecol. 51:13-21.

Takashima, Y., and others. 2007. Development and applica-tion of quantitative detection of cyanophages phylogeneti-cally related to cyanophage Ma-LMM01 infecting Microcys-tis aeruginosa in fresh water. Microbes Environ. 22:207-213.

Tamura, K., J. Dudley, M. Nei, and S. Kumar. 2007. MEGA4:Molecular evolutionary genetics analysis (MEGA) softwareversion 4.0. Mol. Biol. Evol. 24:1596-1599.

Thompson, J. D., T. J. Gibson, F. Plewniak, F. Jeanmougin, andD. G. Higgins. 1997. The ClustalX windows interface: flex-ible strategies for multiple sequence alignment aided by

quality analysis tools. Nucleic Acids Res. 25:4876-4882.Wang, K., and F. Chen. 2008. Prevalence of highly host-

specific cyanophages in the estuarine environment. Envi-ron. Microbiol. 10:300-312.

Wilhelm, S. W., M. J. Carberry, M. L. Eldridge, L. Poorvin, M.A. Saxton, and M. A. Doblin. 2006. Marine and freshwatercyanophages in a Laurentian Great Lake: Evidence frominfectivity assays and molecular analyses of g20 genes.Appl. Environ. Microbiol. 72:4957-4963.

———, and A. R. Matteson. 2008. Freshwater and marine viri-oplankton: a brief overview of commonalities and differ-ences. Freshw. Biol. 53:1076-1089.

Wommack, K. E., T. Sime-Ngando, D. M. Winget, S. Jamindar,and R. R. Helton. 2010. Filtration-based methods for thecollection of viral concentrates from large water samples, p.110-117. In S. W. Wilhelm, M. G. Weinbauer, and C. A. Sut-tle [eds.], Manual of Aquatic Viral Ecology. ASLO.

Yuryev, A. [ed.]. 2007. PCR primer design. Humana Press.Zeidner, G., and others. 2003. Molecular diversity among

marine picophytoplankton as revealed by psbA analyses.Environ. Microbiol. 5:212-216.

Zhong, Y., F. Chen, S. W. Wilhelm, L. Poorvin, and R. E. Hod-son. 2002. Phylogenetic diversity of marine cyanophageisolates and natural virus communities as revealed bysequences of viral capsid assembly protein gene g20. Appl.Environ. Microbiol. 68:1576-1584.

92

Introduction

Table 1 shows a list of viruses infecting eukaryotic algae andnonphotosynthetic protists that have been characterizedusing typical clonal virus isolates. So far, viruses infectious toChlorophyceae, Chrysophyceae, Pelagophyceae, Prasinophyceae,

Haptophyceae, Dinophyceae, Raphidophyceae, Bacillario-phyceae, Bicosoecophyceae (Mastigophorea), Acanthamoe-bidae, and Thraustochytriaceae have been isolated.

Research on algal and protistan viruses has now passed afirst stage of development and each year a number of pub-lished reports disclosing their roles in natural environments,their host relationships, their molecular and biological char-acteristics, and other information, including genome infor-matics and nanostructures, are published. Many of these stud-ies, which are often advanced, have been possible only thanksto successful isolation of novel viruses (Table 1). Virus isola-tion itself is not at all a very advanced activity, but its successis often critically dependent on the researchers’ skill and expe-rience. A number of different approaches and strategies havebeen used in various laboratories. All of these techniques areuseful (i.e., they have led to successful isolation of a virus), butbecause no comparative studies are available it is impossible tostate that one method is better or more efficient than theother. Nevertheless, summarizing the methods described, aswell as additional information, notes, and observations thatare scattered in the literature but relevant for a rational andsuccessful virus isolation protocol, may be instructive for allworkers in the field. A typical isolation procedure is shown inFig. 1. Appropriate modifications must be made to meet

Isolation of viruses infecting photosynthetic andnonphotosynthetic protistsKeizo Nagasaki1* and Gunnar Bratbak2†

1National Research Institute of Fisheries and Environment of Inland Sea, Fisheries Research Agency, 2-17-5 Maruishi,Hatsukaichi, Hiroshima 739-0452, Japan2University of Bergen, Department of Biology, Box 7800, N-5020 Bergen, Norway

AbstractViruses are the most abundant biological entities in aquatic environments and our understanding of their eco-

logical significance has increased tremendously since the first discovery of their high abundance in naturalwaters. About 40 viruses infecting eukaryotic algae and 4 viruses infecting nonphotosynthetic protists have so farbeen isolated and characterized to different extents. The isolated viruses infecting phytoplankton(Chlorophyceae, Prasinophyceae, Haptophyceae, Dinophyceae, Pelagophyceae, Raphidophyceae, andBacillariophyceae) and heterotrophic protists (Bicosoecophyceae, Acanthamoebidae, and Thraustochytriaceae)are all lytic. Some of the brown algal phaeoviruses, which infect host spores or gametes, have also been found ina latent form (lysogeny) in vegetative cells. Viruses infecting eukaryotic photosynthetic and nonphotosyntheticprotists are highly diverse both in size (ca. 20–220 nm in diameter), genome type (double-strand deoxyribonu-cleic acid [dsDNA], single-strand [ss]DNA, ds–ribonucleic acid [dsRNA], ssRNA), and genome size [4.4–560 kb]).Availability of host–virus laboratory cultures is a necessary prerequisite for characterization of the viruses and forinvestigation of host–virus interactions. In this report we summarize and comment on the techniques used forpreparation of host cultures and for screening, cloning, culturing, and maintaining viruses in the laboratory.

*Corresponding author: E-mail: [email protected]†[email protected]



The authors are grateful to the following people for their fruitfuladvice: Hiroyuki Mizumoto, Yoshitake Takao, and Yuji Tomaru fromJapan; Mikal Heldal from Norway; and Olga Stepanova from Ukraine.These extremely knowledgeable and experienced people provided uswith much insight and guidance for which we are grateful.

ISBN 978-0-9845591-0-7, DOI 10.4319/mave.2010.978-0-9845591-0-7.92Suggested citation format: Nagasaki, K., and G. Bratbak. 2010. Isolation of viruses infecting pho-tosynthetic and nonphotosynthetic protists, p. 92–101. In S. W. Wilhelm, M. G. Weinbauer, andC. A. Suttle [eds.], Manual of Aquatic Viral Ecology. ASLO.


MANUALof


Nagasaki and Bratbak Isolation of aquatic viruses

93

Table 1. Viruses infecting eukaryotic algae and the methods used for their isolation.

Virus Host (class) Size, nm Genome SourceViruses infecting unicellular algae

AaV (BtV) Aureococcus anophagefferens (Pelagophyceae) 140 dsDNA Water sampleWater sample

CbV Chrysochromulina brevifilum (Haptophyceae) 145–170 dsDNA Water sampleCdebDNAV Chaetoceros debilis (Bacillariophyceae) 30 ssDNA, fragmented? Water sampleCeV Chrysochromlina ericina (Haptophyceae) 160 dsDNA, 510 kbp

Chlorella virus (e.g., ATCV-1, ATCV-2) Chlorella SAG 3.83 (Chlorophyceae) 140–190 dsDNA, 288 kbp Water sample(symbiont of Acanthocystis turfacea)

Chlorella virus (e.g., PBCV-1, NY-2A, AR158) Chlorella NC64A (Chlorophyceae) 150–190 dsDNA, 331–369 kbp Water sample(symbiont of Paramecium bursaria)

Chlorella virus (e.g., MT325, FR483) Chlorella Pbi (Chlorophyceae) 140–150 dsDNA, 314–321 kbp(symbiont of Paramecium bursaria)

CsNIV Chaetoceros salsugineum (Bacillariophyceae) 38 (ss+ds)DNA, 6.0 kb Sediment sample

CspNIV Chaetoceros cf. gracilis (Bacillariophyceae) 25 — Water sampleCsfrRNAV Chaetoceros socialis f. radians (Bacillariophyceae) 22 ssRNA, 9.5 kb Water sampleCtenRNAV Chaetoceros tenuissimus (Bacillariophyceae) 31 ssRNA, 8.9 and 4.3 kb Water sampleCwNIV Chaetoceros cf. wighamii (Bacillariophyceae) 22–28 — Water sampleEhV Emiliania huxleyi (Haptophyceae) 170–200 dsDNA, 410–415 kbp Water sample,†

mesocosm with host bloomHaNIV Heterosigma akashiwo (Raphidophyceae) 30 — Water sample, integrated

from 0–75-m depthHaV Heterosigma akashiwo (Raphidophyceae) 202 dsDNA, 294 kbp Water sample, host bloom†

HaRNAV Heterosigma akashiwo (Raphidophyceae) 25 ssRNA, 9.1 kb Water sample, Chl maximum

HcRNAV Heterocapsa circularisquama (Dinophyceae) 30 ssRNA, 4.4 kb Water sample

HcV Heterocapsa circularisquama (Dinophyceae) 197 dsDNA, 356 kbp Water sampleMpRV Micromonas pusilla (Prasinophyceae) 50–60 dsRNA, 24.6 kbp in total Lysed host culture

(segmented)MpV Micromonas pusilla (Prasinophyceae) 115 dsDNA, 200 kbp Water sample

MpVN1 Micromonas pusilla (Prasinophyceae) 110–130 (ds?)DNA Water sampleMpVN2 Micromonas pusilla (Prasinophyceae) 110–130 (ds?)DNA Water sampleOIs1 Heterosigma akashiwo (Raphidophyceae) 30 and 80 dsDNA, 20 and 130 kbp,

respectively for 30 and 80 nm

PgV-102P Phaeocystis globosa (Haptophyceae) 98 dsDNA, 176 kbp Water samplePgV Group I Phaeocystis globosa (Haptophyceae) 150 dsDNA, 466 kbp Water samplePgV Group II Phaeocystis globosa (Haptophyceae) 100 dsDNA, 177 kbp Water samplePoV Pyramimonas orientalis (Prasinophyceae) 180–220 dsDNA, 560 kbp

PpV Phaeocystis pouchetii (Haptophyceae) 130–160 dsDNA, 485 kbp Water sample, host bloom†

RsRNAV Rhizosolenia setigera (Bacillariophyceae) 32 ssRNA, 11.2 kb Water sample, nonhost bloomTampV Teleaulax amphioxeia (Cryptophyceae) 203 — Water sample, nonhost bloom

Viruses infecting multicellular algaeEsV Ectocarpus siliculosus(Phaeophyceae) 130–150 dsDNA, 336 kbp Plants showing symptoms of infection

EfasV Ectocarpus fasciculatus (Phaeophyceae) 135–140 dsDNA, 340 kbp Plants showing symptoms of infectionFlexV Feldmannia simplex (Phaeophyceae) 120–150 dsDNA, 170 kbp Plants showing symptoms of infectionFirrV Feldmannia irresguralis (Phaeophyceae) 140–170 dsDNA, 180 kbp Plants showing symptoms of infectionFsV Feldmannia species (Phaeophyceae) 150 dsDNA, 158 and 178 kbp Plants showing symptoms of infectionHincV Hincksia hinckiae (Phaeophyceae) 140–170 dsDNA, 220 kbp Plants showing symptoms of infectionMclaV Myriotrichia clavaeformis (Phaeophyceae) 170–180 dsDNA, 340 kbp Plants showing symptoms of infectionPlitV Pilayella littoralis (Phaeophyceae) 161 dsDNA, 280 kbp Plants showing symptoms of infection

Viruses infecting marine heterotrophic protistsCroV (BV-PW1) Cafeteria roenbergensis (Bicosoecophyceae/ 230–300 dsDNA, 730 kbp Water sample

Mastigophorea) (reported as Bodo sp.)Mimivirus Acanthamoeba polyphaga (Acanthamoebidae) 750 dsDNA, 1.2 Mb Viral lysateSssRNAV Aurantiochytrium sp. NIBH N1-27 (Thraustochytriaceae) 25 ssRNA, 10.2 kb Water sample

(reported as Schizochytrium sp. NIBH N1-27)SmDNAV Sicyoidochytrium minutum (Thraustochytriaceae) 140 dsDNA, 250 kbp Water sample

*PC, polycarbonate membrane filters; GF, Glassfiber filter; PVD, polyvinylidene difluoride filters (Durapore, Millipore); kbp, kilobase pair; kb, kilobase; NC, nitrocellulose membrane; CA, cellulose acetate filters (Schleicher & Schuell); HT, HT Tuffryn low-protein-binding polysulphonate membrane.†UV treatment applied but not essential.


94

Virus sample filtration, Virus concentration Principal references, unpublished data, or pore size and type* and inoculation Lysate filtration personal communication

0.2-µm PC Ultracentrifugation 105,000g, 3 h 0.2-µm filter Garry et al. 1998; Gastrich et al. 20040.2-µm polypropylene Ultrafiltration 30,000 MW cutoff 0.2-µm; GV low-protein Gobler et al. 2004, 2007; Rowe et al. 2008

filter capsules (MSI) binding filter from Millipore1.2-µm GF and 0.22 or 0.45 μm PVD Ultrafiltration 30,000 MW cutoff 0.8-µm GF and 0.22-µm PVD Suttle and Chan 19950.2-µm PC Non, 20% v/v 0.1-µm PC Tomaru et al. 20080.45-µm Supor filters (Gelman) Plankton concentrated by 0.2-µm Meditron syringe filters Sandaa et al. 2001; Thyrhaug et al. 2003; Monier et al. 2008

continuous flow centrifugation (Schleicher & Schuell)0.45-µm NC Plaque assay after 2-week incubation Bubeck and Pfitzner 2005; Fitzgerald et al. 2007c

in liquid culture0.4-µm PC 0.4-µm PC Van Etten et al. 1983, 1991, 2002; Van Etten and Meints 1999;

Yamada et al. 1999, 2006; Fitzgerald et al. 2007bReisser et al. 1988a,b; Van Etten et al. 1991;

Yamada et al. 2006; Fitzgerald et al. 2007a0.7-µm GF and 0.2-µm Non, 20% v/v 0.1-µm PC Nagasaki et al. 2005c

Dismic-25cs filters (Advantec)0.22-µm Tangential-flow membrane Ultrafiltration 30,000 MW cutoff Bettarel et al. 20050.2-µm (DISMIC-25, Advantec) Non, 33% v/v 0.1-µm PC Tomaru et al. 20090.2-µm PC Non, 20% v/v 0.1-µm PC Shirai et al. 2008

Ultrafiltration 30 000 MW cutoff 0.22-µm Eissler et al. 2009Non, 10% v/v 0.2-µm Syringe filter Castberg et al. 2002; Wilson et al. 2002, 2005;

(FP30/0.2CA-S [Schleicher & Schuell]) Schroeder et al. 2003; Thyrhaug et al. 2003; Allen et al. 20061.2-µm GF and 0.4-µm Ultrafiltration 30,000 MW cutoff 0.22-µm filter Lawrence et al. 2001

PC cartridge filter0.2-µm PC Non, 4% v/v 0.2-µm (DISMIC-25, Advantec) Nagasaki and Yamaguchi 1997; Nagasaki et al. 1999, 2005b;

Tarutani et al. 2000; Tomaru et al. 2004b1.2-µm GF and 0.2-µm Ultrafiltration 30,000 MW cutoff 0.22-µm GV Durapore filter (Millipore) Tai et al. 2003; Lang et al. 2004

PC cartridge filter0.8-µm PC Non, 40% v/v 0.1-µm PC Tomaru et al. 2004a; Nagasaki et al. 2004a, 2005a, 2006;

Mizumoto et al. 2007, 20080.2-µm PC Non, 50% v/v 0.2-µm PC Tarutani et al. 2001; Nagasaki et al. 2003, 2005b, 2006

Non, 10% v/v 0.45- and 0.2-µm CA, and 0.1-µm filter Brussaard et al. 2004b; Attoui et al. 2006

0.2-µm PC or 0.45-µm PVD Cottrell and Suttle 1991; 1995; Mayer and Taylor 1979; Waters and Chan 1982

0.22-µm HT Non, 10–0.01% Zingone et al. 20060.22-µm HT Non, 10–0.01% Zingone et al. 2006

Lawrence et al. 2006; Lawrence unpubl. data

0.2-µm Non, 2% v/v 0.2-µm Wilson et al. 20060.7-µm GF Non, 10–20% v/v 0.2-µm CA Brussaard et al. 2004a, 2007; Baudoux and Brussaard 20050.7-µm GF Non, 10–20% v/v 0.2-µm CA Brussaard et al. 2004a, 2007; Baudoux and Brussaard 20050.45-µm Supor filters (Gelman) Plankton concentrated by 0.2-µm Meditron syringe filters Sandaa et al. 2001; Thyrhaug 2003; Monier et al. 2008

continuous flow centrifugation (Schleicher & Schuell)Plankton concentrated by Jacobsen et al. 1996; Bratbak et al. 1998; Yan et al. 2005;

continuous flow centrifugation, 1% v/v Monier et al. 20080.2-µm PC Non, 25% v/v 0.1-µm PC Nagasaki et al. 2004b; Shirai et al. 20060.2-µm PC Non, 50% v/v 0.2-µm PC Nagasaki et al. 2009

Müller 1991; Lanka et al. 1993; Müller et al. 1996, 1998; Van Etten et al. 2002

Müller et al. 1996, 1998Friess-Klebl et al. 1994; Müller et al. 1998Kapp et al. 1997; Müller et al. 1998Henry and Meints 1992; Müller et al. 1998; Meints et al. 2008Kapp et al. 1997; Müller et al. 1998Kapp et al. 1997; Müller et al. 1998Maier et al. 1998; Müller et al. 1998

1.2-µm GF and 0.2 or 0.45μm PVD Ultrafiltration 30 000 MW cutoff 0.2-µm PC Garza and Suttle 1995; Suttle pers. comm.

No filtration; sequential centrifugation Sucrose cushion centrifugation La Scora et al. 2003; Raoult et al. 20040.2-µm PC Non, 40% v/v 0.2-µm PC Takao et al. 2005, 2006

0.2-µm PC Non, 40% v/v 0.2-µm PC Takao et al. 2007


95

requirements and characteristics of the host organisms. Herewe discuss some of the practical issues involved in each step.

Preparation of host culturesPreparation of host cultures is essential for isolation of

viruses. Use of unialgal and axenic host cultures is preferablebecause bacterial contaminants can cause sudden and unex-pected cell lysis in the cultures. Contaminating bacteria maypossibly also introduce their associated phages into down-stream applications and analysis, creating havoc with theresults. To establish axenic microalgal cultures, micropipetting

or other washing methods are commonly employed, often incombination with use of various antibiotics (Paasche 1971;Stein 1973; Lee 1993; Imai and Yamaguchi 1994; Connell andCattolico 1996).

For nonphotosynthetic protists only a few examples of suc-cessful virus isolation have been reported in the literature. Alarge double-stranded deoxyribonucleic acid (dsDNA) virus(CroV) infecting Cafeteria roenbergensis was isolated by Garzaand Suttle (1995) (reported as BV-PW1 infecting Bodo sp., C.Suttle pers. comm.). Cultures of C. roenbergensis and other bac-terivors will inevitably be difficult or impossible to makeaxenic. Takao et al. (2005, 2007) succeeded in isolating a single-stranded ribonucleic acid (ssRNA) virus infectingSchizochytrium sp. and a dsDNA virus infecting Sicyoidchytriumminutum. To make the host cultures axenic these investigatorsused liquid and solid media containing chloramphenicol(0.2% w/v), and because these protists are able to formcolonies on agar plates, they can easily be isolated and puri-fied by picking single colonies.

Generally, viruses of eukaryotic algae and nonphotosyn-thetic protists are strain specific as well as species specific; i.e.,virus sensitivity spectra can differ among host clones (e.g.,Sahlsten 1998; Tarutani et al. 2000; Tomaru et al. 2004a,b; Zin-gone et al 2006). Thus, the use of several different clones ofthe same host species for isolating viruses is often advanta-geous. In most cases, exponentially growing cultures tend tobe more sensitive to viral infection than stationary phase cul-tures (e.g., Nagasaki et al. 2003). Considering that viruses usethe biosynthetic apparatus of the host cell for DNA and pro-tein synthesis and that the host cells in exponential growthhave a higher biosynthetic activity, use of vigorously growinghost cultures is also important.

Screening for virusesInformation on the screening methods used for isolation of

various algal and protist viruses (including considerationssuch as origin and sample preparation) is included in Table 1.

Microalgal viruses are found not only in seawater but alsoin marine sediments (Lawrence et al. 2002; Nagasaki et al.2004a), and samples from both have been used as inoculumsfor virus isolation (Table 1). The most straightforward methodfor isolating viruses infecting algae or protists from naturalseawater is to add (10% to 50% by volume) a filtered (0.2–0.8-µm pore size) or unfiltered water sample potentially contain-ing viruses to a prospective host culture (see Table 1 for refer-ences). Two more elaborate methods have also been used: thevirus concentration method and the virus induction by ultra-violet (UV)-treatment method. In the former method, theviruses in a large water sample (typically 10–100 L) are sepa-rated from larger particles by filtration (0.2- or 0.45-μm poresize) and concentrated by ultrafiltration (typically 30,000 MWcutoff); the concentrated virus-size fraction is then used asinoculum (Suttle et al. 1991). Many viruses have been success-fully isolated by this technique, including viruses infecting

Fig. 1. Summarized scheme of virus isolation procedure. A typical pro-tocol for isolation of virus includes the following steps: (1) Preparation ofinoculum: a natural water sample, or the supernatant left after low-speedcentrifugation of a sediment suspension, is filtered through a 0.2-µm(0.22-, 0.45-, or 0.8-µm) nuclepore filter to obtain a virus sized fraction.(2) Propagation: the filtrate with the virus sized fraction is inoculated intovigorously growing monoclonal algal host cultures and incubated underappropriate conditions. (3) Purification: when algal lysis is observed byoptical microscopy, the lytic agent (virus) is purified using an extinctiondilution procedure or plaque assay method (if possible). (4) Storage: theestablished monoclonal virus culture is stored under appropriate condi-tions. See main text for details.


96

Micromonas pusilla (MpV) (Cottrell and Suttle 1991),Chrysochromulina brevifilum (CbV) (Suttle and Chan 1995),Heterosigma akashiwo (HaV, HaNIV, HaRNAV) (Nagasaki andYamaguchi 1997; Lawrence et al. 2001; Tai et al. 2003), and C.roenbergensis (CroV, reported as BV-PW1 infecting Bodo sp.)(Garza and Suttle 1995; C. Suttle pers. comm.). A detailed pro-tocol of the ultrafiltration technique is given by Suttle (1993).

The intention of the UV-treatment technique is to causeinduction and production of latent viruses in natural phyto-plankton populations. Briefly, natural seawater containing thetarget phytoplankton species is collected and concentrated bycentrifugation (Jacobsen et al. 1996). The resulting cell con-centrate is then poured into a petri dish and exposed to UVirradiation for an appropriate time. Next, the treated cell sus-pension is incubated overnight in the dark, centrifuged toremove cell debris, and then inoculated into a growing hostculture. By using this method viruses infectious to Emilianiahuxleyi (EhV) (Bratbak et al. 1996) and Phaeocystis pouchetii(PpV) (Jacobsen et al. 1996) have been isolated. Assuminglatent viruses to be present, use of other inducing agents, e.g.,mitomycin C, may also be possible for boosting virus produc-tion. However, the efficiency of the UV-treatment techniquehas not been verified, and no lysogenic viruses infecting uni-cellular algae or protists have so far been reported.

Preincubation of natural water samples has in some casesalso been observed to increase the abundance of free virusesand facilitate virus isolation. Brussaard et al. (2004a) reportedthat nutrient addition and incubation of natural seawater fora week at in situ temperatures promoted the isolation ofviruses infecting Phaeocystis globosa. Nagasaki et al. (1994)found an increased fraction of virus containing H. akashiwocells after incubation of the collected samples for 26 h. Themechanisms involved may include increased abundance ofhost cells supporting an increased production of associatedviruses and maturation and lysis of already infected cells. Bothof these mechanisms bring about an increase in free viral par-ticles and thus an increase in the probability for successfulvirus isolation after filtration of the samples.

Mussel mantel fluid and material from gills of fish and mus-sels have successfully been used as inoculum for isolating novelviruses infecting Tetraselmis viridis and Phaeodactylum tricornu-tum (O. A. Stepanova pers. comm., Institute of Biology of theSouthern Seas, National Academy of Sciences of Ukraine, Sev-astopol). The concept behind this approach is that particles,including viruses and infected cells, are retained and concen-trated on the gills of fish and filter feeders and that the proba-bility for isolating a virus will be greater using this material asinoculum rather than just seawater. Further detailed informa-tion concerning this new method is of much interest.

Virus in sediments can be extracted by shaking with a VoltexShaker (VR-36, Taitec) (e.g., 400 rpm × 30 min) or vortexing sed-iment samples with a suitable medium such as SWM3 (Chen etal. 1969), phosphate-buffered saline, or sodium pyrophosphate(Lawrence et al. 2002), followed by low-speed centrifugation

(e.g., 600g × 10 min) and filtration (0.2–0.45 µm) to separateviruses from bacteria, sediment particles, and debris (Marangerand Bird 1996; Danovaro and Serresi 2000; Lawrence et al. 2002).

Prior to inoculation of the virus suspension to host cul-tures, bacteria and larger organisms (including zooplankton,phytoplankton, nanoflagellates, and fungi) should beremoved by filtration. Considering the size of microalgalviruses (20–220 nm in diameter), 0.2-, 0.22-, or 0.45-µm pore-size polycarbonate membrane filters (Nuclepore) are suitable.The resulting filtrate (viral fraction) is then added to vigor-ously growing potential host cultures. Growth of hosts ismonitored and compared to that of control cultures withoutvirus inoculation.

When the host algae are able to grow and form a lawn onsolid media, viruses may also be isolated by means of plaqueassay. Cultivation of algal flagellates on solid media may bedifficult, especially in the case of marine forms (Nagasaki andImai 1994), but the approach has nevertheless been appliedwith success on a number of occasions, e.g., Chlorella viruses(van Etten et al 1991), M. pusilla (Cottrell and Suttle 1991), E.huxleyi (Bratbak et al. 1996), and Chaetoceros ceratosporum(Sakata et al. 1991). As mentioned above, fungoid protistssuch as Schizochytrium sp. (Aurantiochytrium sp.) can grow wellon agar plate (Y. Takao pers. comm., Fukui Prefectural Univer-sity, Japan); thus, plaque assay is a plain option. For virusesthat replicate very slowly and form small plaques that may behard to observe, it may help to reduce the initial cell densityin the soft agar overlay or, in case of algal hosts, put the platesin the dark for a couple of days. This may allow the plaques todevelop and not become overgrown by cells that grow fasterthan the virus. Moreover, as long as it is possible, plaquepurification and cloning of viruses is also much faster and eas-ier than end-point dilution.

One possible method conceivable on the basis of a recentreport by Mizumoto et al. (2007) is the use of a gene gun. Inthe case of HcRNAV (Table 1), purified genome RNA extractedfrom virions caused a regular infection when transfected intoits host cell by using a gene gun. Based on these data, isolationof unknown viruses might be possible if environmental viralRNA is extracted from sediments or seawater and physicallyinjected into potential host cells. This method would be akind of “enrichment culturing” of viruses, in which onlyviruses that can replicate inside the cells and produce descen-dant virions that infect the same cells are selectively propa-gated. No successful results have been obtained and reportedwith this method so far, however.

Cloning and maintenance of microalgal virusesWhen decay (i.e., bleaching, decrease in chlorophyll a flu-

orescence, clearing, etc.) of the tested host algal culture isdetected in the screening procedure, the lytic factor should becloned as soon as possible. In many cases clones have beenobtained by using an extinction dilution procedure (e.g., Sut-tle 1993; Tomaru et al. 2004a,b). Briefly, the culture lysate is


97

diluted with an adequate liquid medium in a series of 10-folddilution steps. Aliquots (100 µL) of each dilution are added to8 wells in cell-culture plates with 96 round-bottom wells,mixed with 150 µL of exponentially growing host culture, andincubated under the conditions suitable for the host’s growth.Lysed cultures are removed from the most diluted wells inwhich lysis occurred, and the entire procedure is repeated. Thelysate in the most diluted wells of the second assay is sterilizedby filtration through 0.1-µm (for ssRNA, ssDNA, or dsRNAviruses) or 0.2-µm pore size polycarbonate membrane filters(phycodnaviruses or large dsDNA viruses) and transferred intoan exponentially growing host culture; certification of thelytic activity of the lysate is essential. After cell debris isremoved by low-speed centrifugation, the supernatant is usedas the clonal pathogen suspension.

Microalgal viruses are diverse in terms of stability, and asuitable protocol for maintaining infective viruses must be setup in each case. Chlorella viruses (PBCV-1) are so durable thatsignificant decreases in titer are rarely seen as long as theviruses are kept refrigerated. In contrast, the titers of HaV, HcV,and PpV gradually decrease even when stored at 4˚C in thedark. Isolated viruses may also be maintained in culture byroutine transfer of the viral lysate to fresh host cultures. Lossof infectivity, however, caused for example by defective inter-fering particles (Bratbak et al. 1996), is a possible risk thatshould be taken seriously. Cryopreservation may, at least insome cases, be an alternative, and HaV and PpV have forexample been stored at –196˚C in 10–20% dimethyl sulfoxideand at –70˚C in 10–20 % sucrose, respectively (Nagasaki andYamaguchi 1999; Nagasaki 2001).

CommentsFiltration of the inoculum used for isolating viruses through

0.2–0.45-µm filters and later of the obtained lysates through0.1–0.2-µm filters is a common procedure (see Table 1). Severalworkers have noted that some viruses do not pass or lose infec-tivity when filtered through certain types of filters and throughfilers with small pore size (i.e., 0.2 µm) (Van Etten et al. 1981;1983; Suttle et al. 1991; Bratbak unpubl. data). Exceptionallylarge viruses such as the Mimivirus (~750 nm; La Scora et al.2003; Xiao et al. 2005) will also be lost during filtration.

With use of the extinction dilution method for cloning, itshould be noted that only the most abundant virus showinglytic activity to the host culture will be isolated. In otherwords, the less dominant viruses (if there are any) will be lost.

In most cases, lysis of the host culture will be a key criterionfor detection of virus infection. Recovery and regrowth of thehost in lysed cultures appears to be a common occurrence,however, and if this phenomenon occurs lysis may pass unno-ticed if cultures are infected with a too high virus dose thatresults in weak lysis and rapid regrowth, or if cultures are leftfor too long before being inspected for lysis (Thyrhaug et al.2003). Moreover, Mizumoto et al. (2008) have recentlydemonstrated that microscopic observation alone may be

inadequate for detecting viral lysis, and infected cultures maybe erroneously disposed of if the symptoms of lysis are weak.

Another issue that should be considered when preparinghost cultures for isolation of virus is that viral susceptibilitymay change between life cycle stages with different ploidy lev-els in unicellular eukaryotes. Frada et al. (2008) has recentlydemonstrated that the haploid phase of E. huxleyi is resistantto EhVs that kill the diploid phase and that exposure ofdiploid E. huxleyi to EhVs induces transition to the haploidphase. The ensuing hypothesis, that the ploidy level of thehost cultures may explain earlier unsuccessful attempts to iso-late viruses and that viral induced transition between hostploidy levels may result in an apparent loss of infectivity dur-ing isolation, should be tested.

The lytic activity differs between various host–virus systemson the species level and also between various clonal combina-tions of host and virus within the same species (Nagasaki andYamaguchi 1998; Mizumoto et al. 2008). The isolation proce-dures used so far may have selected for viruses having stronglytic activity because researchers inadvertently may preferhost–virus systems showing massive cell lysis. Viruses that areslow, latent, cause chronic infections, or are produced andreleased without killing or lysing the host may be hard to iso-late but have an important ecological impact. The conceptionof viruses in natural ecosystems may thus be miscalculated ifbased only on the properties of the possibly “extreme”host–virus systems available in culture.

Brown algal phaeoviruses infect only the host spores orgametes and are reproduced in the sporangia. Virus particlesare not present in vegetative cells, but the genome of the FsVvirus infecting Feldmannia sp. has recently been found inte-grated into the host genome in vegetative cells (Meints et al.2008). Protocols for isolation and culturing of viruses infect-ing these macroalgae are described by Lanka et al. (1992) andMüller (1996).

List of materials and reagents:• Bucket or water sampler (e.g., Van Dorn water sampler)• Sediment sampler (e.g., Ekman bottom grab sampler)• Centrifuge and centrifuge tubes (e.g., 15- or 50-mL Falcon

tubes)• Sterilized filter holder and membrane filters (0.2, 0.22, 0.45

or 0.8µm)• Vacuum pump• Sterile medium for algal culture (e.g., SWM-III, f/2)• Sterile test tubes, pipette with tips and vortexer (for serial

dilution)• 24-well and 96-well cell culture plates (e.g., Falcon)• 8-channel pipette and sterile reservoir tray for filling (for

extinction dilution procedure)• Plastic tape (for sealing the culture plates to avoid drying)• Incubator (with light and temperature control)• Inverted microscope• Refrigerator, freezer, deep freezer, or liquid nitrogen container


98

Tips and hints

Detailed outline of the extinction dilution method:All work should be done in a clean bench.Preparation of dilution series:

(1) Prepare 9 tubes (numbered #1–#9) with 4.5 mL medium.(2) Add 500 μL of the virus filtrate (virus size fraction) to

tube #1 and vortex.(3) Change the pipette tip and transfer 500 μL of suspension

#1 to tube #2 and vortex.(4) Repeat the procedure to tube #8.

Preparation of cell culture plates:Use an 8-channel pipetter filled from a reservoir tray (an

ordinary pipette may be used but makes the work moretedious). When starting with the most diluted samples it is notnecessary to change tips or reservoir tray while working withthe same virus.(5) Pour vigorously growing algal host culture into the reser-

voir tray, fill the pipetter and add 150 μL culture to eachwell in lines 1–9 of a 96-well cell culture plate. Empty thetray.

(6) Pour dilution tube #9 (control medium, no virus) intothe reservoir tray, fill the pipetter and add 100 μL of toeach well in line 9. Empty the tray.

(7) Repeat procedure (6) for dilution tube #8–#1 and fill therespective well lines in the culture plate.

(8) Put on the lid and seal tightly with plastic tape to avoiddrying.

(9) Incubate under appropriate conditions.Inspection:

(10) Use an inverted microscope and inspect the cultureplates for signs of cell lysis at regular intervals.

(11) Mark wells where lysis is observed and continue the incu-bation with daily inspections until no more lysis occurs.

(12) Prepare a second extinction dilution with virus from themost-diluted well.

(13) Propagate virus clone from the most diluted well in alarger volume and store appropriately.

ReferencesAllen, M. J., D. C. Schroeder, M. T. Holden, and W. H. Wilson.

2006. Evolutionary history of the coccolithoviridae. Mol.Biol. Evol. 23:86-92.

Attoui, H., F. M. Jaafar, M. Belhouchet, P. Micco, X. Lambal-lerie, and C. P. D. Brussaard. 2006. Micromonas pusillareovirus: A new member of the family Reoviridae assigned toa novel proposed genus (Mimoreovirus). J. Gen. Virol.87:1375-1383.

Baudoux, A. C., and C. P. Brussaard. 2005. Characterization ofdifferent viruses infecting the marine harmful algal bloomspecies Phaeocystis globosa. Virology 341:80-90.

Bettarel, Y., and others. 2005. Isolation and characterisation ofa small nuclear inclusion virus infecting the diatom Chaeto-ceros cf. gracilis. Aquat. Microb. Ecol. 40:103-114.

Bratbak, G., W. Wilson, and M. Heldal. 1996. Viral control ofEmiliania huxleyi blooms? J. Mar. Syst. 9:75-81.

———, A. Jacobsen, M. Heldal, K. Nagasaki, and F. Thingstad.1998. Virus production in Phaeocystis pouchetii and its rela-tion to host cell growth and nutrition. Aquat. Microb. Ecol.16:1-9.

Brussaard, C. P. D., S. M. Short, C. F. Frederickson, and C. A. Sut-tle. 2004a. Isolation and phylogenetic analysis of novel virusesinfecting the phytoplankton Phaeocystis globosa (Prymnesio-phyceae). Appl. Environ. Microbiol. 70:3700-3705.

———, A. A. M. Noordeloos, R. A. Sandaa, M. Heldal, and G.Bratbak. 2004b. Discovery of a dsRNA virus infecting themarine photosynthetic protist Micromonas pusilla. Virology319:280-291.

———, G. Bratbak, A. C. Baudoux, and P. Ruardij. 2007. Phaeo-cystis and its interaction with viruses. Biogeochemistry83:201-215.

Bubeck, J. A., and A. J. P. Pfitzner. 2005. Isolation and charac-terization of a new type of chlorovirus that infects anendosymbiotic Chlorella strain of the heliozoon Acanthocys-tis turfacea. J. Gen. Virol. 86:2871-2877.

Castberg, T., R. Thyrhaug, A. Larsen, R.-A. Sandaa, M. Heldal,J. L. Van Etten, and G. Bratbak. 2002. Isolation and charac-terization of a virus that infects Emiliania huxleyi (Hapto-phyceae). J. Phycol. 38:767-774.

Chen, L. C. M., T. Edelstein, and J. McLachlan. 1969. Bon-nemaisonia hamifera Hariot in nature and in culture. J. Phy-col. 5:211-220.

Connell, L., and R.A. Cattolico. 1996. Fragile algae: Axenic cul-ture of field-collected samples of Heterosigma carterae. Mar.Biol. 125:421-426.

Cottrell, M. T., and C. A. Suttle. 1991. Wide-spread occurrenceand clonal variation in viruses which cause lysis of a cos-mopolitan, eukaryotic marine phytoplankter Micromonaspusilla. Mar. Ecol. Prog. Ser. 78:1-9.

———, and ———. 1995. Genetic diversities of algal viruseswhich lyse the photosynthetic picoflagellate Micromonaspusilla. Appl. Environ. Microbiol. 61:3088-3091.

Danovaro, R., and M. Serresi. 2000. Viral density and virus-tobacterium ratio in deep-sea sediments of the EasternMediterranean. Appl. Environ. Microbiology 66:1857-1861.

Eissler, Y., K. Wang, F. Chen, K. E. Wommack, and D. W. Coats.2009. Ultrastructural characterization of the lytic cycle ofan intranuclear virus infecting the diatom Chaetoceros cf.wighamii(Bacillariophyceae) from Chesapeake Bay, USA. J.Phycol. 45:787-797.

Fitzgerald, L. A., M. V. Graves, X. Li, T. Feldblyum, J. Hartigan,and J. L. Van Etten. 2007a. Sequence and annotation of the314-kb MT325 and the 321-kb FR483 viruses that infectChlorella Pbi. Virology 358:459-471.

———, ———, ———, ———, W. C. Nierman, and J. L. VanEtten. 2007b. Sequence and annotation of the 369-kb NY-2A and the 345-kb AR158 viruses that infect ChlorellaNC64A. Virology 358:472-484.


99

———, ———, ———, J. Hartigan, A. J. P. Pfitzner, E. Hoffart,and J. L. Van Etten. 2007c. Sequence and annotation of the288-kb ATCV-1 virus that infects an endosymbioticchlorella strain of the heliozoon Acanthocytis turfacea. Virol-ogy 362:350-361.

Frada, M., I. Probert, M.J. Allen, W.H. Wilson, and C. de Var-gas. 2008. The “Cheshire Cat” escape strategy of the coc-colithophore Emiliania huxleyi in response to viral infec-tion. Proc. Natl. Acad. Sci. USA 105:15944-15949.

Friess-Klebl, A. K., R. Knippers, and D. G. Müller. 1994. Isola-tion and characterization of a DNA virus infecting Feld-mannia simplex (Phaeophyceae). J. Phycol. 30:653-658.

Garry, R. T., P. Hearing, and E. M. Cosper. 1998. Characteriza-tion of a lytic virus infectious to the bloom-formingmicroalga Aureococcus anophagefferens (Pelagophyceae). J.Phycol. 34:616-621.

Garza, D. R., and C. A. Suttle. 1995. Large double-strandedDNA viruses which cause the lysis of a marine het-erotrophic nanoflagellate (Bodo sp.) occur in natural marineviral communities. Aquat. Microb. Ecol. 9:203-210.

Gastrich, M. D., J. A. Leigh-Bell, C. J. Gobler, O. R. Anderson,S. W. Wilhelm, and M. Bryan. 2004. Viruses as potentialregulators of regional brown tide blooms caused by thealga, Aureococcus anophagefferens. Estuaries 27:112-119.

Gobler, C. J., S. Deonarine, J. Leigh-Bell, M. D. Gastrich, O. R.Anderson, and S. W. Wilhelm. 2004. Ecology of phyto-plankton communities dominated by Aureococcus anophag-efferens: The role of viruses, nutrients, and microzooplank-ton grazing. Harmful Algae 3:471-483.

———, O. R. Anderson, M. D. Gastrich, and S. W. Wilhelm.2007. The ecology of viruses infecting the harmful browntide alga, Aureococcus anophagefferens. Aquat. Microb. Ecol.47:25-36.

Henry, E. C., and R. H. Meints. 1992. A persistent virus infec-tion in Feldmannia (Phaeophyceae). J. Phycol. 28:517-526.

Imai, I., and M. Yamaguchi. 1994. A simple technique forestablishing axenic cultures of phytoflagellates. Bull. Jpn.Soc. Microb. Ecol. 9:15-17.

Jacobsen, A., G. Bratbak, and M. Heldal. 1996. Isolation andcharacterization of a virus infecting Phaeocystis pouchetii(Prymnesiophyceae). J. Phycol. 32:923-927.

Kapp, M., R. Knippers, and D. G. Müller. 1997. New membersto a group of DNA viruses infecting brown algae. Phycol.Res. 45:85-90.

La Scola, B., and others. 2003. A giant virus in amoebae. Sci-ence 299:2033.

Lang, A. S., A. I. Culley, and C. A. Suttle. 2004. Genomesequence and characterization of a virus (HaRNAV) relatedto picorna-like viruses that infects the marine toxic bloom-forming alga Heterosigma akashiwo. Virology 320:206-217.

Lanka, S. T., M. Klein, U. Ramsperger, D. G. Müller, and R.Knippers. 1993. Genome structure of a virus infecting themarine brown alga Ectocarpus siliculosus. Virology 193:802-811.

Lawrence, J. E., A. M. Chan, and C. A. Suttle. 2001. A novel virus(HaNIV) causes lysis of the toxic bloom-forming alga Het-erosigma akashiwo (Raphidophyceae). J. Phycol. 37:216-222.

———, ———, and ———. 2002. Viruses causing lysis of thetoxic bloom-forming alga Heterosigma akashiwo (Raphido-phyceae) are widespread in coastal sediments of BritishColumbia, Canada. Limnol. Oceanogr. 47:545-550.

———, C. P. D. Brussaard, and C. A. Suttle. 2006. Virus-specificresponses of Heterosigma akashiwo to infection. Appl. Envi-ron. Microbiol. 72:7829-7834.

Lee, J. J. 1993. General techniques for the isolation and cultureof marine protists from estuarine, littoral, psammolittoral,and sublittoral waters, p. 41-50. In P. F. Kemp, B. Sherr, E.Sherr, and J. J. Cole [eds.], Handbook of methods in aquaticmicrobial ecology. Lewis Publishers.

Maier, I., S. Wolf, N. Delaroque, D. G. Müller, and H. Kawai.1998. A DNA virus infecting the marine brown algaPilayella littoralis (Ectocarpales, Phaeophyceae) in culture.Eur. J. Phycol. 33:213-220.


Mayer, J. A., and F. J. R. Taylor. 1979. A virus which lyses themarine nanoflagellate Micromonas pusilla. Nature 281:299-301.

Meints, R. H., J. L. Van Etten, D. Kuczmarski, K. Lee, and B.Ang. 1981. Viral infection of the symbiotic chlorella-likealga present in Hydra viridis. Virology 113:698-703.

———, R.G. Ivey, A.M. Lee, and T.-J. Choi. 2008. Identificationof two virus integration sites in the brown alga Feldmanniachromosome. J. Virol. 82:1407-1413.

Mizumoto, H., Y. Tomaru, Y. Takao, Y. Shirai, and K. Nagasaki.2007. Intraspecies host specificity of a single-stranded RNAvirus infecting a marine photosynthetic protist is deter-mined at the early steps of infection. J. Virol. 81:1372-1378.

———, ———, ———, ———, and ———. 2008. Diverseresponses of the bivalve-killing dinoflagellate Heterocapsacircularisquama to infection by a single-stranded RNA virus.Appl. Environ. Microbiol. 74:3105-3111.

Monier, A., J. B. Larsen, R.-A. Sandaa, G. Bratbak, J. M.Claverie, and H. Ogata. 2008. Marine mimivirus relativesare probably large algal viruses. Virol. J. 5:12.

Müller, D. G. 1991. Mendelian segregation of a virus genomeduring host meiosis in the marine brown alga Ectocarpussiliculosus. J. Plant Physiol. 137:739-743.

———. 1996. Host-virus interactions in marine brown algae.Hydrobiologia 21:21-28.

———, M. Sengco, S. Wolf, M. Brautigam, C. E. Schmid, M.Kapp, and R. Knippers. 1996. Comparison of two DNAviruses infecting the marine brown algae Ectocarpus siliculo-sus and E. fasciculatus. J. Gen. Virol. 77:2329-2333.

———, M. Kapp, and R. Knippers. 1998. Viruses in marinebrown algae. Adv. Virus Res. 50:49-67.

Nagasaki, K. 2001. Domestication of eucaryotic microalgal


100

viruses from the marine environments. Microbes Environ.16:3-8.

———, M. Ando, S. Itakura, I. Imai, and Y. Ishida. 1994. Viralmortality in the final stage of Heterosigma akashiwo (Raphi-dophyceae) red tide. J. Plankton Res. 16:1595-1599.

———, and I. Imai. 1994. Solid-phase culture of marinephytoflagellates. Bull. Jpn. Soc. Microb. Ecol. 9:37-43. Mostof article in Japanese.

———, and M. Yamaguchi. 1997. Isolation of a virus infectiousto the harmful bloom causing microalga Heterosigmaakashiwo (Raphidophyceae). Aquat. Microb. Ecol. 13:135-140.

———, and ———. 1998. Effect of temperature on the algici-dal activity and the stability of HaV (Heterosigma akashiwovirus). Aquat. Microb. Ecol. 15:211-216.

———, K. Tarutani, and M. Yamaguchi. 1999. Growth charac-teristics of Heterosigma akashiwo virus and its possible use asa microbiological agent for red tide control. Appl. Environ.Microbiol. 65:898-902.

———, and M. Yamaguchi. 1999. Cryopreservation of a virus(HaV) infecting a harmful bloom causing microalga, Het-erosigma akashiwo (Raphidophyceae). Fisheries Sci. 65:319-320.

———, Y. Tomaru, K. Tarutani, N. Katanozaka, S. Yamanaka,H. Tanabe, and M. Yamaguchi. 2003. Growth characteristicsand intra-species host specificity of a large virus infectingthe dinoflagellate Heterocapsa circularisquama. Appl. Envi-ron. Microbiol. 69:2580-2586.

———, ———, K. Nakanishi, N. Hata, N. Katanozaka, and M.Yamaguchi. 2004a. Dynamics of Heterocapsa circularisquama(Dinophyceae) and its viruses in Ago Bay, Japan. Aquat.Microb. Ecol. 34:219-226.

———, ———, N. Katanozaka, Y. Shirai, K. Nishida, S. Itakura,and M. Yamaguchi. 2004b. Isolation and characterizationof a novel single-stranded RNA virus infecting the bloom-forming diatom Rhizosolenia setigera. Appl. Environ. Micro-biol. 70:704-711.

———, Y. Shirai, Y. Takao, H. Mizumoto, K. Nishida, and Y.Tomaru. 2005a. Comparison of genome sequences of sin-gle-stranded RNA viruses infecting the bivalve-killingdinoflagellate Heterocapsa circularisquama. Appl. Environ.Microbiol. 71:8888-8894.

———, ———, Y. Tomaru, K. Nishida, and S. Pietrokovski.2005b. Algal viruses with distinct intraspecies host speci-ficities include identical intein elements. Appl. Environ.Microbiol. 71:3599-3607.

———, Y. Tomaru, Y. Takao, K. Nishida, Y. Shirai, H. Suzuki,and T. Nagumo. 2005c. Previously unknown virus infectsmarine diatom. Appl. Environ. Microbiol. 71:3528-3535.

———, ———, Y. Shirai, Y. Takao, H. Mizumoto. 2006.Dinoflagellate-infecting viruses. J. Mar. Biol. Assoc. U.K.86:469-474.

———, J-J. Kim, Y. Tomaru, Y. Takao, S. Nagai. 2009. Isolationand characterization of a novel virus infecting Teleaulax

amphioxeia (Cryptophyceae). Plankton Benthos Res. 4:122-124.

Paasche, E. 1971. A simple method for establishing bacteria-free cultures of phototactic flagellates. J. Cons. Int. Explor.Mer. 33:509-511.

Rowe, J. M., J. R. Dunlap, C. J. Gobler, O. R. Anderson, M. D.Gastrich, and S. W. Wilhelm. 2008. Isolation of a non-phage-like lytic virus infecting Aureococcus anophagefferens.J. Phycol. 44:71-76.

Reisser, W., T. Klein, and B. Becker. 1988a. Studies on phy-coviruses. I. On the ecology of viruses attacking Chlorellaeexsymbiotic from an European strain of Paramecium bur-saria. Arch. Hydrobiol. 111:575-583.

———, D. E. Burbank, S. M. Meints, R. H. Meints, B. Becker,and. J. L. Van Etten. 1988b. A comparison of viruses infect-ing two different chlorella-like green algae. Virology167:143-149.

Sahlsten, E. 1998. Seasonal abundance in Skagerrak-Kattegatcoastal waters and host specificity of viruses infecting themarine photosynthetic flagellate Micromonas pusilla. Aquat.Microb. Ecol. 16:103-108.

Sakata, T., Y. Fujita, and H. Yasumoto. 1991. Plaque formationby algicidal Saprospira sp. on a lawn of Chaetoceros ceratospo-rum. Nippon Suisan Gakkaishi 57:1147-1152.

Sandaa, R.-A., M. Heldal, T. Castberg, R. Thyrhaug, and G.Bratbak. 2001. Isolation and characterization of two viruseswith large genome size infecting Chrysochromulina ericina(Prymnesiophyceae) and Pyramimonas orientalis (Prasino-phyceae). Virology 290:272-280.


Shirai, Y., Y. Takao, H. Mizumoto, Y. Tomaru, D. Honda, and K.Nagasaki. 2006. Genomic and phylogenetic analysis of asingle-stranded RNA virus infecting Rhizosolenia setigera(Stramenopiles: Bacillariophyceae). J. Mar. Biol. Assoc. U.K.86:475-483.

———, Y. Tomaru, Y. Takao, H. Suzuki, T. Nagumo, K.Nagasaki. 2008. Isolation and characterization of a single-stranded RNA virus infecting the marine planktonic diatomChaetoceros tenuissimus Meunier. Appl. Environ. Microbiol.74:4022-4027.

Stein, J.R. [ed.]. 197. Handbook of phycological methods: Cul-ture methods and growth measurements. Cambridge Univ.Press.

Suttle, C.A. 1993. Enumeration and isolation of viruses, p.121-137. In P. F. Kemp, B. Sherr, E. Sherr, and J. J. Cole[eds.], Handbook of methods in aquatic microbial ecology.Lewis Publishers.

———, A. M. Chan, and M. T. Cottrell. 1991. Use of ultrafil-tration to isolate viruses from seawater which arepathogens of marine phytoplankton. Appl. Environ. Micro-biol. 47:721-726.

———, and ———. 1995. Viruses infecting the marine Prym-


101

nesiophyte Chrysochromulina spp.: isolation, preliminarycharacterization and natural abundance. Mar. Ecol. Prog.Ser. 118:275-282.

Tai, V., J. E. Lawrence, A. S. Lang, A. M. Chan, A. I. Culley, andC. A. Suttle. 2003. Characterization of HaRNAV, a single-stranded RNA virus causing lysis of Heterosigma akashiwo(Raphidophyceae). J. Phycol. 39:343-352.

Takao, Y., K. Nagasaki, K. Mise, T. Okuno, D. Honda. 2005. Iso-lation and characterization of a novel single-stranded RNAvirus (SssRNAV) infectious to a marine fungoid protistSchizochytrium sp. (Thraustochytriaceae, Labyrinthulea).Appl. Environ. Microbiol. 71:4516-4522.

———, K. Mise, K. Nagasaki, T. Okuno, D. Honda. 2006. Com-plete nucleotide sequence and genome organization of asingle-stranded RNA virus (SssRNAV) infecting the marinefungoid protist Schizochytrium sp. J. Gen. Virol. 87:723-733.

———, K. Nagasaki, D. Honda. 2007. Squashed ball-likedsDNA virus infecting a marine fungoid protist Sicyoid-chytrium minutum (Thraustochytriaceae, Labyrinthulea).Aquat. Microb. Ecol. 49:101-108.

Tarutani K., K. Nagasaki, and M. Yamaguchi. 2000. Viralimpacts on total abundance and clonal composition of theharmful bloom-forming phytoplankton Heterosigmaakashiwo. Appl. Environ. Microbiol. 66:4916-4920.

———, ———, S. Itakura, and M. Yamaguchi. 2001. Isolationof a virus infecting the novel shellfish-killing dinoflagellateHeterocapsa circularisquama. Aquat. Microb. Ecol. 23:103-111.

Thyrhaug. R., A. Larsen, T. F. Thingstad, G. Bratbak. 2003. Sta-ble coexistence in marine algal host-virus systems. Mar.Ecol. Prog. Ser. 254:27-35.

Tomaru, Y., N. Katanozaka, K. Nishida, Y. Shirai, K. Tarutani, M.Yamaguchi, and K. Nagasaki. 2004a. Isolation and charac-terization of two distinct types of HcRNAV, a single-strandedRNA virus infecting the bivalve-killing microalga Hetero-capsa circularisquama. Aquat. Microb. Ecol. 34:207-218.

———, K. Tarutani, M. Yamaguchi, and K. Nagasaki. 2004b.Quantitative and qualitative impacts of viral infection on aHeterosigma akashiwo (Raphidophyceae) bloom inHiroshima Bay, Japan. Aquat. Microb. Ecol. 34:227-238.

———, and others. 2007. Ecological dynamics of the bivalve-killing dinoflagellate Heterocapsa circularisquama and itsinfectious viruses in different locations of western Japan.Environ. Microbiol. 9:1376-1383.

———, Y. Shirai, H. Suzuki, T. Nagumo, K. Nagasaki. 2008. Iso-lation and characterization of a novel single-stranded DNAvirus infecting a cosmopolitan marine diatom Chaetocerosdebilis. Aquat. Microb. Ecol. 50:103-112.

———, Y. Takao, H. Suzuki, T. Nagumo, K. Nagasaki. 2009. Iso-

lation and characterization of a single-stranded RNA virusinfecting the bloom-forming diatom Chaetoceros socialis.Appl. Environ. Microbiol. 75:2375-2381.

Van Etten, J. L., R. H. Meints, D. E. Burbank, D. Kuczmarski, D.A. Cuppels, and L. C. Lane. 1981. Isolation and characteri-zation of a virus from the intracellular green alga symbioticwith Hydra viridis. Virology 113:704-711.

———, D. E. Burbank, Y. Xia, and R. H. Meints. 1983. Growthcycle of a virus, PBCV-1 that infects Chlorella-like algae.Virology 126:117-125.

———, L. C. Lane, and R. H. Meints. 1991. Viruses and virus-like particles of eukaryotic algae. Microbiol. Rev. 55:586-620.

———, and R. H. Meints. 1999. Giant viruses infecting algae.Ann. Rev. Microbiol. 53:447-494.

———, M. V. Graves, D. G. Müller, W. Boland, and N.Delaroque. 2002. Phycodnaviridae—large DNA algal viruses.Arch. Virol. 147:1479-1516.

Waters, R. E., and A.T. Chan. 1982. Micromonas pusilla virus:the virus growth cycle and associated physiological eventswithin the host cells; host range mutation. J. Gen. Virol.63:199-206.

Wilson, W. H., G. A. Tarran, D. Schroeder, M. Cox, J. Oke, andG. Malin. 2002. Isolation of viruses responsible for thedemise of an Emiliania huxleyi bloom in the English Chan-nel. J. Mar. Biol. Assoc. U.K. 82:369-377.

———, and others. 2005. Complete genome sequence andlytic phase transcription profile of a coccolithovirus. Sci-ence 5737:1090-1092.

———, D. C. Schroeder, J. Ho, and M. Canty. 2006. Phyloge-netic analysis of PgV-102P, a new virus from the EnglishChannel that infects Phaeocystis globosa. J. Mar. Biol. Assoc.U.K. 86:485-490.

Yamada, T., N. Chuchird, T. Kawasaki, K. Nishida, and S. Hira-matsu. 1999. Chlorella viruses as a source of novel enzymes.J. Biosci. Bioeng. 88:353-361.

———, H. Onimatsu, and J.L. Van Etten. 2006. Chlorellaviruses. Adv. Virus Res. 66:293-336.

Yan, X., P. R. Chipman, T. Castberg, G. Bratbak, and T. S. Baker.2005. The marine algal virus PpV01 has an icosahedral cap-sid with T=219 quasisymmetry. J. Virol. 79:9236-9243.

Xiao, C., P. R. Chipman, A. J. Battisti, V. D. Bowman, P. Renesto,D. Raoult, and M. G. Rossmann. 2005. Cryo-electronmicroscopy of the giant Mimivirus. J. Mol. Biol. 353:493-496.

Zingone, A., F. Natale, E. Biffali, M. Borra, G. Forlani, and D.Sarno. 2006. Diversity in morphology, infectivity, molecu-lar characteristics and induced host resistance between twoviruses infecting Micromonas pusilla. Aquat. Microb. Ecol.45:1-14.

102

Introduction

Viruses are the most numerous biological entities in aquaticecosystems, typically on the order of 107 mL–1 (Suttle 2007).Many studies have contributed to the acknowledgment thatviruses are active and diverse players in freshwater and marineecosystems (e.g., Brussaard et al. 2008b; Suttle 2007). Viralactivity profoundly impacts ecosystem function and structureby affecting host population dynamics, species succession,biodiversity, and global biogeochemical cycles.

To detect specific viruses or virus subpopulations, the use ofantibodies, plaque assay, or dilution to extinction in the pres-

ence of the appropriate host and approaches based on molec-ular markers have proven very useful (Larsen et al. 2007; Mühlinget al. 2005; Schroeder et al. 2003; Short and Suttle 2002). Tocount total free viruses in aqueous samples, transmission elec-tron microscopy (TEM), epifluorescence microscopy (EFM),and flow cytometry (FCM) are most often used. TEM has theadvantage of providing specific information about the mor-phology and size of the virus particles, but TEM is time- consuming and costly and, although information-rich, hasinherently low throughput. Over the past two decades,the introduction of highly sensitive fluorescent nucleicacid–specific dyes (e.g., SYBR Green) in combination withaffordable EFM have greatly facilitated the detection andquantification of viruses in a broad range of aquatic ecosys-tems. Although the same sensitive nucleic acid–specific stainscan be used in combination with FCM, its powerful analyticalcapabilities allow sensitive detection, accurate quantification,and rapid analysis of viruses relative to other conventionaltechniques such as TEM and EFM. FCM is a high-throughputmethod that, in addition, permits the discrimination of vari-ous virus populations based on their fluorescence and scattersignal after staining (Brussaard et al. 2000; Jacquet and Bratbak2003). This is of benefit for spatial and seasonal analysis ofviral dynamics and structure in a large number of natural sam-ples (Brussaard et al. 2008a; Li and Dickie 2001; Payet and Suttle 2008).

This article describes critically and in detail the protocol toenumerate aquatic viruses by FCM based on the methodologypreviously developed by Marie and co-workers (1999a) andoptimized by Brussaard (2004). Its application for samplesfrom different environments is discussed and compared toresults based on EFM.

Quantification of aquatic viruses by flow cytometryCorina P. D. Brussaard1*, Jérôme P. Payet2, Christian Winter 3, and Markus G. Weinbauer 4

1NIOZ Royal Netherlands Institute for Sea Research, PO Box 59, 1790 AB Den Burg, Texel, The Netherlands2Department of Earth and Ocean Sciences, University of British Columbia, Vancouver, BC, Canada3Department of Marine Biology, University of Vienna, Vienna, Austria4Microbial Ecology & Biogeochemistry Group, Université Pierre et Marie Curie-Paris 6, and CNRS, Laboratoire d’Océanographiede Villefranche, Villefranche-sur-Mer, France

AbstractFor many laboratories, flow cytometry is becoming the routine method for quantifying viruses in aquatic sys-

tems because of its high reproducibility, high sample throughput, and ability to distinguish several subpopula-tions of viruses. Comparison of viral counts between flow cytometry and epifluorescence microscopy typicallyshows slopes that are statistically not distinguishable from 1, thus confirming the usefulness of flow cytometry.Here we describe in detail all steps in the procedure, discuss potential problems, and offer solutions.




We gratefully thank A. Culley, A. Comeau, C. Pedrós-Alió, C. Lovejoy,and C. Martineau for their efforts in field sampling; and A. Ortmann, A.Chan, and officers and crew of the CCGS Amundsen for their supportduring the Canadian Arctic Shelf Exchange Study (CASES) expedition.We furthermore thank J. Brandsma for setting up the C. calcitrans experi-ments and J. Martínez Martínez, U. Wollenzien, and A. Noordeloos forproviding algal cultures.

ISBN 978-0-9845591-0-7, DOI 10.4319/mave.2010.978-0-9845591-0-7.102Suggested citation format: Brussaard, C. P. D., J. P. Payet, C. Winter, and M. G. Weinbauer. 2010.Quantification of aquatic viruses by flow cytometry, p. 102–109. In S. W. Wilhelm, M. G.Weinbauer, and C. A. Suttle [eds.], Manual of Aquatic Viral Ecology. ASLO.


MANUALof


Brussaard et al. Flow cytometric virus counting

103


An outline of the FCM assay for detection and quantifica-tion of aquatic viruses is presented in Fig. 1.

Reagents and solutions—FCM detection and enumeration ofviruses requires high-quality reagents. Water samples are pre-served using 25% electron microscopy (EM)-grade glutaralde-hyde (Sigma; storage at 4°C). The EM-grade glutaraldehyde isfree of polymers and other contaminants and, hence, is opti-mal for fixing the samples. To avoid cross-contamination ofsamples in pipetting, it is important to prepare small aliquotsof the fixative solution.

The use of ultrapure sheath fluid is essential, since oneworks close to the limits of detection of the instrument. Animproved FCM signal is obtained when using Milli-Q water(ultrapure deionized water with resistivity of 18.2 MΩ cm–1)instead of the commercially available sheath fluids (e.g., FACSFlow). Working stain solution of SYBR® green I(10,000× concentrate in DMSO; Invitrogen, MolecularProbes; storage at –20°C) is usually prepared by diluting thecommercial stock (1:200) in either autoclaved Milli-Q orgrade molecular water. Working solution stain can be reusedbut it is best to limit the number of freeze–thaw cycles (twoor three) to prevent the loss of staining efficiency over time.Thus, small aliquots (1 mL) of working stain solution shouldbe prepared. The commercial stock is supplied in DMSO, butfurther dilution in DMSO typically increases noise levelsupon addition to the samples. Occasionally, the fluorescentdye seems responsible for generating noise, and a brief spin

(~20,000g ) of the stock solution in a microfuge generallyreduces the noise levels.

Samples are diluted in sterile TE-buffer, pH 8.0 (10 mM Tris-hydroxymethyl-aminomethane, Roche Diagnostics; 1 mMethylenediaminetetraacetic acid, Sigma-Aldrich) to avoid elec-tronic coincidence (e.g., see below). The use of any diluents(e.g., phosphate-buffered saline, Milli-Q water, etc.) other thanTE-buffer was found to negatively affect flow cytometric sig-natures of stained viruses (Brussaard 2004).

TE-buffer should be autoclaved directly after preparation tomaintain low background fluorescence (check pH before useand adjust if needed using HCl). The quality of Tris may differdepending on the supplier, and thus, it is likely to affect thequality of TE-buffer (Brussaard unpubl. data). In principle, fil-tration before first-time use of the TE-buffer should not be nec-essary. But, once opened, small batches (i.e., ~50 mL) of the TE-buffer should be prefiltered (sterile FP30/0.2 µm; Schleicher &Schnell) just before use. Change the filter for each batch of theTE-buffer. Filtration may result in enhanced noise level,depending on the filter type. Filtration of the TE-bufferthrough 30-kDa molecular weight cutoff filters would be idealbut time-consuming. Instead, use a new batch of sterile TE-buffer and carefully check the noise level by running a stainedblank (see “Blank and reference”) before use.

Sampling and storage—Proper storage and preservation ofaquatic samples is crucial to prevent loss of virus particles.Typically, there is no need to filter or treat the natural watersamples before fixation. Filtration of the samples before fixa-tion may result in substantial loss of viruses (data not shown).

Fig. 1. Different processes, accompanying methodology, and critical notes for flow cytometric enumeration of aquatic viruses.


104

Samples of 1 mL are usually taken (replicate sampling isadvised), transferred into 2-mL cryovials, and fixed at a finalconcentration of 0.5% glutaraldehyde for 15–30 min at 4°C inthe dark. After fixation, the samples are flash frozen in liquidnitrogen. Flash-freezing is very important, as fixed samplesstored at 4°C show significant and rapid reductions in viruscounts (Brussaard 2004; Wen et al. 2004). For the same reason,it is important to minimize the fixation period to less than anhour (15–30 min is optimal). Once frozen in liquid nitrogen,the samples should be stored at –80°C (testing storage for 6months showed no detectable virus decay [Brussaard 2004]).Note that after a field expedition, frozen samples should besent either in dry ice or in a liquid nitrogen dry-shipper tokeep samples deep-frozen during transport.

FCM setup—Not all FCMs have equal sensitivity to detectand enumerate aquatic viruses. Whereas some FCMs willdetect only the higher green fluorescent virus subpopulations,others may not be sensitive enough to detect viruses at all. The488-nm argon laser benchtop FCMs of Becton Dickinson (e.g.,BD-FACScalibur) provide high sensitivity for virus detection.The advantage of benchtop FCMs is that the machines can beeasily taken on board ship.

Virus particles are too small to scatter light of standardbenchtop FCMs. The use of nucleic acid–specific stains, suchas SYBR Green I, is thus essential for virus detection. FCM sig-natures of stained viruses from natural samples can partiallyoverlap with background fluorescence generated by FCM. Ulti-mately, it is important to work with a clean FCM with lowbackground noise to obtain high-quality, reproducible data.Moreover, several blanks should be run before analysis tocheck whether the FCM and the reagents are clean or not (see“Blank and reference”).

Use maximum voltage for the green fluorescence photo-multiplier tube (PMT) at which no electronic or laser noise isdetected. This can be obtained by running freshly preparedMilli-Q water as sample and increasing the voltage for thegreen PMT until noise is detected; the maximum voltage thatcan be used is just below this. In some instances, the machinemay seem to be clean but after running a stained blank highlevels of noise are observed. Try running TE-buffer for sometime, followed by another stained blank to check. If still dirty,a useful remedy can be to purge (prime) a few times and cleanonce more. Once the FCM is clean and ready for use, try toanalyze the samples in one series, not interrupted by analysisof other organisms and use of other dyes. Bacterial enumera-tion can also be done from the same sample using a slightlydifferent setting and staining protocol (Marie et al. 1999b)with no interference for virus counts. In case of a high-event-rate sample (i.e., >1000 events s–1), rinse shortly with TE-bufferor Milli-Q before the next sample.

Sample dilution—Typically, samples need to be diluted beforeanalysis to minimize electronic coincidence of the virus particles(i.e., two or more virus particles pass the analysis window of thelaser simultaneously, reducing accuracy of the analysis). For virus

samples, this coincidence is minimized at event rates <1000events s–1 (Marie et al. 1999b). Salts present in the water samplescan strongly interfere with the efficiency of the stain SYBRGreen I, resulting in inaccurate quantification of the viruses.Consequently, the final dilution factor should be greater than10-fold, with a sample volume ≥50 µL used for the dilution. Foreach sample, a serial dilution of three to four different dilutionsof 500 µL final volume is usually optimal to obtain an event ratewithin 200–800 events s–1). Subsequently, the rest of the sam-ples can be analyzed using this optimal dilution factor.

Sample staining—The samples are stained with SYBR GreenI at a final concentration of 0.5 × 10–4 of the commercial stock(i.e., add 5 µL working stain solution to 500 µL sample). Thesamples are then incubated at 80°C for 10 min in the dark, fol-lowed by a cooling period at room temperature in the dark for5 min before analysis. Heating of fixed samples significantlyenhances the staining efficiency of viruses (Brussaard 2004;Marie et al. 1999a).

Blanks and references—Control blanks, consisting of TE-buffer with autoclaved 0.2-µm-filtered (or 30-kDa ultrafil-tered) sample at the same dilution factor as the natural sam-ples, should be used before FCM analysis of the samples.Filtering natural sample through a 0.02-µm pore-size filterinstead of autoclaving is not advised, as this may generate sub-stantial background noise.

Blanks are diluted, stained, and processed identically to thesamples. Very low coincidence (0–15 events s–1) and backgroundfluorescence levels should be detected before proceeding withsample analysis. Blanks ideally show a total amount of 400–1100events in 1 min of acquisition at a flow rate of ca. 40 µL min–1.During the analysis, always add one to two blanks to every batchof samples to monitor whether the noise level stays low.

An internal reference can be used not only to normalize thefluorescent signal of the stained virus populations, but moreimportantly to detect deviations of the FCM from standardbehavior. Highly diluted and well-mixed fluorescent micros-pheres (FluoSpheres carboxylate modified yellow-green fluo-rescent microspheres; 1.0 µm diameter; Invitrogen, MolecularProbes; F8823; stored at 4°C) may be used as reference. An ini-tial brief sonication of the primary stock (1% vol/vol, storageat 4˚C) is recommended to disrupt the aggregates. Workingbead solutions are then prepared by diluting the primary stockin sterile Milli-Q water (i.e., add 10 µL stock in 2.5 mL Milli-Qwater) every day.

Acquisition and data analysis—The appropriate settings fordetection of stained virus particles are specific for each FCM.Fluorescence and scatter signals are collected on a logarithmicscale (4-decade dynamic range) for best results. The trigger fordetection is set on green fluorescence, and data are acquiredon a dot plot displaying green fluorescence versus side scattersignal (Fig. 2). Commercial benchtop FCMs come with a cer-tain minimum threshold. This standard instrument thresholdlevel (typically 52 for BD-FACScalibur) should be used duringacquisition of the data.


105

A medium flow rate between 30 and 50 µL min–1 is ade-quate to detect viruses. FCMs with a sample injection port(e.g., BD-FACScalibur) should have the outer sleeve cleanedbetween samples to prevent cross-contamination (wipe withKimwipes® tissue). Samples should be mixed by hand beforeanalysis, as vortexing may result in decay of viruses (reductionof 15% for natural coastal seawater, data not shown). Allowthe flow rate to stabilize before analyzing the sample. Acquisi-tion time is typically 1 min.

Data analysis of the raw data collected in list-mode files canbe performed using a wide array of software (either suppliedwith the FCM or freeware from the internet; e.g., CytoWin orWinMDI). For optimal reproducibility and to include the verylow green fluorescent virus particles in the data analysis, thegating should always be set to include all the particles (Fig. 2).Importantly, virus counts in the sample should be correctedfor particles counted in the blanks (Fig. 3) before calculatingvirus concentrations.

AssessmentStaining—FCM analysis of the stained aquatic viruses gener-

ally discriminates two or three viral subpopulations (V1–V3)with different green fluorescence properties (Fig. 4). A fourthviral subpopulation (V4) may be observed (Fig. 5), commonlyrepresenting large dsDNA algal viruses (Brussaard et al. 2000;Jacquet and Bratbak 2003). Although most of the bacterio-phages (i.e., viruses infecting bacteria) are thought to be

included in the lower fluorescent viral subpopulations (V1 andV2 windows, Fig. 4), it was recently found that some eukaryoticalgal viruses displayed similar low fluorescence upon staining(Brussaard and Martínez Martínez 2008). Similarly, some pro-and eukaryotic algal viruses were also found in the V3 window(Brussaard et al. 2000). Furthermore, the level of nucleicacid–specific fluorescence is not indicative of the viral genomesize. There was no linear relationship between the viralgenome size and green fluorescence properties upon stainingwith a nucleic acid–specific stain (Brussaard et al. 2000).

SYBR Green I has a strong affinity for dsDNA but can alsostain ssDNA and RNA, according to the manufacturer (Invit-rogen). Several tests using various types of viruses indicatedthat these ssDNA and RNA viruses can be stained with SYBRGreen I (Brussaard et al. 2000). Nevertheless, some RNA-viruspopulations may not be fully separated from the backgroundnoise fluorescence; using other acid-specific dyes such as SYBRGreen II (higher quantum yield when bound to RNA than todsDNA) or SYBR Gold did not improve the detection of theseviruses (Brussaard et al. 2000; Brussaard 2004).

SYBR Gold, a fluorescent dye, detects DNA and RNA and ismore sensitive than SYBR Green I and can also be used as analternative of SYBR Green I for FCM detection of viruses (Chenet al. 2001). However, FCM data revealed significantly highercounts of viruses stained with SYBR Green I than with SYBRGold (Brussaard 2004). Thus, SYBR Green I seems best for opti-

Fig. 3. Cytogram of SYBR Green I–stained blank (using autoclaved 0.2-µm pore-size or 30-kDa prefiltered seawater instead of natural sample)according to protocol described herein (all events obtained plotted, i.e.,a total of 840, of which 222 were in the window used to discriminateviruses). The diagonal streak of dots outside and on the right side of thevirus window is due to the TE-buffer in combination with the fluorescentdye (SYBR Green I). r.u., relative units.

Fig. 2. Cytogram of SYBR Green I–stained viruses in typical naturalaquatic sample according to protocol described herein (10,000 eventsplotted). For optimal reproducibility and to include the very low greenfluorescent virus particles in the data analysis, the gating should always beset to include all the particles. r.u., relative units.


106

mal staining and detection of viruses in aquatic environ-ments. It might be useful, however, to test whether other flu-orescent stains or combination of stains can improve detec-tion when working with specific viruses.

Reproducibility—A critical question for the FCM user is howreproducible the analysis is and how representative of the “cor-rect” concentration. Usual practice is to include replicate countsin a random order. Standard deviations should be smaller than5%. Samples should possible be run in small batches (i.e., 6–10samples) to prevent poor reproducibility due to virus decay inthawed samples. Once thawed, the samples can be stored at 4°Cfor at most a few hours. Refreezing and reanalysis of samplesmust be avoided due to extensive loss in virus counts.

Accuracy is improved by regular calibrations of the sampleflow rate. Weighing the sample before and after a known timeperiod of running at one of the flow rates provides good esti-mates of the flow rate. However, this cannot be achieved whenon board a ship. Instead, preweighed and sealed tubes con-taining Milli-Q water can be used as an alternative, and theflow rate can be determined once the tubes are weighed backin the laboratory. Another rough estimate of the flow ratewhile on board can be obtained using back-pipetting: a knownvolume is dispensed in the tube, the remaining volume isback-pipetted after the run, and the actual volume taken up bythe FCM can be estimated by dividing change in volume overtime. Running a sample of fluorescent beads of known con-centration for determination of flow rate is not advised, sincethis may be unreliable due to clumping of beads.

Comparison of FCM versus EFM counts—A large data set (n =259, Table 1) from distinct marine environments was used tocompare viral counts obtained by FCM with EFM (using theprotocol of Hennes and Suttle 1995). Overall, total viruscounts ranged from <1 to 200 × 106 mL–1, with highest countsin Southern North Sea (e.g., 107–108 viruses mL–1). Linear least-squares regression analysis indicated a strong correlationbetween FCM and EFM counts (FCM = 1.08 × EFM + 0.65, r 2 =0.80, n = 259). Regression slopes and intercept values were notsignificantly different from a 1:1 regression line with a slopeof 1 and an intercept of 0 (slope: t-test = 0.143, P = 0.886;intercept: t-test = 0.069, P = 0.945).

Additionally, regression slopes ranged from 0.97 to 1.70and were not significantly different between the environ-ments (analysis of variance [ANOVA] on ranks, P > 0.05).Highest slope values were found for North Atlantic andCuraçao samples. The deep samples (>500 m, n = 8) are likelyto explain this result for North Atlantic; ratio of FCM to EFMof those samples are high and ranged from 4 to 6 (2500–4350m, n = 4). The high slope value for Curaçao samples is likelydue to the small number of samples leading to a nonsignifi-cant regression (r 2 = 0.36, P < 0.28, n = 5). Coastal and offshoremarine samples displayed similar regression slope values(Table 1). Moreover, the depth of sampling did not influenceregression slopes (Table 1). In the Arctic, samples were col-lected over a seasonal cycle at different stations, but no sea-sonal and/or spatial trends were observed in the FCM versusEFM regressions (Table 2).

Fig. 5. Cytogram of SYBR Green I–stained viruses in natural aquatic sam-ple according to protocol described herein (10,000 events plotted). Afourth subpopulation with enhanced side-scatter signal may be observed.This subpopulation, V4, commonly represents large dsDNA algal viruses.r.u., relative units.

Fig. 4. Cytogram of SYBR Green I–stained viruses in typical naturalaquatic sample according to protocol described herein (10,000 eventsplotted). Virus subpopulation with lowest green fluorescence is namedV1, with midlevel fluorescence V2 and highest fluorescence V3. r.u., rela-tive units.


107

Discussion and recommendations

FCM versus EFM—For bacterial samples, FCM counts are gen-erally identical to EFM counts (Monfort and Baleux 1992, Payetand Suttle 2008). Furthermore, quantitative intercomparisonbetween FCM and EFM counts of large dsDNA algal viruses alsoshowed a strong correlation (Marie et al. 1999a). We show here(Fig. 6) that also for natural marine virus samples, typically dom-inated by lower fluorescent bacteriophages, counting viruses byFCM and EFM gave similar results (FCM = 1.08 × EFM + 0.65, n =259). FCM allows high-speed detection and enumeration ofviruses and may represent a better alternative than EFM.

The method presented here should be taken into accountfor FCM detection and enumeration of aquatic marine viruses.Importantly, nonfrozen samples and low dilution factors willresult in unreliable virus counts. The potentially higher totalvirus count obtained by flash-freezing compared to nonfrozensamples (Brussaard 2004) is not an artifact caused by lysis ofinfected organisms and subsequent release of viruses. Testswith pure virus cultures and also with 0.2-µm-filtered naturalsamples (to remove all organisms) systematically showedhigher total virus counts upon flash-freezing (data not shown).

In contrast to EFM, FCM is more sensitive and has theadvantage of discriminating different virus populations, thusproviding more information about the community structure ofviral communities in a broad range of aquatic ecosystems. Fur-thermore, the high throughput of FCM and the ability to dis-criminate particular large dsDNA algal viruses has permittedthe execution of detailed experiments enhancing our insight ofvirus–host interactions and the impact of viruses on algalbloom population dynamics (Brussaard et al. 2005; Jacobsen etal. 2007; Jacquet et al. 2002; Larsen et al. 2001, 2007).

Application in different environments—Although FCM detectionof viruses was initially developed for marine samples (Marie et al.1999a), it was further applied to other environments such asfreshwater and sediments (Chen et al. 2001; Danovaro et al.2001; Duhamel et al. 2006; Duhamel and Jacquet 2006; Goddardet al. 2005; Lymer et al. 2008). Virus counts obtained from 13

lake sediment samples were 2.5-fold higher using FCM than EFM(Duhamel and Jacquet 2006); however, the FCM signatures ofthese virus samples are distinctly different from aquatic samples.The extraction of viruses from the sediment is a critical step.Although Danovaro and Middelboe (2010, this volume) presentan optimized protocol for enumeration of viruses in sedimentsusing EFM, they also highlight the importance of improvingmethods for dislodging viruses from particles in different typesof sediments. Preliminary tests using different types of sedimentsamples in combination with FCM (Brussaard unpubl. data)showed that high background levels of very low fluorescent par-ticles, probably colloids, can occasionally interfere with thedetection of viruses. We believe that further assessment of theFCM assay for (diverse) sediment samples is needed.

Virus detection—The term virus-like particles (VLPs) is typi-cally used for virus detection by TEM, where the morphology ofthe viruses is used as the discriminator. The detection of particlesby FCM, however, is not based on morphology but on nucleicacid–specific fluorescence and side-scatter signal. Comparison toEFM, TEM, and end-point dilution has shown that the green flu-orescent particles counted by FCM (Fig. 2) are indeed viruses.Particles such as gene transfer agents (GTAs) (Lang and Beatty2007) may be confused with virus particles after staining with anucleic acid–specific dye, but they seem to represent only a verysmall portion of the total viral particle pool (Lang pers. comm.).Only in axenic algal cultures have we observed nonvirus con-taminants with the same signature on a scatter and fluorescenceplot. Addition of bacteria (natural seawater) to the axenic algalcultures (Chaetoceros calcitrans and Micromonas pusilla) resultedin a steep decrease of these interfering particles to undetectablelevels, most likely due to decomposition of the organic matterreleased upon lysis of infected algal host cells. Thus, under natu-ral conditions (with heterotrophic bacteria present), the abun-dance of these contaminants will be insignificant and of no con-sequence for virus enumeration.

FCM detection of viruses requires the use of fluorescent dye;however, not all viruses are readily stainable by currently avail-able fluorescent dyes (Brussaard 2004). Dye penetration and effi-

Table 1. Linear least-squares regression analysis for viral abundance determined by flow cytometry and epifluorescence microscopy(FCM = slope × EFM + intercept).

Origin Slope Intercept (× 106 mL–1) r 2 P value n Depth range, m

Mackenzie shelf 1.00 1.29 0.89 <0.0001 59 0–526

Mackenzie river plume 0.97 1.25 0.94 <0.0001 26 0–240

Amundsen Gulf 1.06 0.07 0.92 <0.0001 51 0–530

Franklin Bay 0.95 0.79 0.90 <0.0001 94 0–223

Curaçao coral reef 1.50a 1.48a 0.36a 0.2821a 5 3–40

Southern North Sea 0.97 13.49 0.66 0.0013 12 5

North Atlantic 1.70 0.42 0.77 0.0002 12 5–4350

Offshore 1.11 0.76 0.84 <0.0001 98 0–4350

Coastal 1.08 0.45 0.80 <0.0001 161 0–526

Marine 1.08 0.65 0.80 <0.0001 259 0–4350aNot relevant (r 2 < 0.5) regression.


108

ciency of staining can greatly change depending on the type ofvirus; in some instances, some viruses may have genomes toosmall for optimal detection (Brussaard et al. 2000). In turn, thismay cause underestimation of total virus counts in aquatic envi-ronments. Advances in stain sensitivity and FCM technology incoming years will likely allow a better evaluation of these verylow fluorescent virus particles in aquatic virus communities.

The dsRNA virus MpRV exemplifies that the optimized assaypresented here is not necessarily most favorable for all viruses(Brussaard et al. 2004). Currently, FCM cannot properly distin-guish MpRV viruses because their low fluorescent intensitiesinterfere with the background fluorescence. The reason for the

poor detection is not related just to genome size (25.6 kb, Attouiet al. 2006), as other viruses with similar genome size were clearlydetected on the FCM (Brussaard et al. 2000). MpRV belongs to thenonturreted reoviruses, containing several concentric protein lay-ers (inner and outer capsid layers), which may prevent properpenetration of the fluorescent dye. Using the BD FACSaria (newerFCM), a better detection of MpRV virus was found, but not for theentire population (Brussaard unpubl. data). Furthermore, detec-tion of MpRV was significantly improved by dilution of the viruswith Milli-Q water before fixation and flash-freezing; however,predilution of a natural marine sample in Milli-Q water resultedin lower detection of the major virus populations.

Standards—The use of true standards adds to consistency inmethodology and allows optimal comparison of results. So far,green fluorescent 1-µm microspheres have been used as inter-nal reference to normalize fluorescence and track instrumentperformance. These beads have a relatively high fluorescenceand are not effective standards for detection efficiency, stainingcharacteristics, and quantitative evaluation. Smaller beads withfluorescence and side-scatter signals comparable to low fluo-rescent viruses may be useful for quantitative standards, butparticles with staining properties similar to those of viruses arepreferred to check for the efficiency of the staining procedure.The use of a known (natural) virus sample as standard is notideal in the long run, as we do not know yet if aquatic samplescan be safely stored for prolonged times without virus decay.

On the whole, FCM is an accurate and highly reproduciblemethod for virus enumeration and discrimination of mainvirus subpopulations in aquatic environments. A majoradvantage is the high throughput, a key issue for analysis of alarge number of samples. There is an increasing interest inaquatic viral ecology, and with benchtop FCMs becomingmore affordable, soon many laboratories will be able to rou-tinely perform virus enumerations by FCM.

ReferencesAttoui, H., F. M. Jaafar, M. Belhouchet, P. De Micco, X. De

Lamballerie, and C. P. D. Brussaard. 2006. Micromonas

Fig. 6. Linear least-square regression of total viral abundance of aquaticsamples determined by FCM and EFM. The line represents the linear least-squares regression. Note that a double-logarithmic scale is used.

Table 2. Linear least-squares regression analysis for viral abundance in Arctic waters determined by FCM and EFM (FCM = slope × EFM +intercept).

Origin Season Slope Intercept (× 106 mL–1) r 2 n

Coastal arctic Fall 0.95 0.78 0.91 34

Winter 0.85 1.04 0.88 52

Spring 0.98 0.51 0.92 24

Summer 0.95 2.06 0.89 46

All 1.00 0.67 0.93 156

Offshore arctic Fall 0.90 0.95 0.84 33

Spring 1.30 –0.02 0.97 12

Summer 1.03 0.91 0.91 29

All 1.00 0.79 0.89 74

Arctic All 1.00 0.71 0.92 230

For all regression analyses, the P value was <0.0001.


109

pusilla reovirus: A new member of the family Reoviridaeassigned to a novel proposed genus (Mimoreovirus). J. Gen.Virol. 87:1375-1383.

Brussaard, C. P. D. 2004. Optimization of procedures forcounting viruses by flow cytometry. Appl. Environ. Microb.70:1506-1513.

———, D. Marie, and G. Bratbak. 2000. Flow cytometric detec-tion of viruses. J. Virol. Methods 85:175-182.

———, A. A. M. Noordeloos, R.-A. Sandaa, M. Heldal, and G.Bratbak. 2004. Discovery of a dsRNA virus infecting themarine photosynthetic protist Micromonas pusilla. Virology319:280-291.

———, B. Kuipers, and M. J. W. Veldhuis. 2005. A mesocosmsstudy of Phaeocystis globosa population dynamics. I. Regula-tory role of viruses in bloom control. Harmful Algae 4:859-874.

———, and J. Martínez Martínez. 2008. Algal bloom viruses.Plant Viruses 2:1-13.

———, K. R. Timmermans, J. Uitz, and M. J. W. Veldhuis. 2008a.Virioplankton dynamics in the waters southeast of the Kerguelen (Southern Ocean). Deep-Sea Res. Pt. II 55:752-765.

———, and others. 2008b. Global scale processes with ananoscale drive: The role of marine viruses. ISME J 2:575-578.

Chen, F., J.-R. Lu, B. J. Binder, Y.-C. Liu, and R. E. Hodson.2001. Application of digital image analysis and flow cytom-etry to enumerate marine viruses stained with SYBR Gold.Appl. Environ. Microbiol. 67:539-545.

Danovaro, R., A. Dell’anno, A. Trucco, M. Serresi, and S.Vanucci. 2001. Determination of virus abundance inmarine sediments. Appl. Environ. Microbiol. 67:1384-1387.

———, and M. Middelboe. 2010. Separation of free virus par-ticles from sediments in aquatic systems, p. 74-81. In S. W.Wilhelm, M. G. Weinbauer, and C. A. Suttle [eds.], Manualof Aquatic Viral Ecology. ASLO.

Duhamel, S., I. Domaizon-Pialat, S. Personnic, and S. Jacquet.2006. Assessing the microbial community dynamics andthe role of bacteriophages in bacterial mortality in LakeGeneva. Rev. Sci. l’Eau 19:115-126.

——— and S. Jacquet. 2006. Flow cytometric analysis of bacteria- and virus-like particles in lake sediments. J. Micro-biol. Methods 64:316-332.

Goddard, V. J., and others 2005. Temporal distribution ofviruses, bacteria and phytoplankton throughout the watercolumn in a freshwater hypereutrophic lake. Aquat.Microb. Ecol. 39:211-223.


Jacobsen, A., A. Larsen, J. Martínez-Martínez, P. G. Verity, andM. E. Frischer. 2007. Susceptibility of colonies and colonialcells of Phaeocystis pouchetii (Haptophyta) to viral infection.Aquat. Microb. Ecol. 48:105-112.

Jacquet, S., M. Heldal, D. Iglesias-Rodriguez, A. Larsen, W. Wil-son, and G. Bratbak. 2002. Flow cytometric analysis of an

Emiliania huxleyi bloom terminated by viral infection.Aquat. Microb. Ecol. 27:111-124.

———, and G. Bratbak. 2003. Effects of ultraviolet radiationon marine virus-phytoplankton interactions. FEMS Micro-biol. Ecol. 44:279-289.

Lang, A. S., and J. T. Beatty. 2007. Importance of widespreadgene transfer agent genes in α-proteobacteria. TrendsMicrobiol. 15:54-62.

Larsen, A., and others 2001. Population dynamics and diver-sity of phytoplankton, bacteria and viruses in a seawaterenclosure. Mar. Ecol. Prog. Ser. 221:47-57.

Larsen, B. J., A. Larsen, R. Thyrhaug, G. Bratbak, and R.-A. San-daa. 2007. Marine viral populations detected during anutrient induced phytoplankton bloom at elevated pCO2

levels. Biogeosci. Disc. 4:3961-3985.Li, W. K. W., and P. M. Dickie. 2001. Monitoring phytoplank-

ton, bacterioplankton, and virioplankton in a coastal inlet(Bedford Basin) by flow cytometry. Cytometry 44:236-246.

Lymer, D., J. B. Logue, C. P. D. Brussaard, A.-C. Baudoux, K.Vrede, and E. S. Lindström. 2008. Temporal variation infreshwater viral and bacterial community composition.Freshw. Biol. 53:1163-1175.

Marie, D., C. P. D. Brussaard, R. Thyrhaug, G. Bratbak, and D.Vaulot. 1999a. Enumeration of marine viruses in cultureand natural samples by flow cytometry. Appl. Environ.Microbiol. 65:45-52.

———, F. Partensky, D. Vaulot, and C. P. Brussaard. 1999b.Enumeration of phytoplankton, bacteria, and viruses inmarine samples, p. 11.11.11-11.11.15. In J.P. Robinson[managing ed.], Z. Darzynkiewicz, P. N. Dean, A. Orfao, P. S.Rabinovitch, C. C. Stewart, H. J. Tanke, L. L. Wheeless[eds.], Current protocols in cytometry. Wiley.

Monfort, P., and B. Baleux. 1992. Comparison of flow cytom-etry and epifluorescence microscopy for counting bacteriain aquatic ecosystems. Cytometry 13:188-192.

Mühling, M., and others 2005. Genetic diversity of marineSynechococcus and co-occurring cyanophage communities:Evidence for viral control of phytoplankton. Environ.Microbiol. 7:499-508.

Payet, J. P., and C. A. Suttle. 2008. Physical and biological cor-relates of virus dynamics in the southern Beaufort Sea andAdmunsen Gulf. J. Mar. Syst. 74:933-945.


Short, S. M., and C. A. Suttle. 2002. Sequence analysis ofmarine virus communities reveals that groups of relatedalgal viruses are widely distributed in nature. Appl. Envi-ron. Microbiol. 68:1290-1296.

Suttle, C. A. 2007. Marine viruses: Major players in the globalecosystem. Nature Rev. Microbiol. 5:801-812.


110

Introduction

Two procedures, direct enumeration and concentration ofviruses, form the core methodological basis of most ecologicalinvestigations of natural viral assemblages. Indeed, nearly all ofthe methods outlined in this volume contain enumeration or

viral concentration as a component procedure. Because theseprocedures are especially critical, aquatic viral ecologists havecontinually sought improvements in sample throughput, pre-cision, and yield. As a consequence, today there are dozens oftechnical approaches to viral enumeration and concentration.This plethora of techniques can be confusing to a new investi-gator, and in too few cases, has thorough consideration beengiven to methods comparison. This report will briefly reviewapproaches for the concentration of viruses from large (>1 L tohundreds of liters) water samples, outline an example method,and provide guidance for the assessment of concentrationmethods geared to a particular analytical outcome.

Early methods for viral concentration from water samplesrelied on adsorption of viruses to a solid matrix such as a fiber-glass (Sobsey et al. 1977), membrane filter (Farrah et al. 1976;Katayama et al. 2002), or diatomaceous earth (Farrah et al.1991) followed by elution of viruses into a small volume ofbuffer. Whereas the details of these methods vary widely, theyessentially rely on increasing the concentration of cations anddecreasing the pH of the water sample, which encourages pos-itively charged viral particles to adsorb to negatively chargedsurfaces. Because of the extensive water conditioning requiredfor negatively charged membranes, electropositive filters havealso been employed and have shown more consistent recoveryof polioviruses (Sobsey and Glass 1980). Subsequently,

Filtration-based methods for the collection of viral concentratesfrom large water samplesK. Eric Wommack1,2*, Télesphore Sime-Ngando3, Danielle M. Winget1, Sanchita Jamindar2, and Rebekah R. Helton1

1University of Delaware, College of Marine and Earth Studies, Delaware Biotechnology Institute, 15 Innovation Way, Newark,DE 197112University of Delaware, Dept. of Plant and Soil Sciences, Delaware Biotechnology Institute, 15 Innovation Way, Newark, DE197113Université Blaise Pascal, UMR CNRS 6023, 63177 Aubière Cedex, France

AbstractEcological investigations rely on data describing the biomass, diversity, and composition of living things. In

the case of microbial communities, these data are primarily gathered using microscopy and molecular geneticapproaches. The diminutive size of viruses means that obtaining genetic material sufficient for molecularapproaches for examining the diversity and composition of aquatic viral assemblages can be challenging.Moreover, in procedures for the isolation and cultivation of novel viruses from natural waters, high-density viralinocula provide the best chance for success. To address the need for samples containing a high-density of virus-es, investigators have used tangential-flow filtration (TFF) to concentrate viruses from large-volume (>20 L)water samples. This report outlines procedures for the preparation of viral concentrates from large volume watersamples using TFF and discusses the effect of concentration procedures on viral recovery and downstreammolecular genetic analyses.




The authors would like to thank those who have spent countlesshours watching water move from bottle to bottle and back to the sea,in particular: Kui Wang, Sharath Srinivasiah, Kurt Williamson, ShellieBench, Shannon Williamson, Matt Simon, and Tim Mills. Work con-tributing to this manuscript was supported through National ScienceFoundation grants (MCB-0132070 and OCE-0221825) to K.E.W. and anNSF predoctoral fellowship to D.M.W.

ISBN 978-0-9845591-0-7, DOI 10.4319/mave.2010.978-0-9845591-0-7.110Suggested citation format: Wommack, K. E., T. Sime-Ngando, D. M. Winget, S. Jamindar, and R.R. Helton. 2010. Filtration-based methods for the collection of viral concentrates from largewater samples, p. 110–117. In S. W. Wilhelm, M. G. Weinbauer, and C. A. Suttle [eds.], Manualof Aquatic Viral Ecology. ASLO.


MANUALof


Wommack et al. Virioplankton concentration methods

111

adsorbed viruses are eluted (removed) from the solid matrixinto a small volume of a positively charged buffer such as 50mM glycine or dilute beef extract. Despite the speed and effi-ciency of absorption-elution methods for the concentration ofviruses, those researchers focused on the ecology ofautochthonous aquatic viral assemblages have not adoptedthese methods. The primary reason is that because virusescan differ in biophysical characteristics, not all viruses areconcentrated with equal efficiency through adsorption-elu-tion. Indeed, all adsorption-elution methods for viral concen-tration have focused on the detection of specific pathogenicviruses within freshwater, and to a lesser extent, seawater sam-ples. Moreover, differences in the characteristics of a givenwater sample can influence viral recovery, and eluant bufferssuch as beef extract can be incompatible with downstreamanalyses such as molecular genetic assays and microscopy(Williamson et al. 2003).

Aquatic viral ecology studies conducted over the past twodecades have avoided the inherent limitations of adsorption-elution techniques through the use of two distinctapproaches for the concentration of viruses from water sam-ples. Ultrafiltration has been primary among these and is thefocus of this review; however, direct collection of viruses byultracentrifugation has also been used (Short and Short 2008;Steward et al. 2000). The primary technical challenge in con-centrating sub-micron particles from large aqueous samples isthe prevention of filter clogging. While a number ofapproaches have been developed to avoid clogging of ultra-filtration membranes, the most widely adopted has been tan-gential-flow filtration (TFF).

In TFF, the process fluid (in our case, a water sample) flowsalong a parallel tangent to the filter surface. Application ofhydrodynamic pressure to the process flow (usually throughrestriction of the flow at one end of the filter) causes water andparticles smaller than the pore size of the ultrafiltration mem-brane to flow through the membrane (Fig. 1). The water flow-

ing across the membrane is known as the retentate, while thewater flowing through the filter is the permeate. The speed offiltration is controlled by varying the amount of back pressureon the retentate flow (Fig. 1). Although back pressure can becontrolled by both the retentate flow rate and the amount ofrestriction on the retentate flow; in practice, the flow rate isheld constant while the restriction to the flow is used to con-trol the overall filtration rate. Higher flow rates will moreeffectively prevent filter clogging as will more modest levels ofback pressure. Manufacturers of TFF systems provide opera-tional limits in terms of back pressure and often recommendoptimal flow rates to prevent filter clogging. Thus, TFF repre-sents a balance between the prevention of filter clogging (highretentate flow and low back pressure) and the speed of filtra-tion (high permeate flow and high back pressure). In the spe-cial case of concentrating viruses from natural water samples,it is unlikely that an investigator will really challenge the oper-ational limits of most TFF systems with sheet-type filter mem-branes; however, hollow fiber-based systems can be less toler-ant to excessive back pressure. Nevertheless, the prudentinvestigator will carefully monitor TFF operating conditionsand check filter integrity. Poor operating procedures or inap-propriate cleaning and storage procedures can ruin theintegrity of the filter. Such loss of integrity will allow virusesto pass through the filter resulting in a reduction in viral con-centration efficiency.

There are at least three manufacturers that supply TFF sys-tems suitable for the concentration of viruses from naturalwater samples. Millipore Corporation (www.millipore.com) isthe largest supplier of TFF technologies to the research marketand manufactures TFF platforms in two reusable cartridge for-mats including spiral-wound (e.g., Helicon and Prep-Scale)and flat plate (e.g., Pellicon). These systems are designed invarious sizes to allow for scaling process volumes from ~100mL up to thousands of liters. Smaller TFF systems in a flat-plate format are supplied by Sartorius Stedim Biotech

Fig. 1. Schematic diagram of tangential flow filtration cartridge


112

(www.sartorius-stedim.com; Vivaflow) and Pall Corporation(www.pall.com; Ultrasette) and are rated for initial filtrationvolumes of up to 5 L and can be expanded by parallel con-nection of multiple filters.

In most cases, initial concentration of viruses from large(10 to hundreds of liters) water samples by TFF in the field isfollowed by laboratory procedures designed to further con-centrate viruses into a small volume (e.g., <10 mL). Initial TFF-based concentration of viruses within large water samples ispreceded by microfiltration to remove bacterial cells and pro-tists. Most TFF filtration housings can withstand inclusion ofparticles up to 100 µm in size; however, most investigatorschoose to remove all particles and cells larger than 0.22 µmprior to TFF so as to ensure the purity of viral concentrates.This report will outline and evaluate the filtration stepsneeded for the preparation of samples (viral concentrates)containing a density of viral particles concentrated from largevolumes of natural water samples.

Materials and proceduresI. List of materials

A. Large-scale concentration.i. Tubing: silicone, various sizesii. Tubing: PharMed for pump heads onlyiii. Connectors to tubing and filtersiv. Diaphragm pump (Jabsco Industrial Diaphragm

pump #31801, 12 L per min capacity)v. Two 50-L plastic carboysvi. 25-L carboyvii. 25-µm wound polypropylene sediment filter

(pool filter) and housingviii. Large peristaltic pump with 2 pump headsix. 0.22-µm TFF filterx. 30- to 50-kD TFF filterxi. 0.22-µm syringe filtersxii. 60-mL syringesxiii. 50-mL conical tubes

B. Small-scale concentration.i. Tubing: silicone, various sizesii. Tubing: PharMed for pump heads onlyiii. Connectors to tubing and filtersiv. Small, reversible peristaltic pump with 1 pump

headv. Two 2-L polycarbonate bottlesvi. 500-mL polycarbonate bottlevii. 30- to 50-kD compact spiral filter

C. Postconcentration.i. Ultracentrifugeii. Swing bucket rotoriii. Polyallomer tubesiv. Balancev. Waste containervi. Final collection tubes of needed sizevii. 200-µL pipettman and tips

II. List of reagentsA. Sterile 60% glycerol

III. General protocol for the concentration of virioplanktonfrom a large volume (>20 L) water sample

Note: All filters should be cleaned and rinsedaccording to manufacturer’s recommendations beforeuse in any application during the concentrationprocedures. (Fig. 2A and B)

A. Prefiltration.i. Prefilter ambient water sample with a 25-µm

wound polypropylene sediment filter before anyconcentration (Fig. 2A). Use a diaphragm pump tofilter raw water into a 50-L plastic carboy.

ii. Rinse the 50-L carboy three times with a few litersof the prefiltered water before filling with finalsample.

B. Tangential flow microfiltration to remove particlesand cells >0.22 µm (Fig. 2B).i. Once a 50-L carboy is full of 25 µm prefiltered

ambient water, attach the 0.22-µm TFF filter feedand retentate tubing. Run the feed tubing throughone of the large peristaltic pump heads (Caution:check for the correct flow direction before makingconnections!). Attach the permeate tubing to a 25-L carboy to contain the viral concentrate. Thepermeate valve on the 0.22-µm TFF filter shouldbe in the OFF position.

ii. If the peristaltic pump head has an occlusionsetting, set it to maximal occlusion to help primethe 0.22-µm TFF filter and remove all air. Increasethe pump speed to 20%. Once the tubing andfilter are completely filled with water, adjust theocclusion knob to a looser setting to preventexcessive tubing wear. Slowly increase the pumpspeed to 45%. Once the system is runningsmoothly, partially open the permeate valve onthe 0.22-µm TFF filter to the second tick mark(~20° open) and collect the >0.22 µm permeateinto a 25-L carboy.

iii. Cells and particulates between 25 µm and 0.22 µmwill concentrate within the 50-L carboy as TFFthrough the 0.22-µm filter proceeds.

C. Concentration of virioplankton.i. Large-scale (Fig. 2B)

a. When the 25-L carboy is more than half fullwith <0.22 µm permeate, prepare the 30-kDTFF filter for viral concentration. Slow thelarge peristaltic pump speed to 10% (Do notstop the pump as this encourages adherence ofviruses to filter matrix). Attach the feed andretentate tubing to the appropriate ports onthe 30-kD TFF. Make sure that the backpressureknob on the 30-kD TFF filter is completelyopen (counter-clockwise). Direct the permeate


113

tubing from the 30-kD TFF filter into a cleancarboy to collect ultrafiltrate (UF; virus-freewater). Carefully work the feed tubing into thesecond pump head of the large peristalticpump.

b. Increase the large peristaltic pump speed to20%. Again, tighten the occlusion knob on thepump head to prime the 30-kD TFF filter andremove all air. Once the filter and lines arefully primed loosen the pump head occlusionknob. Slowly increase the pump speed to 45%.Slowly close the backpressure knob on the 30-kD TFF filter until permeate begins to flow.Monitor the level of <0.22 µm water in the 25-L carboy. The level should be maintained athalf full until the prefiltered ambient water isnearly gone. Collect 1 L UF in a 2-L bottle forrinsing of the 30-kD TFF filter.

c. When ~5 L of 25 µm filtered ambient waterremains in the large carboy, release the pumphead and remove the 0.22-µm TFF filter feedtubing from the large peristaltic pump.Continue to run the 30-kD TFF filter until ~1 Lof <0.22 µm water remains in the 25 L carboy.Avoid entry of air into the feed line as thelevel of retentate water nears this minimum.Stop the large peristaltic pump and drain the30-kD TFF filter into the retentate carboy.Slowly prime the 30-kD TFF filter from a 1-Lstock of UF and then recirculate the UF at 30%pump speed for 5 min. Be careful to avoid airbubbles. Stop the pump, release the pumphead, and drain all tubing into the 2-L UFbottle. (Note: Recirculation of UF after primaryviral concentration has been reported by themanufacturer to substantially improverecovery of retained molecules and viruses[Millipore 2003]).

d. Pool the recirculated UF with the ~1 L ofretentate from the primary viral concentrationthrough the 30-kD TFF filter. The volume ofthe ~2 L viral concentrate can be furtherreduced by using a small-scale TFF ultrafilter.To avoid excessive degradation of viruses, it isadvisable to store the first stage VC at 4°C andperform a second stage small-scaleconcentration a soon as possible (i.e., withinno more than 1 d).

ii. Small-scaleThe steps involved in the concentration of virusesfrom smaller water samples (i.e., <2 L) are similarto those for large samples. However, the TFFfilters used for concentration of viruses fromsmall-scale samples are usually 10-fold smaller in

filter area. The smaller size of these filters, andthe tubing connected to them, results in acoordinately smaller minimum hold-up volume.In our experience, small-scale TFF filtration resultsin viral concentrates of ca. 250 mL in volume.Small-scale concentration of viruses from ambientwater samples will require prefiltration to removecells and particulates larger than 0.22 µm in size.If small-scale TFF filtration is used as a secondstep viral concentration following a large-scale(50 L to 2 L) procedure no prefiltration isrequired. After this two step process, theoreticalviral concentration ratios of 200 fold (50 L to0.25 L) can be achieved. Actual viralconcentration ratio will depend on overallfiltration efficiency. Oftentimes, large- and small-scale TFF concentration of viruses is performed inthe field, and the final 250 mL concentrate isfrozen (preferably snap frozen in LN2) fortransport back to the lab. Some investigators havereported better viral preservation by addingglycerol to a final concentration of 10% prior tosnap freezing (Glass and Williamson pers. comm.).

iii. Postconcentration proceduresOftentimes, it is desirable to reduce the volume ofviral concentrates below the ca. 250 mL minimumhold-up volume of most small-scale TFF apparatus.In particular, fingerprinting of viral assemblagesby pulsed-field gel electrophoresis (Steward 2001;Wommack et al. 1999) or preparation of viralconcentrates for metagenomic sequencing (Benchet al. 2007; Breitbart et al. 2002) requires viralparticle densities of ≥109 mL–1.By and large, investigators have produced high-density viral concentrates using either spin filters,which rely on centrifugal force to push waterthrough a 30- to 50-kD molecular sieve, or anultracentrifuge to pellet viruses followed byresuspension in a smaller volume of UF or buffer.Disposable spin filters are provided by a numberof manufacturers and require only a benchtopswinging bucket rotor for filtration (Bench et al.2007). Pelleting of viruses can only be done in anultracentrifuge at centrifugal forces exceeding100,000g, a requirement that can be costprohibitive or unavailable at smaller researchfacilities (Wommack et al. 1992). Recently,Colombet and coworkers (2007) adaptedpolyethylene glycol (PEG) precipitation of viruses,for postconcentration of viruses within 1 L viralconcentrates derived from 20-L water samples. ThePEG protocol showed a greater than 2-foldincrease in the recovery efficiency of virusparticles as compared with the ultracentrifugation


114

procedure. The lower cost of this procedure iswelcomed; however, we have found thatpostconcentration procedures can have a dramaticeffect on the quality of PFGE virioplanktonfingerprints. In contrast to the reportedimprovement in fingerprints after PEGprecipitation (Colombet et al. 2007), we foundthat postconcentration of viral concentrates byultrafiltration consistently produces the bestresolved PFGE fingerprints (Fig. 3). The source ofloss in PFGE band clarity and sharpness forsamples processed by PEG precipitation or spinfiltration is not known and did not appear to beattributable to sample loading as the PEG andCentricon samples were loaded with 2-fold lessand 3-fold more viruses, respectively, than theultracentrifuged sample.

Assessment

Because each processing step can result in the loss of virusesit is important to optimize filtration procedures for maximalviral recovery. Essentially, there are three major factors that canhave a substantial effect on viral recovery: i) filter material; ii)flow rates; and iii) the physical means by which the processfluid travels through the filtration apparatus. An experiment toassess prefiltration and TFF methods is illustrative of the com-bined impact of these factors on viral recovery as assessedthrough viral direct counts by epfluorescence microscopy(Fig. 4). At least 60 L estuarine water samples (~30‰ salinity)collected at the entrance to the Delaware Bay near Lewes, DE(38°48′N; 75°07′W) were subjected to five methods for removalof larger particles and cells: gravity filtration through a 30-µmNitex screen; pump-driven filtration through a 25-µmpolypropylene sediment filter; vacuum filtration through a

Fig. 2. Schematic diagrams of prefiltration (panel A) and TFF procedures (panel B) for the concentration of viruses from large volume water samples.Abbreviations are as follows: Ret, retentate; Perm, permeate; kD, kilodalton.


115

glass fiber (GF/C) impact filter, and filtration through the 25-µm sediment filter followed by GF/C filtration. Ambient viraland bacterial abundances were typical of mid-Atlantic coastalwaters at ca. 8 × 107 and 3.4 × 106 mL–1, respectively. Amongthese methods, GF/C filtration had the worst viral recovery at27% and was significantly lower than ambient water and 30µm Nitex-filtered water (one-way ANOVA, Tukey post hoc, P <0.05) whereas all other methods produced no significant dif-ferences in viral abundances. It is important to remember thatthe influence of prefiltration procedures on viral recovery seenin this experiment is highly correlated to the physiochemicalconditions of the modestly estuarine waters at the entrance tothe Delaware Bay. Oligotrophic pelagic waters or those con-taining greater levels of particulate matter would likely show adifferent outcome in the same experiment.

Subsequently, seawater samples prefiltered through the25-µm sediment pool filter were subjected to four treatmentsfor the removal of cells and small particulates: 1 and 2) vacuum-driven impact filtration (a.k.a., dead-end filtration) through0.22-µm or 0.45-µm filters; 3 and 4) TFF through 0.22-µm or0.45-µm TFF filters. In the case of TFF through 0.22-µm and0.45-µm TFF filters, three cross-filter flow rates were tested as

measured by the ratio of retentate to permeate flow. In thisscenario, the high flow rate treatment was a retentate to per-meate flow ratio of 4 to 1 (i.e., in a given amount of time,4-fold more water flowed through the retentate than thepermeate). Whereas statistically significant differences inviral recovery were not found among these filtration proce-dures, clear trends were apparent (Fig. 5). In generalmedium cross-filter flow rates (a 2.5:1 retentate to permeateflow ratio) resulted in better viral recovery for TFF proce-dures. Impact filtration and TFF performed similarly; how-ever, because limited sample volumes were tested, the influ-ence of impact filter clogging on viral recovery was notassessed. A key factor in deciding whether to use impact orTFF procedures for removal of cellular material is the degreeto which filter clogging will influence viral recovery for agiven volume of sample water. In productive coastal watersthat support high levels of bacterial abundance, impact fil-tration is less advisable as significant clogging can occur forrelatively low sample volumes (e.g., ~5 L). Moreover, impactfiltration through 0.45-µm filters tended to pass more bac-terial cells into the filtrate as compared with impact filtra-tion through 0.2-µm filters or tangential flow filtrationwith either 0.45- or 0.22-µm filters.

Assessment of final viral concentration by TFF using a 30-kDspiral cartridge (Helicon, Millipore) showed no significant dif-ferences in viral abundance resulting from TFF at each of three

Fig. 3. Effect of final viral concentration procedures on pulsed-field gelelectrophoretic analysis of Chesapeake Bay virioplankton assemblages.Lanes are as follows: 1, 9, and 10) molecular weight markers (band sizein kilobases of DNA are shown on the left); 2) viral concentrate withoutadditional concentration procedure (0.1 × 109 viruses); 4) ultracentrifu-gation pelleting and resuspension of viral concentrate (1.6 × 109 viruses);6) polyethylene glycol precipitation and resuspension of viral concen-trate (0.8 × 109 viruses); 8) ultrafiltration of viral concentrate using Cen-tricon Plus 70 (Millipore) (5 × 109 viruses). All initial viral concentrateswere 0.22 µm filtered before final concentration procedures or loadingon PFGE gel.

Fig. 4. Effect of prefiltration procedures on recovery of viral particlesfrom seawater. Treatments are as follows: Ambient, seawater withoutprefiltration; Nitex, 30-µm Nitex screen; Pool, pump-driven filtrationthrough a 25-µm wound polypropylene sediment filter; GF/C, glass fiberfilter C nominal pore size (0.7 µm); Pool + GF/C, GF/C filtration follow-ing pool filter. Letters above bars denote significance groupings by Tukeypost hoc tests (P < 0.05). Error bars are standard deviations of replicatemeasurements.


116

cross-flow rates (one-way ANOVA, P > 0.05; data not shown),although concentration efficiency tended to be highest for thehigh flow rate TFF procedure. Thus, among the three filtrationsteps used in preparation of viral concentrates from large vol-ume water samples, the greatest potential for loss of virusesoccurs with the removal of cells and particulates of ≥0.22 µm.

Discussion

Tangential-flow filtration procedures create high-densityviral concentrates that are clear of contaminating cells andparticles larger than 0.22 µm. These concentrates can directlyfeed a number of downstream analyses common to viral eco-logical investigations such as isolation of new viral-host sys-tems (Suttle and Chan 1995; Wang and Chen 2008) andassessing the impact of increased viral predation on host phys-iology (Suttle et al. 1990); detection of gene targets by PCR(Wang and Chen 2004; Zhong et al. 2002); characterization ofwhole virioplankton assemblages by randomly amplified poly-morphic DNA-PCR (Winget and Wommack 2008) or PFGEprofiling (Wommack et al. 1999). In the case of PCR proce-dures, many investigators have chosen to treat viral concen-trates with nucleases to remove free DNA and ensure that PCRamplicons are derived from only capsid-enclosed viralgenomic DNA (Winget and Wommack 2008). This considera-tion is also critical in the creation of viral metagenomelibraries where procedures such as CsCl purification andnuclease digestion have been used prior to high-throughputsequencing (Breitbart et al. 2002).

The relative scale of viral concentration necessary for eachprocedure can differ. PCR-based procedures may require onlymodest levels of viral concentration (10 to 50 times concen-tration), whereas PFGE profiling or metagenomic analysesmay require viral densities in excess of 100-fold that of ambi-ent viral abundance. Thus, it is important to match the filtra-tion apparatus to the required degree of viral concentrationand initial sample volume. The end-point consideration forthe investigator is the surface area of TFF filters. Filter surfacearea is directly proportional to filtration rate, thus, larger fil-ters can process larger volumes of process fluid in a fixedamount of time. One caveat, however, is that minimum hold-up volumes will also increase with larger TFF filters ensuringthat a second, smaller scale TFF system will be needed to fur-ther boost viral density within concentrates.

While viral concentration methods have been essential toadvancing our understanding of viruses in aquatic environ-ments, these methods also present significant limitations. Therequired separation of viruses and cells means that viruseswith capsid sizes exceeding 220 nm are lost from endpointviral concentrates. For example, DNA polymerase sequencesfrom phycoviruses are commonly detected in whole microbialcommunities from the Chesapeake Bay, but are absent fromviral concentrates prepared from Bay waters (Chen pers.comm.). As a consequence of prefiltration, extant geneticdiversity of large viruses, such as those within the Phycod-naviridae and nucleo-cytoplasmic large DNA viruses(NCLDVs), has been under-sampled within virioplanktonmetagenome libraries (Monier et al. 2008).

Among the aquatic viral ecology research community, it iswidely believed that TFF filtration methods lack the inherentand possibly systematic biases in viral concentration that

Fig. 5. Effects of prefiltration on viral recovery (upper panel) and bac-terial cell removal (lower panel). Category designations are as follows:Ambient, initial unprocessed water samples; 25-µm pool filter, filtrationof ambient water through wound polypropylene sediment filter; IF,impact filtration; TFF, tangential-flow filtration; High, Medium, and Low,4:1, 2.5:1, and 1:1 retentate to permeate flow ratio, respectively; Pellicon,0.1-m2, 0.22-µm TFF filter. Error bars are standard deviations of replicatemeasurement.


117

might result from adsorption-elution methods (Suttle et al.1991). Thus, TFF viral concentration procedures operate onthe assumption that all viruses are concentrated with equalefficiency. To our knowledge, however, this assumption hasnot been thoroughly tested nor compared with alternate pro-cedures such as adsorption-elution. Although it is a tediousand difficult experiment, certainly more extensive testing forsystematic loss of virus groups during TFF is warranted. A sec-ondary alternative to such an extensive test would be a quickand inexpensive means to monitor concentration efficiency.At present, few investigators report the efficiency of their viralconcentration procedures. As shown in Figs. 4 and 5, evenslight alterations in filtration media or filtration conditionscan dramatically affect the efficiency of viral concentration.Ironically, simply increasing initial water sample volume 5-fold (from 50 to 250 L) lowered concentration efficiency by~10-fold (50% to 5%) for Chesapeake Bay water samples (datanot shown) and eliminated possible gains in the concentra-tion of viruses from processing a larger sample volume.

ReferencesBench, S. R., and others. 2007. Metagenomic characterization

of Chesapeake Bay virioplankton. Appl. Environ. Micro-biol. 73:7629-7641.

Breitbart, M., and others. 2002. Genomic analysis of uncul-tured marine viral communities. Proc. Natl. Acad. Sci U.S.A.99:14250-14255.

Colombet, J., A. Robin, L. Lavie, Y. Bettarel, H. M. Cauchie, andT. Sime-Ngando. 2007. Virioplankton ‘pegylation′: use ofPEG (polyethylene glycol) to concentrate and purify virusesin pelagic ecosystems. J. Microbiol. Methods 71:212-219.

Farrah, S. R., C. P. Gerba, C. Wallis, and J. L. Melnick. 1976.Concentration of viruses from large volumes of tap waterusing pleated membrane filters. Appl. Environ. Microbiol.31:221-226.

———, D. R. Preston, G. A. Toranzos, M. Girard, G. A. Erdos,and V. Vasuhdivan. 1991. Use of modified diatomaceousearth for removal and recovery of viruses in water. Appl.Environ. Microbiol. 57:2502-2506.

Katayama, H., A. Shimasaki, and S. Ohgaki. 2002. Develop-ment of a virus concentration method and its applicationto detection of enterovirus and Norwalk virus from coastalseawater. Appl. Environ. Microbiol. 68:1033-1039.

Millipore. 2003. Protein concentration and diafiltration bytangential flow filtration. Millipore.

Monier, A., J. M. Claverie, and H. Ogata. 2008. Taxonomic dis-tribution of large DNA viruses in the sea. Genome Biol. 9:R106.


Sobsey, M. D., C. P. Gerba, C. Wallis, and J. L. Melnick. 1977.Concentration of enteroviruses from large volumes of tur-bid estuarine water. Can. J. Microbiol. 23:770-778.

———, and J. S. Glass. 1980. Poliovirus concentration fromtap water with electropositive adsorbent filters. Appl. Envi-ron. Microbiol. 40:201-210.

Steward, G. F. 2001. Fingerprinting viral assemblages by pulsedfield gel electrophoresis, p. 85-103. In J. H. Paul [ed.],Marine microbiology. Academic Press. (Methods in micro-biology; V. 30).


Suttle, C. A., A. M. Chan, and M. T. Cottrell. 1990. Infectionof phytoplankton by viruses and reduction of primary pro-ductivity. Nature (London) 347:467-469.

———, ———, and ———. 1991. Use of ultrafiltration to iso-late viruses from seawater which are pathogens of marinephytoplankton. Appl. Environ. Microbiol. 57:721-726.

———, and ———. 1995. Viruses infecting the marine Prym-nesiophyte Chrysochromulina spp.: Isolation, preliminarycharacterization and natural abundance. Mar. Ecol. Prog.Ser. 118:275-282.

Wang, K., and F. Chen. 2004. Genetic diversity and populationdynamics of cyanophage communities in the ChesapeakeBay. Aquat. Microb. Ecol. 34:105-116.

——— and ———. 2008. Prevalence of highly host-specificcyanophages in the estuarine environment. Environ.Microbiol. 10:300-312.

Williamson, K. E., K. E. Wommack, and M. Radosevich. 2003.Sampling natural viral communities from soil for culture-independent analyses. Appl. Environ. Microbiol. 69:6628-6633.

Winget, D. M., and K. E. Wommack. 2008. Randomly ampli-fied polymorphic DNA PCR as a tool for assessment ofmarine viral richness. Appl. Environ. Microbiol. 74:2612-2618.

Wommack, K. E., R. T. Hill, M. Kessel, E. Russek-Cohen, and R.R. Colwell. 1992. Distribution of viruses in the ChesapeakeBay. Appl. Environ. Microbiol. 58:2965-2970.

———, J. Ravel, R. T. Hill, and R. R. Colwell. 1999. Populationdynamics of Chesapeake Bay virioplankton: Total commu-nity analysis using pulsed field gel electrophoresis. Appl.Environ. Microbiol. 65:231-240.

Zhong, Y., F. Chen, S. W. Wilhelm, L. Poorvin, and R. E. Hod-son. 2002. Phylogenetic diversity of marine cyanophageisolates and natural virus communities revealed bysequences of viral capsid assembly protein g20. Appl Envi-ron. Microbiol. 68:1576-1584.

118

Introduction

It was not until the late 1980s when the use of electronmicroscopy had revealed the presence of very large abun-dances (106-108 mL–1) of viruses in marine pelagic environ-ments that it was realized that viruses were quantitativelyimportant players in natural ecosystems (e.g., Bergh et al.

1989; Proctor and Fuhrman 1990; Suttle et al. 1990). Untilthen, studies of viruses in marine environments had been lim-ited to examination and quantification of specific bacteria-virus systems, based on relatively few culturable bacterialhosts and co-occurring bacteriophages (e.g., Moebus 1991,1992). These studies, which were based on numbers of infec-tive units, generally showed that specific viruses were presentin low densities (hundreds to a few thousands per milliliter)and did not suggest the significant role of viruses in themarine ecosystem, which was demonstrated a few years later.

The realization that marine viruses were significant agentsof mortality for both heterotrophic bacteria and cyanobacteriawith large impact also on microbial diversity, populationdynamics, and nutrient cycling has accelerated the scientificeffort in aquatic viral ecology over the past 15 years (e.g., Wein-bauer 2004; Suttle 2007; Brussaard et al. 2008; Middelboe2008). Since the development and refinement of techniques toquantify total viral abundance (e.g., Hennes and Suttle 1995;Noble and Fuhrman 1998) and production (e.g., Steward et al.1992; Wilhelm et al. 2002; Winget et al. 2005), a large researcheffort has been allocated to describe the dynamics and impactsof the total viral community, based mainly on enumeration oftotal viral abundance, total viral production, and frequency ofinfected cells. However, little is still known about the dynam-ics, specificity, evolution, and ecological impact of the mostbasic property of the aquatic viral community: the interaction

Isolation and life cycle characterization of lytic viruses infectingheterotrophic bacteria and cyanobacteriaMathias Middelboe1*, Amy M. Chan2, and Sif K. Bertelsen1

1Marine Biological Laboratory, University of Copenhagen, Strandpromenaden 5, DK-3000 Helsingør, Denmark2Department of Earth and Ocean Sciences, The University of British Columbia, 6339 Stores Road,Vancouver, BC, V6T 1Z4,Canada

AbstractBasic knowledge on viruses infecting heterotrophic bacteria and cyanobacteria is key to future progress in

understanding the role of viruses in aquatic systems and the influence of virus–host interactions on microbialmortality, biogeochemical cycles, and genetic exchange. Such studies require the isolation, propagation, andpurification of host–virus systems. This contribution presents some of the most widely used methodologicalapproaches for isolation and purification of bacteriophages and cyanophages, the first step in detailed studiesof virus–host interactions and viral genetic composition, and discusses the applications and limitations of dif-ferent isolation procedures. Most work on phage isolation has been carried out with aerobic heterotrophic bac-teria and cyanobacteria, culturable both on agar plates and in enriched liquid cultures. The procedures pre-sented here are limited to lytic viruses infecting such hosts. In addition to the isolation procedures, methods forlife cycle characterization (one-step growth experiments) of bacteriophages and cyanophages are described.Finally, limitations and drawbacks of the proposed methods are assessed and discussed.

*Corresponding author: E-mail: [email protected]; Phone: +45 3532 19 91, Fax: +45 35 32 19 51



This work was supported by the Danish Natural Sciences ResearchCouncil, the Directorate for Food and Agri Business and the CarlsbergFoundation (MM); and by the University of British Columbia and theNatural Research and Engineering Council of Canada (AMC). The com-ments of two anonymous reviews were appreciated and helped toimprove the manuscript.

ISBN 978-0-9845591-0-7, DOI 10.4319/mave.2010.978-0-9845591-0-7.118Suggested citation format: Middelboe, M., A. M. Chan, and S. K. Bertelsen. 2010. Isolation andlife cycle characterization of lytic viruses infecting heterotrophic bacteria and cyanobacteria, p. 118–133. In S. W. Wilhelm, M. G. Weinbauer, and C. A. Suttle [eds.], Manual of Aquatic ViralEcology. ASLO.


MANUALof


Middelboe et al. Isolation and characterization of phages

119

between a specific virus and its hosts. As the overall viral activ-ity represents the sum of all the virus–host systems at any giventime, the role of viruses is therefore a property of the total col-lection of virus–host interactions.

Viral infection of specific hosts affects the composition,and therefore the biogeochemical properties of the microbialcommunity, as well as the fluxes of carbon and nutrients inthe ecosystem. Viral infection also acts as a driving force inmicrobial evolution by selecting for virus resistant mutantsand by mediating genetic exchange between their hosts. Con-sequently, understanding virus–host interactions at the levelof individual cells and populations is a prerequisite for obtain-ing fundamental insight on the role of viruses for essentialbiogeochemical and evolutionary processes.

Assessing the impact of viruses on an ecosystem scalerequires the inclusion of viruses in models of food web inter-actions and biogeochemical cycling. To constrain values ofviral parameters in models of virus–host interactions, it isessential to obtain detailed information about the fundamen-tal properties of the virus–host interaction, e.g., adsorptionrate, life cycle, host range, resistance development, etc, andhow these properties are influenced by the environmentalconditions. Such models again provide a frame for under-standing and predicting the behavior, dynamics, and evolu-tion of specific virus–host systems in more complex naturalenvironments.

More recently, genomic characterization of specific viruses(e.g., Fuller et al. 1998; Mann et al. 2003) has opened a newapproach to identify and track genes shared among groups ofviruses and explore the genetic diversity and distribution ofviruses in the oceans. For instance, specific molecular probes,arising from genomic studies, are now being employed totrack populations of both hosts and viruses within environ-mental samples (e.g., Short and Short 2008) and thus deter-mine how these interactions influence the dynamics of virusesand their hosts.

Consequently, despite the fact that individual viralstrains/types and their bacterial or cyanobacterial hosts prob-ably constitute a very small fraction of the total microbialcommunity, there is a lot to learn from studying this particu-lar level of microbial processes. Also, it links very well with thecurrent genomic studies of viral diversity, dynamics, and dis-tribution in a research area that allows the combination ofviral ecology, biodiversity, biogeochemistry, and genomics.Common for these types of studies is that they depend on theisolation of specific virus–host systems for the further charac-terization of their properties.

Isolation of viruses is thus the first step in detailed studiesof virus–host interactions, and the present article describesbasic methodological approaches for isolation of two groupsof viruses, bacteriophages and cyanophages, and discusses theapplications and limitations of different isolation procedures.It should be emphasized that viruses targeted by the presentedprotocols are restricted to (1) lytic viruses and (2) viruses

infecting culturable hosts, thus at the same time excludingprobably the majority of marine viruses, i.e., viruses withunculturable hosts and temperate or chronic lifestyles.

Materials and proceduresNatural virus communities from the aquatic environment

(e.g., seawater, rivers, lakes, ponds, sediments, water sur-rounding cyanobacterial mats, etc.) can provide the source ofpotential virus isolates. For the isolation of bacteriophages orcyanophages, samples for screening should be prefiltered orcentrifuged before using. For example, use 0.8 µm to 1.2 µmpore-size glass fiber or PCTE filters to remove larger particlesand organisms, followed by 0.2 µm or 0.45 µm pore-size lowprotein-binding PVDF filters (Millipore Durapore) to removethe remaining bacteria and phytoplankton that could inter-fere with the isolation assays. The drawback of such pre-screening, however, is that filtration and centrifugation mayalso remove a fraction of the phages in the sample, and thusreduce the chance of finding lytic phages against the targetbacteria.

Store the filtered samples at 4°C, in the dark until use. If theoverall virus abundance is suspected to be low, such as onewould expect from oligotrophic environments (e.g., polarregions), virus concentrations in the sample can be increasedseveral hundred fold using ultrafiltration/tangential flowmethodologies (e.g., Suttle et al. 1991; Wommack et al. 2010,this volume). For sediment samples, procedures to extractviruses from particles (e.g., Danovaro and Middelboe 2010,this volume) should be applied prior to phage isolation. Muchof the following procedures were adapted from earlier tech-niques described for the isolation of bacteriophages andcyanophages from environmental samples (e.g., Adams 1959;Safferman and Morris 1963; Eisenstark 1967; Berg 1987).

Described below are methods that are used routinely to iso-late, purify, and characterize bacteriophages and cyanophagesfrom aquatic environments. Although there are a number ofsimilarities between the methods used, the specific practicalprotocols for bacteriophages and cyanophages differ from oneanother. We have therefore chosen a full step-by-step presen-tation of the proposed protocols for each group of phages,rather than combining the two, and focus on the specific partswhere the procedures differ from each other. Our approachresults in some overlap between the two sections, but on theother hand, increases the clarity of the individual protocols. Itis assumed that culturing methodology has been determinedfor the target hosts of interest, and if host culture conditionsare not optimized, this should be the first step in the isolationprocedure for any type of phage.

Isolation and host range characterization of bacteriophages—Bacteriophage isolation by spotting on target host cells: Phagelysis of host bacteria can be visualized by plaque formation onlawns of host cells in soft agar overlayed on agar plates(Adams, 1959; Sambrook et al. 1989). This principle can beused for the detection and subsequent isolation of specific


120

lytic phages in environmental samples (e.g., Carlson 2005).Spotting environmental samples on lawns of a target host cellwould thus reveal the presence of a lytic phage for that par-ticular host cell in the given sample:1. The host cells are grown overnight in liquid cultures con-

taining an appropriate growth medium for the organism(i.e., a medium that will yield a visible lawn of bacteria insoft agar when plated on an agar plate). Rich media suchas Luria Broth (10 g tryptone, 5 g yeast extract, 10 g NaClin 1 L distilled water) or MLB (0.5 g Casamino acids, 0.5 gpeptone, 0.5 g yeast extract, 3 mL glycerol in 800 mL pre-filtered [GF/C] seawater and 200 mL distilled water) can beused to culture many varieties of marine bacteria.

2. Measure optical density spectrophotometrically at 525 nm(OD525) and adjust OD to 0.3–0.5 with growth medium.This ensures sufficient bacterial density as well as thecapacity for further bacterial growth during the followingplate incubation.

3. Soft agar (0.5 to 0.6% agar in media of choice) is melted ina water bath or a microwave oven, then distributed inaliquots of 4 mL to sterile culture tubes and kept just abovesolidification temperature until use. The solidificationtemperature of soft agar depends on the agar or agaroseused and is about 45°C for common agar. If the host bac-teria cannot survive exposure to 45°C soft agar, low-melt-ing-point agaroses are available for a range of lower tem-peratures (see “Cyanophage” section).

4. 200–300 µL bacterial culture is added to the 4 mL tubeswith melted soft agar. The bacteria-soft agar mixture isthen vortexed and immediately poured onto an agar platewith an agar that supports growth of the host bacterium(e.g., Zobell agar [5 g tryptone, 1 g yeast extract, 15 g agarin 800 mL GF/C filtered sea water and 200 mL distilledwater]), and distributed evenly on the plate, which isplaced on a flat surface.

5. When the soft agar containing the target bacteria has solidi-fied, triplicate aliquots of 5–10 µL of each of the environ-mental water samples from which phages should be isolatedare spotted on top of the soft agar. Before spotting, the sam-ples should be filtered (e.g., 0.2 µm or 0.45 µm syringe filters)or centrifuged (e.g., 10,000g, 10 min) to remove bacteria.These procedures would minimize bacterial contaminationin the spotting zone, which may hide clearing zones. Asmentioned above, filtration and centrifugation may alsoremove a fraction of the phages in the sample, and thusreduce the chance of finding lytic phages against the targetbacteria. As a negative control 5–10 µL phage buffer (e.g., SMbuffer: 450 mM NaCl, 50 mM MgSO4, 50 mM Tris, 0.01 %Gelatin, pH = 8) or 0.02 µm filtered sample water is spottedin triplicate on the soft agar. If the abundance of specificphages is expected to be low, the phages can be concentratedby various procedures prior to spotting on the target bacte-ria. Concentrating procedures are described elsewhere in thisspecial issue (Wommack et al. 2010, this volume).

6. The plates are incubated for 1–3 d depending on thegrowth rate of the bacteria, and the presence of lytic phagesin the sample is detected as a clearing zone (plaque) in thespotted area of the lawn of bacteria that develops over timeon the plate. A single or a few phages added will result inonly small plaques in the zone, whereas many phages inthe spotted sample will yield a large clearing zone.

7. If clearing zones appear in the spotted area, this indicatesthe presence of lytic phages, which can be isolated andpurified (see below). To confirm that the clearing is due tophage lysis and not some other growth-inhibiting factor inthe original sample, a dilution series of the sample can beperformed and spotted on the target bacteria. Diluted suf-ficiently, phages would appear as single plaques in thespotted zone rather than a gradual reduction in growthinhibition as would be the case if the clearing was causedby some chemical factor. Alternatively, heat-killed (e.g.,90°C for 5 min) or 0.02 µm filtered controls can be used toverify that a clearing zone is caused by a biological com-ponent and not a chemical. Presence of phages can also beverified by SYBR staining and subsequent detection by epi-fluorescence microscopy (Noble and Fuhrman 1998; Suttleand Fuhrman 2010, this volume).

8. Once detected as clearings in the spotting zone, phages arefurther isolated and purified from the plates as describedbelow.

The main advantage of this procedure is that presence oflytic viruses is visually apparent as clearing zones on host bac-terial lawns, and that subsequent isolation of phages is there-fore fast, as the phages can be isolated directly from theplaques.

Bacteriophage isolation using enrichment cultures: A gen-erally more efficient way of isolating lytic phages from marineenvironments is by the use of enrichment cultures. In thisapproach, the prefiltered water sample that is to be screenedfor phages against a given target bacterium, is enriched with abacterial growth medium and amended with that target bac-teria (Eisenstark 1967; Carlson 2005). This allows any lyticphages present in the sample to infect the target bacteria andpropagate in the cultures, and subsequently, be isolated andpurified. The two main advantages of the enrichmentapproach are 1) that it allows for screening for phages in amuch larger volume of sample (typically 25–50 mL, ratherthan 5–10 µL), thus increasing the probability of isolating rarephages, and 2) that it allows the combination of different tar-get hosts (e.g., different strains of a specific bacteria of inter-est) in the same incubation, again increasing the possibility ofphage isolation.

The following is the standard procedure in our lab whensearching for phages against specific target bacteria in envi-ronmental samples. The sample volume and number of hoststrains used may be varied according to the sample investi-gated and the purpose of the phage isolation.1. As for the spot test procedure, the potential host cells are


121

grown overnight in liquid cultures containing a richgrowth medium (e.g., MLB) and adjusted to an opticaldensity measured at 525 nm (OD525) of 0.3–0.5.

2. Approximately 25 mL water sample is filtered (0.2 µm or0.45 µm syringe filters) to minimize the risk of bacterialcontamination of the enrichment cultures (note also herethat the filtration may result in loss of a fraction of thephages present in the original sample).

3. Transfer filtered samples to triplicate sterile 100 mL cultureflasks and add 3 mL 10× growth medium (i.e., 10 timesconcentrated medium).

4. The enrichment cultures are now started by adding 1 mLof each of the target host strains of interest (e.g., 1–6 dif-ferent strains for each incubation) to the culture flask.

5. Incubate the cultures on a shaking table at a temperatureand period that is appropriate for the host bacteria (typi-cally 1–5 d). A control culture is established where theenvironmental sample is replaced by 25 mL artificial sea-water (or 0.02 µm filtered water sample) to verify bacterialgrowth in absence of phages.

We recommend that bacterial growth during incubation isexamined by OD525 measurements, which can give an indica-tion of whether bacterial growth is inhibited by phage lysis. Ifonly a single target bacterial strain is inoculated in the enrich-ment culture, the presence of a lytic phage against that partic-ular strain will often result in clearing of the culture. However,if the lytic potential of the phage is limited, and/or if severalstrains are used, phage lysis may be difficult to detect by visualinspection of the culture, as some strains may be resistant toinfection by the present phages. In that case, even a smalldecrease in OD relative to the control culture may indicate thepresence of lytic phages.1. If phage production is detected (in fact, phage production

may have occurred even if lysis is not detectable byreduced OD values [See “Assessment.”]), the culture istransferred to a 50-mL centrifuge tube and the bacteria arepelleted (10,000g, 10 min).

2. The supernatant is then sterile filtered (0.2 µm or 0.45 µmfiltered) and kept at 4°C until further analysis. A few dropsof chloroform will preserve the sample, however, it alsointroduces the risk of eliminating lipid-containing phages.

3. To verify the presence of lytic phages in the enrichmentculture filtrates, 5–10 µL aliquots of the filtrates are spot-ted on lawns of host bacteria as described above.

Cleared liquid cultures and clearing zones on plates ofimmobilized host bacteria may potentially contain severaltypes of infective phages that were present in the original sam-ple and which propagated in the enrichment culture. Subse-quent steps of isolation and purification are therefore neces-sary to obtain stocks of specific phages (see below).

A more general search for phages, which are lytic to co-occurring bacteria in the water sample, requires a differentprocedure. In that case, 25 mL unfiltered water sample isamended with 1 mL 10× growth medium and incubated

overnight (or longer depending on incubation temperatureand type of target bacteria). During this incubation, lyticphages will potentially propagate by infecting indigenous bac-teria that are favored by the given substrate and incubationconditions. As for the incubations above, phages producedduring incubation are obtained after centrifugation and sterilefiltration.

Concomitant with the enrichment culture incubations,potential bacterial host cells are then isolated after spreading100 µL subsamples of the original water sample on agar platescontaining a growth medium that is similar to the medium inthe enrichment cultures. Single colonies are picked andrestreaked on new agar plates, and subsequently transferred toliquid medium. New phage–host systems can then beobtained by spotting aliquots of filtered enrichment cultureon lawns of bacterial isolates in soft agar, as above, andinspected for clearing zones.

Obtaining pure phage stock: As mentioned earlier, clearingzones on lawns of host bacteria potentially contain all lyticphages against the given host that were present in the originalwater sample. Little is still known about the occurrence anddiversity of phages infecting specific hosts, but generally iso-lation procedures as those described above may reveal multi-ple phages against specific target bacteria (e.g., Comeau et al.2006; Holmfeldt et al. 2007; Stenholm et al. 2008). It is there-fore necessary to further isolate the phages to obtain specificstocks of single phages:1. Transfer phages from the clearing zone on the plate to

1 mL phage buffer or sterile sea water in a sterile tube byscraping off the surface layer of the soft agar containingthe phages using a sterile loop. Alternatively, use a Pasteurpipette to harvest a plug of the soft agar.

2. Allow the phages to diffuse into the medium overnight at4°C.

3. Vortex the tube and centrifuge the sample (10,000g, 10min) to remove bacteria and agar.

4. Transfer the supernatant to a new tube. This sample willtypically contain 106–108 phages mL–1.

5. To isolate single phages, dilutions of this concentrateshould be done, followed by plaque assay, and subsequentisolation of phages from single plaques. Different plaquemorphologies may be selected as an indication of the pres-ence of different phages. Again, the phages are transferredto 1 mL phage buffer, vortexed, and centrifuged, and sub-sequently, the supernatant containing the phages aretransferred to a new tube.

6. Usually, this procedure is repeated 3 times to dilute outany contaminant phage associated with the phage of inter-est and increase the probability that only one specificphage is present in the final phage stock.

7. In the end, the phage concentrate is 0.2 µm filtered andkept in the fridge. If the phage is insensitive to chloroform,preservation with a few drops of chloroform will prolongthe life span of the phage stock. A viability test should,


122

however, be carried out before adding chloroform to thesample. Stocks of specific phages in a buffer can remaininfective for years.

If the relative abundance of individual phages in a phageassemblage obtained from a single clearing zone vary signifi-cantly (more than 10–100-fold), it will be very difficult to iso-late the least abundant types as they would be diluted out inthe attempt to obtain single plaques on the plates. Conse-quently, the method selects for isolation of the dominant frac-tion of lytic phages against a certain host bacterium at thetime of sampling.

Life cycle characterization of bacteriophages—One-step growth experiments: The life cycle of phages can

be characterized by one-step growth experiments, which aredesigned in a way that allows only a single infection cycle totake place (i.e., no re-infections occurring by phages pro-duced during the experiment). Originally developed by Ellisand Delbrück (1939), the one-step growth experiment meas-ures the latent period and the burst size of a given phage ona given host (e.g., Adams 1959; Carlson 2005). Latent periodand burst size are essential parameters in a description ofphage properties and varies between phages and hosts andalso with host growth conditions. The latent period is theminimum length of time it takes from adsorption of thephages to a host cell to lysis of the host with release of prog-eny viruses (Fig. 1). The burst size is the average number ofphages released per infected host cell. The one-step experi-ment can be adapted to test the effects of different environ-mental factors on the infection process. For example, theburst size can be affected by the growth rate of the host (e.g.,Middelboe 2000); they are expected to be higher when thehost cell is nutrient replete and growing exponentially whileone might expect a decrease in burst size if the host cells areunder nutrient limitation.

To limit the phage–host interaction in the experiment to asingle infection cycle, phages and hosts have to be mixed inthe right ratio. Prior to experiment, it is therefore necessary todetermine the titer of the phage stock and to know the rela-tion between cell density and optical density (i.e., obtain cor-responding numbers of cells mL–1 and OD) of the host. Infec-tion should be done at low MOI (multiplicity of infection =ratio of phage to host) e.g., between 0.1 and 0.01. At higherMOI, the probability of cells infected by more than one phagewould increase and the total estimate of infected cellsbecomes less than the phage input.1. 200 µL overnight culture is inoculated in 100 mL culture

flask with 50 mL growth medium (e.g., LB), and incubatedon a shaking table until the density in the culture hasreached cell density of ~5 × 108 CFU mL–1 (correspondingto an OD525 of ~0.3). This may take from a few hours to aday.

2. 1 mL aliquots of the bacterial culture are mixed with sub-samples of the phage stock in triplicate microfuge tubes atan multiplicity of infection (MOI) of approximately 0.01

(i.e., ~ 5 × 108 CFU mL–1 and 5 × 106 PFU mL–1 (final con-centrations), and incubated for 10 min to allow the phagesto adsorb to the host cells. At this point, the infectioncycle of the adsorbed phages is assumed to begin, whichmarks the initiation of the experiment (T = 0)

3. Centrifuge the cells (6000g, 10 min).4. Remove the supernatant (removes unadsorbed viruses)

and resuspend the pellet in 1 mL growth medium (e.g., LB).5. Repeat step 3 and 4 to wash out any further unadsorbed

phages.6. Transfer 50 µL of the resuspended culture (bacteria and

adsorbed phages) to 50 mL growth medium in a 100 mLculture flask and mix well. Assuming that most of thephages have adsorbed to host cells during step 2, the con-centration of adsorbed phages in the 50-mL flask is ~5 ×103 PFU mL–1.

7. Transfer 1 mL to a microfuge tube (note the time) andincubate the triplicate 50 mL cultures on a shaking table.

8. Determine the number of PFU (total infectious centers) byplaque assay in the collected sample (see below).

9. Continue to collect samples for PFU over time for 6–8 hours.It is recommended to carry out a preliminary experiment

with just a few time points over a large time span (6–8 h), toget a first idea of time between adsorption and burst. Thisexperiment should then be followed by a more detailed exper-iment with more frequent samplings (every 10–20 min)around the time when the burst is expected. A successful one-step growth experiment shows a period of constant virusabundance, which reflect the period from when the cell isinfected and until mature phages are released. The latentperiod (Fig. 1) is followed by a single burst of phages fromwhich the burst size can be calculated as the ratio between thenumber of phages before and after the burst (Fig. 1). Thehighly dilute bacterial culture reduces contact rate betweenthe virus and host so that no re-infection will occur during theone-step experiment.

Fig. 1. An example of the development in the number of plaque- formingunits during a one-step growth experiment with a bacteriophage, and thedefinition of the parameters “Latent period” and “Burst size.”


123

One step growth experiments are often difficult to get towork for new phage host systems and adjustments to the stan-dard procedure (e.g., the number of added phages, length ofthe experiment, sampling frequency, etc.) are often requireddepending on host growth rate, phage adsorption rate, infec-tion efficiency, etc.

Isolation of cyanophages—Isolation of cyanophages by liquid bioassays: As mentioned

earlier, there are some basic similarities between the isolationprocedures for bacteriophages and cyanophages. However, thespecific protocols for isolation of cyanophages differ substan-tially from the procedures presented in the previous sectionand are described in detail in the following sections. Figure 2shows possible strategies one could follow, depending on thesuspected titer of cyanophages in the sample of interest. Theseassays assume that the target cells are unialgal and clonal. Ifcultures are not unialgal or clonal, complete lysis of the cul-ture may not occur or plaques could be obscured by contami-nating bacteria. Axenic cultures of the host are preferred forthe plaque assay, but not necessary for liquid assays. Manystrains of marine cyanobacteria can be purchased from culturecollections. Alternatively, new hosts can be isolated from thenatural environment of interest. However, it can take a lot oftime and effort to produce clonal cultures. As well, manycyanobacteria do not grow well on solid substrate. Unless thetarget host of interest is already cultured on solid substrate,the simplest method for isolation of novel cyanophages wouldbe via the liquid bioassay. It is simple, inexpensive, and thehost need not be axenic.

In principle, a small volume of a water sample is added tothe host culture and monitored over time for signs of infec-tion. The treated cultures are compared with control cul-tures by eye for obvious signs of viral infection such as totallysis (clearing) of the culture, decrease or change in overallpigmentation of the culture, or clumping and settling ofcells to the bottom of the culture vessel. This approach hasbeen used to isolate and detect cyanophages from seawateras well as marine sediment samples (Suttle and Chan 1993;Waterbury and Valois 1993; Wilson et al. 1993; Suttle 2000).Multiwell plates (e.g., Corning brand 24- or 96-well poly-styrene plates with lids) are the culture vessel of choice.They are conducive to the screening of many samples,require minimal culture volumes, and take up little incuba-tor space. Glass culture tubes with screw caps (e.g., 13-mmor 25-mm diameter) may be substituted and are useful forscreening larger sample volumes. These glass tubes allownondestructive monitoring of the in vivo chlorophyll fluo-rescence of the cultures using a fluorometer (e.g., TurnerDesigns TD700) or similar. Any lysis of the cells would resultin a decrease in relative fluorescence (rf) compared with thecontrol cultures.

Liquid assays can be used for all aquatic cyanobacteria andfor screening all types of samples, including sediments, andhost cells need not be axenic. Samples to be tested are not sub-

jected to the possibility of elevated temperatures encounteredwhen using plaque assays.

Multiwell plates are convenient for isolating cyanophagesfrom environmental samples using the liquid bioassayapproach (Table 1). Below is a typical procedure (96-well assay)used to detect and isolate cyanophages from marine samplesthat lyses Synechococcus sp. strain DC2 (also known asCCMP1334 or WH7803). Using this method, greater than 105

lytic phages per milliliter of seawater have been detected inthe Gulf of Mexico (Suttle and Chan 1994) that lyses this per-missive target host:1. Collect ca. 50 mL seawater sample in a clean, acid-

washed plastic (HDPE, PP, or PC) container or sterile Fal-con tube.

2. Rinse container 3 times with the sample before filling (ifnot filtering right away, keep sample cold and in the dark).

3. Remove phytoplankton and bacteria from the water sam-ple by filtration.i. Glass fiber filter (e.g., GC50, Advantec); this step may

help to reduce premature clogging of the next filter.ii. Followed by 0.2 µm or 0.45 µm PVDF filter (e.g., Milli-

pore Durapore filters). As for the bacteriophage isola-tion (above), this filtration step may also remove someof the larger viral particles.

4. Store filtered seawater sample in the dark at 4°C (or on ice)until use.

5. Have ready, a culture of host cells in exponential growth(approximately 106 cells/mL).

Fig. 2. A flow chart suggesting various strategies of cyanophage isola-tion depending on type of sample and host characteristics.


124

6. Dilute cells ca. 10-fold with sterile media such as F/2 media(Guillard 1975) (to about 1 × 105 cells/mL).

7. Allow a minimum of 30 mL host cells for every 96-well plate.8. Using a multichannel pipette, aliquot cells into wells,

cover plate with the lid and set aside in the incubator (seeTable 1 for suggested volumes).

9. Warning: make sure that the total volume of host and sam-ple does not exceed the maximum capacity of the wells.There should be ca. 1–2 mm clearance from the top; exces-sive volume would cause overflow of the contents and sub-sequent cross-contamination of the wells.

10. Prepare 10-fold serial dilutions of the seawater sample (e.g.,0.5 mL sample added to 4.5 mL media in a 15-mL Falcontube, up to 5 dilution levels) using sterile media as the dilu-ents. Note: triplicate dilutions series are recommended fordetermining the titer of lytic cyanophages in the sample.

11. Add diluted samples to wells: for example, 50 µL to 16wells (2 rows of 8 wells each) for each dilution level; with5 dilution levels, that would leave 2 rows for the negativecontrols if using 96-well plates.

12. Replace the seawater sample with same volume of mediafor negative controls.

13. Cover plate with lid; carefully seal the lid to the plate usingeither parafilm or thin strips of plastic film.

14. Incubate plates at ca. 25°C, between 10 to 25 µmol quantam–2s–1

15. Compare color development in the wells with the controlwells; clear wells can be discerned from pigmented wells in4 to 7 d.

16. Monitor wells daily for signs of lysis for 10 to 14 d (couldbe longer for slow growing host cells).

17. Choose a clear well from the highest dilution.18. Transfer the lysate to a microtube, centrifuge the lysate for

5 min ca. 12,000g to pellet cell debris.19. Store the supernatant (about 250 µL) at 4°C and use it for

further rounds of purification (via liquid or plaque assay).Tips and Tricks:

1. Use neutral density screens (gray or black window screen-ing) to attenuate the light. Low light levels enhance devel-opment of pigments, which allows for easier discrimina-tion of lysed versus unlysed cultures in the wells.

2. Condensation forming on lids can occur due to tempera-ture changes in drafty incubators. Excessive condensationcan make it difficult to visualize the wells. Sandwich the

full plates between a layer of empty plates to insulate thecultures from temperature shifts.

3. Sealing the lid to the bottom of the plate helps to slowdown evaporation of well contents, particularly the oneslocated at the plate perimeter.

Isolation of cyanophages by liquid enrichment assay: If lowtiters are expected, the viruses in the sample can be concen-trated via TFF to make a virus concentrate (VC) (Suttle et al.1991; Wommack et al. 2010, this volume). To increase thedetection limit, several different VCs can be combined andadded to the same culture. Another convenient way to screenlarger sample volumes is to perform liquid enrichment cul-tures (Suttle 1993). By this approach, larger volumes of watersamples can be screened for cyanophages, thus enablingdetection of “rare” viruses. Similar to the use of enrichmentcultures for bacteriophages, the disadvantages include the factthat lysis of the host cells are not always obvious, especially ifthe initial titer is low or if the host culture is not clonal orunialgal. Also, more steps are required to dilute out nonrepli-cating viruses to obtain pure clonal isolates. Because it is anend-point dilution assay, only the most abundant phages willbe isolated. The principle is the same as for the isolation ofbacteriophages mentioned earlier except that the sample vol-umes screened are in the order of liters instead of milliliters.Here is a typical procedure for cyanophage liquid enrichmentcultures.1. Prefilter at least 3 L water sample through a glass fiber fil-

ter (e.g., Advantec type GC50, Whatman type GF/C, orGelman type A/E), followed by a 0.2 µm or 0.45 µm lowprotein binding PVDF filter (e.g., Millipore Durapore filter).

2. Dispense the filtered water samples (e.g., 0.5 L or more)into culture vessels, e.g., 1 L or larger Erlenmeyer flasks.

3. Add nutrients to the filtered water to support growth ofthe target cells. For example, F/2 nutrients for marinecyanobacteria or BG-11 nutrients (Rippka et al. 1979) forfreshwater cyanobacteria.

4. Seed the filtered water with 1% to 10% v/v of host culture.Target cells must be in exponential growth to avoid loss ofpotential viral infection (e.g., viruses adsorbed to dead ordying cells will not cause infection and subsequent pro-duction of progeny virus).

5. As a control, replace the filtered environmental samplewith virus-free (0.02 µm filtered or heat-killed) water sam-ple. The volume of this culture need not be as large as theexperimental flasks. This control is to make sure that thereis not anything in the water sample that would inhibitgrowth of the target cells (e.g., chemical inhibition).

6. Incubate the flasks at the temperature and light conditionsappropriate for the cyanobacteria and look for signs oflysis. This could take 2 to 3 weeks, depending on thegrowth rate of the host as well as the initial titer ofcyanophages. It is recommended that the in vivo chloro-phyll fluorescence be monitored regularly. A small

Table 1. Suggested volumes of target cells and sample to usefor the liquid bioassay.

Cell Sample Max. total Culture vessel volume volume volume

Plate, 96-wells 200–250 µL 50–100 µL 300 µL

Plate, 24-wells 2.5–3 mL 0.2–0.5 mL 3.5 mL

Tube, 13 × 100 mm 3.5–4 mL 0.5–1 mL 4.5 mL

Tube, 25 × 150 mm 30–35 mL 2–10 mL 40 mL


125

decrease in relative fluorescence could indicate the pres-ence of a lytic virus.

7. Remove an aliquot of the enrichment culture, and pelletremaining cells by centrifugation (e.g., 20 minutes at6000g).

8. Filter the supernatant through a 0.22 µm or 0.45 µm PVDFfilter, and store the lysate at 4°C until further analysis.

9. Verify the presence of lytic phages by liquid assay (orplaque assay).

To propagate/amplify the lytic agent, the liquid bioassay isrepeated using the putative lytic agent as the test sample:1. Set up bioassay using 5 mL or larger culture tubes in tripli-

cates. Add between 5 to 50 µL of each sample below to tar-get cells in log phase.a. Whole lysate (unfiltered)b. Filtered lysate (0.22 or 0.45 µm)c. Negative control (no addition, or use filtered media)

2. Monitor in vivo chlorophyll fluorescence for about 1 week,look for decrease in relative fluorescence compared withcontrol cultures.

3. If the cultures lyse, then the lytic agent is most probably avirus.

4. Propagate the lytic agent several times to dilute out non-replicating viruses.

5. Filter the lysate and use it to obtain pure clonal stocks (seebelow).

Isolation of cyanophages by plaque assays: More than 40years ago, Safferman and Morris (1963) used the plaque assaymethod to isolate the first cyanophage that infects a freshwa-ter filamentous cyanobacterium, Plectonema boryanum. Sincethen, this approach has been used successfully to detect andisolate a number of different phages infecting marine Syne-choccocus and Prochlorococcus (e.g., Suttle and Chan 1993;Waterbury and Valois 1993; Wilson et al. 1993; Sullivan et al.2003).

The advantages of this method are that results are easilyinterpreted as plaques formed on pigmented lawns can be eas-ily identified. Since a plaque is the result of a single infectionevent, the virus can be easily purified and cloned. The disad-vantages include the following: The target cells must be ableto grow on solid media; small bacteria that pass through thefilter can interfere with lawn formation of slow growingcyanobacteria while some bacteria can also cause plaques oncyanobacteria lawns. The higher temperature of the moltensoft agar can inhibit or destroy temperature sensitive virusesor inhibit growth of the host, and the sample volume that canbe tested is limited (<1 mL). The first part of the proceduresdescribed as follows: (1) preparation of base plates and (2)preparation of top agar/agarose are applicable also for bacte-riophage plaque assays.1. Prepare base plates: For example, purified agar or agarose

(1% w/v) is added to your media of choice and autoclaved.This will provide a support base for the top agar/agaroseoverlay as well as nutrients for the host cells. For best

results, use plates within 1 week of pouring.a. Add 5 g purified agar or agarose to 500 mL culture

media in a 1-L Erlenmeyer or media bottle.b. Gently stir to disperse the agar/agarose.c. Autoclave for 20 to 25 min to sterilize.d. When cooled to about 60°C, dispense 15 to 20 mL per

plate.e. To reduce condensation forming on the insides of the

lids, leave lids slightly ajar to allow escape of steam orstack the plates immediately after pouring

f. Invert plates once the agar has solidified to prevent con-densation from dripping onto the surface of the agar.

g. Plates can be used about 12 h after pouring if the agarsurface is not wet; a longer time is needed if conditionsare humid.

h. Warning: If the surface of the bottom agar is too moist,the top agar/agarose will not stick to the bottom plateand will slide off when the plate is inverted.

i. Tip: plates can be fast-tracked: dry plates at 37°C; leavelids slightly ajar; monitor closely to prevent over drying.

Considerations: Depending on the composition of the mediaused, the addition of solidification agents (in particular thecombination of high salinity seawater-based media and com-mon agar such as Bacto Agar) can often result in the formationof precipitates when autoclaved together. These “flocks” cansometimes interfere with interpretation of the plaque assay.Moreover, impurities in common agar can negatively affectthe growth of the host cells. Here are some suggestions onhow to reduce the formation of these precipitates. Some test-ing may be required to determine the best combination to usefor your particular situation.

a. Do not use common agar; rule of thumb—the whiterthe agar, the “cleaner” it is.

b. Use commercially available purified agar or agarose; orclean common agar using a washing procedure such asthe one outlined in Waterbury and Willey (1989).

c. Reduce the salinity of seawater media with purifiedwater; e.g., to 20–25 psu.

d. Add purified agar or agarose to autoclaved media asep-tically and then melt the agar/agarose in themicrowave (bring to a short boil 2–3 times to com-pletely dissolve the agar/agarose).

e. For cells that will grow in artificial media, prepare mediaand gelling agent at 2× concentration and autoclave sep-arately. When cooled to ca. 60°C, gently mix the gellingagent into the media and dispense immediately.

f. In the case of artificial media, add agar/agarose to fil-ter-sterilized media and melt the gelling agent in themicrowave

2. Prepare top agar/agarose: Prepare 100 mL portions of 0.4 to0.5% (w/v) of purified agar, agarose or low-melting point(LMP) agarose (i.e., Invitrogen #15517-022) in your mediaof choice. Although LMP agarose can be quite expensive, itis recommended for temperature sensitive samples and


126

cells, since it solidifies at ca. 25°C. Purified agars, as well aslow-melting point agars and agaroses are available for arange of lower temperatures (consult the following web-sites for more details: www.sigmaaldrich.com andwww.invitrogen.com)a. Autoclave or microwave sterilize on the day of the assay.b. Dispense 2.5 to 3 mL into 13-×-100–mm glass dispos-

able culture tubes (Fisher Scientific #1496127).c. Transfer tubes to a water bath or dry heat block set at

the appropriate temperature, (e.g., ca. 30 to 32°C forLMP agarose, ca. 40–42°C for purified agar or agarose).

d. Cover tubes with foil or cap, allow for temperature toequilibrate.

e. For each water sample, prepare triplicate tubes; controltubes containing cells are only used to monitor lawngrowth.

f. For best results (smooth lump-free top agar), use freshlyprepared top agar/agarose since repeated re-melting ofsolidified agar/agarose can give inferior results.

3. Prepare target (indicator) cells: Grow the cyanobacteria inliquid media, harvest in exponential growth and adjustcell density to about 107 to 108 cells/mL. If necessary, cellscan be concentrated by gentle centrifugation and resus-pended in media. Preliminary testing may be required todetermine the best cell density to use for your particularhost organism. The objective is to start with a lawn of cellsthat will have the capacity for additional growth duringthe length of the assay. Depending on the growth rate ofthe target cells, one can expect plaques to appear on thelawn as early as 3 to 4 d to weeks after infection. The ini-tial lawn of cells will be very faint in color. However, thelawn will develop into an evenly distributed dense layer ofcells within 7 to 10 d. If the lawn is too thin, plaques willgo undetected. If the lawn is too thick, the cells could runout of nutrients prematurely which may result in poorlydeveloped plaques.

4. Prepare the sample: Environmental samples should be pre-filtered as described earlier. If high titers are expected,serial dilutions of the sample may need to be performed.

5. The assay:a. Adsorb 50 to 100 µL sample (as is, and 10-fold serial

dilutions, up to ca. 3 levels) to 0.5 mL target cells underthe usual culturing conditions (e.g., for Synechococcussp. strain DC2, constant 5–25 µmol quanta m–2s–1, at25°C), agitate occasionally to encourage adsorption ofphage to host.

b. After 1 h, transfer virus: host mixture to 2.5 mL softagar. Quickly and gently vortex the mixture and pourthe entire tube contents onto the surface of the agarplate. Working rapidly, gently rock and swirl the plateto spread the mixture evenly onto the plate surfacebefore the agar starts to gel. Set aside on a flat surfaceto harden (about 1 h). For best results, the total volumeof cells + virus + soft agar is between 3 to 4 mL. Larger

volumes would make it easier to pour, but is not rec-ommended as the top layer would be too thick, andplaques could form on top of one another.

c. Prepare a control plate containing only cells; this platewill allow you to monitor cell growth.

d. Seal plate with parafilm, flip plates upside down. Incu-bation of plates under constant low light conditions (5to 25 µmol quanta m–2s–1) will produce darker lawnsthus allowing for easier detection of plaques. Plaqueswill appear within 1 to 2 weeks, depending on thegrowth rate of the host cells.

e. Note the number of plaque forming units (PFUs),plaque size, and morphology.

f. Choose a well-isolated plaque on a plate that containsless than 100 PFUs.

g. Harvest the plaque using a Pasteur pipette: gently pressthe tip of the pipette into the plaque to the bottomagar; using gentle suction, remove the plug.

h. Transfer the plug to 1 mL media and vortex briefly tobreak it up.

i. Place the tube at 4°C and allow the phages particle toelute from the plug overnight to form a plaque lysate.

j. Vortex and centrifuge the sample (ca. 12,000g for 10min) to pellet cyanobacteria and agar.

k. Transfer the supernatant to a new tube; typical titer ofthe plaque lysate can be 104 to 105 PFU mL–1.

l. Repeat steps f to k for a minimum of 3 plaques. Choos-ing plaques with different morphologies may result inthe isolation of different phages.

Obtaining pure cyanophage stocks (liquid assay):1. Determine the titer of the lysate using the 96-well assay as

described earlier.a. Prepare end-point dilution series (10-fold serial dilu-

tions, 5 to 6 levels).b. Monitor plates for lysis every few days, recording the

number and position of clear wells on the plate.c. When clear or nearly clear wells no longer appear for 1

week, record the final “score” for each dilution level.d. Use the MPN Assay Analyser program (Passmore et al.

2000) to determine the most-probable-number (Taylor1962) of infective phages in the lysate.

2. Once the cyanophage titer is determined for the stock tube,proceed to purify a clonal virus:a. Use 13-×-100–mm culture tubes (or 24-well plates).b. Prepare exponentially growing target cells (e.g., >100 mL).c. Dilute some of the titered lysate to 1 infective virus/mL.d. Add 0.2 mL (0.2 infectious units) to each of 20 tubes of

susceptible host cells.e. Monitor tubes for 2 to 3 weeks.f. Cultures in which lysis occurs are assumed to be the

result of a single-virus infection; the probability thatmore than 1 infective unit occurred in a given cultureis 0.0176.

g. If lysis occurs in 4 tubes or less of 20, it is assumed that


127

lysis in each tube was caused by one infectious unit,therefore each tube would contain a separate phageclone.

h. Propagate an aliquot from all the tubes to confirm theresults.

i. If lysis occurs in more than 4 tubes, repeat the cloneout procedure by reducing the volume of dilutedlysate added to the 20 tubes (e.g., add 0.1 mL insteadof 0.2 mL)

j. Scale up each phage clone to make primary phagestocks; e.g., add 5µL of the lysate to 40 mL of cells

k. Centrifuge, filter, and titer the stock, store at 4°C in thedark.

Obtaining pure cyanophage stocks (plaque purification):1. Make a dilution series of the lysate (assume 104 to 105 PFU

per mL in the plaque lysate), and use this to perform a sec-ond round of plaque assays (steps 1 to 11) to purify thephage.

2. Repeat the plaque purification procedure 2 more times toensure that the cyanophage isolated is clonal.

3. Finally, prepare a primary cyanophage stock using lysatefrom the final purification via one of the following methods:i. Liquid amplification: add some of the lysate to target

host in liquid culture. After the culture has lysed,remove cell debris via centrifugation, filter sterilize thestock, and store at 4°C until further analysis.

ii. Plate amplification: Prepare plaque assays (see above)with a dilution series of lysate from the final purifica-tion. Plates with confluent lysis of the host lawn (typi-cally ca. 104 PFUs) can then be used to obtaincyanophage stocks by elution of phages from theplates. Add 5 mL sterile seawater to the plate, scrape offthe top agar layer into the seawater, and leave at 4°Covernight. Remove agar and cell debris by centrifuga-tion, filter sterilize the stock and store at 4°C until fur-ther analysis.

4. Titer the final stock via plaque assay.5. Cyanophage stocks stored at 4°C in the dark are stable for

at least a year.Life cycle characterization of cyanophages—Adsorption of phage to cyanobacteria: The first step in the

life cycle of a virus is adsorption to host cells. If the virus doesnot adsorb to a viable host, infection will not take place.Where bacteriophage adsorption rates are often in the order ofminutes, cyanophages usually adsorb to their host cells at amuch slower rate; and not all contacts result in an infection.Suttle and Chan (1993) found that it took 45 min for 80% ofcyanophage BBC1-P1 to be adsorbed to its host. Some factorsthat can affect adsorption rates are (1) host abundance (a min-imum of 104 cells per mL are needed to have a sufficientlyhigh rate of adsorption); (2) physiological state of the host(which could affect the availability of receptor sites); (3) phys-ical environment (temperature, viscosity); (4) chemical envi-ronment (ions, salts, co-factors); (5) light (adsorption rate is

light dependent in some species (Clokie et al. 2006); and (6)host strain. If every contact results in an infection, this wouldhave tremendous ecological impact. Understanding how thesefactors affect the rate of virus adsorption would help onedesign better experiments, interpret data, and construct bettermodels for virus–host cell relationships.

Several methods have been presented by Adams (1959) todetermine adsorption kinetics of bacteriophages. One way tomeasure the adsorption efficiency of cyanophages is to assayfor free unadsorbed phages (Suttle and Chan 1993). The prin-ciple of this assay is to add a known quantity of viruses to hostcells (e.g., at an MOI = 0.01 to 0.1). Over a period of 1 to 2 h,small subsamples of the virus:host solution are removed ca.every 15 min and diluted 100-fold to stop further adsorption.The host is then separated from free viruses by centrifugation,and the number of free viruses remaining in the supernatantdetermined by plaque or end-point dilution assays. A plot ofthe abundance of free viruses remaining in solution as a func-tion of time should produce a straight line. The slope of thisline is then used to calculate the adsorption rate.

Under constant environmental and cultural conditions ofthe host cell, the rate of adsorption can be described using thefollowing equation:

K = 2.3/(B)t × log (p0/p)

where B = concentration of cyanobacteria (cells mL-1), p0 =phage assayed at time zero, p = phage not adsorbed at time t(min), K is the velocity constant (ml min-1).

Because of the short sampling intervals, it is very importantto have everything ready before you begin. Having mediaequilibrated to the appropriate temperature, pre-labeled agarplates and tubes, tables to record the time etc will facilitateacquisition of better data.

This procedure can be adapted for other host–virus systems.The example given is for host BBC1 and cyanophage BBC1-P1(Suttle and Chan 1993). 1. Have everything ready to perform plaque assay (allow for

8 time points, in triplicate).a. Plating cells, aliquoted, and set aside.b. Bottom plates labeled (24 + plates).c. Top agar aliquoted and temperature equilibrated.d. Dilution tubes—these contain 1.5 mL media, labeled

and kept on ice.e. Cyanophage stock diluted into 1–5 mL media.

2. Set up a table to record times such as Table 2.3. Set up adsorption cultures (e.g., 250 mL polycarbonate

Erlenmeyer flasks with screw cap). Cyanobacteria shouldbe in exponential growth, e.g., for BBC1, about 106 to 107

cells/mL (The actual numbers should be predetermined bymicroscopy).

4. Fill flask with 100 mL host cells.5. Add cyanophage stock of known titer to host at an MOI of

ca. 0.01 and quickly mix to disperse the virus. For exam-ple, for a host concentration of 1 × 107 cells mL–1 (i.e., total


128

number in 100 mL = 109 cells), a total of 107 infectiousviruses is needed to achieve an MOI = 0.01. Thus, if thevirus stock is 1 × 109 infectious units mL–1, add 0.01 mL tothe host culture. If the virus stock is highly concentrated,we recommend diluting the virus into a larger volumebefore adding to the host cells. This will enhance rapid dis-persal of the viruses.

6. Immediately remove a subsample and dilute 100× for timezero: Transfer 15 µL to a tube containing 1.5 mL of ice coldmedia, vortex to mix, pellet host for 5 min at ca. 16,000gand 4°C; note the time.

7. Carefully remove a small aliquot (50 µL) of the super-natant to a new tube and keep cold for plaque assay; notetime.

8. Place adsorption cultures under usual conditions (e.g.,light and temp)

9. Repeat sampling at 15 min intervals for 1 to 1.5 h.10. Determine the concentration of viruses remaining in the

supernatant for each time point by plaque assay.One-step growth experiments (cyanophages): The proce-

dure is similar to that for bacteriophages with one major dif-ference. Where bacteriophage growth is measurable in theorder of minutes, cyanophage growth curves are measured interms of hours. The burst sizes are similar, being in the tens tohundreds.

As for bacteriophages, infection should be done at a MOIbetween 0.1 and 0.01. In this instance, the total infective cen-ter = the phage input because the proportion of multiply-infected host cells is small. At higher MOI, the probability ofcells infected by more than one virus would increase and thetotal infective centers (TIC = total number of infected cells +free viruses) become less than the phage input.

Described below is the procedure used to perform a one-step growth curve for cyanophage BBC1-P1 via plaque assayon Synechococcus sp. BBC1 (Suttle and Chan 1993).

Procedure—1. Set up triplicate adsorption tubes (AT) in 1.5 mL micro-

tubes: Add phages of a known titer to the cyanobacteriahost in exponential growth (cell concentration determinedby microscopy; Cyanoinput) at an MOI of approximately

0.02 (e.g., 0.9 mL hosts [1 × 107 cells mL–1] + 0.1mL phages[2 × 106 PFUs mL–1] = MOI of ~0.02).

2. Allow the phages to adsorb to the hosts for 60 min. atroom temperature and an illumination of 25µmol quantam–2s–1. Flick tubes a couple of times at 30 min.

3. Set up control tubes (CT) as above, excepta. positive controls: replace host cells with media (this

gives input phage numbers).b. negative controls: omit virus.

4. After 60 min, remove unadsorbed phages (T = 0).a. centrifuge briefly to pellet host cells (e.g., 5 min,

16,000g, 4°C).b. remove supernatant, resuspend cells in fresh media.c. repeat washing step.d. assay washes to determine the number of unadsorbed

phages (Free PhageT=0) for calculation of efficiency ofadsorption.

5. Prepare nine 15-mL centrifuge tubes containing 10 mLalgal growth media (6 labeled “FGT” [first growth tube]and 3 labeled “SGT” [secondary growth tube]).

6. Add 100 µL from AT to FGT (10–2 dilution). Mix andremove 0.1 mL sample for T = 0. Perform plaque assayimmediately to determine total infective centers (TICT=0).

7. Add 100 µL from FGT to SGT (10–4 dilution).8. Incubate FGT and SGT tubes at room temperature and ca.

25 µmol quanta m–2s–1.9. Remove 100 µL samples every 3 h from FGT (at 3, 6, 9, 12,

16 h) and SGT (from 12 h onwards) for ca. 30 h and deter-mine the TIC via plaque assay.

10. Add 100 µL from positive control tube (CT) to FGT-control(10–2 dilution) and determine the total phage input(Phageinput).

The dilution factor in going from the adsorption tube toFGT is chosen so that a reasonable number of plaques (ca. 50-200) will form on the plates during the latent period. The dilu-tion factor in going from the FGT (first growth tube) to SGT(second growth tube) is chosen on the basis of the expectedincrease in plaques at the end of the rise period so that laterplatings of SGT samples will also yield countable numbers. Asmentioned in the bacteriophage section, it is recommended to

Table 2. Suggested tabler for recording sampling times and details for adsorption kinetics experiments.

Time point (min)

Replicate Time: subsampled from tube

Time:centrifuged

Time: tubeplaced on ice

Time: added toplating cells

Volumetitered

Results: # ofPFUs

Average #of PFUs

T=0 A

B

C

T=15 A

B

C

Etc.


129

perform a preliminary one-step experiment to estimate thepossible length of the latent period and burst size. Then repeatthe experiment with more frequent sampling with adjust-ments to the dilution factors to gain more precision. Samplingtimes may need to be adjusted for each phage–host system.

In summary:1. Determine total phage input (Phageinput) and total

cyanobacteria input (Cyanoinput).2. At T = 0 determine the titer of unadsorbed phages (Free

PhageT=0) and total infective centers (TICT=0 = initial totalinfected cells).

3. During the latent period and rise, determine the titer oftotal infective centers (TIC = total infected cells + anyphages released).

4. At the end determine the titer of total progeny (TICfinal).Calculations.Adsorbed phages = Phageinput – Free PhageT=0

Percent adsorption = Adsorbed phages/ Phageinput × 100Average MOI = Adsorbed phages/Cyanoinput

Burst size = (TICfinal/TICT=0)

AssessmentDetection limit of the methods—In principle, the presence of

a single infective virus in an enrichment culture or in a sam-ple that is spotted on a bacterial or cyanobacterial lawn shouldbe detected by the proposed methods, as a single infectiontheoretically is sufficient to initiate propagation of a givenvirus, which could then be isolated. However, not all infec-tions are successful, and infectivity depends on a suite of con-ditions, such as the encounter rate between viruses and theirhosts, the adsorption rate, host susceptibility, host growthconditions, virus decay rates etc; all factors which may deter-mine whether a given virus will propagate in a given hostcommunity. Moreover, there exists, to our knowledge, no sys-tematic study on the efficiency and detection limit of the cur-rent methods for isolating environmental phages.

We have done a series of experiments to evaluate the effi-ciency of the spot assay and the enrichment culture approachin detecting low densities of specific bacteriophages in a sam-ple. Obviously, the enrichment approach has the advantage ofthe ability to screen a large volume of sample for viruses, buthere we also wanted to test whether there are systematic dif-ferences in the two methods’ ability to detect the presence ofa given amount of viruses spotted to a lawn or added to anenrichment culture, respectively.

A dilution series of a stock of the specific phage (Vir#12)infecting the marine Cellulophaga sp group (Holmfeldt et al.2007) was performed and the number of infectious units wasdetermined in triplicate in each dilution by both plaque assayand spot assay on lawns of Cellulophaga sp. #12. The numberof PFU in the Vir#12 dilutions ranged from 2.9 × 104 ± 1.4 ×103 PFU mL–1 (10–2 dilution) to 30 ± 60 × 102 PFU mL–1 (10–5

dilution) obtained in the plaque assay and from 9.3 × 102 ±3.5 × 102 (10–3 dilution) to no clearing observed (10–5 dilution)

in the spot assay (Table 3). From these dilutions, a series ofdilution culture experiments were established in duplicate 50mL batch cultures with MLB medium inoculated with 10 µL ofeach of the dilutions (Table 3), as well as a positive controlwith undiluted virus stock and a negative control withoutaddition of viruses. This corresponded to the addition of arange of infectious units from 0 PFU (negative control) to~300 PFU (10–2 dilution) and ~30,000 PFU (positive control),corresponding to initial phage concentrations in the enrich-ment cultures of 0 to 6 PFU mL–1 (Table 3).

Two sets of experiments were set up in duplicate: In Exper-iment 1, the phage dilutions were added together with 1.5 mLof an overnight culture of the host bacterium Cellulophaga #12(originally used to isolate Vir#12), corresponding to a start celldensity of approximately 1.5 × 107 cells mL–1. In Experiment 2,the phage dilutions were added to a mixture of 3 Cellulophagasp strains in equal densities (Cellulophaga #12, Cellulophaga #3,and Cellulophaga NN16038). As for Experiment 1, the initialtotal bacterial density was 1.5 × 107 cells mL–1, however thetwo new strains had reduced susceptibility to Vir#12 relativeto Cellulophaga #12 (not shown). In Experiment 1, 10 µLundiluted Vir#12 stock (approximately 30,000 PFU) wasadded to another culture of the host bacterium as a positivecontrol to be certain to see the effect of viral addition on ODmeasurements.

Samples were collected every 3-8 h during the 47-h incuba-tion for OD measurements and for detection of phages by spotassay. For the spot assay, 10 µL sample was spotted on a lawnof Cellulophaga #12, and incubated for at least 24 hours fordetection of a clearing zone.

The results from the spot assay showed that in Experiment1, which contained only the most susceptible host, phageswere propagating relatively fast. Here phages reacheddetectable numbers already after 3 h in cultures with an initialphage concentration of 6 PFU mL–1 and after 11 h with initialconcentrations of 0.6 and 0.06 PFU mL–1 (Fig. 3). This meansthat the phage concentration had increased from ~6 × 10–2

PFU mL–1 to > 100 PFU mL–1 (i.e., corresponding to >1 phagein the 10 µL spotted) in 11 h in the culture where 3 phageshad been added initially (Fig. 3). Addition of 30,000 PFU (thepositive control) had a significant controlling effect on thebacteria, and after 9 h, no net increase in OD was observed inthe culture (Fig. 4). In Experiment 2 with a combination ofhosts with variable susceptibility to the phage, spot detectionwas first observed in cultures with initial phage concentra-tions of 0.6 and 0.06 PFU mL–1 after 17 h incubation (Fig. 3).In neither experiment did an initial phage density of 0.006PFU mL–1 result in any systematic phage production, and con-sequently, the phage did not build up significant populationsin these cultures during the incubation.

Overall, the data showed that a concentration of approxi-mately 0.06 phages mL–1, corresponding to the presence of 3phages in enrichment cultures with potential host cells wassufficient to detect the phage by the enrichment assay, and


130

subsequently, isolate it by spot assay. The detection limit was,in this case, independent on host strain composition; how-ever, the faster propagation of phages in Experiment 1 indi-cated that the composition and susceptibility of potential hoststrains in the culture may influence the detection limit ofviruses for other virus–host systems. Similar, for the spot assayapproach, it was shown that the 10–4 dilution (3 phages in 10µL) resulted in positive reactions in the spot assay, whereas10–5 dilution did not produce a clearing zone (Table. 3). Con-sequently, the two methods had very similar detection limitsand were in principal in both cases capable of detecting a sin-gle infectious unit.

Interestingly, viral proliferation in the enrichment culturesas verified by the spot assay did not result in a reduction in ODin all cases (Fig. 4). In fact, in Experiment 2, only the highestinitial virus concentration (6 PFU mL–1) resulted in a signifi-cant OD reduction, whereas in Experiment 1, initial concen-trations as low as 0.6 PFU mL–1 affected OD in cultures relative

to the control (Fig. 4). These results clearly demonstrate thepoint made above, that OD measurements are not necessarilysufficiently sensitive to detect the presence of viruses in enrich-ment cultures and that viruses therefore may be isolated fromcultures even without observations of a reduction in OD. Thisis particularly the case if the host cells only have a low suscep-tibility to the phages in the sample or if a mixture of host cellsis used for viral isolation. However, we recommend OD mea-surements as a rapid way of getting a first indication aboutwhether cell lysis is occurring in the enrichment cultures.

DiscussionAn essential property of a virus isolation procedure is which

fraction of the total viral assemblage that is targeted by theprocedure and what is the detection limit of the procedure(i.e., how many of a specific virus are needed in the originalwater sample to detect it by the various methods). Varioustypes of studies may require different knowledge about the

Positive spot registration

Time (h)

0 3 11 17 20 23 26 29.5 33.5 45 48

Exp. 1, Pos. control nd nd nd nd

Exp. 1, 6 PFU/mL

Exp. 1, 0.6 PFU/mL

Exp. 1, 0.06 PFU/mL

Exp. 1, 0.006 PFU/mL

Exp. 2, 6 PFU/mL

Exp. 2, 0.6 PFU/ml

Exp. 2, 0.06 PFU/mL

Exp. 2, 0.006 PFU/mL

Fig. 3. Results from the 10 µL spot assays performed over time in Experiment 1 and 2 to detect the presence of viruses in the dilution cultures. Whitesquares indicate that plaques were not detected, gray squares indicate positive plaque formation (horizontal lines indicate that plaques were onlyobtained in one of the replicate cultures). nd, not done.

Table 3. The number of infectious units obtained in a dilution series of a specific phage (Vir#12) by plaque assay and spot test, respec-tively. From these dilutions, 10 L were added to a series of dilution culture experiments in duplicate 50 mL batch cultures with the hoststrain (Cellulophaga #12) corresponding to the addition of a range of infectious units from 0 PFU (negative control) to ~300 PFU. Thus,the initial phage concentrations in the enrichment cultures ranged from 0 to 6 PFU mL–1 and a positive control containing 600 PFU mL–1.

Number of infectious phages (PFU mL–1)Dilutions 100 (Positive control) 10–2 10–3 10–4 10–5 Negative controlPlaque assay 2.9 × 104 2.7 × 103 280 30 0

Spot assay 9.3 × 102 70 0 0

Initial phage concentration in enrichment cultures 600 6 0.6 0.06 0.006 0


131

actual targets of the isolation procedures. If the purpose is sim-ply random isolation of heterotrophic virus-bacteria systems,the methods described here are usually quite efficient with asuccess rate of 10-50% (i.e., it is usually possible to isolate lyticviruses against 10-50% of culturable bacteria isolated from thesame water sample). However, if one is looking for, or wants toquantify, naturally occurring viruses against specific bacterialhosts, it is important to know the detection limit of the pro-cedure. In that context, it is relevant to know the recovery effi-ciency of the method, i.e., the extent of loss of phage particlesand/or phage infectivity during the various steps in the con-centration procedures, etc.

The first and most obvious limitation of the presented iso-lation procedures is that the method only works with bacteriaand cyanobacteria that are culturable. By far, most of the workon viral isolation has been carried out with aerobic het-erotrophic bacteria and cyanobacteria, culturable both on agarplates and in enriched liquid cultures. In principle, however,it is possible to isolate viruses under anaerobic conditions orfor bacteria that cannot grow in enriched cultures or on plates(e.g., the SAR 11 cluster), however, this requires modificationsand adaptations of the standard protocols described here.Consequently, the methods presented here may need adjust-ments to the type of hosts or phages that are targeted. A sec-ond limitation is that the procedure selects for lytic viruseswhereas isolation of temperate viruses is less straightforwardand requires quite different techniques to obtain in culture(e.g., Dillon and Parry 2007).

It is, therefore, important to be aware that the method doesnot provide a general screening for lytic phages but rather ascreening for phages infecting specific target hosts, or in thebroader approach, hosts that are culturable under the givenset of growth conditions and present in significant numbersand activities to allow propagation of a viral population. Also,the isolation procedure introduces a competition betweeninfective phages for the applied host, and will therefore favorspecific viruses (e.g., broad host range phages, phages withhigh affinity for the host, high burst size, etc.), at the expenseof low-efficiency viruses, and will therefore not necessarilyprovide a representative selection of the viruses that are ableto infect a given target bacterium.

The methods for characterization of adsorption kineticsand life cycles of phages are tedious and often require someadaptation for individual phage–host systems. However, wehave presented some basic procedures that are known to workon certain types of phages and may function as a startingpoint in the development of more specific procedures forcharacterization of phage–host systems of interest.

ReferencesAdams, M. H. 1959. Bacteriophages. Interscience Publ.Berg, G. 1987. Methods for recovering viruses from the envi-

ronment. CRC Press.Bergh, Ø., K. Y. Børsheim, G. Bratbak, and M. Heldal. 1989.

High abundance of viruses found in aquatic environments.Nature 340:467-468.

Brussaard, C. P. D., and others. 2008. Global scale processeswith a nanoscale drive—the role of marine viruses. ISME J2:575-578.

Carlson, K. 2005. Appendix: Working with bacteriophages:Common techniques and Methodological approaches, p.437-494. In E. Kutter and A. Sulakvelidze, [eds.], Bacterio-phages. Biology and applications. CRC Press.

Clokie, M. R. J., A. D. Millard, J. Y. Metha, and N. H. Mann.2006. Virus isolation studies suggest diel cycles in viralabundance in natural cyanophage population of the Indian

Fig. 4. Measurements of optical density (OD525) in dilution cultures ofhost strains after addition of different numbers of the lytic phage #12(range 0-6 PFU mL–1. Experiment 1: Dilution cultures contained only thehighly susceptible host strain Cellulophaga #12. Experiment 2: Dilutioncultures contained a mixture of 3 Cellulophaga sp. strains in equal densi-ties (Cellulophaga #12, Cellulophaga #3, and Cellulophaga NN16038) ofwhich two strains (Cellulophaga #3 and Cellulophaga NN16038) hadreduced susceptibility to the phage relative to Cellulophaga #12.


132

Ocean. J. Mar. Biol. Assoc. U.K. 86:499-505.Comeau, A. M., A. M. Chan, and C. A. Suttle. 2006. Genetic

richness of vibriophages isolated in a coastal environment.Environ. Microbiol. 8:1164-1176.


Dillon, A., and J. D. Parry. 2007. Characterization of temperatecyanophages active against freshwater phycocyanin-richSynechococcus species. Freshw. Biol. 53(6):1253-1261.

Eisenstark, A. 1967. Bacteriophage techniques, p. 449-524. InK. Maramorosch and H. Koprowski [eds.], Methods in virol-ogy. V. 1. Academic Press.

Ellis, E. L., and M. Delbrück. 1939. The growth of bacterio-phage. J. Gen. Physiol. 22:365-384.

Fuller, N., W. H. Wilson, and N. H. Mann. 1998. Occurrence ofT4 gp20 homologues in marine cyanophages and theirapplication to PCR-based detection and quantificationtechniques. Appl. Environ. Microbiol. 64:2051-2060.

Guillard, R. R. L. 1975. Culture of phytoplankton for feedingmarine invertebrates, pp. 29-60.. In W. L. Smith and M. H.Chanley [eds.], Culture of marine invertebrate animals.Plenum Press.

Hennes, K., and C. A. Suttle. 1995. Direct counts of viruses innatural waters and laboratory cultures by epifluorescencemicroscopy. Limnol. Oceanogr. 40:1050-1055.

Holmfeldt, K., M. Middelboe, O. Nybroe, and L. Riemann.2007. Large variability in host strain susceptibility andphage hot range govern interactions between lytic marinephages and their Flavobacterium hosts. Appl. Environ.Microbiol. 73:6730-6739.

Mann, N. H., A. Cook, A. Millard, S. Bailey, and M. Clokie.2003. Bacterial photosynthesis genes in a virus. Nature424:741.

Middelboe, M. 2000. Bacterial growth rate and marine virus-host dynamics. Microb. Ecol. 40:114-124.

———. 2008. Microbial disease in the sea: Effects of viruses onmarine carbon and nutrient cycling, pp. 242-259. In R. S.Ostfeld, F. Keesing, and V. T. Eviner [eds.], Infectious diseaseecology: effects of ecosystems on disease and of disease onecosystems. Princeton Univ. Press.

Moebus, K. 1991. Preliminary observations on the concentra-tion of marine bacteriophages in the water around Hel-goland. Helgo. Meeresunters. 45:411-422.

———. 1992. Further investigations on the concentration ofmarine bacteriophages in the water around Helgoland,with reference to the phage-host systems encountered.Helgo. Meeresunters. 46:275-292.


Passmore, R., and others. 2000. MPN assay analyzer.<http://www.ocgy.ubc.ca/~suttle/MPN.zip>.


Rippka, R., J. Deeruelles, J. B. Waterbury, M. Herdman, and R.Y. Stanier. 1979. Generic assignments, strain histories andproperties of pure cultures of cyanobacteria. J. Gen. Micro-biol. 111:1-61.

Safferman, R. S., and M. E. Morris. 1963. Algal virus: Isolation.Science 140:679-680.

Sambrook, J., E. F. Fritsch, and T. Maniatis. 1989. Molecularcloning: a laboratory manual, 2nd ed, vol. 1. Cold SpringHarbor Laboratory Press.


Stenholm, A. R., I. Dalsgaard, and M. Middelboe. 2008. Isola-tion and characterization of bacteriophages infecting thefish pathogen Flavobacterium psychrophilum. Appl. Environ.Microbiol. 74:4070-4078.

Steward, G. F., J. Wikner, W. P. Cochlan, D. C. Smith, and F.Azam. 1992. Estimation of virus production in the sea: I.Method development. Mar. Microb. Food Webs 6:57-78.

Sullivan, M. B., J. B. Waterbury, and S. W. Chisholm. 2003.Cyanophage infecting the oceanic cyanobacteriumProchlorococcus. Nature 424:1047-1051.

Suttle, C.A. 1993. Enumeration and isolation of viruses, pp121-134. In P. F. Kemp, B. F. Sherr, E. B. Sherr, and J. J. Cole[eds.], Handbook of methods in aquatic microbial ecology.Lewis.

———. 2000. Cyanophages and their role in the ecology ofcyanobacteria, p. 563-589. In B. A. Whitton and M. Potts[eds.], The ecology of cyanobacteria: Their diversity in timeand space. Kluwer Academic.

———. 2007. Marine viruses – major players in the globalecosystem. Nat. Rev. Microbiol. 5:801-812.

———, ———, and M. T. Cottrell. 1990. Infection of phyto-plankton by viruses and reduction of primary productivity.Nature 347:467-469.

———, ———, and M. T. Cottrell. 1991. Use of ultrafiltration toisolate viruses from seawater which are pathogens to marinephytoplankton. Appl. Environ. Microbiol. 57:721-726.

——— and ———. 1993. Marine cyanophages infectingoceanic and coastal strains of Synechococcus: abundance,morphology, cross-infectivity and growth characteristics.Mar. Ecol. Prog. Ser. 92:99-109.

———, and ———. 1994. Dynamics and distribution ofcyanophages and their effect on marine Synechococcus spp.Appl. Environ. Microbiol. 60:3167-3174.

———, and J. A. Fuhrman. 2010. Enumeration of virus parti-cles in aquatic or sediment samples by epifluorescencemicroscopy, p. 145-153. In S. W. Wilhelm, M. G. Wein-bauer, and C. A. Suttle [eds.], Manual of Aquatic Viral Ecol-ogy. ASLO.

Taylor, J. 1962. Estimation of numbers of bacteria by tenfolddilution series. J. Appl. Bact. 25:54-61.


133

Waterbury, J. B., and M. Willey. 1988. Isolation and growth ofmarine planktonic cyanobacteria. Methods Enzymol167:100-105.

———, and F. W. Valois. 1993. Resistance to co-occurringphages enables marine Synechococcus communities to coex-ist with cyanophages abundant in seawater. Appl. Environ.Microbiol. 59:3393-3399.

Weinbauer, M. 2004. Ecology of prokaryotic viruses. FEMSMicrobiol. Rev. 28:127-181.

Wilhelm, S., S. Brigden, and C. A. Suttle. 2002. A dilution tech-nique for the direct measurement of viral production: Acomparison in stratified and tidally mixed coastal waters.Microb. Ecol. 43:168-173.

Wilson, W. H., I. R. Joint, N. G. Carr, and N. H. Mann. 1993.Isolation and molecular characterization of five marinecyanophages propagated on Synechococcus sp. strainWH7803. Appl. Environ. Microbiol. 59:3736-3743.

Winget, D., K. Williamson, R. Helton, and K. E. Wommack.2005. Tangential flow diafiltration: an improved techniquefor estimation of virioplankton production. Aquat. Microb.Ecol. 31:221-232.


134

Introduction

By way of a preface, this is not a detailed list of step-by-stepmethods on how to sequence a virus genome. Sequencing proj-ects, particularly for large DNA viruses (100 kb–1200 kb), aresignificant undertakings for any lab-based project, and themagnitude of such an onerous task is often underestimated.Here, a sequencing project is defined as all the steps from

obtaining a clonal virus isolate through to generation of a com-pletely annotated virus genome. The first question to ask iswhether it is financially feasible to get your virus of choicesequenced by a professional facility, i.e., large-scale, highthroughput sequencing and bioinformatics facilities. It takesseveral skill sets and some dedicated expensive equipment todo the job efficiently, a facility not often afforded by a standardaquatic virology laboratory. Exceptions to this are smaller-scalesequencing projects (1 kb–50 kb) such as the small RNA-virusgenomes (Lang et al. 2004; Shirai et al. 2008) or some of thesmaller, straightforward (e.g., no extensive repeat regions) DNAviruses (Rohwer et al. 2000). With an increase in new tech-nologies such as 454 (www.454.com), high throughput masssequencing is becoming more accessible. However, it is stillexpensive to sequence a single virus genome unless you worktogether with other researchers to get several viruses on a sin-gle run or can negotiate with a facility to use up spare capacityon a high throughput run. The first swathe of sequence dataare only a starting point, and although you can get 99% of thegenome sequence very quickly, i.e., within 1–2 weeks (Landerand Waterman 1988), finishing the genome can take up to95% of the time and budget of an entire sequencing project. Inthis chapter, we will discuss some of the options when consid-ering the practicality of finishing a virus genome. Clearly, therewill be projects on a tight budget that will want to attempt toglean basic sequence information to help develop hypotheseson their viruses. For this, we will provide some basic protocols

Sequencing and characterization of virus genomesWilliam H. Wilson1* and Declan Schroeder 2*1Bigelow Laboratory for Ocean Sciences, 180 McKown Point Road, POB 475, West Boothbay Harbor, Maine, 04575, USA2Marine Biological Association of the UK, Citadel Hill, Plymouth, PL1 2PB, UK

AbstractBy unraveling the genetic code of viruses, genome sequencing offers a new era for aquatic virus ecology giv-

ing access to ecological function of viruses on an unprecedented scale. Although this chapter starts with the sug-gestion to that virus genome sequencing should be conducted professionally if financially feasible, we essen-tially try and guide the reader through some of the procedures that will direct a novice through a genomesequencing project. Arguably, the most important recommendation is to start with as high purity virus nucleicacid as possible. We use the adage, junk in equals junk out. Once sequence information is obtained, there isplenty of free, user-friendly software available to help build, annotate, and then compare sequence data.Acquiring metadata is another important aspect that is not often considered when embarking on a genomeproject. A new initiative by the Genomic Standards Consortium has introduced Minimum Information about aGenome Sequence (MIGS) that allows standardization of the way the data are collected to make it useful fordownstream post-genomic analyses. Most viruses sequenced to date have produced surprises, and there is moreto come from the other 1031 viruses still to be sequenced. This chapter focuses on sequencing purified virus iso-lates rather than virus metagenomes.

*Corresponding author: E-mail: [email protected] [email protected]



Research discussed here was funded partly from the NaturalEnvironment Research Council of the UK (NERC) Environmental Genomicsthematic program (NE/A509332/1 and NE/D001455/1) and partly fromthe US National Science Foundation (EF0723730). DCS is funded byNERC and through the NERC core strategic research programmeOceans2025 (R8-H12-52). We thank Mike Allen for providing Table 1.

ISBN 978-0-9845591-0-7, DOI 10.4319/mave.2010.978-0-9845591-0-7.134Suggested citation format: Wilson, W. H., and D. Schroeder. 2010. Sequencing and characteriza-tion of virus genomes, p. 134–144. In S. W. Wilhelm, M. G. Weinbauer, and C. A. Suttle [eds.],Manual of Aquatic Viral Ecology. ASLO.


MANUALof


Wilson and Schroeder Virus sequencing

135

with Web links and citations to similar projects. This chapterwill also explore some of the thinking behind why sequencinga genome should be considered, we will provide an assessmentof some of the current techniques and explore how sequencedata can be verified once a basic genome annotation has beenconducted. Guidelines for collecting metadata associated withthe genomes will also be provided as metadata is increasinglyimportant as more viral genomes are sequenced and compara-tive analyses become feasible. Verification of functional assign-ment (annotation) is essentially how characterization isdefined in this chapter—a much more significant process thanthe grunt of obtaining the sequence data.

Materials and proceduresBefore embarking down the road of virus genome sequenc-

ing and annotation, a number of key validation checks shouldbe made:

• Make every effort to start with a clonal and axenic isolate.Why clonal? Mutation rates in viruses are known to bevery high, especially when compared with its cellularhosts, and because the progeny of a single infection eventmay lead to the production of many variants of the origi-nal, it is, therefore, imperative not to complicate thesequence analysis even further by starting off with a mix-ture of similar genotypes. This can be done by performingeither dilution to extinction (Nagasaki and Bratbak 2010,this volume) or plaque assay (Schroeder et al. 2002) exper-iments. Why axenic? Bacterial sequence contamination isa major problem when sequencing novel genomes. Manysequencing programs have come unstuck because of thisvery issue. Consequently, give yourself every opportunityof generating sequencing information of your virus bystarting off with a well-defined clean system.

• Aim to get a large starting quantity of virus. Sequencingprotocols are very wasteful, so it is of utmost importancethat large quantities of virus, and thus genomic material,be produced (a minimum of 100 ng, though ideally aimfor up to 10 µg). Ideally, this needs to be done in one sin-gle event from the same starting virus inoculum (i.e., froma clonal virus preparation such as a plaque resuspension).This is to avoid amplifying variants of the original gener-ated by successive and continuous re-inoculation.

• Determine an efficient virus concentration protocol(Lawrence and Steward 2010, this volume). Before pro-ceeding to the next step of extracting virus nucleic acids,it is best to test whether you have any carry over of cellu-lar genomes. This can be easily achieved by using univer-sal ribosomal DNA primer sets. If these PCRs produce pos-itive results, treat your virus concentrate withcommercially available nucleases. Since virus nucleicacids are still protected by their capsid, this treatment willhave little or no effect on its genome. However, doremember to inactivate the enzymes before proceeding tothe nucleic acid extraction phase.

• Nucleic acid quality assessment. Test the quality of thenucleic acids generated using either PCR or restrictiondigestion. This will help determine if your nucleic acid issuitable for downstream sequencing.

Once the initial validation has been done, you can proceedwith the assurance that you have done everything possible tomitigate sources of producing junk sequence data. The nextfew steps entail the extraction and manipulation of nucleicacids, depending on the format available to you:

• Nucleic acid preparation. The choice of nucleic acidextraction will depend on the type of virus genome (RNAor DNA, ss or ds), the quantity and quality produced bythe method, and the budget available to you. There are anumber of commercially available kits (e.g., Qiagen) thatwill perform a perfectly adequate job. Alternatively, theuniversally tried and tested phenol-chloroform methodfor nucleic acid extraction (Lawrence and Steward 2010,this volume) normally delivers good results.

• Nucleic acid random fragmentation. Depending upon thesize of the genome and sequencing strategy used, frag-mentation may be necessary. For larger dsDNA viruses(>100 kb), the nucleic acids will need to be fragmentedinto smaller clonable sizes (e.g., 1–4 kb) for shotguncloning and Sanger-based sequencing. This can be doneenzymatically by controlled DNaseI treatment (Rohwer etal. 2000) or physically by sonication (Wilson et al. 2005).

DNaseI treatment—Generation of DNA fragments by DNasedigestion involves digesting DNA for a range of times, thenpicking the time that gives optimal-sized DNA fragments (typ-ically 1000–4000 bp). In a 50 µL reaction volume, resuspend 8µg DNA in 50 mM Tris·HCl (pH 7.6), 10 mM MnCl2, 100 µgmL–1 bovine serum albumin, and 0.01 SU mL–1 DNase I.Remove 5 µL aliquots (adding to 45 µL TE buffer, pH7.6) 0,0.5, 1, 2, 5, 10, 15, and 30 min after addition of the digestionmixture and immediately transfer to a tube containing 25 µLTris-buffered (pH 7.0) phenol (typically the shorter incuba-tions, up to 2 min, give optimally sized fragments). After aphenol:chloroform (1:1) and two chloroform extractions, pre-cipitate the fragmented DNA, wash with 70% ethanol, anddry. Resuspend fragmented DNA in 23 µL of Blunt-ending Mix(100 µM dNTPs, 1 × T4 DNA Pol Buffer) and heat at 65°C for30 min to resuspend DNA and inactivate any DNase I that wascarried over. After cooling to room temperature, add 2.5 UKlenow fragment and 5 U T4 DNA polymerase then incubatethe reaction at 37°C for 1 h. The fragmented and blunt-endedvirus DNA can be run on a 1% agarose gel prior to excisingfragments in the 1000–4000 bp range using a standard gelextraction procedures before downstream cloning (NB do notexcise fragments smaller than 1000 kb, as downstream cloningwill preferentially clone the smaller fragments).

Sonication—Generation of DNA fragments by sonication isperformed by placing a microcentrifuge tube containing thebuffered DNA sample into an ice-water bath in a cup-hornsonicator. Sonication is conducted for a varying number of


136

10-s bursts using maximum output and continuous power.Exact conditions for sonication should be empirically deter-mined for a given DNA sample before a preparative sonicationis performed. Typically, 100 µg DNA in TE buffer is split into10 aliquots of 35 µL; 5 are subjected to sonication for increas-ing numbers of 10 s bursts. Aliquots from each time point arerun on an agarose gel to determine optimal-sized DNA frag-ments (1–4 kb). Once optimal sonication conditions are deter-mined, the remaining 5 aliquots (approximately 8 µg) are son-icated according to those predetermined conditions. DNA canbe blunt-ended and size-selected as above prior to downstreamcloning.

• Cloning. If newer “next generation sequencing” (seebelow) options are chosen, any nucleic acid fragmenta-tion or cloning will be conducted by the sequencing facil-ity. Fragmented DNA can be cloned into a wide range ofcommercially available cloning vectors (e.g.,www.promega.com/vectors/cloning_vectors.htm) and/orcloning kits that are available from a wide range of molec-ular reagents companies (e.g., Promega, Invitrogen, NewEngland Biolabs) that make cloning almost fool-proof.However, remember that cloning procedures work bestwith high purity insert DNA. It is worth checking thequality of DNA inserts with A260/280 ratios prior tocloning with specific spectrophotometry devices such asNanoDrop (www.nanodrop.com). Smaller DNA or RNAgenomes can be cloned whole in specifically designedbacterial artificial chromosome (BAC)- or Yeast ArtificialChromosome (YAC)-based cloning and sequencing vec-tors, EPICENTRE Biotechnologies is one company thatprovides a range of options for this (www.epibio.com).Clones can then be sequenced directly, usually usingSanger-based technology (for basic explanation and ani-mation, see http://www.dnai.org/text/mediashowcase/index2.html?id = 552).

• Sequencing. For laboratory-based ‘in-house’ projects, clonelibraries (or PCR products) are typically run with Sanger-based technology. A search of Web sites reveals numeroustips and protocols for improving reads and reducing thecost of sequencing (typically by diluting sequencingenzyme) (e.g., www.nucleics.com). Note the first tip onthis site is “Use clean DNA” (junk in = junk out!). If thesequencing project can afford to send DNA directly to asequencing facility, there are numerous options currentlyavailable, from Sanger-based sequencing services to thenew generation high throughput services:

• Sanger-based sequencing. This technology utilizes DNApolymerase and chain terminating fluorescently bases tocreate four series of labeled DNA fragments. Sequencingplatforms (e.g., ABI & Beckman) capable of resolving thesefragments can accurately and efficiently resolve on aver-age ~700–800 bp. Therefore, M13 E. coli based vectors areroutinely used to create ~1 kb size clone libraries. Theamount of sequence generated is dependent on the size of

the virus genome. The general rule of thumb is to gener-ate sequence data of at least 8-fold coverage of yourgenome. This approach will provide up to 99% coverageof your genome, and then finishing approaches will berequired to obtain a full genome (see below).

• 454 Sequencing Technology (Roche’s Genome SequencerFLX system; www.454.com). Pyrosequencing is based on amethod developed by Ronaghi et al. (1996) and usesenzyme-mediated luminescence during DNA synthesis.The 454 sample preparation first relies on fragmentationof the genomic DNA followed by binding each fragmentto a microscopic bead. Pyrosequencing is then used todetermine the sequence of each bead-associated DNAfragment. This technology can produce sequence reads ofaround 200–400 bp, providing around 80–120 Mb perrun. The plate platform design of 454 allows the utility ofsplitting and/or dividing plates, a cost-saving measurethat does not appear to be available for other technolo-gies. 454 has been successfully used for finishing orgenome assembly purposes (see below).

• Illumina’s Solexa technology (www.illumina.com). The sec-ond commercially available next-generation technologyalso fragments genomic DNA, which is then ligated to aglass surface where bridge amplification creates multipleclusters of identical sequences (Bentley et al. 2008). Thesethen go on to be sequenced by a synthesis step using fluo-rescently labeled terminators with imaging following eachsuccessive base addition. The read lengths are much shorterthan those from 454’s instrument (~35 bp), therefore thistechnology is not appropriate for de novo sequencing.However, it comes into its own if you have a genome toalign the sequence against as each run (flowcell with 8channels) produces many times more reads than 454.

• SOLiD technology (ABI) (http://www3.appliedbiosystems.com/AB_Home/applicationstechnologies/SOLiDSystemSequencing/index.htm). This most recent addition to the next-gener-ation sequencing platforms amplifies fragmentedgenomic DNA by emulsion PCR on beads, followed bycyclic array (polony) sequencing (Shendure et al. 2005).To date, this technology has yet to be thoroughlyexploited by sequencing enthusiasts.

• Assembly. Assembly is a crucial component of the annota-tion process. The major sequencing facilities (e.g., SangerInstitute, Genoscope, DOE Joint Genome Institute) are con-tinually developing and optimizing their own assembly pro-grams. Many assembly programs are however freely availablefor download, e.g., Phred/Phrap (http://www.phrap.org/),Staden (http://staden.sourceforge.net/overview.html), CeleraAssembler (http://apps.sourceforge.net/mediawiki/wgs-assembler), AMOS (http://amos.sourceforge.net/), andARACNE (http://www.broad.mit.edu/science/programs/genome-biology/computational-rd/computational-research-and-development), all being suited for virus genome assem-bly. Many commercial packages such as DNASTAR


137

(http://www.dnastar.com/products/seqmanpro.php) andDNA BASER (http://www.dnabaser.com/) arguably producemore user friendly software, much more suited for the part-time assembler.

• Finishing. To finish or not to finish, that is the question!The next phase, the finishing strategies are arguably themost onerous and time-consuming tasks. Recent develop-ments, i.e., next generation sequencing, are being toutedas the savior of many aspiring genome assembler. As thesetechnologies are for those with substantial research budg-ets, closing physical gaps are mainly achieved throughcumbersome phases of PCR amplifications on genomicDNA with primer pairs positioned on independent contigends. However, if there is a hole in the sequence, whichmay be due to a physical barrier introduced by the frag-mentation/cloning or some troublesome spot forsequencing, it may be best to focus on identifying thearrangement of the contigs with respect to one anotherand then “filling in” the holes with PCR amplificationsacross gaps and subsequent sequencing of those products.Other strategies include sequencing libraries with muchlarger inserts using cosmids or fosmids (around 40 kbinserts). These vectors provide greater cloning flexibilityand construct stability, however, if larger gaps need fill-ing, BACs and YACs can be used:

• Bacterial Artificial Chromosome (BAC). BAC vectors canaccommodate up to 300-kb size DNA fragments. Theselarge fragments are usually generated by restrictionendonuclease digestion, separated on a pulse field gel(Sandaa et al. 2010, this volume), excised and gel purified.Electroporation is the transformation method of choicewhen creating BAC libraries.

• Yeast Artificial Chromosome (YAC): YAC vectors canaccommodate similar size ranges as BAC vectors; however,yeast cells can offer some advantages over cloning in bac-terial cells. DNA fragments containing repeat sequencesare difficult to propagate in bacterial cells becauseprokaryotes do not have such extensive DNA elements intheir genomes. Since yeasts are eukaryotes, they toleratesuch sequences better. This is an important point as largedsDNA viruses contain extensive repetitive repeats(Schroeder et al. 2009; Wilson et al. 2005), a problem notalways easy to resolve (Delaroque et al. 2003).

• Metadata and MIGS. All authors of genome sequence datamust consider the corresponding metadata associated withthe organism or virus that is being sequenced. This requiresstandardization of the way the data are collected to make ituseful for downstream post-genomic analyses e.g., compar-ative genomics. To address the issues surrounding develop-ment of better metadata descriptions of genomic investiga-tions (including whole genome sequencing andmetagenomics), the Genomic Standards Consortium (GSC)(http:// gensc. org/gc_wiki/index.php/Main_Page) wasrecently formed. GSC introduced the Minimum Informa-

tion about a Genome Sequence (MIGS) specification (Fieldet al. 2008), an ongoing process that has the intent of pro-moting community participation in its development anddiscussing the resources that will be required to developimproved mechanisms of metadata capture and exchange.It is worth checking the GSC Web portal (URL above) forcontinuous updates in MIGS implementation. Examplemetadata required for MIGS compliancy include (but is notlimited to) environmental parameters (e.g., location of iso-lation, physicochemical parameters, type of habitat), patho-genicity parameters; propagation conditions, treatment dur-ing collection, nucleic acid extraction procedures, andsequencing procedures (e.g., see example MIGS compliantreport in Figure 1).

An excellent example of the application of MIGS and use ofmetadata for comparative analysis and annotation of all pub-licly available genomes (including viruses) can be found at theIntegrated Microbial Genomes (IMG) Web portal(http://img.jgi.doe.gov/cgi-bin/w/main.cgi) (Markowitz et al.2007). It is a user-friendly interface allowing navigation ofgenome data along its three key dimensions (genes, genomes,and functions), and groups together the main comparativeanalysis tools.

• Annotation. This is often the most intimidating aspect ofthe process—making sense of all that sequence! Arguablythe main reason for the anxiety is the enormity of the taskat hand. To alleviate this unwelcome uncertainty of “will Ibe able to recognize any of the genes?” or “what if I missgenes?” in my new uncharted virome, the ever-growingfield of bioinformatics has come to our rescue—or has it?Unfortunately, bioinformatics has created its own share ofmayhem and confusion, i.e., what software package is best?As pointed out at the start of this chapter, “leave it to theprofessionals,” i.e., get someone onboard who has someexperience in this area. That said, if you go back to thebasics, the task will not be as intimidating as first thought.A good start is a book called Bioinformatics for Dummies(Claverie and Notredame 2003), which as its cover states is“a painless and thorough introduction to the field.”

As novices, you will require a software package that can:• Identify open reading frames (ORFs). This is classically

defined as a string of sequence starting with a start (AUG)and ending with a stop codon (UAG, UAA, or UGA). Assome organisms do not exclusively recognize AUG as astart codon, you might want your software to allow youto look at ORFs between two stop codons.

• View relevant features in all six possible frames. Thisallows you to quickly see where the larger ORFs arelocated on the genome.

• Upon selecting an ORF, provide a predicted amino acidsequence. This is an important feature as it gives you theflexibility to search (BLAST) Web-based genomic data-bases (NCBI, EBI, etc.) for DNA or protein homologs(though you do not need a putative amino acid sequence


138

to conduct homology searches, NCBI will allow you to doall permutations).

• Give you these aforementioned features and can performall these functions in real time, i.e., easily accessible (pointand click), integrated software with automated tools anduser friendly output formats.

We have personal experience with two software packages,namely Artemis (www.sanger.ac.uk/Software/Artemis/) andDNASTAR (www.dnastar.com) (Schroeder et al. 2009; Wilsonet al. 2005). Both are more than suited for custom annotationof viromes, as are many other packages in the market place, sothe choice is ultimately up to the user. An alternate strategyfor ORF identification is to use automated software such asAMIGene (www.genoscope.cns.fr/agc/tools/amiga/form.php),GLIMMER (www.cbcb.umd.edu/software/glimmer/), and Gen-eMark (http://exon.biology.gatech.edu/)—all of which havebeen used to annotate the latest giant virus, Feldmannia sp.virus 158 (Schroeder et al. 2009).

An important resource to BLAST against is the increasingvolume of virus metagenomics data that is now available.Without doubt, metagenomics has revolutionized the studyof microbiology and has revealed an incredible amount of

genetic diversity, particularly in the marine environment,and in particular among virus communities. The variousmethodologies of metagenomics have been discussed andreviewed extensively (Allen and Wilson 2008; Delwart 2007;Hall 2007; Handelsman 2004; Kunin et al. 2008; Riesenfeldet al. 2004; Streit and Schmitz 2004; Tringe and Rubin 2005;Wommack et al. 2008). A useful starting point for metage-nomics data can be found at the Community Cyberinfra-structure for Advanced Marine Microbial Ecology Researchand Analysis (CAMERA) Web portal (http://camera.calit2.net/). CAMERA is making raw environmental sequence dataaccessible along with associated metadata, pre-computedsearch results, and high-performance computationalresources.

• Verification of functional assignment (annotation)I) From ORFs to coding sequences (CDS) with putative

function:• Phylogeny. ORFs that have a BLAST homolog is likely to

be coding for a gene of a putative function, i.e., codingsequences (CDS). CDS homologs identified by the variousBLAST outputs should only be considered as indicators ofpossible function. An important first step in verify puta-

Fig. 1. Example MIGS-compliant report for EhV-86. Adapted from supplementary table supplied with Field et al. (2008). Modified screenshots takenfrom the Genome Catalogue (http://gensc.sf.net). Read-only information is imported from the NBCI Genome Projects Database and the GenomesOnline Database (http://www.genomesonline.org) to place each record into context. Values are only given for fields in MIGS marked as “minimal” (M,or mandatory) although more information may be available online. Figure reproduced by permission of M. Allen.


139

tive gene identities is by performing a phylogenetic analy-sis. This sort of analysis adds weight to the BLAST analy-sis by identifying its closest neighbor and whether itgroups with rigorously, well-characterized, and peerreviewed homologs. Many phylogenetic software arefreely available for use (e.g., PHYLIP—http://evolution.genetics.washington.edu/phylip.html). The NationalCenter for Biotechnology Information (www.ncbi.nlm.nih.gov/) provides a phylogenetic analysis for BLAST hitsearch outputs at a click of a button. Alternatively, specificWeb pages have dedicated links for identifying virus CDSsbased on phylogenetic profiling (www.igs.cnrs-mrs.fr/phydbac/Mimi/indexvirus.html).

II) From CDS to Gene:• Reverse transcriptase PCR (RT-PCR): The ability to detect

the mRNA of a CDS is an important validation step, i.e.,proof that the CDS is being expressed, of it performing aparticular function. Many commercial kits such as thoseprovided by Quantace (Sensimix™—www.quantace.com/country.asp) and Qiagen (OneStep RT-PCR—http://www1.qiagen.com/Products/Pcr/QiagenReverse Transcrip-tases/OneStepRtPcr.aspx) allow for rapid detection of CDStarget. Other uses of this technology are to link levels ofexpression with time point of infection or infectious phe-notype, i.e., quantitative PCR or real time RT-PCR). Quan-titative PCR utilizes the five or so cycles where DNA mol-ecules are synthesized logarithmically from scarcelydetectable to the log-linear phase. Therefore, the analysisis characterized by the threshold cycle (Ct)—the soonerthe threshold is reached the higher the starting number oftarget molecules. Using standard curve samples (DNA forDNA molecules and RNA for RNA molecules) with knownconcentrations, it is possible to determine the copy num-ber of the molecule in question.

• Microarrays. RT-PCR looks at the expression of one geneat a time. Microarrays can carry thousands of gene-spe-cific probes to detect multiple targets in a single sample(Allen and Wilson 2006; Allen et al. 2010, this volume).Sample cDNA is hybridized to a platform (e.g., a micro-scope slide) containing spots of DNA (60-70-mer oligonu-cleotide probes, each diagnostic for a target of interest). Intranscriptional microarrays, positive hybridization indi-cates up-regulation of a gene, hence confirming transcrip-tion of a CDS. As an example, microarrays were employedin the annotation of the EhV-86 genome (Wilson et al.2005). Functional information from preliminary expres-sion results can be used to determine correct readingframes for disputed CDSs. In addition, it can be used tohelp to identify new and unannotated CDSs. The primaryuse of the microarray is to assign virus transcripts intokinetic classes with the distinct aim of helping to deter-mine the function of coordinately expressed genes withno database homologues (Allen et al. 2006a). An oppor-tunistic use of the microarray is to use it as a tool for

genome diversity analysis (Allen et al. 2007). Very simply,the array can be used as a hybridization tool to determinepresence or absence (or highly divergent) of genes ingenomes of related coccolithoviruses. Rather than focus-ing on a single gene, the microarray will allow the forma-tion of a diversity index based on whole genomes withoutthe need to sequence these genomes. This can help toreveal core coccolithovirus genes and identify variableand absent genes between coccolithovirus genomes (Allenet al. 2006c).

III) From Gene to Protein:• Proteomics. Analysis and characterization of the complete

set of proteins (proteome) of a virus is one way to determineif structural genes are eventually translated into proteins(Allen et al. 2008; Clokie et al. 2008). Currently, the method isnot commonly used for aquatic viruses, however it is a prom-ising tool which combines 2D-gel electrophoresis followed byquantitative or semiquantitative mass spectrometry-basedanalysis of virus proteomes. Analysis of interactions betweenvirus and host proteomes is leading to the new field of inter-actomics, an emerging area that uses genomic and proteomictools to determine the full set of interactions between virusesand their hosts (Viswanathan and Fruh 2007).

Assessment and discussionThe fundamental basis of life is written in nucleic acid. For

viruses, this code is written in a variety of forms, be it single- ordouble-stranded, RNA or DNA. To date, genomic analysis ofviruses has had the biggest impact on the study of virus diver-sity. Regardless of the nature of the genome, the pinnacle ofassessing any biological entity’s genetic diversity is to sequenceits entire genome, then crucially, compare it to other genomesto assess the magnitude of the changes. The best (and mostcomprehensive) assessment of virus diversity would be tosequence the entire genome of every virus on the planet. This,of course, is beyond the realms of possibility and plausibility.At the time of writing, the viral genomes page on NCBI (www.ncbi.nlm.nih.gov/genomes/GenomesHome.cgi?taxid = 10239)contained links to 3235 reference sequences for 2178 virusgenomes and 41 reference sequences for viroids. The vastmajority of these viruses are medically or agriculturally related,essentially reflecting the levels of funding for these importantareas of research. Only a fraction (1%–2%) of the genomes arefrom aquatic viruses (Table 1). However, it is clear from thissmall but diverse collection of viruses that their hosts areequally diverse and include bacteria, archaea, algae, amoeba,invertebrates, and vertebrates, with virus genomes ranging insize from a few thousand bases to over a million bases.

These fully sequenced viruses represent only a minute frac-tion of the estimated 1031 viruses in aquatic environments,yet they have revealed a plethora of novelty and have alteredour view of viruses as simple ‘bags of genes’. It is common toidentify genes involved in core virus functions such as RNApolymerase, DNA polymerase, and structural proteins, yet it


140

Table 1. Selection of viruses of aquatic origin listed at NCBI that have completed genomes (as of February 2009). NB. The list is notexhaustive, the focus is on viruses that infect primary producers, and the microbial components of aquatic ecosystems. It is worth not-ing that descriptors for search terms are limited, and it is currently not possible to pull out virus genomes that are of aquatic (eithermarine or freshwater) origin. As increasing numbers of virus genomes are added to databases, it is important that appropriate meta-data, such as proposed by MIGS (Field et al. 2008), is made available to allow researchers to refine searches for specific groups ofgenomes. Although the aim of this table is to point the reader in the direction of aquatic virus reference genomes and a virus genomeWeb portal, the sequencing pipeline for each type of virus (i.e., RNA or DNA) is essentially the same once you have access to an ade-quate concentration of good quality nucleic acid. Acquiring enough high purity nucleic acid is arguably the biggest challenge of anyaquatic virus sequencing project.

Virus Family Accession* Size Reference

DNA virusesAcanthamoeba polyphaga Mimivirus Mimiviridae NC_006450 1,181,404 (Raoult et al. 2004)Emiliania huxleyi virus 86 Phycodnaviridae NC_007346 407,339 (Wilson et al. 2005)Paramecium bursaria Chlorella virus NY2A Phycodnaviridae NC_009898 368,683 (Fitzgerald et al. 2007b)Paramecium bursaria Chlorella virus AR158 Phycodnaviridae NC_009899 344,691 (Fitzgerald et al. 2007b)Ectocarpus siliculosus virus 1 Phycodnaviridae NC_002687 335,593 (Delaroque et al. 2001)Paramecium bursaria Chlorella virus 1 Phycodnaviridae NC_000852 330,743 (Li et al. 1997)Paramecium bursaria Chlorella virus FR483 Phycodnaviridae NC_008603 321,240 (Fitzgerald et al. 2007a)Paramecium bursaria Chlorella virus MT325 Phycodnaviridae DQ491001 314,335 (Fitzgerald et al. 2007a)Shrimp white spot syndrome virus Nimaviridae NC_003225 305,107 (Yang et al. 2001)Cyprinid herpesvirus 3 Herpesviridae NC_009127 295,146 (Aoki et al. 2007)Acanthocystis turfacea Chlorella virus 1 Phycodnaviridae NC_008724 288,047 (Fitzgerald et al. 2007c)Prochlorococcus phage P-SSM2 Myoviridae NC_006883 252,401 (Sullivan et al. 2005)Synechococcus phage S-PM2 Myoviridae NC_006820 196,280 (Mann et al. 2005)Crocodilepox virus Poxviridae NC_008030 190,054 (Afonso et al. 2006)Ostreococcus virus OsV5 Phycodnaviridae NC_010191 185,373 (Derelle et al. 2008)Synechococcus phage syn9 Myoviridae NC_008296 177,300 (Weigele et al. 2007)Prochlorococcus phage P-SSM4 Myoviridae NC_006884 178,249 (Sullivan et al. 2005)Microcystis phage Ma-LMM01 Myoviridae NC_008562 162,109 (Yoshida et al. 2008)Feldmannia species virus Phycodnaviridae NC_011183 154,641 (Schroeder et al. 2009)Thermus phage P23-45 Siphoviridae NC_009803 84,201 (Minakhin et al. 2008)Synechococcus phage P60 Podoviridae NC_003390 47,872 (Chen and Lu 2002)Cyanophage Syn5 Podoviridae NC_009531 46,214 (Pope et al. 2007)Vibriophage VpV262 Podoviridae NC_003907 46,012 (Hardies et al. 2003)Prochlorococcus phage P-SSP7 Podoviridae NC_006882 44,970 (Sullivan et al. 2005)Archaeal BJ1 virus Siphoviridae NC_008695 42,271 (Pagaling et al. 2007)Phormidium phage Pf-WMP4 Podoviridae NC_008367 40,938 (Liu et al. 2007)Roseobacter phage SIO1 Podoviridae NC_002519 39,898 (Rohwer et al. 2000)Pseudoalteromonas phage PM2 Corticoviridae NC_000867 10,079 (Mannisto et al. 1999)Penaeus merguiensis densovirus Parvoviridae NC_007218 6,321 (Sukhumsirichart et al. 2006)

RNA VirusesBeluga Whale Coronavirus SW1 Coronaviridae NC_010646 31,686 (Mihindukulasuriya et al. 2008)Micromonas pusilla reovirus Reoviridae NC_008171 – NC_008181 25,563 (Attoui et al. 2006)

segmentedDolphin morbillivirus Paramyxoviridae NC_005283 15,702 (Rima et al. 2005)Salmon pancreas disease virus Togaviridae NC_003930 11,919 (Weston et al. 2002)Chaetoceros tenuissimus RNA virus Unclassified AB375474 9,431 (Shirai et al. 2008)Marine RNA virus JP-A Seawater sample† NC_009757 9,236 (Culley et al. 2007)Marine RNA virus JP-B Seawater sample† NC_009758 8,926 (Culley et al. 2007)Heterosigma akashiwo RNA virus SOG263 Marnaviridae NC_005281 8,587 (Lang et al. 2004)Seal picornavirus type 1 Picornaviridae NC_009891 6,718 (Kapoor et al. 2008)Chaetoceros salsugineum Nuc Incl virus Unclassified NC_007193 6,000 (Nagasaki et al. 2005b)Marine RNA virus SOG Seawater sample† NC_009756 4,449 (Culley et al. 2007)Heterocapsa circularisquama RNA virus Unclassified NC_007518 4,375 (Nagasaki et al. 2005a)

*Most data obtained from NCBI Genome www.ncbi.nlm.nih.gov/genomes/GenomesHome.cgi?taxid = 10239†Complete genome sequence obtained from a metagenomic analysis of RNA extracted from seawater.


141

has becoming increasingly obvious that most viruses also har-bor the ability to alter and manipulate host metabolism inhighly specific ways to maximize the chance of successfulinfection. For example, genes involved in sphingolipid pro-duction, photosynthesis, and carbon metabolism have allbeen identified on virus genomes in recent years (Clokie andMann 2006; Lindell et al. 2005; Lindell et al. 2004; Mann etal. 2005; Mann et al. 2003; Sandaa et al. 2008; Wilson et al.2005; Yoshida et al. 2008). Complete genome sequencingprojects are becoming commonplace, yet these projects arestill restricted to relatively small numbers of isolates makingcomparative genomics of viruses still a young science. Pre-liminary comparative genomic analyses are making promis-ing progress in helping to determine evolutionary and taxo-nomic status of certain groups of viruses (Allen et al. 2006c;Hendrix et al. 1999; Iyer et al. 2006; Rohwer and Edwards2002). Full genome comparisons of aquatic viruses have beencompleted recently, particularly with algal virus andcyanophage genomes (Allen et al. 2006b; Delaroque et al.2003; Fitzgerald et al. 2007a; Fitzgerald et al. 2007b; Fitzger-ald et al. 2007c; Sullivan et al. 2005), this can help determinethe selective advantage of containing certain genes, especiallyif a gene is missing in a closely related virus isolate. Thisallows downstream hypothesis testing with virus host sys-tems to help determine the true function of observed differ-ences between genomes.

Despite falling costs and the obvious scientific advantagesof having complete coverage, it is often difficult forresearchers to justify sequencing closely related viruses (Allenet al. 2006b; Delaroque et al. 2003). However, subtle differ-ences at the genetic level can have profound effects on theinfection success of closely related virus isolates. This diversityoften remains hidden within a genome and is not immedi-ately obvious until full genomic sequences become available.

A quick analysis of the size spectrum of genomes reveals asignificant gap between the 407kb EhV-86 and the 1,181kbMimivirus (Table 1). One explanation is simply lack of isolatedrepresentatives in this size range and here lies a methodologi-cal conundrum. Viruses in this genome size range are likely tobe too large to pass through a 0.2 µm filter. Standard proce-dure includes passing a water sample through a 0.2µm filterand looking for viruses in the filtrate. However, most largeviruses will not be filterable through such a pore size, so evenat this early sampling stage, researchers often introduce a sizebias in their sampling strategy. Numerous Mimivirus sequencehomologs have been identified in the Venter Sargasso Seaenvironmental database (Ghedin and Claverie 2005; Monieret al. 2008a; Monier et al. 2008b). Indeed, these authors sug-gest that their data are indicative of high concentrations ofMimivirus ‘relatives’ in the ocean. Only a concerted samplingeffort to specifically isolate giant viruses in these environ-ments will identify new giant viruses that will fill the gap atthe top end of the genome size range.

Certainly the biggest dilemma when trying to choose new

viruses to sequence is trying to determine where to start andhow to justify some viruses over others? Options for justifica-tion could be global importance of the host (clear implicationsfor biogeochemical cycling and gene transfer processes);whether the host is sequenced (useful for downstream post-genomic analysis of virus-host interactions); extraordinary sizeof virus (help explain why the virus is so large); other virus rel-atives already sequenced (useful for comparative genomics andevolutionary determination); unusual host niche (does thegenetic signature provide clues of a selective advantage in thatniche) or exploitation opportunities (viruses from e.g., extremeenvironments). Whichever isolates are chosen, all will help toanswer specific hypotheses as well as give novel information tohelp generate new questions and hypotheses for future projects.

Future advances will involve using tools such as microar-rays, proteomics, and interactomics to help determine func-tionality of unknown genes. Sequence information should beconsidered as a starting point for asking questions and devel-oping hypotheses about the role of viruses. It is an excitingnew era for virus ecology and when used in combination withmore traditional approaches, virus genomics will give usaccess to their ecological function on an unprecedented scale.

Comments and recommendationsStart with as high purity nucleic acid preparation as you

can manage. Do not rush it, proceed with care. If your budgetcan stand it, get as much professional help as you can, partic-ularly with bioinformatics. However, if your budget is limited,there is plenty of user friendly software now available to helpbuild, annotate, and then compare sequence data. But do notgive up. Most viruses sequenced to date have produced sur-prises in their genomes, and there are more to come from theother 99.9999% (or 1031) viruses yet to be sequenced.

ReferencesAfonso, C. L., and others. 2006. Genome of crocodilepox

virus. J. Virol. 80:4978-4991.Allen, M. J., and W. H. Wilson. 2006. The coccolithovirus

microarray: an array of uses. Briefings in FunctionalGenomics and Proteomics 5:273-279.

Allen, M. J., T. Forster, D. C. Schroeder, M. Hall, D. Roy, P.Ghazal, and W. H. Wilson. 2006a. Locus-specific geneexpression pattern suggests a unique propagation strategyfor a giant algal virus. J. Virol. 80:7699-7705.

———, D. C. Schroeder, A. Donkin, K. J. Crawfurd, and W. H.Wilson. 2006b. Genome comparison of two Coccol-ithoviruses. Virol. J. 3:15.

———, ———, M. T. G. Holden, and W. H. Wilson. 2006c.Evolutionary history of the Coccolithoviridae. Mol. Biol.Evol. 23:86-92.

———, J. Martinez-Martinez, D. C. Schroeder, P. J. Somerfield,and W. H. Wilson. 2007. Use of microarrays to assess viraldiversity: from genotype to phenotype. Environ. Microbiol.9:971-982.


142

———, J. A. Howard, K. S. Lilley, and W. H. Wilson. 2008. Pro-teomic analysis of the EhV-86 virion. Proteome Sci. 6:11.

———, and W. H. Wilson. 2008. Aquatic virus diversityaccessed through omic techniques: A route map to func-tion. Curr. Opin. Microbiol. 11:226-232.

———, B. Tiwari, M. E. Futschik, and D. Lindell. 2010. Con-struction of microarrays and their application to virusanalysis, p. 34-56. In S. W. Wilhelm, M. G. Weinbauer, andC. A. Suttle [eds.], Manual of Aquatic Viral Ecology. ASLO.

Aoki, T., and others. 2007. Genome sequences of three koi her-pesvirus isolates representing the expanding distribution ofan emerging disease threatening koi and common carpworldwide. J. Virol. 81:5058-5065.

Attoui, H., F. M. Jaafar, M. Belhouchet, P. De Micco, X. DeLamballerie, and C. P. D. Brussaard. 2006. Micromonaspusilla reovirus: a new member of the family Reoviridaeassigned to a novel proposed genus (Mimoreovirus). J. Gen.Virol. 87:1375-1383.

Bentley, D. R., and others. 2008. Accurate whole humangenome sequencing using reversible terminator chemistry.Nature 456:53-59.

Chen, F., and J. R. Lu. 2002. Genomic sequence and evolutionof marine cyanophage P60: a new insight on lytic and lyso-genic phages. Appl. Environ. Microbiol. 68:2589-2594.

Claverie, J. M., and C. Notredame. 2003. Bioinformatics fordummies. Wiley.

Clokie, M. R. J., and N. H. Mann. 2006. Marine cyanophagesand light. Environ. Microbiol. 8:2074-2082.

———, and others. 2008. A proteomic approach to the identi-fication of the major virion structural proteins of the marinecyanomyovirus S-PM2. Microbiology-Sgm 154:1775-1782.

Culley, A. I., A. S. Lang, and C. A. Suttle. 2007. The completegenomes of three viruses assembled from shotgun librariesof marine RNA virus communities. Virology J. 4:69.

Delaroque, N., D. G. Muller, G. Bothe, T. Pohl, R. Knippers,and W. Boland. 2001. The complete DNA sequence of theEctocarpus siliculosus virus EsV-1 genome. Virology 287:112-132.

———, W. Boland, D. G. Muller, and R. Knippers. 2003. Com-parisons of two large phaeoviral genomes and evolutionaryimplications. J. Mol. Evol. 57:613-622.

Delwart, E. L. 2007. Viral metagenomics. Rev. Med. Virol.17:115-131.

Derelle, E., and others. 2008. Life-cycle and genome of OtV5,a large DNA virus of the pelagic marine unicellular greenalga Ostreococcus tauri. PLoS ONE 3:e2250.

Field, D., and others. 2008. The minimum information abouta genome sequence (MIGS) specification. Nat. Biotech.26:541-547.

Fitzgerald, L. A., M. V. Graves, X. Li, T. Feldblyum, J. Hartigan,and J. L. Van Etten. 2007a. Sequence and annotation of the314-kb MT325 and the 321-kb FR483 viruses that infectChlorella Pbi. Virology 358:459-471.

———, ———, ———, ———, W. C. Nierman, and J. L. Van

Etten. 2007b. Sequence and annotation of the 369-kb NY-2Aand the 345-kb AR158 viruses that infect Chlorella NC64A.Virology 358:472-484.

———, ———, ———, J. Hartigan, A. J. P. Pfitzner, E. Hoffart,and J. L. Van Etten. 2007c. Sequence and annotation of the288-kb ATCV-1 virus that infects an endosymbioticchlorella strain of the heliozoon Acanthocystis turfacea.Virology 362:350-361.

Ghedin, E., and J. M. Claverie. 2005. Mimivirus relatives in theSargasso Sea. Virol J. 2:62.

Hall, N. 2007. Advanced sequencing technologies and theirwider impact in microbiology. J. Exp. Biol. 210:1518-1525.

Handelsman, J. 2004. Metagenomics: application of genomicsto uncultured microorganisms. Microbiol. Mol. Biol. Rev.68:669-685.

Hardies, S. C., A. M. Comeau, P. Serwer, and C. A. Suttle. 2003.The complete sequence of marine bacteriophage VpV262infecting Vibrio parahaemolyticus indicates that an ancestralcomponent of a T7 viral supergroup is widespread in themarine environment. Virology 310:359-371.

Hendrix, R. W., M. C. M. Smith, R. N. Burns, M. E. Ford, andG. F. Hatfull. 1999. Evolutionary relationships amongdiverse bacteriophages and prophages: All the world’s aphage. Proc. Natl. Acad. Sci. U.S.A. 96:2192-2197.

Iyer, L. M., S. Balaji, E. V. Koonin, and L. Aravind. 2006. Evo-lutionary genomics of nucleo-cytoplasmic large DNAviruses. Virus Res. 117:156-184.

Kapoor, A., and others. 2008. A highly divergent picornavirusin a marine mammal. J. Virol. 82:311-320.

Kunin, V., A. Copeland, A. Lapidus, K. Mavromatis, and P.Hugenholtz. 2008. A bioinformatician’s guide to metage-nomics. Microbiol. Mol. Biol. Rev. 72:557-578.

Lander, E. S., and M. S. Waterman. 1988. Genomic mappingby fingerprinting random clones: a mathematical analysis.Genomics 2:231-239.

Lang, A. S., A. I. Culley, and C. A. Suttle. 2004. Genomesequence and characterization of a virus (HaRNAV) relatedto picoma-like viruses that infects the marine toxic bloom-forming alga Heterosigma akashiwo. Virology 320:206-217.

Lawrence, J. E., and G. F. Steward. 2010. Purification of virusesby centrifugation, p. 166-181. In S. W. Wilhelm, M. G.Weinbauer, and C. A. Suttle [eds.], Manual of Aquatic ViralEcology. ASLO.

Li, Y., Z. Lu, L. Sun, S. Ropp, G. F. Kutish, D. L. Rock, and J. L.Van Etten. 1997. Analysis of 74 kb of DNA located at theright end of the 330-kb chlorella virus PBCV-1 genome.Virology 237:360-377.

Lindell, D., M. B. Sullivan, Z. I. Johnson, A. C. Tolonen, F.Rohwer, and S. W. Chisholm. 2004. Transfer of photosyn-thesis genes to and from Prochlorococcus viruses. Proc.Natl. Acad. Sci. U.S.A. 101:11013-11018.

———, J. D. Jaffe, Z. I. Johnson, G. M. Church, and S. W.Chisholm. 2005. Photosynthesis genes in marine virusesyield proteins during host infection. Nature 438:86-89.


143

Liu, X. Y., M. Shi, S. L. Kong, Y. Gao, and C. C. An. 2007.Cyanophage Pf-WMP4, a T7-like phage infecting the freshwa-ter cyanobacterium Phormidium foveolarum: Complete genomesequence and DNA translocation. Virology 366:28-39.

Mann, N. H., A. Cook, A. Millard, S. Bailey, and M. Clokie.2003. Marine ecosystems: Bacterial photosynthesis genes ina virus. Nature 424:741.

———, and others. 2005. The genome of S-PM2, a “photosyn-thetic” T4-type bacteriophage that infects marine Syne-chococcus strains. J. Bacteriol. 187:3188-3200.

Mannisto, R. H., H. M. Kivela, L. Paulin, D. H. Bamford, and J.K. H. Bamford. 1999. The complete genome sequence ofPM2, the first lipid-containing bacterial virus to be isolated.Virology 262:355-363.

Markowitz, V. M., and others. 2008. The integrated microbialgenomes (IMG) system in 2007: data content and analysistool extensions. Nucleic Acids Res. 36(Database issue):D528-D533 [doi:10.1093/nar/gkm846].

Mihindukulasuriya, K. A., G. Wu, J. S. Leger, R. W. Nord-hausen, and D. Wang. 2008. Identification of a novel coro-navirus from a beluga whale by using a panviral microarray.J. Virol. 82:5084-5088.

Minakhin, L., and others. 2008. Genome comparison and pro-teomic characterization of Thermus thermophilus bacterio-phages P23-45 and P74-26: Siphoviruses with triplex-form-ing sequences and the longest known tails. J. Mol. Biol.378:468-480.

Monier, A., J. M. Claverie, and H. Ogata. 2008a. Taxonomicdistribution of large DNA viruses in the sea. Genome Biol.9:R106 [doi:10.1186/gb-2008-9-7-r106].

———, J. Larsen, R.-A. Sandaa, G. Bratbak, J.-M. Claverie, andH. Ogata. 2008b. Marine mimivirus relatives are probablylarge algal viruses. Virol. J. 5:12.

Nagasaki, K., Y. Shirai, Y. Takao, H. Mizumoto, K. Nishida, andY. Tomaru. 2005a. Comparison of genome sequences of sin-gle-stranded RNA viruses infecting the bivalve-killingdinoflagellate Heterocapsa circularisquama. Appl. Environ.Microbiol. 71:8888-8894.

———, Y. Tomaru, Y. Takao, K. Nishida, Y. Shirai, H. Suzuki,and T. Nagumo. 2005b. Previously unknown virus infectsmarine diatom. Appl. Environ. Microbiol. 71:3528-3535.

———, and G. Bratbak. 2010. Isolation of viruses infectingphotosynthetic and nonphotosynthetic protists, p. 92-101.In S. W. Wilhelm, M. G. Weinbauer, and C. A. Suttle [eds.],Manual of Aquatic Viral Ecology. ASLO.

Pagaling, E., and others. 2007. Sequence analysis of anArchaeal virus isolated from a hypersaline lake in InnerMongolia, China. BMC Genomics 8:13.

Pope, W. H., and others. 2007. Genome sequence, structuralproteins, and capsid organization of the cyanophage Syn5:A “horned” bacteriophage of marine Synechococcus. J.Mol. Biol. 368:966-981.

Raoult, D., and others. 2004. The 1.2-megabase genomesequence of mimivirus. Science 306:1344-1350.

Riesenfeld, C. S., P. D. Schloss, and J. Handelsman. 2004.Metagenomics: genomic analysis of microbial communi-ties. Ann. Rev. Genet. 38:525-552.

Rima, B. K., A. M. J. Collin, and J. A. P. Earle. 2005. Comple-tion of the sequence of a cetacean morbillivirus and com-parative analysis of the complete genome sequences of fourmorbilliviruses. Virus Genes 30:113-119.

Rohwer, F., A. Segall, G. Steward, V. Seguritan, M. Breitbart, F.Wolven, and F. Azam. 2000. The complete genomic sequenceof the marine phage Roseophage SIO1 shares homology withnonmarine phages. Limnol. Oceanogr. 45:408-418.

———, and R. Edwards. 2002. The phage proteomic tree: agenome-based taxonomy for phage. J. Bacteriol. 184:4529-4535.

Ronaghi, M., S. Karamohamed, B. Pettersson, M. Uhlen, and P.Nyren. 1996. Real-time DNA sequencing using detection ofpyrophosphate release. Anal. Biochem. 242:84-89.

Sandaa, R. A., M. Clokie, and N. H. Mann. 2008. Photosyn-thetic genes in viral populations with a large genomic sizerange from Norwegian coastal waters. FEMS Microbiol.Ecol. 63:2-11.

Sandaa, R.-A., S. M. Short, and D. C. Schroeder. 2010. Finger-printing aquatic virus communities, p. 9-18. In S. W. Wil-helm, M. G. Weinbauer, and C. A. Suttle [eds.], Manual ofAquatic Viral Ecology. ASLO.

Schroeder, D. C., J. Oke, G. Malin, and W. H. Wilson. 2002.Coccolithovirus (Phycodnaviridae): Characterisation of anew large dsDNA algal virus that infects Emiliania huxleyi.Arch. Virol. 147:1685-1698.

———, and others. 2009. Genomic analysis of the smallestgiant virus - Feldmannia sp. virus 158. Virology 384:223-232.

Shendure, J., and others. 2005. Accurate multiplex polonysequencing of an evolved bacterial genome. Science309:1728-1732.

Shirai, Y., Y. Tomaru, Y. Takao, H. Suzuki, T. Nagumo, and K.Nagasaki. 2008. Isolation and characterization of a single-stranded RNA virus infecting the marine planktonic diatomChaetoceros tenuissimus Meunier. Appl. Environ. Microbiol.74:4022-4027.

Streit, W. R., and R. A. Schmitz. 2004. Metagenomics—the keyto the uncultured microbes. Curr. Opin. Microbiol. 7:492-498.

Sukhumsirichart, W., P. Attasart, V. Boonsaeng, and S. Panyim.2006. Complete nucleotide sequence and genomic organi-zation of hepatopancreatic parvovirus (HPV) of Penaeusmonodon. Virology 346:266-277.

Sullivan, M. B., M. L. Coleman, P. Weigele, F. Rohwer, and S.W. Chisholm. 2005. Three Prochlorococcus cyanophagegenomes: Signature features and ecological interpretations.PloS Biol. 3:790-806.

Tringe, S. G., and E. M. Rubin. 2005. Metagenomics: DNAsequencing of environmental samples. Nat. Rev. Genet.6:805-814.

Viswanathan, K., and K. Fruh. 2007. Viral proteomics: global

evaluation of viruses and their interaction with the host.Expert Rev. Proteomics 4:815-829.

Weigele, P. R., and others. 2007. Genomic and structural analy-sis of Syn9, a cyanophage infecting marine Prochlorococcusand Synechococcus. Environ. Microbiol. 9:1675-1695.

Weston, J., and others. 2002. Comparison of two aquaticalphaviruses, salmon pancreas disease virus and sleepingdisease virus, by using genome sequence analysis, mono-clonal reactivity, and cross-infection. J. Virol. 76:6155-6163.

Wilson, W. H., and others. 2005. Complete genome sequence

and lytic phase transcription profile of a coccolithovirus.Science 309:1090-1092.

Wommack, K. E., J. Bhavsar, and J. Ravel. 2008. Metagenomics:Read length matters. Appl. Environ. Microbiol. 74:1453-1463.

Yang, F., J. He, X. H. Lin, Q. Li, D. Pan, X. B. Zhang, and X. Xu.2001. Complete genome sequence of the shrimp white spotbacilliform virus. J. Virol. 75:11811-11820.

Yoshida, T., and others. 2008. Ma-LMM01 infecting toxicMicrocystis aeruginosa illuminates diverse cyanophagegenome strategies. J. Bacteriol. 190:1762-1772.

144


145

Evidence for high abundances of viruses in aquatic systemsdate back more than 40 years (Anderson et al. 1967; Torrellaand Morita 1979); however, it was not until 20 years ago that

more quantitative estimates of abundance were made bytransmission electron microscopy (TEM) (Bergh et al. 1989;Børsheim et al. 1990; Proctor and Fuhrman 1990). Methodsusing TEM are time consuming, requiring specialized trainingand access to expensive equipment with limited availability.Moreover, TEM-based methods are subject to a large numberof artifacts that can lead to relatively inaccurate and low- precision estimates of viral abundance (Hennes and Suttle1995; Weinbauer and Suttle 1997). As a consequence, effortswere made to develop accurate, rapid, and affordable tech-niques of viral enumeration based on epifluorescencemicroscopy (EFM). Today, EFM is routinely used to estimatevirus abundances in aquatic samples and sediments.

It must be emphasized that estimates of viral abundanceby TEM or EFM are operationally defined by the techniqueused, that is, virus-like particles (TEM) or small pinpricks oflight (EFM). The techniques provide no information oninfectivity or the biological nature of the particles. Althoughthe large majority of observed particles are likely infectiousviruses, they may also include defective viruses, gene trans-fer agents, viruses damaged by solar radiation, or otherunknown particles. As well, there are RNA and single-stranded DNA viruses that are difficult, or in some casesimpossible, to see using the methods presented below. Theselimitations should always be kept in mind when followingthese methods.

Enumeration of virus particles in aquatic or sediment samples byepifluorescence microscopyCurtis A. Suttle1* and Jed A. Fuhrman2*1Departments of Earth & Ocean Sciences, Microbiology & Immunology, and Botany, University of British Columbia,Vancouver, BC, Canada2Marine Environmental Biology Section, University of Southern California, Los Angeles CA, USA

AbstractMicrobes and microbial processes are crucial and quantitatively important players in aquatic environments,

and viruses as major agents of microbial mortality and nutrient cycling are a key component of aquatic systems.These roles have led to the need to routinely quantify viral abundance as the part of many investigations.Electron microscopy was first used to demonstrate high viral abundances in aquatic samples; by the mid-1990s,the greater accuracy and higher precision of estimates of viral abundance made by epifluorescence microscopy(EFM) were evident. Initially, DAPI (6-diamidino-2-phenylindole) was the stain used to enumerate virus parti-cles in natural samples, but this dye was soon superseded by a new generation of much brighter fluorochromes.This article outlines detailed protocols for enumerating virus particles in aquatic or sediment samples usingSYBR Green, SYBR Gold, and Yo-Pro-1. Each of these stains has advantages and disadvantages, but for naturalwater samples they produce indistinguishable estimates of viral abundance when the appropriate protocols arecarefully followed.

*Corresponding author: E-mail: [email protected]; [email protected]



The authors are exceedingly grateful for the many people within theirrespective laboratories who have worked on the development of epifluores-cence microscopy for estimating viral abundance, especially L. M. Proctor,K. P. Hennes, M. G. Weinbauer, R. T. Nobel, A. C. Ortmann, A. M. Chan, K. Wen, and A. Patel. Comments by A. M. Chan, J. P. Payet, and threeanonymous reviewers were helpful in improving the manuscript. Researchon developing the methods for enumerating virus abundance have beensupported over the years by the US Office of Naval Research (C.A.S.), theUS National Science Foundation (C.A.S., J.A.F.), and the Natural andEngineering Research Council of Canada (C.A.S.). The authors alsoacknowledge the support of the Scientific Committee for OceanographicResearch for supporting working group 126 (marine virus ecology).

ISBN 978-0-9845591-0-7, DOI 10.4319/mave.2010.978-0-9845591-0-7.145Suggested citation format: Suttle, C. A., and J. A. Fuhrman. 2010. Enumeration of virus particlesin aquatic or sediment samples by epifluorescence microscopy, p. 145–153. In S. W. Wilhelm, M.G. Weinbauer, and C. A. Suttle [eds.], Manual of Aquatic Viral Ecology. ASLO.


MANUALof


Suttle and Fuhrman Enumeration of virus particles by epifluorescence microscopy

146

EFM-based methods to estimate viral abundance were adaptedfrom approaches in which fluorescent dyes with high bindingcoefficients for nucleic acids are used to stain heterotrophicprokaryotes, which are then captured on filters and enumeratedusing EFM (Daley and Hobbie 1975; Porter and Feig 1980). Ini-tially, DAPI (4′,6-diamidino-2-phenylindole) was used for esti-mating viral abundance in natural samples (Hara et al. 1991,1996; Proctor and Fuhrman 1992; Ricciardi-Rigault et al. 2000;Suttle 1993; Suttle et al. 1990), and although estimates were accu-rate and precise relative to TEM (Hara et al. 1991; Weinbauer andSuttle 1997), the small amount of DNA in viruses, relatively lowfluorescence yield of DAPI, and rapid fading made counting dif-ficult (Bettarel et al. 2000). The advent of nucleic-acid stains withmuch higher fluorescent yields circumvented this limitation, andcounts of viruses in natural waters and sediments can now rou-tinely be made in any laboratory with access to an epifluores-cence microscope and the appropriate filter sets.

A new generation of very bright, cyanine-based nucleic-acid stains, Yo-Pro-1, SYBR Green 1, and SYBR Gold, which areexcited by blue light (491 to 495 nm) and emit green or yel-low light (509 to 537 nm), has superseded the use of DAPI.Studies with Yo-Pro-1 (Hennes and Suttle 1995) and SYBRGreen 1 (Noble and Fuhrman 1998) on natural water samplesyield highly reproducible estimates of viral abundance. More-over, estimates of viral abundance made by epifluorescencemicroscopy have much greater precision and are routinelyhigher than estimates made by TEM (Chen et al. 2001; Hennesand Suttle 1995). Yo-Pro has a high fluorescence yield, a strongbinding coefficient to DNA, and stable fluorescence; however,samples need to be stained for 48 h and cannot be preservedwith aldehyde-based fixatives. Treatment with microwaveradiation has been recommended as a means to overcomethese limitations (Bettarel et al. 2000; Xenopoulos and Bird1997) and has been used in a number of studies (Bird et al.2001; Juniper et al. 1998). In a study by Bettarel et al. (2000),samples prepared in this manner had a higher coefficient ofvariation than the original method and give slightly lowerestimates of viral abundance, although the difference was notsignificant given the small sample size examined.

A widely adopted alternative to Yo-Pro-1 is SYBR Green 1(Noble 2001; Noble and Fuhrman 1998; Patel et al. 2007) or SYBRGold (Chen et al. 2001; Shibata et al. 2006). The primary advan-tages of SYBR dyes over Yo-Pro-1 are a much shorter staining time(<30 min), lower cost, and less sensitivity to aldehyde fixatives.Yo-Pro-1 has the advantages of extremely bright and long-lastingfluorescence and does not require the use of an antifade reagent.When bound to DNA, SYBR Green 1 has less stable fluorescencethan Yo-Pro-1 (Suzuki et al. 1997) or SYBR Gold (Chen et al.2001), which has caused concerns as to the reproducibility andaccuracy of viral abundance estimates made using SYBR Green(Bettarel et al. 2000; Chen et al. 2001). However, it is clear that ifprocedures are carefully followed, SYBR and Yo-Pro stains yieldstable brightly fluorescent viruses and estimates of viral abun-dance that are indistinguishable (Wen et al. 2004).

There are several constraints to using EFM to enumerateviruses in natural samples. High humic content can result inan unacceptable level of background fluorescence. As well,other microorganisms and some detrital particles are alsostained. Viruses can generally be distinguished from bacteriaand detritus by their staining characteristics and shape,although some very small bacteria may be counted as viruses.Because there are typically more than 10 times as manyviruses as bacteria, however, even if all bacteria in a naturalsample were counted as viruses, the introduced error wouldtypically be relatively small (Hennes and Suttle 1995). It ismuch more likely that bacterial abundances will be overesti-mated by counting some viruses as bacteria.

Another potential pitfall, which went unrecognized for sev-eral years, is that stainable viral particles decline very quicklyin aldehyde-based fixatives, although loss rates decreasemarkedly with time (Wen et al. 2004). For example, Wen et al.found that ~16% of the particles were lost within the firsthour, whereas about half the counts remained after 2 days.Many earlier EFM counts that used SYBR staining were doneon fixed samples that were refrigerated for days or weeksbefore the slides were made, and are likely significant under-estimates of actual abundances. As aldehyde-based fixativesare not compatible with Yo-Pro staining, slides have to bemade on unfixed samples as soon as possible after collection;hence, decay in preserved samples is not an issue. Accurateestimates of viral abundance require that slides are madeimmediately after sample collection or that the samples arefixed and then flash-frozen in liquid nitrogen.

Estimates of viral abundance made using Yo-Pro-I, SYBRGreen I, and SYBR Gold give comparable estimates of viralabundance, and detailed procedures for sample preparation,staining, and counting have been published for aquatic(Hennes and Suttle 1995; Noble 2001; Noble and Fuhrman1998; Ortmann and Suttle 2009; Patel et al. 2007; Shibata et al.2006; Wen et al. 2004) and sediment (Danovaro et al. 2001;Fischer et al. 2005; Hewson et al. 2001; Maranger and Bird1996; Ortmann and Suttle 2009) samples. Despite the brighterand more stable fluorescence of Yo-Pro, SYBR stains are mostwidely used because of the much shorter staining times andlower cost. The fluorescence of SYBR Gold–stained samples isreported to be more stable than that of samples stained withSYBR Green (Shibata et al. 2006), and as the procedures are thesame and SYBR Gold is less expensive, it a good choice formost samples. However, SYBR Gold emits at a longer wave-length (537 versus 520 nm), and the gold color can make itdifficult to count viruses in samples with a high humic con-tent or a lot of detritus, which is often autofluorescent in asimilar wavelength.

This article draws on previous publications, in particularPatel et al. (2007) and Ortmann and Suttle (2009), and theauthors’ experience to provide detailed protocols for usingepifluorescence microscopy to determine the abundance ofviral particles in aqueous and sediment samples.


147


Equipment—• Epifluorescence microscope equipped with the following:

�≥100 W Hg-vapor or 150 W xenon lamp (xenon lampsare richer in the blue spectrum and have longer lifetimesand lower lifetime cost)

�100× fluorescence objective (phase contrast objectivestypically reduce brightness)

�blue excitation filter (wide bandpass preferred for maxi-mum brightness, e.g., 450–480 or 460–500 nm; excita-tion peaks 497, 495, and 491 nm for SYBR Green 1, SYBRGold, and Yo-Pro-1, respectively)

�dichroic mirror (beam splitter), typically 500 or 510 nm

� long-pass sharp cutoff filter (typically 515 nm; emissionpeaks 520, 537, and 509 nm for SYBR Green 1, SYBRGold, and Yo-Pro-1, respectively; a long-pass filter allowsmaximum brightness)

�ocular reticule divided into 100 grid squares

� stage micrometer• Filtration unit, to hold 25-mm-diameter filters• Vacuum pump• Vacuum flask• Pipettes suitable for dispensing 1 µL to 2 mL• Event counter• Filter forceps• Sonicating bath (sediment samples only)• Squeeze bottle (or similar) containing 0.2-µm filtered

MilliQ water that will be used to wet the underlay filter.Reagents and solutions—• (SYBR only) SYBR Green or SYBR Gold nucleic-acid gel

stain, 10,000× concentrate in anhydrous DMSO (Invitro-gen)

• (Yo-Pro only) Yo-Pro-1, 1 mM stock solution in a 1:4 solu-tion of dimethyl sulfoxide and water (Invitrogen)

• (SYBR only) Antifade solution: p-phenylenediamine dihy-drochloride or 1,4-phenylenediamine dihydrochloride (seSigma P-1519 and not P-6001 [Patel et al. 2007] and storein a tightly capped container away from light)

• Spectrophotometric-grade glycerol• Phosphate-buffered saline (PBS): 0.05 M Na2HPO4, 0.85%

NaCl (wt/vol), pH 7.5)• (Sediment samples only) Pyrophosphate (10 mM)• 0.02-µm filter autoclaved MilliQ H2O• Tris EDTA (TE) buffer, pH 8 (for acidic samples)• (Yo-Pro only) Aqueous NaCl solution (0.3% wt/vol)• (SYBR only) 25% EM-grade glutaraldehyde, kept at 4°C• Ethanol• DF- or FF-grade immersion oil (refractive index 1/4, 1.516;

Olympus).Cautions: Nucleic acid stains, phenylenediamine dihy-drochloride and aldehyde-based fixatives are toxic; greatcare should be taken to avoid inhalation or contact withskin.

Disposable supplies—• Anodisc Al2O3 filters, 0.02 µm pore size, 25 mm diameter,

with support ring (Whatman)Note: At the time of writing, GE Healthcare, whichrecently purchased Whatman, is considering discontinu-ing production of the Anodisc membrane. Currently,there are no other membranes known that are suitable forEFM counts of virus particles. For polycarbonate filters,the background fluorescence is too high (even if stainedblack) and the porosity is too low to filter an adequate vol-ume of water.

• Nitrocellulose membrane filters, 0.45 µm pore size, 25mm diameter

• Precleaned glass microscope slides, 25 × 75 mm• Glass coverslips (25 × 25 mm) of proper thickness (each

microscope objective has an optimal coverslip thickness—the cover glass is part of the optics; for example, an Olym-pus infinity-corrected 100× UVPlanApo is imprinted“∞/0.17,” which indicates a 0.17-mm-thick cover glass,also known in the US as #1 1/2)

• 2.0 mL sterile polypropylene microcentrifuge tubes• 2.0 mL screw-cap cryovials• 10 µL, 200 µL, and 1 mL sterilized pipette tips• Polypropylene centrifuge tubes (15 or 50 mL)• Petri plates• Kimwipes (Kimberly-Clark) or other lint-free paper wipes• 9-cm-diameter paper filters (Whatman #1)• Additional supplies for long-term storage of samples for

SYBR staining:

�Liquid nitrogen

�Nylon stockings.Preparation of reagents—Reagents must be made in freshly

prepared deionized 0.02-µm filtered water to prevent virusparticles being introduced into the samples and causing highblanks.

Stock stain solution: Stains should always be handled inlow light to prevent photodegradation. Because the stains aresensitive to repeated freezing and thawing, the stains shouldbe aliquoted as stock solutions in small volumes. For SYBRstains, make a secondary stock by diluting the concentrateddye supplied by the manufacturer 10-fold with 0.02-µm fil-tered deionized water (dH2O) and dispense into polypropylenescrew-cap microcentrifuge tubes. For Yo-Pro-1, dilute to 50 µMin an aqueous solution of 2 mM NaCN to prevent any micro-bial growth during the 48-h staining period. Because the fluo-rescence of the dyes is very pH sensitive, it can be helpful todilute the stain in TE buffer (pH 8) if processing strongly acidor basic samples (Chan, pers. comm.). The diluted stainsshould be stored at –20°C, and ideally should be used withina week; the dye should be checked before use to make surethat it has not precipitated or adsorbed to the walls of thestorage tube. Adsorption of the stain to the tube walls is min-imized when stored in polypropylene. Each filter requires 2 µLSYBR stain; hence, 40 µL dispensed into each tube provides


148

enough stain for 20 filters. For Yo-Pro, freeze 800-µL aliquotsfor enough dye to stain 10 filters.

Glycerol/PBS solution: Prepare a solution of 50% glyceroland 50% PBS. The solution should be shaken or vortexed toensure complete mixing, 0.02 µm filtered, and stored as a liq-uid at –20°C. Alternatively, for long-term storage, 990 µL ofthe filtered mixture can be dispensed into microfuge tubesand frozen until ready for use.

Antifade (SYBR only): Prepare a 10% stock solution of theantifade reagent by diluting 1 g phenylenediamine (PDA) in10 mL of 0.02-µm filtered autoclaved dH2O. The PDA shouldgo completely into solution, producing a colorless liquid. Ifthe stock solution is tea-colored or darker, it has oxidized andshould not be used. Dispense 500-µL aliquots of the workingsolution into microcentrifuge tubes and store at –20°C to min-imize freezing and thawing. The frozen reagent should bewhite; if it has a brownish tint it should not be used. It is pos-sible to use other antifade reagents such as 0.5% (wt/vol)ascorbic acid in 50% (vol/vol) glycerol/PBS or SlowFade (Invit-rogen), but they may provide less protection against fading(Noble and Fuhrman 1998). In contrast, DABCO (1,4-diazabi-cyclo[2.2.2]octane) in TE/glycerol is reported to be a superiorantifade to PDA (Ortmann, pers. comm.). Immediately beforepreparing the slides, make a 0.1% working solution of theantifade by adding the 10% phenylenediamine solution to theglycerol-PBS mixture. Estimate 50 µL of reagent per slide.

Sample collection and preparation—Aqueous samples: Collect the samples in sterile containers

that are rinsed three times with the sample water. Polypropy-lene centrifuge tubes or bottles work well as sample contain-ers. The range of viral abundances suitable for enumerationis ~105 to 107 mL–1; hence, dilution may be necessary for veryproductive natural samples or cultures. If necessary, dilutethe sample with 0.02-µm filtered water, ideally preparedfrom the same or very similar water from which the samplewas obtained.

Sediment samples: Undisturbed sediment samples shouldbe collected with a piston corer. The sediment–water interfacecan be sampled with a wide-bore serological pipette to mini-mize disruption to the sediment surface. Samples from deeperwithin the core can be obtained by carefully pushing the coreup from the bottom of the core barrel and slicing the sedimentat the desired depth. The sediment should be sampled fromthe center of the core, leaving a well-defined area of sedimentaround the periphery to ensure the sample is not contami-nated with sediment smeared along the side of the core barrel.Remove a 0.5-cm3 subsample of sediment from the center ofthe core slice and transfer it into 4 mL of 0.02-µm filtered sea-water and 1.0 mL pyrophosphate (10 mM final concentra-tion). Sonicate the mixture for 3 min and centrifuge at 800gfor 1 min (Ortmann and Suttle 2009). The supernatant canthen be diluted and the slides prepared as outlined below. Fordifferent types of sediments and soils, the amount ofpyrophosphate and length of sonication may need to be

tested and optimized. Potassium citrate has been reported tobe superior to sodium pyrophosphate for extracting phagefrom soils (Williamson et al. 2003) and may be an alternativefor aquatic sediments, as well.

Preservation of samples (SYBR only): If slides cannot bemade in the field on freshly collected samples, preservationwill be necessary. Many earlier estimates of viral abundancehave been made on preserved samples; however, several stud-ies have shown rapid and significant decay of viral particlesin aldehyde-fixed samples (Brussaard 2004; Danovaro et al.2001; Wen et al. 2004), although there is evidence that thedecay is lessened when larger-volume samples are preserved(Patel et al. 2007). Samples preserved in the field should beflash-frozen in liquid nitrogen. Even freezing with dry ice(frozen CO2) or immediately freezing in a –87°C freezer is notadequate for preservation, and to our knowledge, freezing ina dry ice/ethanol slurry (–78°C) has not been tested. Samplesto be preserved should be fixed in 0.5% (final concentration)EM-grade glutaraldehyde and flash-frozen in liquid N2. Forrelatively productive samples with virus abundances ≥106

mL–1, add 30 µL of 25% EM-grade glutaraldehyde into prela-beled 2-mL cryovials, add 1470 µL of the sample to becounted, and mix. These volumes can be halved for sampleswith virus abundances ≥107 mL–1 and increased to 4 mL in 5-mL cryovials for very oligotrophic samples. Allow the sam-ples to stand for at least 15 min but no longer than 30 min,and then freeze immediately in liquid N2. To facilitate easyretrieval, vials can be placed in women’s nylon stockingsbefore freezing. (We find black sheer ones the most aestheti-cally pleasing.) Once frozen, the vials can be transferred to–80°C for long-term storage, until the slides can be prepared.For counting, the frozen samples are thawed in a 37°C waterbath and immediately stained with SYBR Green I or SYBRGold, as outlined below.

In the field, if it is impossible to make slides or preserve inliquid N, then it probably is best to collect the samples in aslarge volumes as is practical and maintain them under in situconditions for as short a time as is possible until slides can bemade.

Filtration and staining of sample—1. Prepare slide labels with critical information such as the

date, sample location, and volume filtered to keep track ofthe samples once they have been filtered and stained. Thelabels should be affixed when each slide is prepared.

2. For every set of four samples to be stained, use a perma-nent pen to mark the bottom of a plastic Petri plate intofour labeled sections.

3. (SYBR only) For each filter that is to be prepared, add a 78-µL drop of 0.02-µm filtered dH2O on each section of themarked Petri plates. Note that the efficacy of the stain ispH dependent; hence, if samples are from an acidic envi-ronment, diluting the stain in pH 8 TE buffer rather thandH2O has been found to result in more stable fluorescence(Chan, pers. comm.).


149

4. (SYBR only) Thaw a 40-µL vial of the stock SYBR solutionand add 2 µL stock solution to each drop (78 µL) of steriledH2O or buffer. Mix the stain by gently pipetting up anddown. If a large number of filters are to be processed, it iseasier to prepare a working solution of the stain and trans-fer 80-µL drops onto the plates. Place the Petri plates in thedark so that the stain is not bleached.

5. (SYBR only) Prepare the antifade solution in a clean, steril-ized 2-mL microcentrifuge tube. About 40 µL of antifadesolution will be required for each filter. Dilute the 10%(wt/vol) stock of p-phenylenediamine 1:100 using glyc-erol/PBS as the diluent. For example, add 1 µL stock p-phenylenediamine per 99 µL of 1:1 glycerol:PBS solu-tion. Keep the solution on ice and protect it from light.

6. (Yo-Pro only) For each filter that is to be prepared, add a80-µL drop of thawed Yo-Pro working solution on eachsection of the marked Petri plates. Place a 9-cm-diameterfilter paper soaked with 3 mL aqueous NaCl solution (0.3%wt/vol) in the lid of the Petri plates to prevent evaporationof the stain.

7. Connect a filtration unit for 25-mm filters to a vacuumsource, ensuring the vacuum is ≤13 kPa. A stronger vac-uum will likely crack the filter.

8. Place a 0.45-µm nitrocellulose backing filter on each filtersupport, and overlay it with a thin layer of dH2O. If the fil-ter does not wet evenly (if there are areas or spots thatremain white), replace the filter with another one. Backingfilters can be reused if they are smooth, are wet evenlywith water, and have no holes.

9. Carefully pick up a 0.02-µm Anodisc filter by its plasticring and lay it over the wet backing filter, with the plasticring facing upward. The Anodiscs are ceramic and breakeasily; make sure it is not cracked and that air is nottrapped between the filters. If necessary, pull excess waterthrough the filters using the vacuum, but make sure thefilters remain wet.

10. (SYBR only) If the sample has been preserved and frozen asdescribed above, thaw it in a 37°C water bath. For a sam-ple that has just been collected, fix it with 0.5% glu-taraldehyde for 15–30 min at 4°C before preparing slides.In addition, prepare duplicate control samples by fixing 1mL of the 0.02-µm filtered water that was used to dilutethe SYBR stain. For some samples, fixation may improvethe fluorescence of the particles and make counting easier.Also, some virus particles are prone to breakage if not fixed(Chan, pers. comm.).

11. (Yo-Pro only) Because divalent cations interfere with thebinding of the stain, seawater samples should be diluted to<7 psu with 0.02-µm filtered dH2O before filtration.

12. It is a good idea to make test slides (including a controlwith no sample added) to be sure an appropriate volume isfiltered, that the procedure is working, and that the filtersand reagents do not have viruses on or in them (somebatches of Anodiscs have been covered with bacteria and

viruses). For most lake and coastal seawater samples,which have viral abundances of ~107 mL–1, 0.8–1.0 mLsample is added to the surface of the Anodisc filter whilethe vacuum is off. A filtration tower is not needed, as sur-face tension will hold the water on the surface of the filter.Make sure the entire volume is within the plastic ring, orthe sample will be pulled under the edge of the filter. Turnon the vacuum and suck the sample through the filter.

13. For oligotrophic or very deep ocean samples, it may benecessary to filter 4 mL or more. If the volume to be fil-tered is slightly greater than 1 mL, the additional volumecan be added while the sample is filtering, being careful toensure the entire filter surface is continuously coveredwith liquid during filtration. For larger volumes, a sterilefiltration tower can be used; if the inside diameter of thetower is less than the filter diameter inside the plastic ring,it must be taken into consideration when calculating viralabundance (see below). Filter towers must be cleanedbetween samples. Rinse the towers with 0.02-µm filtereddH2O followed by ethanol. Dry with lint-free paper (e.g.,Kimwipe).

14. Once the sample is filtered, remove the Anodisc with thevacuum still on. There should not be any liquid on the sur-face of the filter. Touch only the plastic ring, so as not tocrack the membrane. To assist in lifting the filter, a 10-µLpipette tip can be cut at a slight angle and slid under thefilter edge. It also helps to place the Anodisc filter about amillimeter off center on the 0.45 underfilter, to provide anedge you can grasp with forceps. Allow the filter to air-dry(typically a minute or less), until the surface is visibly dry.

15. Place the Anodisc, sample side up, on a drop of stain in thePetri dish.

16. (SYBR only) Allow the filter to stain for 15 min in the dark.17. (Yo-Pro only) Allow the filter to stain for 48 h in the dark,

at room temperature.18. Add a drop of dH2O on the backing filter, lay the stained

Anodisc on top, and use the vacuum to remove anyremaining fluid. Do not use the filter if there is visiblewater on top of the Anodisc when it is done staining, as itis likely cracked.

19. (Yo-Pro samples and samples with high background fluo-rescence) Some samples (e.g., sediments, vent fluid, andhumic waters) may require the filters to be rinsed to reducebackground fluorescence. If so, while the vacuum is stillon and the filter is damp, rinse the filter twice with 1 mLof 0.02-µm filtered dH2O. For some samples, such as thosewith high humic content, or from vent environments,samples can be rinsed with TE buffer to reduce backgroundfluorescence (Chan and Winget, pers. comm.).

20. Remove the Anodisc while the vacuum is on. Place theAnodisc, sample-side up, on a 9-cm filter paper or Kimwipein the dark, and allow the filter to dry until it appearsopaque. The filter paper can be placed inside the lid of aPetri plate, and a foil-lined Petri-plate bottom can be used


150

as a lid. It usually takes about 5 min for the filter to dry,but it can be longer when humidity is high. The processcan be accelerated by laying the filter on a glass slide thatis heated to 35–37°C on a heating block, often very help-ful when humidity is high.

21. (SYBR only) Place 12–15 µL antifade solution on a labeledglass slide and lay the dry Anodisc on top. Add ~20 µLantifade on top of the Anodisc and cover with a coverslip.If the slides are to be frozen, add a little more antifade tocompensate for sublimation.

22. (Yo-Pro only) Place 12–15 µL spectrophotometric-gradeglycerol on a labeled glass slide and lay the dry Anodisc ontop. Add ~20 µL glycerol on top of the Anodisc and coverwith a coverslip.

23. Remove any air bubbles that are trapped under the cover-slip by gently pressing on the surface.

24. The slides can be counted immediately or stored frozen at–20°C for at least 4 months with no decrease in estimates ofviral abundance. Slides can be individually wrapped in aKimwipe and placed in small batches in foil packets beforefreezing. This allows a few slides to be thawed at a time.Once thawed, the slides should be counted immediately.

Determining abundance—1. Count the viruses at 1000× magnification using a 100× oil-

emersion objective. Make sure that the area of the filtercovered by the 10 × 10 ocular reticule has been determinedusing a stage micrometer.

2. Begin by checking the test filters to ensure that thereagents or filters were not contaminated and the filteredvolumes were appropriate. Viral and bacterial particles willappear green (SYBR Green and Yo-Pro) or yellow (SYBRGold) when excited with blue light (Fig. 1).

3. Check each slide before counting to make sure that the fil-ter is evenly stained and that the viruses are on a singleplane of focus and not suspended in the mounting mediumand are evenly distributed across the filter. Using the ocularreticule, select an appropriate number of grid squares sothat there are 10–100 stained viruses in each field. Virusescan generally be distinguished from cells by their stainingcharacteristics. Viruses appear as bright pinpricks of light,whereas cells generally have discernable size (Fig. 1). If the95% confidence intervals (see step 7) overlap for three tran-sects of 20 random fields containing at least 200 viruses, itindicates that 20 fields is adequate to compensate for thevariation in viral abundance among fields.

4. Estimate the abundance of viruses by counting at least 20random fields. Keep a tally of the number of particles ineach field so that the variation in abundance of particlesamong fields can be determined. These data can be used todetermine whether the distribution of particles across thefilter is random. (The data should follow a Poisson distri-bution where the mean equals variance, although 20 fieldsshould resemble a normal distribution.) Particles touchingtwo edges of the grid (e.g., left side and top) should be

counted, whereas particles touching the other two edges(e.g., right side and bottom) should not be counted.

5. For each sample, record the number of particles counted ineach field, the number of fields counted, the area of thefield, and the volume of sample filtered.

6. The abundance of viruses mL–1 (Vt) in the sample = Vc ÷ Fc

× At ÷ Af ÷ S, where Vc = total number of viruses counted,Fc = total number of fields counted, At = surface area of thefilter (µm2) (see note below), Af = area of each field (µm2),and S = volume of sample filtered (mL). (Note that for a 25-mm Anodisc filter, the diameter of the filter inside theplastic ring is 19 mm, which corresponds to an area of283,528,737 µm2. If a filtration tower is used that has adiameter less than that of the surface of the filter, it will benecessary to correct for the smaller filtration surface area.This is most accurately done by using the microscope todetermine the maximum width of the filter across whichvirus particles can be observed [Patel et al. 2007].)

7. The total number of particles counted will determine thesize of the 95% confidence intervals on the estimates ofviral abundance. By assuming a Poisson distribution, the95% confidence intervals can be estimated using the fol-lowing equations (Suttle 1993):

Assessment and discussionThe importance of microbes and microbial processes in

aquatic environments, coupled with the major role of virusesas agents of microbial mortality and nutrient cycling, has ledto the need to quantify viral abundance as part of many inves-tigations. The collection of these data have been greatly facil-itated by the development of epifluorescence microscopy as arelatively inexpensive and quick method to quantify viral par-ticles in natural waters and cultures. There have been manymodifications and improvements to the method since the firstdata were published using DAPI-stained samples (Hara et al.1991; Proctor and Fuhrman 1992; Suttle et al. 1990). However,the development of these methods has also identified chal-lenges that can affect the accuracy of results. These are brieflysummarized below.

Sample preservation—A significant issue that has been prob-lematic for many years is the preservation of samples for lateranalysis. One of the first studies that used epifluorescencemicroscopy to estimate viral abundance in natural waters(Hara et al. 1991) preserved samples in 1% (vol/vol) formalinfor 1–2 weeks before analysis. Similarly, many studies thatused SYBR stains analyzed aldehyde-preserved samples on thepremise that the abundance of viruses would remain stableover time. The same issue did not arise with Yo-Pro staining,as aldehyde fixatives are not compatible with the method.Numerous studies have now shown that the abundance ofstainable viral particles decreases rapidly once fixed (Brussaard

Lower 95% = – 1.96 ( + 0.5) + 1.42V Vc c×

Upper 95% = + 1.96 ( + 1.5) + 2.42V Vc c×


151

2004; Danovaro et al. 2001; Shibata et al. 2006; Wen et al.2004). Consequently, if accurate estimates of viral abundanceare to be obtained, slides must be prepared immediately fol-lowing sample collection, or the samples must be fixed andfrozen in liquid nitrogen, as outlined above.

Distinguishing between stained viruses and other fluorescentparticles—Even if some bacteria are counted as viruses, thisshould be of relatively minor consequence, because the errorintroduced would generally be relatively small even if all thebacteria were counted as viruses (Hennes and Suttle 1995).The typical high-precision viral abundance estimates obtainedwithin laboratories (generally <10%) gives the illusion of accu-racy. However, an interlaboratory comparison made on thesame samples by researchers who routinely estimate viralabundance by epifluorescence microscopy yielded estimatesthat varied by as much as threefold (unpubl. data, SCORWorking Group on Marine Viral Ecology). Yet the repro-ducibility within labs for individual samples was high. Thisindicates that small differences in methodology (e.g., the per-son counting, staining times, etc.) or equipment (e.g., lightsource, filter sets, etc.) can have a significant impact on esti-mates of viral abundance made by epifluorescencemicroscopy. The reasons behind these discrepancies remain tobe resolved, but care must be taken when comparing absoluteestimates of viral abundance made by different investigators.

Lack of a suitable standard—In part, reproducibility of esti-mates among laboratories is exacerbated by the lack of a suit-able standard that can be used to calibrate estimates of viralabundance. Even though it is relatively easy to stain and gen-erate high-precision estimates of abundance for stock culturesof viruses, the accuracy of these estimates is typicallyunknown. Moreover, in natural samples, there is a muchwider spectrum of fluorescent signatures than is found withincultures of viruses, making an appropriate standard difficult todefine. Nonetheless, it is generally accepted that the mostaccurate estimates of viral abundance for natural samples aremade by epifluorescence microscopy. However, until stan-dards that are widely accepted by the community are avail-able, the accuracy of individual estimates of viral abundancewill remain unknown.

All virus particles are not equally stained—Another source oferror is that all viruses are not stained equally, and may notbe visible by epifluorescence microscopy. Current evidencesuggests that the majority of viruses in the ocean are double-stranded DNA with genome sizes >20 kbp (Steward et al. 2000;Wommack et al. 1999), which should be easily stained and

Fig. 1. Epifluorescence micrographs of natural water samples. Viruses areindicated by solid arrows and bacteria by broken arrows. (A), Yo-Pro1–stained slide of a water sample from the Arctic Ocean (courtesy of J. P.Payet). (B), SYBR Green I–stained water sample from the coastal waters ofSouthern California. The large object in the center is a pennate diatom.(C) SYBR Gold–stained sample of a water sample from the ChesapeakeBay (courtesy of F. Chen and K. Wang). The slides emphasize the differ-ences that occur between samples and different staining protocols.


152

resolved by EFM. Nevertheless, it is becoming increasinglyapparent that RNA viruses and single-stranded (ss)DNA virusesare also widely distributed and likely relatively abundantmembers of the viroplankton (Angly et al. 2006; Culley et al.2006; Lang et al. 2009). However, the small genome sizes ofRNA and ssDNA viruses, and the weaker fluorescence yield ofthe dyes when bound to these nucleic acids, makes it unlikelythat these viruses would typically be visible by EFM.

Infectivity remains unknown—EFM provides high precisionand relatively accurate estimates of the abundance of virusparticles in natural samples; however, the proportion of parti-cles that are infectious is unknown, as is the identity of thehosts that they infect. It is reasonable to assume that themajority of virus particles are infectious based on estimates ofcontact rates with host cells (Suttle and Chen 1992) and pho-toreactivation (Wilhelm et al. 1998).

In this contribution we have attempted to bring togetherthe state of methodologies for using EFM to determine theabundance of viruses in aquatic and sediment samples. If pro-tocols are carefully adhered to, EFM can give highly repro-ducible and relatively accurate estimates of the abundance ofviral particles in water samples, sediments, and cultures. Caremust be taken to understand the limitations of the methods,and challenges remain to develop standards that can be usedto obtain absolute and accurate estimates of viral abundance.However, the methods are suitable for obtaining routine esti-mates of viral abundance in a wide range of natural samples.

ReferencesAnderson, N. G., G. B. Cline, W. W. Harris, and J. G. Green.

1967. Isolation of viral particles from large fluid volumes, p.75-88. In G. Berg, ed. Transmission of viruses by the waterroute. US Public Health Service.

Angly, F. E., and others 2006. The marine viromes of fouroceanic regions. PLoS Biol. 4:2121-2131.

Bergh, O., K. Y. Børsheim, G. Bratbak, and M. Heldal. 1989.High abundance of viruses found in aquatic environments.Nature 340:467-468.

Bettarel, Y., T. Sime-Ngando, C. Amblard, and H. Laveran.2000. A comparison of methods for counting viruses inaquatic systems. Appl. Environ. Microbiol. 66:2283-2289.

Bird, D. F., S. K. Juniper, M. Ricciardi-Rigault, P. Martineu, Y. T.Prairie, and S. E. Calvert. 2001. Subsurface viruses and bac-teria in Holocene/Late Pleistocene sediments of SaanichInlet, BC: ODP holes 1033b and 1034b, Leg 169s. Mar.Geol. 174:227-239.

Børsheim, K. Y., G. Bratbak, and M. Heldal. 1990. Enumerationand biomass estimation of planktonic bacteria and virusesby transmission electron microscopy. Appl. Environ. Micro-biol. 56:352-356.

Brussaard, C. P. D. 2004. Optimization of procedures forcounting viruses by flow cytometry. Appl. Environ. Micro-biol. 70:1506-1513.

Chen, F., J. R. Lu, B. J. Binder, Y. C. Liu, and R. E. Hodson.

2001. Application of digital image analysis and flow cytom-etry to enumerate marine viruses stained with SYBR Gold.Appl. Environ. Microbiol. 67:539-545.

Culley, A. I., A. S. Lang, and C. A. Suttle. 2006. Metagenomicanalysis of coastal RNA virus communities. Science312:1795-1798.

Daley, R. J., and J. E. Hobbie. 1975. Direct counts of aquaticbacteria by a modified epifluorescence technique. Limnol.Oceanogr. 20:875-882.

Danovaro, R., A. Dell’anno, A. Trucco, M. Serresi, and S.Vanucci. 2001. Determination of virus abundance inmarine sediments. Appl. Environ. Microbiol. 67:1384-1387.

Fischer, U. R., A. K. T. Kirschner, and B. Velimirov. 2005. Opti-mization of extraction and estimation of viruses in siltyfreshwater sediments. Aquat. Microb. Ecol. 40:207-216.

Hara, S., K. Terauchi, and I. Koike. 1991. Abundance of virusesin marine waters: Assessment by epifluorescence and trans-mission electron microscopy. Appl. Environ. Microbiol.57:2731-2734.

———, I. Koike, K. Terauchi, H. Kamiya, and E. Tanoue. 1996.Abundance of viruses in deep oceanic waters. Mar. Ecol.Prog. Ser. 145:269-277.



Juniper, S. K., D. F. Bird, M. Summit, M. P. Vong, and E. T.Baker. 1998. Bacterial and viral abundances in hydrother-mal event plumes over Northern Gorda Ridge. Deep-SeaRes. Part II 45:2739-2749.

Lang, A. S., M. L. Rise, A. I. Culley, and G. F. Steward. 2009.RNA viruses in the sea. FEMS Microbiol. Rev. 33:295-323.


Noble, R. T. 2001. Enumeration of viruses. Methods Microbiol.30:43-50.

———, and J. A. Fuhrman. 1998. Use of SYBR Green I for rapidepifluorescence counts of marine viruses and bacteria.Aquat. Microb. Ecol. 14:113-118.

Ortmann, A. C., and C. A. Suttle. 2009. Determination of virusabundance by epifluorescence microscopy, p. 87-96. In M.R. J. Clokie and A. M. Kropinski [eds.], Bacteriophages:Methods and protocols. Volume 1, Isolation, characteriza-tion, and interactions. Humana.

Patel, A., R. T. Noble, J. A. Steele, M. S. Schwalbach, I. Hewson,and J. A. Fuhrman. 2007. Virus and prokaryote enumera-tion from planktonic aquatic environments by epifluores-cence microscopy with SYBR Green I. Nat. Protocols 2:269-276.


153

Porter, K. G., and Y. S. Feig. 1980. The use of DAPI for identi-fying and counting aquatic microflora. Limnol. Oceanogr.25:943-948.


———, and ———. 1992. Mortality of marine bacteria inresponse to enrichments of the virus size fraction from sea-water. Mar. Ecol. Prog. Ser. 87:283-293.

Ricciardi-Rigault, M., D. F. Bird, and Y. T. Prairie. 2000.Changes in sediment viral and bacterial abundances withhypolimnetic oxygen depletion in a shallow eutrophic LacBrome (Quebec, Canada). Can. J. Fish. Aquat. Sci. 57:1284-1290.

Shibata, A., Y. Goto, H. Saito, T. Kikuchi, T. Toda, and S.Taguchi. 2006. Comparison of SYBR Green I and SYBR Goldstains for enumerating bacteria and viruses by epifluores-cence microscopy. Aquat. Microb. Ecol. 43:223-231.

Steward, G. F., J. L. Montiel, and F. Azam. 2000. Genome sizedistributions indicate variability and similarities amongmarine viral assemblages from diverse environments. Lim-nol. Oceanogr. 45:1697-1706.

Suttle, C. A. 1993. Enumeration and isolation of viruses, p.121-134. In P. F. Kemp, B. F. Sherr, E. B. Sherr, and J. J. Cole[eds.], Handbook of methods in aquatic microbial ecology.Lewis.

———, A. M. Chan, and M. T. Cottrell. 1990. Infection of phy-toplankton by viruses and reduction of primary productiv-ity. Nature 347:467-469.


Suzuki, T., K. Fujikura, T. Higashiyama, and K. Takata. 1997.DNA staining for fluorescence and laser confocalmicroscopy. J. Histochem. Cytochem. 45:49-53.

Torrella, F., and R. Y. Morita. 1979. Evidence by electron micro-graphs for a high incidence of bacteriophage particles in thewaters of Yaquina Bay, Oregon: Ecological and taxonomicalimplications. Appl. Environ. Microbiol. 37:774-778.

Weinbauer, M. G., and C. A. Suttle. 1997. Comparison of epi-fluorescence and transmission electron microscopy forcounting viruses in natural marine waters. Aquat. Microb.Ecol. 13:225-232.


Wilhelm, S. W., M. G. Weinbauer, C. A. Suttle, R. J. Pledger,and D. L. Mitchell. 1998. Measurements of DNA damageand photoreactivation imply that most viruses in marinesurface waters are infective. Aquat. Microb. Ecol. 14:215-222.

Williamson, K. E., K. E. Wommack, and M. Radosevich. 2003.Sampling natural viral communities from soil for culture-independent analyses. Appl. Environ. Microbiol. 69:6628-6633.

Wommack, K. E., J. Ravel, R. T. Hill, J. Chun, and R. R. Colwell.1999. Population dynamics of Chesapeake Bay virioplank-ton: Total community analysis using pulsed field gel elec-trophoresis. Appl. Environ. Microbiol. 65:231-240.

Xenopoulos, M. A., and D. F. Bird. 1997. Virus a la sauce Yo-Pro: Microwave-enhanced staining for counting viruses byepifluorescence microscopy. Limnol. Oceanogr. 42:1648-1650.

154

The field of viral ecology relies heavily on molecular meth-ods, such as PCR and sequencing, that target viral nucleic acid.The need to extract and purify nucleic acids is therefore nearlyuniversal in the field, and many methods have been describedto accomplish these tasks. Which of these methods is mostappropriate depends on the nature of the starting material,whether one wishes to purify DNA or RNA, and the purityrequired of the nucleic acid. Depending on the application,one may be starting with a purified virus, a partially purifiedvirus assemblage, viruses in a complex assemblage of otherplankton, or viruses in a complex physical matrix such as sed-iments. At the time of extraction, the viruses may be captured

on a filter, present in a pellet, or suspended in solution. Giventhe innumerable possible starting materials and the variedrequirements for the final product, we do not attempt in thisarticle to provide a direct comparison of methods. Instead, wefocus on reviewing the common strategies for nucleic acidextraction and provide a few protocols that have been used inthe field of aquatic viral ecology.

We first present some background information on the com-mon methods for harvesting and storing viruses before extrac-tion. We then consider the two basic steps in the extractionand purification of viral nucleic acids: (1) release of the nucleicacid from the virion and (2) separation of the nucleic acidsfrom other viral structural components. After this backgroundinformation, we present specific step-by-step protocols andassess their advantages and disadvantages. Where appropriate,we also discuss possibilities for adapting these methods to sit-uations other than those explicitly described.

Background informationA. Harvesting viruses for extraction—Prefiltration: For some applications (e.g., genomic and

metagenomic analyses), it may be desirable to prefilter thesample from which viruses will be harvested through a 0.2-µmfilter to remove prokaryotic and eukaryotic cells. The advan-tage of using a 0.2-µm filtered sample is that the majority ofthe nucleic acid extracted from the sample will be viral.Metagenomic analyses of viruses in whole plankton samplesare necessarily limited to the minority of sequences that canbe unambiguously recognized as viral (Williamson et al.

Extraction and purification of nucleic acids from virusesGrieg F. Steward* and Alexander I. CulleyDepartment of Oceanography, University of Hawai‘i at M-anoa, Honolulu, HI, 96822 USA

AbstractResearch on the diversity and ecology of viruses in the environment has been revolutionized by the ability

to detect, fingerprint, and sequence viral genes and genomes. The starting point for these molecular assays isthe release and recovery of the viral nucleic acids. The complexity of this task depends in large part on thenature of the starting material and the purity and quality of the nucleic acids one requires for downstream appli-cations. In some cases, simply heating the sample will suffice; in other cases, a series of organic extractions andpurification in a buoyant density gradient may be required to achieve adequate purity. Our goal in this chapteris to assist the reader in making informed choices from among the many options available. Toward this end, webriefly review the methods that have been used to harvest and store viruses in preparation for extraction, andthe methods by which their nucleic acids may be released and purified. We discuss the general principles uponwhich various commercial extraction kits are based and conclude with the presentation of four step-by-step pro-tocols. We discuss the advantages and disadvantages of these protocols, and the ways in which they may beadapted to various situations.




The writing of this chapter was supported in part by grants from NSFto the authors (OCE-04-42664 and OCE-0826650) and to the Center forMicrobial Oceanography: Research and Education (EF-0424599). Theauthors acknowledge the efforts of two anonymous peer reviewers andtheir suggestions to improve the manuscript.

ISBN 978-0-9845591-0-7, DOI 10.4319/mave.2010.978-0-9845591-0-7.154Suggested citation format: Steward, G. F., and A. I. Culley. 2010. Extraction and purification ofnucleic acids from viruses, p. 154–165. In S. W. Wilhelm, M. G. Weinbauer, and C. A. Suttle[eds.], Manual of Aquatic Viral Ecology. ASLO.


MANUALof


Steward and Culley Extraction and purification of nucleic acids from viruses

155

2008). A serious disadvantage to prefiltration is the variable,and sometimes significant, loss of viruses that can occur (Paulet al. 1991; Steward et al. 1992). Sometimes the losses may beminimal (Suttle et al. 1991; Wommack et al. 1995), but whenlosses do occur, they are likely to be more severe for largerviruses (Brum and Steward, paper accepted 2010). Fractiona-tion in buoyant density gradients is one possible alternative to0.2-µm filtration for separating cells and viruses (Lawrenceand Steward 2010, this volume), since most viruses are moredense than most cells. The separation will not be absolute,however, since the density ranges of cells and viruses overlap.

If one plans to perform virus-specific molecular assays onthe sample (e.g., PCR using virus-specific primers), one may beless concerned about the presence of nonviral nucleic acids. Inthis case, one might consider harvesting the entire microbialassemblage for extraction, especially if one hopes to performquantitative assays. The advantages of extracting the wholecommunity are that biases from prefiltration can be avoidedand all viruses (extra- and intracellular) will be collected. Thedisadvantages are a reduced detection limit and uncertaintiesabout which sequences are truly of viral origin, since themajority of the DNA will be nonviral.

Filtration: Tangential flow filtration (TFF) using an ultrafil-tration membrane (typically 30,000 to 100,000 nominalmolecular weight cutoff) is the method most commonly usedby microbial ecologists to harvest viral assemblages from nat-ural water samples (Proctor and Fuhrman 1990; Paul et al.1991; Suttle et al. 1991; Wommack et al. 2010, this volume).TFF allows one to process a wide range of volumes (tens of mil-liliters to thousands of liters) by scaling the system compo-nents, but the cost and complexity of the equipment and theprocess have usually restricted processing to a single sample ata time. Losses can be significant owing to adsorption of thevirus to the membrane, but methods have been suggested forimproving recoveries (Gerba 1983). A detailed description ofhow to concentrate viruses by TFF may be found elsewhere inthis volume (Wommack et al. 2010).

Viruses may also be harvested by direct (or normal) flow fil-tration using membrane filters with a sufficiently small poresize. Direct flow filtration was impractical with early mem-brane filters because of the very low flow rates (Clive 1967), butnewer aluminum oxide filters having a well-defined pore sizeof 0.02 µm and a high porosity (Furneaux et al. 1989) are wellsuited for virus capture. These filters are available in a syringe-tip housing with a Luer-Lok inlet fitting (Anotop; WhatmanInternational), making them convenient for field sampling.Numerous samples of whole or prefiltered seawater can be eas-ily processed in parallel using a multichannel peristaltic pump.Depending on the nature of the sample, hundreds of milliliters(eutrophic coastal water) to several liters (oligotrophic oceanicwater) can be passed through a single 25-mm-diameter, 0.02-µm-pore-size Anotop filter. This approach has been used suc-cessfully for the analysis of both RNA (Culley and Steward2007) and DNA (Culley et al. 2008) viruses. Existing normal

flow ultrafiltration capsules (OptiScale-25, Millipore; NovasipDV20, Pall) might be suitable alternatives. New nanoporous fil-ter materials are also being developed (e.g., Yang et al. 2008)that may provide additional direct flow filtration options inthe near future.

Centrifugation: For small samples, one can reliably collectviruses by sedimentation in a bucket rotor in an ultracen-trifuge. For much larger volumes, samples may be processedby continuous-flow ultracentrifugation, in which a feedstream is passed through a rotor at high speed. The latter pro-cedure has been used to efficiently harvest viruses from verylarge volumes of seawater (Anderson et al. 1967). In eithercase, batch or continuous flow, the viruses may be pelleted orthey may be banded within a buoyant density gradient in therotor. The latter method both concentrates and partially puri-fies viruses while maintaining them in suspension. Althoughcentrifugation is effective at harvesting viruses, ultracen-trifuges and rotors are expensive (continuous-flow in particu-lar) and not portable. As a consequence, centrifugation hasbeen used infrequently for the initial harvesting of virusesfrom environmental samples. Ultracentrifugation is, however,used frequently for the final concentration and purification ofharvested viruses (Lawrence and Steward 2010, this volume).

Adsorption-elution: A wide variety of virus adsorption- elution (VIRADEL) concentration methods have been devel-oped for monitoring water quality (Percival et al. 2004), andthese are used extensively to screen for low concentrations ofknown viral pathogens in water. VIRADEL-based methods havenot found as much use among aquatic viral ecologists, perhapsbecause of the need in many of these methods to extensivelymanipulate the chemistry of the sample to control the adsorp-tion-elution behavior (Sobsey 1976), and the consequence ofthis for the recovery of total viruses from natural assemblageshas been uncertain. The recovery efficiencies of the variousVIRADEL methods are virus dependent (Percival et al. 2004),suggesting that these methods would lead to biases if appliedto the analysis of complex communities. VIRADEL recoveries,however, are typically measured as infectious units rather thanviral particles. The former may be much lower than the latterif some viruses are inactivated by the procedure. For molecularinvestigations, the loss of infectivity is not a concern, so someof these techniques may turn out to be less biased when evalu-ated in terms of total recovery of virus particles. Recent workon the precipitation of viruses from seawater by adsorption toiron hydroxide appears to result in very high recoveries of totalviruses (John and Sullivan, pers. comm.). Considering the sim-plicity, low cost, and high capacities of some VIRADEL meth-ods, this approach is likely to become more popular amongviral ecologists who need to process many samples under chal-lenging conditions in the field.

Viruses in sediments: Natural viral communities may alsobe harvested from sediments for molecular assays. Viruseshave been separated from unpreserved sediments by squeez-ing with a press (Steward et al. 1996) or by centrifugation


156

(Drake et al. 1998). To facilitate the recovery of viruses, sedi-ments have been diluted and agitated in a buffer (Hewson etal. 2001; Labonté et al. 2009). More aggressive treatments withsodium pyrophosphate (Lawrence et al. 2002; Filippini andMiddelboe 2007; Helton and Wommack 2009) and sonication(Filippini and Middelboe 2007) have been employed to facili-tate recovery viruses more strongly adsorbed to sediment par-ticles. If one wishes to focus exclusively on the viral fractionof the sediment microbial community, additional fractiona-tion such as by filtration (Lawrence et al. 2002; Filippini andMiddelboe 2007; Leroy et al. 2008; Helton and Wommack2009; Labonté et al. 2009) or purification in a density gradient(Filippini and Middelboe 2007; Lawrence and Steward 2010,this volume) would be required to separate the viruses fromthe other microbes. Once the viruses are suspended in liquid,they may then be further concentrated using one of the vari-ous techniques described above. The separation of virusesfrom sediments is discussed in greater detail elsewhere in thisvolume (Danovaro and Middelboe 2010).

B. Storing virus samples before extraction—Freezing: The nucleic acids within harvested viruses can be

preserved by freezing the concentrated material at –80°C. Thiscan result in physical damage to some viruses, and potentiallya loss of infectivity, but both DNA and RNA are well preservedat or below this temperature (Sambrook and Russell 2001). Ifone has decided on an extraction protocol ahead of time, itmay be advisable to freeze the sample in the extraction buffer,as this may help prevent degradation of any nucleic acidsextruded from the virions as a result of the freezing and thaw-ing process. The major drawback of relying on ultra-low- temperature storage is the difficulty of maintaining these tem-peratures in the field.

Preservation solution: Immediate freezing may be unneces-sary with the use of a patented nucleic acid preservation solu-tion (RNALater™; Ambion) that appears to offer protection ofnot only nucleic acids, but also infectivity among the virusestested (Lader 2001). An equal volume of RNALater added toliquid samples protected viral RNA from degradation whenthe samples were subjected to freeze-thaw cycles (Forster et al.2008). This preservation solution, which appears to functionprimarily through ammonium sulfate precipitation of pro-teins and nucleic acids, also preserved the infectivity of RNAviruses in samples stored at room temperature for up to 72 h(Uhlenhaut and Kracht 2005). Although infectious titerdeclined by several logs after 50 days of storage at room tem-perature, the titers were orders of magnitude greater than acontrol sample stored in phosphate-buffered saline for thesame period. These promising results suggest that it would bepossible to preserve concentrates of aquatic viruses (in solu-tion or on filters) under field conditions where immediatefreezing is not possible.

Despite the specificity implied by the name, RNALater pre-serves DNA as well as RNA (Gorokhova 2005), so it shouldwork just as well for the preservation of the genomes of DNA-

containing viruses. If the virion structures are preserved, thenviruses suspended in the preservation solution should berecoverable by centrifugation or filtration before extraction.Small amounts of RNALater are compatible with a variety ofextraction kits and methods (Ambion technical literature), solimited quantities may also be directly extracted.

C. Releasing nucleic acids from viruses—The most commonmethods used to release nucleic acids from virions involve theuse of heat, osmotic shock, detergents, chaotropic salts, ororganic solvents, either alone or in combination, all of whichlead to denaturation of capsid proteins (Ralph and Bergquist1967). The buffer Tris(hydroxymethyl)-aminomethane (orsimply Tris) is commonly used to maintain nucleic acid solu-tions at slightly alkaline pH to minimize chemical hydrolysisof the nucleic acids, but acidic eonditions are sometimes pre-scribed for selective extraction and storage of RNA.

Thermal destabilization in the presence of chelators: Thesimplest method by far to release nucleic acids from virions isto heat the sample (typically to 45–100°C). This alone is suffi-cient for some applications (Richardson et al. 1988), in partic-ular for obtaining nucleic acids from purified viruses wherenuclease contamination is expected to be minimal. If nucle-ases are expected to be present, then heating should be carriedout in the presence of a chelating agent such as ethylenedi-amine-tetraacetic acid (EDTA). Ethylene glycol-bis(2-aminoethylether)-N,N,N′,N′-tetraacetic acid (EGTA) may alsobe added depending on the circumstances. EDTA, and some-times EGTA, is included in buffers used with DNA, becausethey chelate divalent and trivalent cations, which are arequired cofactor for certain nucleases (Adams et al. 1992).Divalent cations (particularly Mg2+ and Ca2+) also contribute tothe stabilization of viral capsids (Brakke 1963; Brady et al.1977; Ruiz et al. 2007). The presence of these chelators there-fore simultaneously facilitates disintegration of the capsid andprotects DNA from degradation. DNase is itself irreversiblyinactivated by heating to 65°C, but the disintegration of viralcapsids can occur faster. Viral DNA can therefore be lost ifviruses are heated to 65°C in the presence of DNase and theabsence of EDTA (G. F. Steward, pers. observation). For exper-iments that rely solely on heat to inactivate DNase in the pres-ence of viruses (e.g., Fuller et al. 1998), viral DNA concentra-tion is likely to be underestimated. It should be kept in mindthat EDTA and EGTA can affect some downstream reactions(such as PCR) that use Mg2+-dependent enzymes if they arecarried over at a concentration that is a significant fraction ofthe Mg2+ concentration in the reaction buffer.

Neither EDTA nor EGTA inactivates RNase, so other RNase-inhibiting agents may be required if one plans to extract RNA-containing viruses by simple thermal destabilization. RNaseinhibitors include ribonucleoside-vanadyl complex (Bergerand Birkenmier 1979) and RNasin (Blackburn et al. 1977), aswell as other commercially available proprietary reagents (e.g.,RNAsecure, Ambion; RNase Out, Invitrogen). The ribonucleo-side-vanadyl complex will also inhibit downstream reactions


157

such as in vitro translation and reverse transcription, so is notrecommended (Farrell 2005).

Osmotic shock: Osmotic shock can be used to disintegratethe capsids of some viruses, but others are resistant to thistreatment (Anderson 1950; Anderson et al. 1953). This phe-nomenon may therefore facilitate some extraction protocols,but is generally not relied on. The sensitivity of some virusesto osmotic shock should be kept in mind when harvesting andpurifying viruses, however, since exchanges of buffers havingvery different osmolarities may lead to unintentional releaseand potential loss of nucleic acids.

Detergent: Sodium dodecyl sulfate (SDS) is an ionic deter-gent frequently added to extraction buffers. SDS solubilizescapsids by disrupting inter- and intraprotein hydrophobicinteractions (Putnam 1948; Reynolds and Tanford 1970). SDSmay be used alone (Sreenivasaya and Pirie 1938), but is typi-cally used in combination with heating (Fraenkel-Conrat et al.1957) and enzymatic digestion of proteins (Sambrook and Rus-sell 2001) to effect the release of nucleic acids. Formamide willalso disrupt phage capsids and has been used as a rapid, simple,but perhaps less effective (Sambrook and Russell 2001), alter-native to treatment with heat, SDS, and proteinase K digestionfor extraction of DNA from viruses (Vega Thurber et al. 2009).

Chaotropic salts: Chaotropic salts such as sodium iodide(NaI) or guanidinium thiocyanate (GTC) can also disrupt cap-sids by denaturing proteins. The guanidinium and thio-cyanate ions of GTC are particularly strong denaturants(Mason et al. 2003) and consequently facilitate disintegrationof viral capsids while simultaneously inactivating nucleases.Because GTC so effectively inactivates RNase, it remains acommon ingredient in RNA extraction protocols since its firstuse in this capacity thirty years ago (Chirgwin et al. 1979). Thenonpolar organic solvent, phenol, also has a long history ofuse in nucleic acid extractions (Kirby 1956; Kirby 1957).Although typically used for its ability to extract proteins fromnucleic acid solutions (Sambrook and Russell 2001), phenolwill simultaneously effect the disruption of viral capsids bydenaturing the proteins (Faulkner 1962).

The above methods and reagents are among the most com-mon, but a wide range of other strategies have been used torelease nucleic acids from viruses. The interested reader willfind a comprehensive review of these earlier efforts in Methodsin Virology, vol. 2 (Ralph and Bergquist 1967).

D. Separating nucleic acids from other macromolecules—Oncenucleic acids have been released from the viruses, it may be nec-essary to separate the nucleic acids from other macromoleculesin the lysate. This may be achieved by exploiting differences insolubility or buoyant density among macromolecules. We willconsider five general approaches to this task: (1) organic extrac-tion, (2) differential precipitation, (3) solid-phase extraction, (4)density gradient fractionation, and (5) electrophoresis.

Organic extraction: In organic extraction, proteins andlipids are extracted from a nucleic acid solution using an alka-line buffer–saturated phenol (Kirby 1957) or phenol plus chlo-

roform (1:1). A small amount of isoamyl alcohol (IAA) is alsocommonly added to the chloroform as an antifoaming agent(phenol:chloroform:IAA; 25:24:1). After emulsification, theaqueous and organic phases are separated by centrifugation.Nucleic acids remain soluble in the upper aqueous phase,which is harvested, whereas lipids and proteins partition to theorganic phase or the interface of the organic and aqueousphases (interphase). Traces of phenol, which can interfere withdownstream enzymatic reactions or assays, are removed fromthe aqueous phase by extraction with chloroform:IAA, andtraces of chloroform can be removed by extraction with water-saturated ether or by alcohol precipitation of the nucleic acids(Sambrook and Russell 2001). In ether extractions, the aqueousphase is on the bottom. After removing the bulk of the etherby pipetting, residual amounts can be easily removed by evap-oration by warming the sample with the lid open.

A modified organic extraction procedure using a mixture ofphenol and guanidine thiocyanate was developed for theextraction and recovery of RNA, DNA, and protein from thesame sample (Chomczynski 1993). In this case, RNA is selec-tively partitioned to an acidic aqueous phase (Kirby 1956)while DNA and protein partition to the interphase andorganic phase. RNA is precipitated from the harvested aqueousphase, and the organic phase is back-extracted with aqueoussolution at a higher pH to solubilize the DNA. DNA is thenprecipitated from the back-extracted aqueous phase and pro-tein is precipitated from the organic phase with acetone. Theorganic extraction mixture and other materials for this proce-dure are commercially available (TRI reagent, MolecularResearch Center; TRIzol, Invitrogen) along with detailed pro-tocols (e.g., www.mrcgene.com/tri.htm).

Differential precipitation: When separating proteins andnucleic acids by differential precipitation, the proteins can be“salted out” directly with ammonium sulfate or precipitated asSDS–protein complexes by the addition of salt to SDS-con-taining lysates (Miller et al. 1988). In either case, the proteinsare removed by centrifugation followed by recovery of theDNA-containing supernatant. Note that if nucleic acids arenot first liberated from viral capsids, ammonium sulfate canresult in the precipitation of the intact virions. This method ofconcentrating viruses has useful applications in molecularbiology (Ziai et al. 1988) and viral ecology (Steward et al.1992) and appears to be the basis of viral preservation inRNALater, as mentioned above.

Instead of precipitating protein, DNA can be selectively pre-cipitated from buffers of low ionic strength with the cationic sur-factant cetyltrimethylammonium bromide (CTAB) (Jones 1953).In this case, the proteins are discarded with the supernatant, andthe DNA in the pellet is resuspended in a high-ionic-strengthbuffer (Sambrook and Russell 2001). Note that in high-ionic-strength buffers, CTAB forms complexes with proteins and poly-saccharides (but not DNA), which has been used to facilitate theremoval of these contaminants by organic extractions with phe-nol and chloroform (Jones and Walker 1963).


158

Solid-phase extraction: One of the most common extractionand purification techniques in use today is a solid-phase extrac-tion, which exploits the selective binding of nucleic acids to sil-ica under conditions of high salt concentration and low pH,and their subsequent elution at low salt concentrations (Vogel-stein and Gillespie 1979; Boom et al. 1990). This phenomenonis the basis for a wide variety of commercial nucleic acid extrac-tion kits, in which the silica is supplied as a fine particle sus-pension (“glass milk”) or a silica-impregnated membrane.

Density gradients: Nucleic acids can be very effectively andcleanly separated from other macromolecules using densitygradient centrifugation. DNA, RNA, proteins, and lipids havesufficiently different buoyant densities that they can be sepa-rated in equilibrium buoyant density gradients in an ultra-centrifuge (Rickwood 1989). CsCl and CsTFA are commonlyused as gradient media for this purpose. DNA can be bandedin gradients of either salt. RNA, because of its high buoyantdensity, will pellet in CsCl gradients (Glisin et al. 1974) butcan be banded in CsTFA (Rickwood 1989). Isopycnic bandingresults in very pure nucleic acids, and this is a reliable way toobtain nucleic acids free of contaminants that can inhibitenzymatic reactions such as PCR. The disadvantages of ultra-centrifugation are the relatively long centrifugation times,the limited number of samples that can be processed simul-taneously, and the effort needed to recover the nucleic acidsfrom the gradient.

Electrophoretic separation: A novel electrophoresis tech-nique has recently been described that, like the density gradi-ents described above, appears to result in highly purified DNA.This method, referred to as synchronous coefficient of dragalteration (SCODA), very effectively separates nucleic acidsfrom contaminants and simultaneously concentrates thembased on their nonlinear response to variable electric fields(Marziali et al. 2005). At present, the instrument used for thistechnique can process only one sample at a time, so through-put is very limited. However, low-throughput, high-puritymethods such as SCODA and ultracentrifugation are invalu-able for special applications in which one must extract nucleicacids from challenging matrices rich in PCR inhibitors(Juniper et al. 2001; Pel et al. 2009).

E. Commercial extraction kits and reagents—Commercialpurification kits or reagents are available that rely on theextraction principles outlined above of selective precipitation(e.g., MasterPure™, Epicenter; Gentra® Puregene®, Qiagen),selective adsorption (e.g., UltraClean® Microbial DNA isolationkit, Mo Bio Laboratories; QIAamp® MinElute®, Qiagen; All-Prep®, Qiagen) or selective solubility (e.g., TRI Reagent®, Molec-ular Research Center; TRIzol®, Invitrogen). Although some kitsare specifically marketed for extraction of nucleic acids fromviruses (e.g., ChargeSwitch® EasyPlex™ Viral RNA/DNA Kit,Invitrogen; QIAamp® UltraSens™ Virus Kit, Qiagen; ArcPure™Viral DNA [or RNA] Isolation and Sample Preparation Kit,Arcxis Biotechnologies), these kits are not inherently selectivefor viruses; rather, they assume a cell-free virus-containing fluid

as the starting material. The underlying extraction principlesare the same as for cell and tissue extraction kits, but the pro-tocols can be simpler, because of the relative ease with whichnucleic acids can be released from viral capsids.

Some kits or reagents discriminate between RNA and DNA(e.g., AllPrep, TRI reagent, and TRIzol) and allow separatepurification of both types of nucleic acid from the same sam-ple. Other kits result in purification of total nucleic acids (e.g.,MasterPure and Puregene), with RNA and DNA being discrim-inated only by selective nuclease digestion. The latterapproach is less desirable if one has a limited amount of mate-rial, since a significant portion of the DNA and RNA must bedestroyed to get pure fractions of each. Compensating for thisdrawback are the simplicity (no special columns) and low tox-icity (no organic solvents) of the approach. One should beaware that, although the names of some kits suggest speci-ficity for DNA or RNA, the procedure may not be selective.One of the authors (A. I. Culley) has found, for example, thata kit marketed for RNA virus extraction (QIamp viral RNAMini kit, Qiagen) works as well for extracting viral DNA. Theproduct literature should be consulted to be sure of the limitsand selectivity of each kit.

Some kits are specifically designed to remove inhibitorsthat may be found in more complex matrices such as soil. Thisis not an issue for many aquatic viral ecology applications, butin some cases (e.g., extracting viral nucleic acids from totalplankton concentrates or from sediments), a kit designed forsoil (e.g., Power Soil Kits, Mo Bio Laboratories) may helpremove substances that can inhibit PCR.

Another option for simultaneously extracting and preserv-ing small-volume virus samples is to spot them on FTA cards(Whatman). These cards are impregnated with buffer, chelat-ing agent, detergent, and uric acid that serve to lyse microbesand protect the nucleic acids (Burgoyne 1996). These cardshave been used for preserving nucleic acids from a wide vari-ety of microorganisms (Rajendram et al. 2006), includingRNA-containing (Li et al. 2004) and DNA-containing (Sud-hakaran et al. 2009) viruses. Nucleic acids have been detectedfrom samples stored for more than 4 years at room tempera-ture with minimal decay (Li et al. 2004). One limitation of thisapproach is that the volume that can be applied is relativelysmall (≤500 µL), since the sample must be absorbed by thepaper without excessively diluting the reagents and then bedried completely.

To decide which commercial kits or reagents are mostappropriate, one needs to consider the type (DNA, RNA),mass, and size of the nucleic acids to be extracted and the finalpurity required. Most kits, particularly those based on selectiveadsorption, have limitations on the mass and size of thenucleic acids that can be efficiently recovered. We do notcover the protocols of these kits here, since the brands arenumerous and the protocols are supplied with each kit.Instead we present a few manual purification protocols for sit-uations not covered by the kits.


159

Protocols

A. Phenol-chloroform extraction with ethanol precipitation—The traditional phenol extraction procedure, based on earlywork by Kirby (1957) and elaborated in most modern protocolcompendia, yields very clean nucleic acids suitable for a vari-ety of downstream applications. This extraction method iscoupled with a routine alcohol precipitation step to allowbuffer exchange, removal of trace amounts of chloroform, andconcentration of nucleic acids (Sambrook and Russell 2001).Materials and equipment:

• fume hood• microcentrifuge (refrigerated if possible)• pipettes and sterile, disposable tips• safety gear (gloves, lab coat, safety glasses)• sterile microcentrifuge tubes• TE buffer (10 mM Tris, 1 mM EDTA; pH 8)• Tris-saturated phenol, pH 8 (see “Warning” below)• CI (chloroform:IAA, 24:1, vol:vol) (see “Warning”)• PCI (phenol:chloroform:IAA, 25:24:1) (see “Warning”)• sodium acetate, 3 M, pH 5.2 (see Sambrook and Russell

[2001] for a discussion of alternative salts that may beused for nucleic acid precipitation and their advantagesand disadvantages)

• Optional: polyacryl carrier (Molecular Research Center)• ethanol, 70% and 100%Warning: Phenol can cause chemical burns if it comes in

contact with bare skin. Phenol and chloroform are volatileand carcinogenic and must be used in a fume hood withproper protection (gloves, lab coat, and safety glasses). PCIand CI preparations that are ready to use can be purchasedfrom a variety of scientific chemical suppliers. Details of howto prepare these solutions for oneself can be found elsewhere(Sambrook and Russell 2001).Steps:

1. To the viral suspension (≤0.6 mL per 2-mL microcen-trifuge tube, scale up for larger volumes) add an equalvolume of PCI and shake to emulsify.

2. Centrifuge at 10,000g for 5 min to facilitate separationof the organic and aqueous phases.

3. Transfer the DNA-containing aqueous phase (upper) toa new tube by aspiration with a pipette, being careful toavoid material at the interface.

4. Repeat steps 1–3 as needed until the interface appears tobe free of extracted material (one extraction may sufficefor relatively pure viral preparations).

5. Add an equal volume of CI to the aqueous phase andshake to emulsify.

6. Centrifuge as in step 2 to separate phases.7. Transfer the aqueous phase (upper) to a new tube.8. Add 1 µL polyacryl carrier (this optional step is unnec-

essary when working with tens of nanograms or moreof DNA, but can improve yields when working withnanogram to subnanogram quantities).

9. Add 1/10 of a volume of sodium acetate and invert tubeor vortex to mix.

10. Add 2 volumes of ethanol and invert tube to mix.11. Incubate sample on ice for 10 min.12. Centrifuge for 10 to 30 min, at 0–4°C if possible.13. Aspirate or decant the supernatant, being careful not to

disturb the pellet (a pellet may not be visible if theamount of DNA is low and no carrier has been added).

14. Add 500 µL ice-cold 70% ethanol.15. Centrifuge at 10,000g for 10 min.16. Decant or aspirate supernatant as completely as possi-

ble, being careful not to disturb the pellet.17. Allow residual liquid in the tube to evaporate by air-dry-

ing with the cap open and the tube upside down or byplacing briefly in a centrifugal vacuum concentrator(e.g., SpeedVac concentrator, Thermo Scientific; Con-centrator plus, Eppendorf). Note that excessive dryingwill make the nucleic acid more difficult to dissolve.

18. Resuspend the dried pellet in a small volume of Tris (10mM, pH 8) or TE buffer. Note that some of the materialwill be on the side of tube, so the appropriate side of thetube should be exposed to the liquid used for resuspen-sion to maximize recovery.

19. The purified, solubilized DNA may be stored at 4°C forshort periods of time, at –20°C for long periods of time,and at –80°C indefinitely. For long-term storage, onemight also consider storing the dried DNA pellet, whichshould remain stable at room temperature or below ifkept dry.

Assessment: This traditional method of extraction is mostcommonly used to extract DNA. When phenol is saturatedwith alkaline buffer (e.g., Tris, pH 8), however, both RNA andDNA will partition to the aqueous phase, so the method canbe used for total nucleic acid extraction. Either DNA or RNAcan be specifically selected by digestion of the recovered totalnucleic acids with RNase or DNase. If targeting RNA, anRNase inhibitor may be included to help ensure stability. Ifone wishes to isolate both RNA and DNA but in separate frac-tions by organic extraction, we recommend the use of thecommercially available reagents TRI Reagent and TRIzol (seeabove).

Although still in use, the popularity of organic extractionhas waned somewhat as new extraction procedures have beendeveloped that do not require the use of toxic organic com-pounds. In addition to the extra precautions that must betaken when handling phenol and chloroform during extrac-tion, the disposal of the resulting organic waste is costly.

B. Release of nucleic acids with heat, chelator, and detergent—Ifone has a purified stock of viruses obtained, for example, bybanding in a buoyant density gradient, or even a relativelypure viral concentrate obtained by size fractionation, it is pos-sible to release the DNA in a high molecular weight form suit-able for some applications (e.g., pulsed-field gel electrophore-sis [PFGE] for sizing or probing, or nucleic acid quantification


160

by fluorescence) relatively simply. This method involvesexchanging the buffer in which the viruses are suspended withone containing EDTA and SDS, followed by heating. Themethod is similar to that described previously (Steward 2001),but with the optional addition of detergent to facilitate disin-tegration of the viral capsid.Materials and equipment:

• Centrifugal ultrafiltration device (30,000 molecularweight cutoff, e.g., Millipore Ultracel YM-30, cat no.42410)

• TE buffer (10 mM Tris, 1 mM EDTA, pH 8) or TEGEDbuffer (10 mM Tris, 1 mM EDTA, 1 mM EGTA)

• Optional: 6× SDS-EDTA loading buffer (1% SDS, 60 mMEDTA, 0.03% bromophenol blue, 0.03% xylene cyanol,60% glycerol)

Steps: 1. Concentrate the viruses by centrifuging in the centrifu-

gal ultrafiltration device at 1000g until only a small vol-ume (ca. 10 µL) remains.

2. Add 100 µL TE (or TEGED).3. Concentrate the sample again to ca. 10 µL.4. Repeat steps 2 and 3 once more.5. Recover the final concentrate.6. Rinse the membrane in the device by adding a small

volume of TE or TEGED (5–10 µL).7. Recover the rinse and pool with the concentrate.8. Optional: If conducting electrophoresis on the sample,

add SDS-EDTA loading dye to a final concentration of1×.

9. Heat the recovered sample (with or without loadingbuffer) to 60°C for 10 min to release the nucleic acid.

Assessment: This method is similar in strategy to the sim-ple protocol for assaying the DNA content of bacteriophage λstocks described by Sambrook and Russell (2001, p.2.45–2.46). SDS is added to facilitate the release of DNA fromthe capsids and to minimize DNA–protein interactions dur-ing electrophoresis. The protocol described here, but withoutSDS, has been used in a number of studies to investigategenome size distributions in viral communities (e.g., Stewardand Azam 2000; Steward et al. 2000; Riemann and Middelboe2002; Jiang et al. 2003; Jiang et al. 2004; Filippini and Mid-delboe 2007). One can carry out essentially the same proce-dure by pelleting viruses in an ultracentrifuge rather thanusing centrifugal ultrafiltration. If one assumes a minimumsedimentation coefficient for viruses (e.g., 80S), the timeneeded to pellet the virus can be determined from the k-factorof the rotor being used (Lawrence and Steward 2010, this vol-ume). The centrifugation time required in a swinging bucketrotor, which will produce the most compact pellet, can varyfrom 30 min (6 × 4 mL sample in a Beckman SW 61 rotor) to3 h (6 × 38.5 mL sample in a Beckman SW 28 rotor). In thiscase, the supernatant is drained completely and carefullyfrom the pellet. Residual liquid on the walls can be removedusing the tip of a twisted lint-free absorbent wipe (e.g.,

KimWipe, Kimberly Clark) or sterile cotton swab. TE is addedto the pellet, and the tube is sealed with plastic wrap to min-imize evaporation and heated to 60°C with occasional gentleagitation for 10 to 15 min.

If the samples are handled carefully (to minimize shearing,pipette slowly, use wide-bore pipette tips, and avoid vortexmixing), the DNA should be of high molecular weight suitablefor sizing by PFGE (Steward et al. 2000; Steward 2001). DNAprepared by the centrifugal ultrafiltration method has resultedin no noticeable shearing of bands up to several hundredthousand base pairs. The alternative ultracentrifugationmethod has sometimes resulted in slight smearing of bands,indicating some shearing. Even in the former case, some smallamount of shearing of the higher molecular weight nucleicacids might be expected from handling them in solution.

The sensitivity of viral DNA to shearing will depend on itssize, composition, and conformation. Most viral genomes aresmall enough that they can be extracted in solution withoutappreciable shearing if handled gently. Bacteriophagegenomes up to 100 kb produced crisp bands with no evidenceof shearing when extracted using a protocol similar to thatdescribed here (Steward et al. 2000). A large algal virus genome(320 kb) was found to be fragmented when subjected to stan-dard phenol-chloroform extraction procedures (Lanka et al.1993). When treated gently, however, Chlorella viruses rang-ing in size up to 380 kb tolerated limited pipetting in liquidand produced crisp single bands by PFGE (Rohozinski et al.1989). If shearing must be minimized to the greatest possibleextent, one should consider embedding the viruses beforeextraction as described in protocol C. Even embedded viralDNA, however, may be susceptible to some degree of frag-mentation (Lanka et al. 1993), perhaps due to premature dis-integration of viruses during the embedding process (see pro-tocol C, “Assessment,” below).

It may be possible to obtain intact viral RNA using the samegeneral approach as described here by simply including anRNase inhibitor in the TE buffer, but we have not explicitlytested this. If more purified nucleic acids are required, the sim-ple release step described here can be followed by purificationby organic extraction (Cottrell and Suttle 1991; Wilson et al.1993; Sambrook and Russell 2001) or purification with any ofa number of commercially available nucleic acid purificationkits. If one wishes to purify both RNA and DNA free from pro-teins, the appropriate extraction buffer from a suitable kit(QIAamp MinElute or UltraSens virus kits, Qiagen; MasterPuretotal nucleic acid extraction kit, Epicenter) can be substitutedfor the TE after concentrating the viruses by centrifugal ultra-filtration or ultracentrifugation.

C. Extracting DNA from viruses embedded in agarose—If onewishes to have a stock of high molecular weight viral DNAthat can be stored for long periods of time with minimalshearing or degradation, the viruses can be embedded inagarose before extraction. Extraction of embedded cells is thestandard procedure for sizing the genomes of bacteria and


161

yeast by PFGE (Sambrook and Russell 2001), and a similar pro-tocol can be used for embedded viruses (Rohozinski et al.1989; Lanka et al. 1993; Wommack et al. 1999; Sandaa et al.2010, this volume).Materials and equipment:

• Agarose (molecular biology grade, low gelling tempera-ture; InCert® Agarose, Lonza)

• SE buffer (75 mM NaCl, 25 mM EDTA, pH 8)• TE buffer (10 mM Tris, 1 mM EDTA, pH 8)• Lysis buffer (TE pH 8; 1% SDS)• Proteinase K• Optional: casting wells• Optional: phenylmethanesulfonylfluoride (PMSF)

Steps: 1. Add agarose to SE buffer for a final concentration of

1.5% (wt/vol).2. Melt the agarose in a microwave oven, then cool and

maintain at 37°C in a water bath.3. Warm the viral concentrate to 37°C in the water bath,

then immediately mix with an equal volume of moltenagarose and quickly transfer the mixture to castingmolds. Casting in rectangular plug molds is preferred ifthe embedded DNA is to be analyzed by electrophore-sis, since this results in plugs of uniform height andthickness that fit the wells without extensive trimming.Electrophoresis using plugs that are not uniform willresult in bands with uneven intensity. Special rectangu-lar plug molds for PFGE are available from BioRad.Alternatively, one can draw the molten mixture into a1-cc syringe that has had the tip cut off, or simplypipette the mixture as drops onto a sheet of plastic wrapor Parafilm M®.

4. Once the agarose has gelled, transfer the plug (or noo-dle from the syringe, or buttons from the parafilm), intoa tube containing 5 volumes of lysis buffer amendedwith proteinase K (1 mg mL–1 final concentration).

5. Incubate at room temperature overnight.6. Decant the lysis buffer, being careful not to lose the

plugs.7. Optional: Rinse the plugs twice, each time adding 25

volumes of fresh TE containing 1 mM PMSF, incubatingfor 1 h with gentle agitation, then decanting the rinsefluid. This step will inactivate the proteinase K, which isrecommended if the DNA in the plug is to be furthermanipulated with enzymes (e.g., digestion with restric-tion endonucleases).

8. Rinse the plugs twice, each time adding 50 volumes offresh TE with no PMSF, incubating for 30 min with gen-tle agitation, then decanting the rinse fluid.

9. Store the plugs at 4°C submerged in TE.Assessment: One of the main purposes of extracting nucleic

acids from embedded viruses or cells is to avoid shearing ofhigh molecular weight DNA. The use of a low-melting/gelling-point agarose in SE buffer is recommended to minimize pre-

mature disruption of viral capsids by thermal and osmoticshock. DNA released before the casting of the gel plugs has thepotential to be sheared during mixing and pipetting.

Use of a low-gelling-temperature agarose also allows one torecover nucleic acids from the agarose plug using an agaraseenzyme (β-agarase, Lonza or New England BioLabs). DNA canbe recovered from other types of agarose using silica-based gelextraction methods, by electroelution, or by organic extrac-tion (Sambrook and Russell 2001), so it is possible to usemolecular biology–grade agaroses with higher gelling temper-atures. In this case, however, the agarose must be maintainedat a higher temperature before mixing with the sample.Gelling temperatures for other pulsed-field grade agaroses arearound 36–42°C, so maintaining at 50–60°C before mixingwith sample should be adequate. One should bear in mindthat some viruses may disintegrate at this temperature. Forsome applications (e.g., shotgun cloning), some fragmenta-tion of the DNA is not an issue. If sheared DNA is not an issuefor one’s application, then one might consider a less cumber-some extraction protocol that results in DNA in solution.

The release of viral DNA in plugs is commonly used for siz-ing of large viral genomes either intact (McCluskey et al. 1992)or after digestion with a restriction endonuclease (Rohozinskiet al. 1989; Lanka et al. 1993). Variations of the above methodhave been used for analyses of genome size distributions inviral assemblages using PFGE (e.g., Wommack et al. 1999;Larsen et al. 2001; Øvreås et al. 2003; Sandaa and Larsen 2006;Parada et al. 2008; Sandaa et al. 2010, this volume). Viral com-munity DNA has also been recovered from agarose plugs forsubsequent sequence analysis by shotgun cloning (Bench et al.2007).

The disadvantages of the method for community genomesize analyses are that the preparation time is longer and theresolution of bands will typically be lower when performingPFGE from viral DNA in plugs (depending on the thickness ofthe plug) compared to that achievable with DNA in solutionprepared by protocol B (Steward 2001). The considerableadvantage of the method is that the DNA appears to be morestable at 4° when embedded in agarose (many months) thanwhen dissolved in buffer (up to a few days), so embedding isrecommended for storing extracted viral DNA that will not beused right away. One of the authors (G. F. Steward) hasobserved that a high molecular weight PFGE standard embed-ded in agarose that was accidentally frozen on dry ice resultedin a banding pattern that was indistinguishable from that ofparallel standard that had never been frozen. Freezing theplugs at –80°C may therefore be useful for long-term archivingof samples. Freezing is not recommended for samples that willbe accessed more than once or twice, since repeated freeze-thaw cycles are likely to degrade the DNA.

D. Extracting nucleic acids from viruses on a filter—This pro-tocol is a minor modification of that reported by Culley andSteward (2007). As the starting point for this protocol, weassume that viruses have been collected on an aluminum


162

oxide 0.02-µm syringe-tip filter (Anotop, Whatman), butother filters capable of capturing viruses may be substituted.Materials and equipment:

• Total nucleic acid extraction kit (MasterPure, Epicenter)• Optional: polyacryl carrier (Molecular Research Center)• Syringes (sterile, disposable, with Luer-Lok tips)• Luer-Lok female–female adapter fittings• Hybridization oven with rotisserie

Steps: 1. Add 1 mL T + C lysis buffer containing 100 µg/ml pro-

teinase K to a low-volume (1–3 cc) syringe that has beenfitted with a female–female Luer-Lok adapter (the injec-tion syringe). It is convenient to use a larger syringe(10–20 cc) as an extraction buffer reservoir. The injec-tion syringe can be easily filled with the proper volumeby connecting it tip to tip with the reservoir syringe viathe adapter. The reservoir can be used to fill multipleinjection syringes if more than one sample is to beextracted.

2. Ensure that there is minimal air in the injectionsyringe-adapter assembly, then connect it to the outletof the filter. Connect a second low-volume syringe tothe filter inlet (the aspiration syringe). Hold the filter-syringe assembly vertically with the injector syringepushing upward from below. Hold the filter securely tothe injection syringe and gently, but firmly, pushextraction buffer into the filter housing until liquid justbegins to appear in the aspiration syringe.

3. Incubate the assembly (filter with two syringesattached) for 15 min at 65°C in a hybridization oven. Itis helpful to connect the syringe-filter assembly to arotisserie so that the entire filter surface is wetted in theevent that bubbles are present in the housing. Thesyringes on either side of the filter can be secured to theclips of the rotisserie with elastic bands.

4. Allow the syringe-filter assembly to cool briefly; thenremove the extract by holding the syringe assembly ver-tically with the aspiration syringe underneath (and thefilter upside down) and gently pulling on the plunger topull the extract into the aspiration syringe.

5. Detach the aspiration syringe; transfer the extract to amicrocentrifuge tube; and chill on ice for 2–3 minutes.

6. Add one-half volume of MPC protein precipitationreagent (supplied in the kit) and vortex for 10 s.

7. Pellet the debris by centrifugation at 10,000g for 10 min. 8. Transfer the supernatant (containing the nucleic acids)

to a sterile microcentrifuge tube; be very careful toavoid the pellet (containing the SDS-protein complex).All or some of the sample may be archived at this pointby freezing at –80°C.

9. Transfer up to 800 µl of the sample to a fresh tube; add1 µl polyacryl carrier, and vortex briefly (carrier isoptional, but can improve yields when working withnanogram to subnanogram quantities of nucleic acid).

10. Add an equal volume of 100% isopropanol and mix byinverting the tube several times.

11. Centrifuge the sample at ≥10,000g for 15–45 min(longer centrifugation times can improve the yields forsmall amounts of nucleic acids, especially in theabsence of carrier).

12. Decant or aspirate the supernatant (use caution, the pel-let can dislodge easily and be lost).

13. Wash the pellet twice, each time adding 70% ethanol,centrifuging for 1 min, and decanting (or aspirating)the ethanol.

14. Air-dry the pellet, then dissolve in 10 µL of 0.02-filtered, sterile 0.5× TE buffer heated to 50°C.

15. If required, DNA or RNA can be selectively removedfrom the total nucleic acid precipitate by enzymaticdigestion with DNase or RNase.

Assessment: One caveat in extracting from aluminumoxide membrane filters is that they can irreversibly bindDNA under certain conditions (Dames et al. 2006). In partic-ular, guanidinium-containing extraction buffers facilitatethe binding of DNA to aluminum oxide (Gerdes et al. 2001)and are likely to result in low yields from the filters. For thisreason, we do not recommend extracting from Anotop filtersusing the lysis buffers from any of the popular silica col-umn–based kits. The Gentra PureGene Kit (Qiagen) is similarto the MasterPure kit and may work as well. These latter kitsare based on a published protocol (Miller et al. 1988) thatcould be adapted for extraction from a filter. A version ofthat protocol designed for simultaneous DNA and RNAextraction (Yu and Mohn 1999) could also be used byemploying heat (65°C, 15 min) instead of bead beating tofacilitate lysis. It is worth noting that SDS and phosphatebuffer appear to inhibit the binding of nucleic acid to alu-minum oxide (Gerdes et al. 2001; Dames et al. 2006) andshould aid in recovery. If one wishes to recover RNA using aself-made recipe, then we would recommend including anRNAse inhibitor (e.g., RNASecure, Ambion) in the extractionbuffer.

We have recovered both viral RNA and DNA suitable forPCR amplification from aluminum oxide filters using essen-tially the procedure as described above (Culley and Steward2007; Culley et al. 2008). A modification added here is theintroduction and removal of the extraction buffer in a direc-tion counter to that of filtration during sample collection (i.e.,backflushing). According to the Anotop specifications, thesefilters are not designed to be backflushed, or to be operated attemperatures above 40°C. We have found that the filter canrupture if too much pressure is applied during backflushing,especially after incubating at 65°C. Injecting the extractionbuffer slowly and removing the extract by gentle aspirationseem to avoid this problem. Although we have not tested theprotocol described here on other filter types, the procedureshould work as well for any direct flow filter capsule capableof retaining viruses.


163

Conclusion

Extraction of nucleic acids from viruses may be achieved bya wide variety of methods. Essentially any protocol that is suit-able for extracting nucleic acids from cells should work as wellfor viruses. Because viruses do not have a cell wall, however,extractions can be much quicker and simpler. The nature ofthe starting material and final purity of the nucleic acidsrequired will vary widely depending on the application, so itis not possible to provide a single recommended protocol. Wehope that the background information and the handful ofexplicit protocols provided here will arm the reader with theinformation necessary to select, adapt, or design a protocolbest suited to their needs and the materials and equipmentthey have available.

ReferencesAdams, R. L. P., J. T. Knowler, and D. P. Leader. 1992. The bio-

chemistry of the nucleic acids, 11th ed. Chapman & Hall.Anderson, N. G., G. B. Cline, W. W. Harris, and J. G. Green.

1967. Isolation of viral particles from large fluid volumes, p.75-88. In G. Berg [ed.], Transmission of viruses by the waterroute. Interscience Publishers.

Anderson, T. F. 1950. Destruction of bacterial viruses byosmotic shock. J. Appl. Phys. 21:70.

———, C. Rappaport, and N. A. Muscatine. 1953. On thestructure and osmotic properties of phage particles. Ann.Inst. Pasteur 84:5-15.

Bench, S. R., and others. 2007. Metagenomic characterizationof Chesapeake Bay virioplankton. Appl. Environ. Micro-biol. 73:7629-7641.

Berger, S. L., and C. S. Birkenmier. 1979. Inhibition ofintractable nucleases with ribonucleoside-vanadyl com-plexes: Isolation of messenger ribonucleic acid from restinglymphocytes. Biochemistry 18:5143-5149.

Blackburn, P., G. Wilson, and S. C. Moore. 1977. Ribonucleaseinhibitor from placenta: Purification and properties. J. Biol.Chem. 252:5904-5910.

Boom, R., C. J. A. Sol, M. M. M. Salimans, C. L. Jansen, P. M.E. Wertheim-van Dillen, and J. Van der Noorda. 1990.Rapid and simple method for purification of nucleic acids.J. Clin. Microbiol. 28:495-503.

Brady, J. N., V. D. Winston, and R. A. Consigli. 1977. Dissoci-ation of polyoma virus by the chelation of calcium ionsfound associated with purified virions. J. Virol. 23:717-724.

Brakke, M. K. 1963. Stabilization of brome mosaic virus bymagnesium and calcium. Virology 19:367-374.

Brum, J. R., and G. F. Steward. Morphological characterizationof viruses in the stratified water column of alkaline, hyper-saline Mono Lake. Microb. Ecol. Accepted 2010.

Burgoyne, L., inventor; Flinders Technologies Pty., Australia,assignee. Solid medium and method for DNA storage. USpatent 5,496,562. 1996 Mar 5.

Chirgwin, J. M., A. E. Przybyla, R. J. MacDonald, and W. J. Rut-

ter. 1979. Isolation of biologically active ribonucleic acid fromsources rich in ribonuclease. Biochemistry 18:5294-5299.

Chomczynski, P. 1993. A reagent for the single-step simulta-neous isolation of RNA, DNA and proteins from cell and tis-sue samples. Biotechniques 15:532-537.

Clive, D. O. 1967. Enterovirus detection by membrane chro-matography, p. 139-141. Transmission of viruses by thewater route. Interscience Publishers.

Cottrell, M. T., and C. A. Suttle. 1991. Wide-spread occurrenceand clonal variation in viruses which cause lysis of a cos-mopolitan, eukaryotic marine phytoplankter, Micromonaspusilla. Mar. Ecol. Prog. Ser. 78:1-9.

Culley, A. I., B. A. Asuncion, and G. F. Steward. 2009. Detec-tion of inteins among diverse DNA polymerase genes ofuncultivated members of the Phycodnaviridae. ISME J.3:409-418.

——— and G. F. Steward. 2007. New genera of RNA viruses insubtropical seawater inferred from polymerase genesequences. Appl. Environ. Microbiol. 73:5937-5944.

Dames, S., L. K. Bromley, M. Erali, R. Smith, and K. V. Voelk-erding. 2006. A single-tube nucleic acid extraction, amplifi-cation, and detection method using aluminum oxide. J.Mol. Diag. 8:16-21.


Drake, L., K. Choi, A. Haskell, and F. Dobbs. 1998. Vertical pro-files of virus-like particles and bacteria in the water columnand sediments of Chesapeake Bay, USA. Aquat. Microb.Ecol. 16:17-25.

Farrell, R. E. 2005. RNA Methodologies: A laboratory guide forisolation and characterization, 3rd ed. Elsevier Academic.

Faulkner, P. 1962. Isolation and analysis of ribonucleic acidfrom inclusion bodies of the nuclear polyhedrosis of thesilkworm. Virology 16:479-484.

Filippini, M., and M. Middelboe. 2007. Viral abundance andgenome size distribution in the sediment and water col-umn of marine and freshwater ecosystems. FEMS Micro-biol. Ecol. 60:397-410.

Forster, J. L., V. B. Harkin, D. A. Graham, and S. J. McCullough.2008. The effect of sample type, temperature andRNAlaterTM on the stability of avian influenza virus RNA. J.Virol. Meth. 149:190-194.

Fraenkel-Conrat, H., B. Singer, and R. C. Williams. 1957. Infec-tivity of viral nucleic acid. Biochim. Biophys. Acta 25:87-96.


Furneaux, R. C., W. R. Rigby, and A. P. Davidson. 1989. Theformation of controlled-porosity membranes from anodi-cally oxidized aluminum. Nature 337:147-149.


164

Gerba, C. P. 1983. Methods for recovering viruses from thewater environment, p. 19-35. In G. Berg [ed.], Viral pollu-tion of the environment. CRC Press.

Gerdes, J. C., J. M. Marmaro, C. A. Roehl, inventors; Xtrana,assignee. Nucleic acid archiving. US patent 6,291,166. 2001Sep 18.

Glisin, V., R. Crkvenjakov, and C. Byus. 1974. Ribonucleic acidisolated by cesium chloride centrifugation. Biochemistry13:2633-2637.

Gorokhova, E. 2005. Effects of preservation and storage ofmicrocrustaceans in RNAlater on RNA and DNA degrada-tion. Limnol. Oceanogr. Methods 3:143-148.

Helton, R. R., and K. E. Wommack. 2009. Seasonal dynamicsand metagenomic characterization of estuarine virioben-thos assemblages by randomly amplified polymorphic DNAPCR. Appl. Environ. Microbiol. 75:2259-2265.


Jiang, S., W. Fu, W. Chu, and J. A. Fuhrman. 2003. The verti-cal distribution and diversity of marine bacteria at a stationoff Southern California. Microb. Ecol. 45:399-410.

Jiang, S. C., G. F. Steward, R. Jellison, W. Chu, and S. Choi. 2004.Abundance, distribution and diversity of viruses in alkaline,hypersaline, Mono Lake, California. Microb. Ecol. 47:9-17.

Jones, A. S. 1953. The isolation of bacterial nucleic acids usingcetyltrimethylammonium bromide (Cetavlon). Biochim.Biophys. Acta 10:607-612.

———, and R. T. Walker. 1963. Isolation and analysis of thedeoxyribonucleic acid of Mycoplasma mycoides var. capri.Nature 198:588-589.

Juniper, S. K., M.-A. Cambon, F. Lesongeur, and G. Barbier. 2001.Extraction and purification of DNA from organic rich sub-surface sediments (ODP Leg 169S). Mar. Geol. 174:241-247.

Kirby, K. S. 1956. A new method for the isolation of ribonucleicacids from mammalian tissues. Biochem. J. 64:405-408.

———. 1957. A new method for the isolation of deoxyribonu-cleic acids: Evidence on the nature of bonds betweendeoxyribonucleic acid and protein. Biochem. J. 66:495-504.

Labonté, J. M., K. E. Reid, and C. A. Suttle. 2009. Phylogeneticanalysis indicates evolutionary diversity and environmen-tal segregation of marine podovirus DNA polymerase genesequences. Appl. Environ. Microbiol. 75:3634-3640.

Lader, E. S., inventor; Ambion, assignee. Methods and reagentsfor preserving RNA in cell and tissue samples. US patent6,204,375. 2001 Mar 20.

Lanka, S. T. J., M. Klein, U. Ramsperger, D. G. Müller, and R. Knip-pers. 1993. Genome structure of a virus infecting the marinebrown alga Ectocarpus siliculosis. Virology 193:802-811.

Larsen, A., and others 2001. Population dynamics and diver-sity of phytoplankton, bacteria and viruses in a seawaterenclosure. Mar. Ecol. Prog. Ser. 221:47-57.

Lawrence, J. E., A. M. Chan, and C. A. Suttle. 2002. Virusescausing lysis of the toxic bloom-forming alga Heterosigmaakashiwo (Raphidophyceae) are widespread in coastal sedi-ments of British Columbia, Canada. Limnol. Oceanogr.47:545-550.

———, and G. F. Steward. 2010. Purification of viruses by cen-trifugation, p. 166-181. In S. W. Wilhelm, M. G. Weinbauer,and C. A. Suttle [eds.], Manual of Aquatic Viral Ecology. ASLO.

Leroy, M., M. Prigent, M. Dutertre, F. Confalonieri, and M.Dubow. 2008. Bacteriophage morphotype and genome diver-sity in Seine River sediment. Freshw. Biol. 53:1176-1185.

Li, C.-C., I. A. Beck, K. D. Seidel, and L. A. Frenkel. 2004. Per-sistence of human immunodeficiency virus type 1 subtypeB DNA in dried-blood samples on FTA filter paper. J. Clin.Microbiol. 42:3847-3849.

Marziali, A., J. Pel, D. Bizzotto, and L. A. Whitehead. 2005.Novel electrophoresis mechanism based on synchronousalternating drag perturbation. Electrophoresis 26:82-90.

Mason, P. E., G. W. Neilson, C. E. Dempsey, A. C. Barnes, andJ. M. Cruickshank. 2003. The hydration structure of guani-dinium and thiocyanate ions: Implications for protein sta-bility in aqueous solution. Proc. Natl. Acad. Sci. U. S. A.100:4557-4561.

McCluskey, K., M. V. Graves, D. K. Mills, and R. H. Meints.1992. Replication of Chlorella virus PBCV-1 and host kary-otype determination studied with pulsed-field gel elec-trophoresis. J. Phycol. 28:846-850.

Miller, S. A., D. D. Dykes, and H. F. Polesky. 1988. A simpledesalting procedure for extracting DNA from human nucle-ated cells. Nucl. Acids Res. 16:1215.

Øvreås, L., and others. 2003. Response of bacterial and viralcommunities to nutrient manipulations in seawater meso-cosms. Aquat. Microb. Ecol. 31:109-121.

Parada, V., A.-C. Baudoux, E. Sintes, M. G. Weinbauer, and G.J. Herndl. 2008. Dynamics and diversity of newly producedvirioplankton in the North Sea. ISME J. 2:924-936.

Paul, J. H., S. C. Jiang, and J. B. Rose. 1991. Concentration ofviruses and dissolved DNA from aquatic environments by vor-tex flow filtration. Appl. Environ. Microbiol. 57:2197-2204.

Pel, J., and others. 2009. Nonlinear electrophoretic responseyields a unique parameter for separation of biomolecules.Proc. Natl. Acad. Sci. U. S. A. 106:14796-14801.

Percival, S. L., R. M. Chalmers, M. Embrey, P. R. Hunter, J. Sell-wood, and P. Wyn-Jones. 2004. Microbiology of waterbornediseases. Elsevier.


Putnam, F. W. 1948. The interactions of proteins and syntheticdetergents, p. 79-122. In M. L. Anson and J. T. Edsall [eds.],Advances in protein chemistry. Academic.

Rajendram, D., R. Ayenza, F. M. Holder, B. Moran, T. Long, andH. N. Shah. 2006. Long-term storage and safe retrieval ofDNA from microorganisms for molecular analysis usingFTA matrix cards. J. Microbiol. Methods 67:582-592.


165

Ralph, R. K., and P. L. Bergquist. 1967. Separation of virusesinto components, p. 463-545. In M. K. Maramorosch and H.Koprowski [eds.], Methods in virology. V. 2. Academic Press.

Reynolds, J. A., and C. Tanford. 1970. Binding of dodecyl sul-fate to proteins at high binding rations: Possible implica-tions for the state of proteins in biological membranes.Proc. Natl. Acad. Sci. U. S. A. 66:1002-1003.

Richardson, K. J., A. B. Margolin, and C. P. Gerba. 1988. Anovel method for liberating viral nucleic acid for assay ofwater samples with cDNA probes. J. Virol. Meth. 22:13-21.

Rickwood, D. [ed.]. 1989. Centrifugation: A practicalapproach, 2nd ed. IRL Press.

Riemann, L., and M. Middelboe. 2002. Stability of bacterialand viral community compositions in Danish coastalwaters as depicted by DNA fingerprinting techniques.Aquat. Microb. Ecol. 27:219-232.

Rohozinski, J., L. E. Girton, and J. L. Van Etten. 1989.Chlorella viruses contain linear non-permuted double-stranded DNA genomes with covalently closes hairpinends. Virology 168:363-369.

Ruiz, M. C., and others 2007. Intracellular disassembly ofinfectious rotavirus particles by depletion of Ca2+

sequestered in the endoplasmic reticulum at the end of thevirus cycle. Virus Res. 130:140-150.

Sambrook, J., and D. W. Russell. 2001. Molecular cloning: A lab-oratory manual, 3rd ed. Cold Spring Harbor Laboratory Press.

Sandaa, R.-A., and A. Larsen. 2006. Seasonal variations invirus-host populations in Norwegian coastal waters: Focus-ing on the cyanophage community infecting marine Syne-chococcus spp. Appl. Environ. Microbiol. 72:4610-4618.

———, S. M. Short, and D. C. Schroeder. 2010. Fingerprintingaquatic virus communities, p. 9-18. In S. W. Wilhelm, M. G.Weinbauer, and C. A. Suttle [eds.], Manual of Aquatic ViralEcology. ASLO.

Sobsey, M. D. 1976. Methods for detecting enteric viruses inwater and wastewater, p. 89-127. In G. Berg, H. L. Bodily, E.H. Lennette, J. L. Melnick, and T. G. Metcalf [eds.], Virusesin water. American Public Health Association.

Sreenivasaya, M., and N. W. Pirie. 1938. The disintegration oftobacco mosaic virus preparations with sodium dodecylsulphate. Biochem. J. 32:1707-1710.

Steward, G. F. 2001. Fingerprinting viral assemblages by pulsedfield gel electrophoresis, p. 85-103. In J. H. Paul [ed.],Marine microbiology: Methods in microbiology. Academic.

———, J. Wikner, D. C. Smith, W. P. Cochlan, and F. Azam.1992. Estimation of virus production in the sea: I. Methoddevelopment. Mar. Microb. Food Webs 6:57-78.

———, D. C. Smith, and F. Azam. 1996. Abundance and pro-duction of bacteria and viruses in the Bering and ChukchiSeas. Mar. Ecol. Prog. Ser. 131:287-300.

———, and F. Azam. 2000. Analysis of marine viral assem-blages. In C. R. Bell, M. Brylinski, and P. Johnson-Green[eds.], Microbial biosystems: New frontiers. Proceedings ofthe 8th International Symposium on Microbial Ecology;

1998 Aug 9–14; Halifax, NS, Canada. Atlantic Canada Soci-ety for Microbial Ecology. p. 159-165.


Sudhakaran, R., and others 2009. A simple non-enzymaticmethod for the preparation of white spot syndrome virus(WSSV) DNA from the haemolymph of Marsupenaeus japon-icus using FTA matrix cards. J. Fish Dis. 32:611-617.

Suttle, C. A., A. M. Chan, and M. T. Cottrell. 1991. Use of ultra-filtration to isolate viruses from seawater which arepathogens of marine phytoplankton. Appl. Environ. Micro-biol. 57:721-726.

Uhlenhaut, C., and M. Kracht. 2005. Viral infectivity is main-tained by an RNA protection buffer. J. Virol. Methods128:189-191.

Vega Thurber, R., M. Haynes, M. Breitbart, L. Wegley, and F.Rohwer. 2009. Laboratory procedures to generate viralmetagenomes. Nat. Protoc. 4:470-483.

Vogelstein, B., and D. Gillespie. 1979. Preparative and analyt-ical purification of DNA from agarose. Proc. Natl. Acad. Sci.U. S. A. 76:615-619.

Williamson, S. J., and others. 2008. The Sorcerer II global oceansampling expedition: Metagenomic characterization ofviruses within aquatic microbial samples. PLoS ONE 3:e1456.

Wilson, W. H., I. R. Joint, N. G. Carr, and N. H. Mann. 1993.Isolation and molecular characterization of five marinecyanophages propagated on Synechococcus sp. strainWH7803. Appl. Environ. Microbiol. 59:3736-3743.

Wommack, K. E., R. T. Hill, and R. R. Colwell. 1995. A simplemethod for the concentration of viruses from natural watersamples. J. Microbiol. Methods 22:57-67.

———, J. Ravel, R. T. Hill, J. S. Chun, and R. R. Colwell. 1999.Population dynamics of Chesapeake bay virioplankton:Total-community analysis by pulsed-field gel electrophore-sis. Appl. Environ. Microbiol. 65:231-240.

———, T. Sime-Ngando, D. M. Winget, S. Jamindar, and R. R.Helton. 2010. Filtration-based methods for the collection ofviral concentrates from large water samples, p. 110-117. InS. W. Wilhelm, M. G. Weinbauer, and C. A. Suttle [eds.],Manual of Aquatic Viral Ecology. ASLO.

Yang, S. Y., J. W. Park, J. Yoon, M. Ree, S. K. Jang, and J. K. Kim.2008. Virus filtration membranes prepared fromnanoporous block copolymers with good dimensional sta-bility under high pressures and excellent solvent resistance.Adv. Funct. Mat. 18:1371-1377.

Yu, Z., and W. W. Mohn. 1999. Killing two birds with onestone: Simultaneous extraction of DNA and RNA from acti-vated sludge biomass. Can. J. Microbiol. 45:269-272.

Ziai, R. M., A. Giordano, E. Armandola, and S. Ferrone. 1988.Purification by ammonium sulfate precipitation of bacte-riophage lambda gt11 DNA for restriction analysis ofcloned cDNA inserts. Anal. Biochem. 171:192-196.

166

Centrifugation can be used to concentrate, analyze, andpurify biological particles ranging in size from whole cells tomacromolecules. Ultracentrifuges have played a vital role invirology because they provide sufficient gravitational force(hundreds-of-thousands to over a million × g) to efficiently sed-iment even the smallest viruses. Depending on the rotor used,viruses can be pelleted from homogeneous water samples aslarge as 40 mL in less than 10 min and from volumes greaterthan a liter in less than an hour. Ultracentrifugation of virusesin density gradients is particularly useful, since it allows one toboth concentrate and purify viruses in solution, thereby avoid-ing the problems of resuspending viruses from a pellet.

There are a number of excellent publications detailing theextensive development of ultracentrifugation technique(Anderson and Cline 1967; Brakke 1967; Rickwood 1984).The physico-chemical properties of relatively few marineviruses have been reported to date, but there is a vast amountof pertinent technical information available from previous

studies of terrestrial animal and plant viruses. It is not ourintent to attempt to replace these invaluable publications,but instead to provide an accessible compilation of the mostpertinent information and relevant techniques forresearchers working in aquatic virology. In this chapter, weprovide a broad overview of ultracentrifugation theory,approaches, and techniques. We also provide details of a fewspecific methods that we expect to be most generally usefulto aquatic virologists. These specific methods are still veryflexible according to the equipment available, sample mate-rial, and required outcome, and should be tailored accordingto each application for optimal results. The technical infor-mation provided below will aid the investigator in modify-ing the methods as required to meet his or her needs.

A primer on centrifugation and density gradient separation

The physics of centrifugation—Centrifugation provides ameans for achieving two goals through one approach: parti-cles can be both concentrated and purified under centrifugalforces. To understand how this is achieved, we must first con-sider the physics of a sinking particle. According to Stokes’Law, the sedimentation velocity, v (m s–1), of a falling sphere isgiven by:

v = [d2 (ρp - ρm)g]/18µ (1)

where d is the diameter of the particle, ρp is the density of theparticle (kg m–3), ρm is the density of the liquid medium (kgm–3), µ is the viscosity (Pa s), and g is gravitational acceleration(m s–2). Under ultracentrifugation conditions, centrifugalforces dwarf, and thereby replace, gravitational forces. There-fore, the velocity of a particle under ultracentrifugationbecomes:

v = [d2 (ρp - ρm) ω2 r] / 18µ (2)

Purification of viruses by centrifugationJanice E. Lawrence1* and Grieg F. Steward2

1Department of Biology, University of New Brunswick, 10 Bailey Drive, Fredericton, NB E3B 5A32Department of Oceanography, University of Hawai`i at Manoa, Honolulu, HI 96822

AbstractUltracentrifugation provides a means to concentrate, analyze, and purify viruses in solution, and therefore

represents an invaluable tool for aquatic virologists. This chapter reviews the theory of ultracentrifugation andpresents the technical knowledge necessary for an investigator to adapt or develop methods to meet his or herneeds. Detailed protocols for the purification of viruses from culture lysates and vial assemblages from naturalwater samples are provided.

*Corresponding author: E-mail: [email protected], phone: (506) 458-7842



The writing of this chapter was supported in part by grants fromNSERC to JEL and from NSF to GFS (OCE-04-42664 and OCE-0826650)and to the Center for Microbial Oceanography: Research and Education(EF-0424599). The authors acknowledge the efforts of the anonymouspeer reviewers and their suggestions to improve the manuscript.

ISBN 978-0-9845591-0-7, DOI 10.4319/mave.2010.978-0-9845591-0-7.166Suggested citation format: Lawrence, J. E., and G. F. Steward. 2010. Purification of viruses by cen-trifugation, p. 166–181. In S. W. Wilhelm, M. G. Weinbauer, and C. A. Suttle [eds.], Manual ofAquatic Viral Ecology. ASLO.


MANUALof


Lawrence and Steward Viral purification by centrifugation

167

where ω is the angular velocity (radians s–1), and r is the dis-tance from the particle to the axis of rotation. From this equa-tion, we see that a particle in an ultracentrifuge is subject totwo dominating forces: buoyancy (ρp – ρm) and sedimentingforces (ω2 r), and is influenced by the viscosity of the fluid andsize of the particle. The balance of these forces governs thebehavior, or sedimentation, of a particle in a centrifugal field.

While the density of a particle is intuitively important indetermining its sedimentation velocity because of its influ-ence on relative buoyancy, another convenient, but less intu-itive expression is the sedimentation coefficient (s):

s = v/ω2 r (3)

Modern expressions of sedimentation coefficient are inSvedbergs (S), which are equivalent to 10–13 s. Used togetherwith density, these terms provide a convenient means forcomparing the sedimentation characteristics of virus particlesand other biological particles (Table 1), and determining themost efficacious type of centrifugal separation for virus purifi-cation, as explained below.

Types of centrifugal separations—Centrifugal separations canbe separated into two basic types: differential pelleting andzonal separations. Differential pelleting is most useful forcrude separations of raw material where purity and yield arenot critical. The method involves sedimenting particles out ofsolution, and either retaining the pellet or supernatantdepending on where the material of interest is located(Fig. 1A). As predicted by the equations above, larger particleswill sediment prior to smaller ones, and more dense particlesprior to less dense ones. In addition, asymmetrical particleswill sediment more slowly than spherical ones of the samemass and density. The separations are not clean, however,since the centrifugal force required to pellet large particlesfrom the top of a sample will also pellet small particles fromthe bottom. The greater the difference in sedimentation ratebetween the particles being separated, the cleaner the prepa-ration will be. The most common practical application of dif-ferential pelleting is removing cell debris from viral lysatesbefore concentrating and purifying viruses in the supernatantby other means.

Viruses can be harvested from a sample by pelleting inorder to concentrate them, separate them from small contam-inants, or exchange buffer, but some viruses may be damagedduring pelleting or resuspension. Furthermore, virus particlesmay be difficult to disaggregate after pelleting, which canadversely affect some subsequent purification steps or analy-ses. The sedimentation rate of the viruses will increase andbecome heterogeneous as a result of aggregation and the num-ber of viruses will be underestimated by infectivity assays orepifluorescence microscopy. If one intends to simply extractviral nucleic acids or proteins, then pelleting of viruses may bea practical harvesting method. If active, undamaged, well-dis-persed viruses free of contaminants are required, then alterna-tive density-gradient techniques should be considered.

There are two types of zonal separations, both of which relyon density gradients: rate zonal and isopycnic. Rate-zonal cen-trifugation separates particles based on differences in theirsedimentation coefficients (s), which, from Eq. 1, we see is afunction of both particle size and density. In practice, differ-ences in size dominate the differences in s among most bio-logical particles, since the range of densities is not large and svaries as the square of particle diameter. Isopycnic separationsdiscriminate among particles based solely on differences inbuoyant density. In both techniques, centrifugation is carriedout in a density gradient, which, among other functions, pre-vents mixing of the sample thereby ensuring that separatedparticles remain separated.

For rate zonal separations, a sample is introduced to the topof a density gradient. When subjected to centrifugal force, thesample components migrate through the gradient accordingto their s. Particles migrate at different speeds, resulting ingreater distance between particles having different s over time(Fig. 1B). Because the particles do not come to rest at equilib-rium in the gradient, care must be taken so that the particlesof interest do not pellet. For effective separations, the initialsample volume should be small (the sample layer should beonly a few millimeters thick), because the sample zone con-tinues to widen over time as a result of diffusion. Therefore,while rate zonal gradients eliminate problems associated withpelleting during the purification, a suitable concentration stepthat does not result in pelleting or aggregation must beemployed prior to using this technique. Many different typesof density gradient media may be employed for rate zonal sep-arations (see “Density Gradient Media”). Choosing the appro-priate medium requires matching the properties of themedium to one’s specific application. In general, it is benefi-cial to employ media preparations of high viscosity for ratezonal separations because viscous forces will magnify differ-ences in settling velocity between similar particles. Centrifu-gation parameters (relative centrifugal force and run-time)must be determined empirically for new virus applications.

In isopycnic (or equilibrium buoyant density) separations,particles migrate through the density gradient until they reachthe point at which their density is equal to that of the sur-rounding medium (Fig. 1C). Media used for this type of sepa-ration must therefore be able to form a solution that is at leastas dense as the viruses that are to be purified. Samples may betop-loaded or bottom-loaded in preformed density gradients,or homogenously mixed with a self-forming gradient mediumbefore centrifugation. As particles approach their equilibriumposition in a gradient, the difference in density between theparticle and the medium decreases and, consequently, so doesthe migration rate of the particle. Particles become increas-ingly focused over time until the focusing force is balanced bydiffusion. Achieving equilibrium, at which point the particlesare most focused, can require long centrifugation runs underhigh g-forces. This method eliminates pelleting and aggrega-tion problems with virus applications, and produces highly


168

concentrated and purified virus preparations at the same time.Isopycnic separations also provide a means for directly deter-mining buoyant density, a commonly reported physico-chem-ical property of viruses. One should be aware, however, thatbuoyant densities are specific only to the media in which theywere determined.

Because the two zonal techniques described above separatebased on partially independent properties (size versus density),they can be used sequentially to separate particles that may notbe separable by either method alone (Fig. 2). Two-dimensionalseparations have been particularly valuable for virus purifica-tion, since most viruses have a combination of sedimentationcoefficient and buoyant density that distinguishes them frommost other cell constituents, including macromolecules andorganelles. This region in the size-buoyant density space isreferred to as the “viral window” (Anderson et al. 1966)

Ultracentrifuge rotors—Of the two major types of centrifugerotors, conventional tube rotors have been used more com-monly for aquatic virus purification than zonal rotors, per-haps because the former are cheaper to buy and simpler to use.In zonal rotors, the entire rotor bowl serves as the separationchamber. Vanes radiating from the axis in the chamber par-tially divide the bowl into compartments, which facilitatesacceleration of the liquid mass while minimizing shear andmixing. With specialized fittings and a pump, fluid can bepumped in and out of the cylindrical cavity while the rotor isspinning (dynamic loading and unloading). Zonal rotors aretechnically challenging to operate, but have the advantage ofpermitting the processing of large volumes of sample as a sin-gle batch and permitting high-resolution separation of allcomponents in a sample. Conventional tube rotors can han-dle volumes around 1/50 to 1/100 of a zonal rotor, but providethe highest yield of a single component in a mixture in theminimum length of time. With the wide range of tube sizesavailable, and the possibility of preconcentrating samples asneeded, conventional tube rotors provide sufficient flexibilityto accommodate most applications. This chapter will onlyaddress conventional tube rotor applications.

Table 1. Density, diameter, and sedimentation coefficient (S ) for subcellular entities.

Sedimentation Diameter Density (g/cm3)

Subcellular entity coefficient (S) (µm) CsCl Sucrose Percoll OptiPrep Metrizamide Ficoll

Nucleus 106 to 107‡ 3–12‡ >1.32* 1.08–1.12* 1.2* 1.23†

Mitochondria 104 to 5 × 104‡ 0.5–4‡ 1.13–1.19* 1.07–1.11* 1.14* 1.14*

Lysosomes 4 × 103 to 2 × 104‡ 0.5–0.8‡ 1.21* 1.04–1.11* 1.12* 1.12*

Peroxisomes 4 × 103‡ 0.5–0.8‡ 1.23* 1.05–1.07* 1.2* 1.24–1.27*

Viruses 42 to >1000 0.02–0.4 1.18–1.51 1.15–1.29 1.06–1.08 1.14–1.22 1.13–1.31 1.07–1.14

Nucleic acids (free) 3.5 to 100 n/a 1.7–1.95 1.6–1.75§ 1.18–1.79*

Ribosomes 80 0.025 1.4

*Hinton and Mullock (1997)†Schmidt (1973)‡Luttmann et al. (2006)§Griffith (1994)

Fig. 1. Differential pelleting in a swinging-bucket rotor (A). Rate zonalcentrifugation through a preformed gradient in a swinging-bucket rotor(B). Isopycnic separation through a self-generating gradient in a swing-ing-bucket rotor (C).


169

There are two main types of tube rotors: swinging-bucketand fixed-angle. In swinging-bucket rotors the sample andtube reorient to 90° from vertical during centrifugation suchthat the gradient is always parallel to the long axis of the tube.Fixed-angle rotors hold the tube in a fixed position and theliquid sample reorients within the tube during rotor accelera-tion and deceleration. Three sub-types of fixed-angle rotors arerecognized based on the angle at which the tube is held. Theterm “fixed-angle” typically is used in reference to rotors inwhich the tube is oriented at an 18°-45° angle relative to ver-tical. Two other special cases of fixed-angle rotors are vertical(0°) and near-vertical (7.5°-9°) tube rotors. The differences intube angle among these rotor types influences both the sedi-mentation path length (i.e., the maximum radial distance aparticle must travel to reach the tube bottom or wall), and therange in magnitude of g-force within the tube.

A very useful descriptor of rotor performance that encapsu-lates particle path-length and centrifugal force is the k factor.Generally, the lower the k factor the shorter the runtime requiredto pellet the same particle. Table 2 provides the maximal k factorsand maximal centrifugal forces for some commonly used rotors.

A very useful resource for modifications and adaptations ofprotocols is the Beckman Coulter Centrifugation ResourceCenter Web site. Applications include a rotor calculator, whichpermits conversions between k factor, centrifugal force, androtations per minute (http://www.beckmancoulter.com/resourcecenter/labresources/centrifuges/rotorcalc.asp), and aruntime converter for different rotors (http://www.beckmancoulter.com/resourcecenter/labresources/centrifuges/run-timecon.asp). Alternatively, centrifugal force can be calculatedusing the following formula:

RCFmax = 1.12 rmax (rpm/1000)2 (4)

where RCFmax is the maximum rated speed of the rotor and rmax

is the maximal radial distance from the centrifugal axis. Forcalculating k factors when operating a rotor at less than max-imal speed (in rpm), the following formula can be applied:

kadj = k (maximum rated speed of rotor / actual speed of rotor)2 (5)

Finally, runtimes can be converted between two rotorsusing the following formula:

k1/t1 = k2/t2 (6)

The k factor can be used to calculate the time (t in hours)required to pellet all of the particles having or exceeding agiven sedimentation coefficient (s in Svedberg units) when therotor is run at its maximum speed:

t = k/s (7)

This calculation, like that for s, assumes that sedimentationtakes place in pure water at 20°C. Corrections can be appliedto account for differences in the viscosity of water at othertemperatures or for sedimentation in solutions other thanpure water, and are best determined empirically.

All of the rotor types noted above can, in principle, be usedfor any of the separations described in the previous section,but best results are obtained when the rotor geometry ismatched to the separation techniques being employed. Belowwe describe the main applications for each of the rotor typesand some of their advantages and disadvantages for differenttypes of separations.

• Swinging-bucket rotors. The path lengths in swingingbucket rotors are long and sedimenting particles haveminimal interaction with the sides of the tube, makingthese rotors an excellent choice among tube rotors forrate-zonal separations. These rotors are also commonlyused for isopycnic separations. In both cases, the lengthof the tube can be exploited to obtain high resolutionamong different types of particles. The large differencein g-force from the top to the bottom of the tube meansthat self-forming gradients are steeper in these rotorsthan in fixed angle rotors. This can be helpful when sep-arating particles that differ greatly in buoyant density.Another advantage of swinging-bucker rotors for gradi-ent applications is minimal disturbance to the gradient,since it does not reorient during the run. There is a rota-tional force on the gradient during acceleration anddeceleration, however, so these are accomplished slowly.Swinging bucket rotors can also be used for pelletingparticles. The potential advantage of pelleting in aswinging bucket rotor is that the pellet is compact andlocated at the very bottom of the tube. The disadvantageis that the time required to pellet material is muchlonger than in a fixed angle rotor.

• Fixed-angle rotors. The major application for fixed-anglerotors is pelleting. Since particles sediment to the wall,rather than the bottom of the tube, the path length is

Fig. 2. Diagram showing the unique signatures of three viruses, V1(σ),V2(υ), and V3(λ). V1 and V2 are not separable on an isopycnic gradient,whereas V1 and V3 are not separable in a rate-zonal gradient. Since buoy-ant density and sedimentation coefficient are partially independent, ratezonal and isopycnic separations may be used sequentially to separate allthree of these viruses from one another.


170

shorter. The angle of the tube causes much of the materialthat reaches the wall to migrate down the wall to the mostdistal point within the tube near the bottom. Pellets aretherefore relatively compact. Because of the wall effects,fixed angle rotors are not recommended for rate-zonalseparations, but they can be used effectively for isopycnicseparations.

• Vertical rotors. The major applications for vertical rotorsare rapid isopycnic and rate-zonal separations. Becausethe gradient reorients by 90° within the tube during cen-trifugation, the path length for a tube in a vertical-rotor isthe diameter rather than the length of the tube. The prac-tical implication is that run times can be much shorterthan in a swinging bucket rotor. The relatively small dif-ference in g-force from rmin to r max, also means that self-forming gradients tend to be shallower than in swingingbucket rotors even at high speeds. A shallow gradient canbe helpful when resolving particles that have small differ-ences in buoyant density, but may be a disadvantage ifone needs to separate particles with widely differing buoy-ant densities. The reorientation of the gradient in a verti-cal tube rotor has implications for the resolution. Gener-ally speaking, the zones or bands in a swinging bucketrotor can be tighter than in a vertical rotor, but the spacebetween bands of differing density can be greater in a ver-tical rotor. Vertical rotors are generally not used for pel-leting, because the pelleted material is distributed alongthe entire length of the tube. This also means that anymaterial that pellets during a gradient separation will bein contact with the entire gradient when the rotor comesto rest after the run.

• Near-vertical rotors. The very low angle of near-verticalrotors provides the benefits of a vertical rotor (short runtimes and minimal wall effects), while improving thepurity of particles separated in a gradient. Contaminantsthat exceed the density range of the gradient during azonal or isopycnic run will either pellet toward the mostdistal part of the tube near the bottom of the tube or floatto the most axial part of the tube near the top, just as ina fixed-angle rotor. Material that is banded in the gradientnear the center of the tube is then not in contact withthese contaminants when the gradient reorients at theend of the run.

Density gradient media—Numerous investigators have iden-tified the criteria for choosing density gradient media for bio-logical separations (Cline and Ryel 1971; Hinton et al. 1974).In summary, the criteria are as follows:

• The media should be inert or at least nontoxic to the spec-imen (minimal osmotic effect, ionic strength, and neutralpH).

• The media should form a solution covering the densityrange for the particular application, and be stable in solu-tion.

• The physical and chemical properties of the media shouldbe known, and it be possible to use one or more proper-ties to determine the precise concentration of the media.

• The solution should not interfere with monitoring ofzones of fractionated material within the gradient.

• It should be easy to separate the sample from gradientmaterial without loss of the sample or its activity.

• The gradient media should be available as a pure com-pound, and be relatively inexpensive.

Table 2. Specifications of swinging-bucket ultracentrifuge rotors commonly used for density gradients.

Rotor Maximum speed (rpm) Maximum force (g) k Factor (@ max g) Rotor capacity (mL)

Beckman Coulter

MLS50 50,000 268,000 71 4 × 5

SW28 28,000 141,000 245 6 × 38.5

SW28.1 28,000 150,000 276 6 × 17

SW32 Ti 32,000 175,000 204 6 × 38.5

SW32.1 Ti 32,000 187,000 228 6 × 17

SW40 Ti 41,000 285,000 137 6 × 14

SW41 Ti 40,000 286,000 125 6 × 13.2

SW55 Ti 55,000 368,000 48 6 × 5

SW60 60,000 485,000 45 6 × 4

TLS55 55,000 259,000 50 4 × 2.2

Sorvall

AH-629 (36 mL buckets) 29,000 151,240 242 6 × 36

AH-629 (17 mL buckets) 29,000 155,850 284 6 × 17

AH-650 50,000 296,010 53 6 × 5

SureSpin 630/36 30,000 166,880 219 6 × 36

SureSpin 630/17 30,000 166,880 268 6 × 17

TH-641 41,000 287,660 114 6 × 13.2

TH-660 60,000 488,580 44.4 6 × 4.4


171

Although viruses are not physiologically active, and there-fore, less susceptible to some stresses than living cells, osmoticand ionic compatibilities are still important for maintainingthe integrity of the virions, especially for membrane-boundviruses (Anderson et al. 1953). Here we summarize the proper-ties and common applications of the most common types ofdensity gradient media noting their advantages and disadvan-tages for specific applications. More detailed information onthe properties of various gradient media can be found in anumber of books and technical manuals (e.g., Rickwood 1984;Griffith 1994).

Salts of alkali metals (cesium, rubidium, lithium, etc.salts)—Ionic media. Alkali metal salts, such as cesium chlo-ride, are most widely used for making isopycnic gradients withany standard technique including preformed or self-forminggradients. Metal salts can provide some of the densest prepa-rations available, have a low viscosity, and their concentrationin solution is easily measured by refractive index. The majordrawbacks of alkali metals lie in their effects on biologicalactivity; salt solutions have high ionic strengths, which dis-rupt protein-protein and nucleic acid-protein bonds, and havehigh osmolarities, affecting particle hydration. These factorsinactivate many sensitive viruses. Therefore, metal saltsshould be limited to use with robust viruses or applicationswhere high yield and purity are needed, but biological activityis not required. Salts can easily be removed by a variety oftechniques, including dialysis and ultrafiltration. The range ofreported densities of viruses in CsCl varies widely, from 1.15to 1.55 g mL–1 (Fauquet et al. 2005). Analyses of marine viralassemblages suggest that DNA-containing viruses in the oceanare dominated by those having densities in CsCl of between1.39 and 1.46 g mL–1 (Steward et al. 2000).

Small hydrophilic organic molecules (sucrose, glycerol, sor-bitol, etc.)—Non-ionic media. Sucrose meets most of the crite-ria of an ideal medium for rate zonal separations, being bio-logically inert, stable, and relatively cheap. Due to itspopularity as such, sucrose is very well characterized withrespect to concentration, viscosity, density, and refractiveindex, making it easy to develop and adapt methods foruncharacterized viruses. While sucrose has little effect onintermolecular bonding and is non-ionic, high osmotic pres-sure may cause shrinkage in enveloped viruses and therebyaffect infectivity in sensitive viruses. The high viscosity ofsucrose at concentrations useful for virus separations may aidin separation between similarly sized particles under ratezonal conditions, but the high viscosity and relatively lowdensity limits the application of sucrose and other smallorganic molecules in isopycnic separations of viruses. Othersugars, notably glycerol and sorbitol, have also been usedeffectively as rate zonal media. These gradients need to be pre-formed as solutions of small organic molecules do not gener-ally form gradients when centrifuged.

High molecular-weight organics (Ficoll, dextran, glycogen,etc.). High molecular-weight polysaccharides do not penetrate

intact biological membranes and have a lower osmolarity thansolutions of monosaccharides. Therefore, these media may beespecially useful when employed with sensitive, membrane-bound viruses. Unfortunately, due to the size of these polysac-charides, they cannot be removed from the sample by dialysisor ultrafiltration, so dilution and high-speed centrifugation aregenerally required, which are contraindicated with sensitivespecimens as discussed above. Since polysaccharide media suchas Ficoll (GE Healthcare) and dextran diffuse slowly, it is nec-essary to preform linear gradients using gradient mixers. Thischaracteristic also ensures that gradients are quite stable onceformed. The high viscosity of these media necessitates longerspin times than those of sucrose gradients.

Colloidal Silica (Percoll, Ludox, etc.). Colloidal silica sus-pensions such as Percoll (GE Healthcare) and Ludox (DuPont)are truly non-ionic media that can be used to rapidly generateself-forming gradients. These media are well characterized,permitting the use of refractive index for examining densityprofiles of gradients since absorption prohibits monitoring byUV light. Percoll density marker bead kits, available from anumber of vendors (e.g., Sigma-Aldrich, product DMB-10), areuseful for visually monitoring gradient profiles. Whereas col-loidal media cannot be effectively filter sterilized, they may beautoclaved before being adjusted for osmolarity and can beused over a wide pH range (5.5-10 for Percoll). Percoll is com-monly used for cell separations, because the suspension of col-loidal silica can be prepared in almost any buffer required tomaintain cell viability. Viruses can also be separated in gradi-ents of colloidal silica (Pertoft et al. 1967), but removal of thegradient medium from the viruses following centrifugation isa challenge, since both the viruses and the medium are col-loidal. Another limitation is that the silica particles may beginto pellet before smaller viruses have time to form discrete,purified bands. To remove Percoll from virus purificationsrequires dilution and high-speed differential centrifugation(i.e., 100,000 g for 2 h in a swinging bucket or 1.5 h in anangled rotor), which may lead to aggregation and deactivationof viruses, as previously discussed.

Iodinated organic compounds (Nycodenz, OptiPrep, andmetrizamide). Iodinated compounds provide an excellentcombination of biological inertness, a wide density range, andlow viscosity, which allows for reduced spin times. These com-pounds, including Nycodenz (Axis-Shield), iodixanol (sold asOptiPrep by Axis-Shield), and metrizamide, are heat stable,autoclavable, and of minimal ionic strength. Working withOptiPrep is simplified as there is a near-linear relationshipbetween concentration and osmolality. The iso-osmoticnature of OptiPrep helps to minimize morphological artifacts,which is useful if the ultrastructure of a virus is to be investi-gated by electron microscopy. Concentrations of Nycodenzabove 30% are hyperosmotic, although less so than forsucrose. In OptiPrep, many viruses have a banding density ofbetween 1.16 and 1.26 g mL–1 (Vanden Berghe 1983). Thereare few analyses of the range of buoyant densities found in


172

natural marine viral communities using iodinated gradientmedia, but one analysis from coastal California seawater indi-cated that most of the viruses Nycodenz were in the rangebetween 1.18 to 1.29, with a peak around 1.24 (Stewardunpubl. data). Stable gradients of iodinated media will self-generate or can be created through the diffusion of step-gradi-ents. The many advantages of iodinated gradient media, andOptiPrep in particular, make them an excellent choice forisopycnic separations of viruses. Extensive resources on theproperties and applications of these gradient media are avail-able from Axis-Shield, including application guides for theirdensity gradient products, a reference database, and specificmethodologies for OptiPrep density gradient purification ofmore than 30 viruses to date (http://www.axis-shield-density-gradient-media.com/).

Gradient design—The shape of a gradient, also known asdesign, refers to the density profile through the gradient tube.When selecting gradient design it is important to first identifythe type of separation and density media to be employed.Once these have been established, the number options aregreatly reduced.

There are two classifications for gradient design: discon-tinuous versus continuous, and preformed versus self-form-ing. Discontinuous gradients have discretely layered zones ofisopycnic media and are preformed before the sample is intro-duced. These steps are created using underlayering or over-layering techniques, and may be distributed in many equal-volume, isopycnic layers throughout the gradient, referred toas a step gradient, or may involve only one or two steps,known as a density-barrier or cushion gradient. Step gradientsare useful for isopycnic separations, and provide an excellentcombined concentration and purification method when onlyone, uniform component in the sample is of interest; the pre-ceding density steps filter unwanted particles of lesser densitywhile the supporting density step collects the particles ofinterest, letting more dense particles pass through. Density-barrier gradients are also very effective for concentratingviruses from simple samples, as they provide a high-densitycushion to concentrate viruses on without forming a hardpellet. The primary considerations when designing a step gra-dient is that the top step is dense enough to support the sam-ple when it is loaded, and the sample volume should notexceed 2% to 3% of the entire gradient volume if the separa-tions are rate zonal, because the sample zone increases duringthe run as a result of diffusion.

In continuous gradients, the density gradually increasesdown the tube, although the slope of the density may be lin-ear, convex, or concave. The simplest shape is linear, wheredensity increases at a constant rate with volume and resolv-ing power decreases with the steepness of the slope. Lineargradients are least affected by diffusion, in terms of theirshape, and may be generated with a simple apparatus or bydiffusion of a step gradient (for details, see “Gradient Prepa-ration”). Self-forming gradients, which can be prepared with

media such as cesium chloride, Percoll, metrizamide, Nyco-denz, or OptiPrep, tend to be nonlinear, and the gradientshape varies depending on the rotor type used. Self-generat-ing gradients form if the sedimentation rate is faster than thediffusion rate. For self-forming, continuous gradients, Percollprovides the most time-efficient option; a Percoll gradientwill form in 30 min at 10-20,000g, although the silica parti-cles continue to pellet after this point, and therefore, the gra-dient is not very stable. Cesium chloride and iodinated com-pounds take hours to self-form at very high speeds(50-100,000g), but if run-time is an issue, these media can beused to make preformed, stable gradients. An additional type,the isokinetic gradient, is beyond the scope of this publica-tion as it requires a priori knowledge of both s and ρ and istherefore generally not applicable to discovery-basedresearch. For rate zonal separations, the best design is a con-tinuous gradient to avoid artifactual sample bands created bydiscrete increases in density. Continuous gradients are alsomost useful for fractionating complex samples since the grad-ual increase in density allows for the greatest resolution ofsample components. When designing continuous gradientsfor isopycnic separations, the maximum density shouldexceed all the particles in a sample, and the slope should beminimized to maximize resolution. Sample volume on isopy-cnic gradients is not as narrowly proscribed as for rate zonalseparations, because the sample particles focus at their isopy-cnic points independent of sedimentation rate.

A few general points should be considered when develop-ing and adapting density gradients. Media concentration doesnot always have a linear relationship with either density orviscosity, but sedimentation rate varies inversely with viscos-ity. Since viscous drag has an increasing effect on sedimenta-tion rates as particles move down a gradient, the viscosity pro-file is more important than density profile for determiningrun-time when working with highly viscous media. Concen-tration profile is the least important factor for consideration.It is important to monitor gradients to ensure that they areoptimized with respect to shape and are reproducible from runto run.

Gradient preparation—Preparing gradient solutions: Onceone has determined the gradient medium and the gradientrange, solutions of the appropriate densities must be prepared.Here we provide reference material to assist in the preparationof CsCl and iodixanol solutions. The densities of these solu-tions vary linearly as a function of concentration (Fig. 3). Theconcentration of the gradient medium required to achieve thedesired solution density can be estimated from the graphs, buta more accurate concentration can be calculated from the lin-ear regression functions for the curves. For CsCl the formula is:

%CsCl = (ρsolution – ρsolvent) × 135 (8)

where %CsCl is the concentration of the CsCl solution (in gper 100 mL), ρsolution is the desired density of the final solution(g mL–1), and ρsolvent is the density of the buffer in which the


173

CsCl is being dissolved (g mL–1). Since CsCl solutions are usu-ally prepared from the solid, Eq. 7 can also be rearranged intoa more convenient form for initial preparation of solutions ofthe desired density:

CsClmass = Vsolution × (ρsolution – ρsolvent) × 1.35 (9)

where CsClmass is the mass of CsCl to be added (in g), and Vso-

lution is the final solution volume (in mL).Iodixanol is supplied as a sterile 60% solution in water

under the name OptiPrep and is usually diluted to ≤ 50% formost applications. To prepare a 54% stock solution in a 1×concentration of a desired buffer, combine 9 parts OptiPrepwith 1 part 10× buffer. The concentration required to achievea desired density can then be calculated as:

%Iodixanol = (ρsolution – ρsolvent) × 186.5 (10)

where ρsolvent is the density of the 1× buffer being used for thedilutions.

Concentrated stocks of iodixanol (or CsCl) can be dilutedto some desired lower density according to the formula:

Vdiluent = V (ρ – ρfinal)/(ρfinal – ρ diluent) (11)

where Vdiluent is the volume of diluent to be added, V and ρ arethe volume and density of the solution to be diluted, ρfinal is thedesired final density, and ρ diluent is the density of the diluent.

The equations above call for the density of the diluent orbuffer in which the gradient medium will be dissolved. Onebuffer commonly used for this purpose is SM (Sambrook and

Russell 2001). Without the gelatin added, this buffer has den-sity at 20°C of approximately 1.003. If one is preparing a gra-dient medium in some other buffer, particularly one with highconcentrations of other solutes, the density can be determinedempirically (see “Measuring fraction densities”) or the equa-tions can be modified accordingly as described in Fig. 3. It isgood practice to verify the density of each solution once it hasbeen prepared, because egregious errors can lead to loss of sam-ple or, in the case of CsCl solutions, catastrophic rotor failure.

Gradient layering and sample loading. Making discontinu-ous gradients by hand is the most cost-effective option, andsimple diffusion of many types of media creates continuousgradients from discontinuous ones. Underlayering each suc-cessively more dense solution results in the least amount ofmixing between layers, providing the most discrete and repro-ducible gradient shape. Overlayering media layers is difficultto reproduce between tubes and runs, resulting in high vari-ability between samples. For the best underlayering results,use a long, wide-bore pipetting needle (16-18 gauge, 4-inchcannula) on a disposable syringe. While resting the tip of theneedle at the bottom of the tube, very slowly dispense the gra-dient media, being careful to not introduce any air bubbles asthese will cause significant mixing (Fig. 4A).

For creating continuous gradients in open-topped tubes,step gradients can be covered and allowed to diffuse at roomtemperature for a couple of hours, in the refrigerator overnight,or even during centrifugation, provided the centrifugationtime is sufficiently long. If the tube is sealed, diffusion in the

Fig. 3. Solution density at 20°C as function of the percent concentration of CsCl (A) or iodixanol (B) dissolved in water (solid squares). Black lines areregression lines for CsCl (y = 0.007416x + 0.9982) and iodixanol (y = 0.005363x + 0.9982). Note that the y-intercept represents the density of the dilu-ent (water at 20°C has a density of 0.9982 g mL–1). The linear regression formulae can be adapted for preparation and dilution of the gradient media ina buffer having a density greater than water by maintaining the slope, but changing the value of the intercept to reflect the density of the buffer. Forexample, dilution in a buffer with a density of 1.05 is illustrated (Panel B, green line; y = 0.0054x + 1.05). For OptiPrep™, which is supplied as a 60%solution of iodixanol in water, the function expressed by the green line assumes a stock solution in 1× buffer is first prepared as described in the text. Ifinstead, the OptiPrep is simply diluted directly in 1× buffer, then both the slope and the intercept change, but the value at 60% remains fixed (Panel B,red line; y = 0.0045x + 1.05). Data for CsCl was obtained from Lide (2009) and that for iodixanol was derived from Axis-Shield PoC product literature(http://www.axis-shield-density-gradient-media.com/brochures.htm).


174

horizontal position greatly speeds diffusion. Gradient mixersare also available from a number of suppliers, providing highlyreproducible continuous gradients in much less time. Instru-ments include the Gradient Master and Gradient Station (Bio-Comp Instruments). The Auto Densi-FlowTM (Labconco) is aninstrument that automates dispensing of gradients when con-nected to an appropriate gradient mixer.

For separations in self-forming gradients, samples are pre-pared as homogeneous solutions in the gradient medium. Forseparations in preformed gradients, samples are usually loadedonto gradients by overlayering. Samples should be overlayeredusing a narrow-tipped pipette or syringe with a narrow gaugeneedle. Keeping the delivery tip in the interface of the gradient,slowly dispense the sample to the surface, and centrifuge thesample immediately after loading to avoid diffusion (Fig. 4B).

Gradient fractionation and sample collection—Once a prepara-tive gradient run has been completed, tubes should be han-dled gently, with as little rotation, tipping, or jarring as possi-ble, and the sample recovered as quickly as practical. There area number of methods for recovering samples from gradients.The method one chooses will depend on the nature of the sep-aration, the type of centrifuge tube being used, and the equip-ment available. Most of these methods can be accomplishedwith little equipment, but the resolution and reproducibilitymay be improved with the use of specialized gradient-harvest-ing devices. Here we present methods for targeted harvestingof an individual band and for fraction collection.

Targeted harvesting of a single band.• Side Puncture. If the desired band can be easily seen in the

gradient, one may wish to selectively and directly harvest

it. If the tube permits it, the tube may be punctured witha needle and syringe just below the band of interest withthe needle hole pointed upward (Fig. 5A). Placing someclear adhesive tape over the desired puncture point beforepuncturing the tube helps maintain a seal around needle.The material is harvested by drawing liquid slowly intothe syringe until the band is no longer visible. For sealedtubes, another puncture must be made near the top of thetube to relieve the vacuum during aspiration. A challengewith this method is making the puncture through the sideof the tube with minimal disturbance. The tube will tendto deform from the pressure while forcing the needlethrough, and air bubbles can be introduced throughpuncture, which will rise through the gradient causingsome mixing. If the needle is pushed directly through thewall of the tube, a plug of plastic may be cored from thewall of the tube and occlude the opening of the needle. To

Fig. 4. Hand-layering a step gradient using the underlayering techniquewith a syringe and pipetting needle (A). Sample loading by overlayeringonto a preformed gradient using a syringe (B).

Fig. 5. Targeted harvesting of a single band in a density gradient by sidepuncture with a needle (A). Direct unloading of gradient fractions fromthe top with a pipette (B). Gravity drip collection with control through abottom stopcock (C). Top unloading of fractions through a plunging,tapered piston (D).


175

avoid this, twist the needle as the tip penetrates the wall.The greatest hazard when using this technique is that theneedle can slide quickly once it breaks through the wall ofthe tube. If one does not exercise control, the needle canpenetrate the opposite side of the tube, and potentiallyone’s finger if it happens to be in the path.

Gradient fractionating.• Direct unloading. The simplest, but crudest method of

collecting fractions from a gradient is to aspirate from thetop of the gradient with a pipette. By keeping the pipettetip as close as possible to the meniscus during each aspi-ration, fractions can be collected with the least densebeing removed first (Fig. 5B). This process can be auto-mated using a device such as the Auto Densi-Flow, whichhas sipping tip with side holes that is attached to a motor-ized mount that moves the tip up and down. A sensormaintains the inlet holes of the sipper just below themeniscus, and a peristaltic pump connected to the tipdrives the aspiration. A similar method is to insert a fine-gauge cannula or tube to the bottom of the gradient anddraw the liquid out, starting with the most dense first.This is not recommended, however, since the gradient isinverted as it moves up the tubing, which leads to con-vective mixing and loss of resolution.

• Bottom puncture. A very effective method for harvestingfractions from a gradient that requires no special equip-ment is to simply puncture a hole in the bottom of thetube and harvest the drops from the bottom. It is mostconvenient to secure the tube above the bench (e.g., witha clamp on a ring stand) to free both hands for the tasks oftube piercing and fraction collection. The tube should beclamped with just enough force to keep it from slipping.Unloading a gradient through a bottom puncture can beaccomplished several ways, the simplest being gravitydrip. For gravity drip with sealed tubes, a hole will have tobe made in the top of the tube to allow the pressure toequalize as the tube empties. This hole should be madebefore the bottom hole to relieve initial pressure differ-ences between the inside and outside the tube. Pressuredifferences may result from squeezing of the tube in aclamp, or from a difference between the current tempera-ture of the liquid in the tube and its temperature whenthe tube was sealed. If the bottom hole is made first, theremay be an uncontrolled flow until the pressure is equal-ized. After the top hole is made, it can be sealed tem-porarily during the bottom puncture to prevent fluid fromimmediately flowing out. One can use a gloved finger tocover and uncover the upper hole to control the flow asfractions are collected by dripping into a series of tubes.A useful alternative means of flow control that also worksfor open top tubes is to pierce the bottom (or near the bot-tom) of the tube with a needle attached to a stopcock (Fig.5C). Use a twisting motion during piercing to ease theneedle through the wall and to avoid plugging the needle

with plastic. A few drops may leak around the needle dur-ing the initial piercing, but these can be collected, ifdesired, by piercing at an angle and positioning a tubedirectly under the point of penetration. Adjust the posi-tion of the needle so that the opening is as near the bot-tom of the tube as possible. The stopcock can then beopened to give the desired flow rate.To provide an extra measure of control over the flow rate,the sample may be drawn from the bottom of the tube viaperistaltic pump rather than relying on gravity. Alterna-tively, the gradient may be pushed out of the bottompiercing by displacement from above. In this case, theneedle used to pierce the top of the tube remains in placeand is connected to tubing passing through a peristalticpump. Either air or pure water is then slowly pumped intothe top of the tube to control flow through the bottomneedle and stopcock assembly. Using this method withopen top tubes would require that the tubes first be sealedwith an appropriate size bung.

• Piston fractionator. An elegant solution for unloadinggradients from open top tubes is to use a piston gradientfractionator (e.g., The Gradient Master). This device hasa piston that seals against the walls of the tube. Whenthe piston is pressed down, the liquid escapes through ahole in the tapered piston head and travels through tub-ing to the collection point (Fig. 5D). This programmableinstrument allows reproducible harvesting of fractionsas thin as 300 µm, and eliminates cross-contaminationbetween fractions by rinsing and blowing-dry the sam-ple tubing between sample points. The other advantagethis instrument provides is by improving the visualdetection of bands within a gradient. The sample holdersupports the centrifuge tube in a bath of water, whichrenders the centrifuge tube transparent, and providesillumination from below. A Plexiglas window on the sideof the sample holder allows the user to visualize andrecord banding patterns.

• Displacement from below. In this method, a liquid moredense than the densest part of the gradient is pumpedinto the bottom of the tube, which displaces liquid outthe top of the tube through a specialized cap. Achievingthis with a homemade set-up is not trivial, so this is typi-cally accomplished with commercial gradient fractionatorsdesigned for the purpose (e.g., Beckman Coulter or Bran-del). The dense solution can be added through a needlepiercing the bottom of the tube, or by inserting a narrowtube to the bottom of the gradient from above. The lattermethod will cause some disturbance to the gradient duringthe insertion of the tube so is recommended only when col-lecting from thick-walled tubes that cannot be pierced.

Measuring fraction densities—It is often desirable to deter-mine the density of a sample, whether to verify the density ofstock solutions, determine the final shape of a gradient, or toestimate the buoyant densities of viruses harvested from gra-


176

dients. The relationship between density and concentration iswell known for many common gradient materials and con-centration can be determined quickly, accurately, and pre-cisely using refractometry (Rickwood 1992). Disadvantages ofthis approach are that it requires specialized equipment andall samples contact the same surface of the instrument sothere is the possibility for cross-contamination. This meansthat subsamples taken for measurement may have to be dis-carded, but the volumes required are not large. If the gradientmedia are prepared in a buffer, then one must correct for therefractive index of the buffer when calculating density. Refrac-tometry is not recommended for measuring fraction densitiesif one is using mixed gradient media (e.g., sucrose and CsCl),because the relative concentrations of the two media will varyamong the fractions after centrifugation, complicating inter-pretation of the refractive index.

Another simple, direct method for measuring density is todetermine the mass of a known volume of sample using a pre-cision balance and a micropipette with disposable tips. Tomake a mass measurement of a fraction, place the tube con-taining a fraction onto the balance and tare the balance. Thenremove a fixed volume (e.g., 100 µL) with a pipette. Retain thevolume that was removed in the pipette tip while the mass isrecorded (as a negative number), then dispense the subsampleback into the same tube. The advantages of this method arethat no volume is lost, repeated measurements can be quicklymade on the same fraction, and only the pipette tip must bechanged between measurements of different fractions.

When determining density by direct mass measurements,one must bear in mind that air-displacement pipettes (themost common type) will systematically under-collect volumeas density increases, leading to increasing underestimation offraction densities. Since the error is systematic, it can be cor-rected for using an empirically determined function. Theproblem can be avoided by using a positive-displacementpipette. In a direct comparison of an air- versus a positive-dis-placement pipette, we found the latter was not only moreaccurate, but also more precise (Fig. 6). Whether the extraaccuracy and precision obtainable with a positive-displace-ment pipette are worth the added cost of the tips will dependon one’s application and the magnitude of the pipetting errorrelative to the many other sources of error (Rüst 1968).

Regardless of the method used to measure the density, thesample should be well mixed before taking a sample to makethe measurement. Another important consideration whenmeasuring density (or refractive index) is the temperature of thesolution. When centrifuging cesium chloride solutions, onemust pay careful attention to the combinations of CsCl con-centration and centrifugation speed and temperature that allowsafe operation. Charts are provided in the instruction manualfor each rotor, but one must remember that the indicated max-imum safe densities for a given speed and run temperature arethe densities as measured at 20°C. If one plans to operate nearthe maximum allowable density, the measurements of initial

density should be made close to 20°C or a correction factorshould be applied to normalize the density to 20°C.

If one wishes to estimate the buoyant density of particularviruses by measuring the density of the fractions in which theyare found at peak concentration, one must account both forthe temperature at which the density measurement is made,and the temperature at which the gradient was centrifuged, Forexample, if a virus is found in a fraction whose density wasmeasured at 24°C, but the gradient was run at 4°C, one mustaccount for the thermal expansion from 4°C to 24°C to deter-mine the density at which the virus banded. The density ofsome viruses may also vary as function of pH or the time ofexposure to CsCl (Rowlands et al. 1971), which could lead tovariable estimates if centrifugation conditions change. Oneshould also remember that the density of a fraction is an aver-age for the entire fraction. The smaller the fractions, the morenarrowly one can constrain the density of a peak. Finally, onemust bear in mind that the diffusivity of CsCl is greater thanthat of viruses, so back-diffusion after the centrifuge hasstopped has the potential to alter the apparent density profileand can lead to errors in estimates of virus buoyant densitythat are based on measurements of fraction density.

Applications in marine viral ecologyPurifying viruses from culture lysates—In the typical culture

lysate, viruses are a homogeneous and very abundant popula-tion of particles that are much smaller than the copious cellu-lar debris, bacterial contaminants, and unlysed cells fromwhich they need to be separated. While filtration is sometimesemployed to clarify culture lysates prior to further purification,the amount of debris associated with lysed cultures oftencauses rapid changes in filter performance (i.e., nominal pore

Fig. 6. Errors in calculated density as determined from mass mea-surements of 100 µL volumes using a positive-displacement or an air-dis-placement pipette. Error bars are standard deviations of triplicate mea-surements using three separate tips for each pipette. The increasing erroras a function of density observed with the air-displacement pipette isattributed to an expected density-dependent undersampling. CsCl solu-tions of known density were prepared by combining measured masses ofCsCl and water according to tables in the Handbook of Chemistry andPhysics (Lide 2009).


177

size) resulting in the extensive loss of virus. Therefore, filtrationis best avoided in favor of differential centrifugation. The spe-cific conditions for clarification are quite variable, but 4000gfor 30 min is a good starting point. If large-volume lysates areto be purified, it may then be necessary to concentrate theviruses prior to further purification. This can be achieved eitherwith ultrafiltration techniques or through polyethylene glycol(PEG) precipitation. It is highly recommended that concentra-tion only be undertaken after initial clarification of the lysate.

Once a relatively pure, concentrated sample of culturedvirus has been obtained, final purification steps are bestachieved through density gradients. A variety of approachesmay be employed, each with relative strengths and weak-nesses depending on the down-stream applications sought.However, if the virus of interest is not well characterized, it iswell worth the effort to obtain infectious particles from thepurification process so the isolated material can be confirmedas the infectious agent before further efforts are expended. Theculture-based purification method described in Protocol Abelow aims toward this end.

Purification of viruses from natural samples—Purifying virusesdirectly from natural marine and freshwater samples is similarin many respects to purifying a virus from a culture lysate; thegoal in both cases is to separate the viruses from as much of thecontaminating material as possible. There are, however, a fewadditional considerations in working with natural samples. Formany applications, one is interested in harvesting and purifyingthe entire viral assemblage, which is comprised of viruses hav-ing a broad range of physico-chemical properties. These diverseviruses must be separated from the heterogeneous collection ofprokaryotic and eukaryotic cells that comprise the plankton.Depending on the purpose to which the purified viruses will beput, it may also be necessary to separate the viruses fromorganic and inorganic particulates and dissolved material. If thepurpose of the purification is to perform molecular analyses ofthe viral nucleic acids, the requirement for purity may be morestringent, and viruses may need to be separated from both cellsand from dissolved nucleic acids. Although viruses usually out-number cells in the plankton by over an order of magnitude,their genomes are so small that their nucleic acids make up onlya small percentage of cellular and dissolved pools.

The most common first step in purifying viruses from natu-ral samples is filtration through a membrane having pore size of0.2 or 0.22 µm. This removes virtually all of the cells, whileallowing the vast majority of viruses to pass in the filtrate. Themethod is not perfect, as there are reports of ultramicrobacteriathat can pass a 0.2 µm filter (reviewed by Velimirov 2001), andsome marine viruses have diameters in excess of 0.2 µm (Brat-bak et al. 1992; Garza and Suttle 1995). Viruses smaller than thepore size can also be trapped on the filter by adsorption or poreocclusion, resulting on occasion in low yields. Filtration is, how-ever, very simple and effective for many purposes.

The alternatives to filtration are few, and each has its prob-lems. One alternative directly relevant to the topic of this

chapter is the separation of cells from viruses in buoyant den-sity gradients. The great advantage of this approach is that, inprinciple, it allows the simultaneous separation of virusesfrom both cellular and dissolved DNA in one procedure, andthere is no inherent selection against large viruses. In practice,the range of buoyant densities of viruses overlaps with that ofcells, meaning that viruses having the lowest buoyant densi-ties (those that are enveloped, contain lipids, or have a lowmass percent of nucleic acid) may not be pure enough forsome purposes. If minimizing cellular contamination is thepriority, then filtration prior to gradient separation is recom-mended. If one is studying large viruses likely to be trapped ona 0.2 µm filter, but which have a buoyant density in CsCl thatis > 1.35 g mL–1, then direct loading of a whole-plankton con-centrate onto a density gradient is an option worth exploring.Following up with other fractionation procedures may berequired in that case to achieve the desired purity.

In the “Protocols” section, we describe a protocol (“Purification of viral assemblages from seawater in CsCl gra-dients”) for the purification of the viral assemblage from anaquatic samples, noting alternative procedures where appro-priate. It is important to note that the exact conditions usedcan vary widely depending on your application. The basicprinciple of density gradient separation can easily accom-modate some variations dictated by equipment availability.Effective separations can be achieved with either swinging-bucket or fixed-angle rotors, for example, although the timerequired for separation and the resolution of the gradient inthe end will vary.

ProtocolsPurification of viruses from culture lysates—

Materials and reagents:• OptiPrep (60% iodixanol solution)—Axis-Shield, Accurate

Chemical and Scientific (Westbury), Progen Biotechnik,or Sigma Aldrich

• Open-topped ultracentrifuge tubes—i.e., Beckman Coul-ter Ultra-Clear

• Ultracentrifuge.• Swing-out Ultracentrifuge Rotor—i.e., Beckman Coulter

SW41, SW28, or MLS50• 30 kDa cutoff disposable centrifugal ultrafiltration

devices—i.e., Millipore• 3-mL syringe with Luer-Lok or Luer-Slip• Pipetting needle—i.e., Cadence Science, stainless-steel 14-

or 16-guage 4-inch cannula with Luer hub or Slip hub• Sterile 1.5 mL microcentrifuge tubes for collecting gradi-

ent fractions• Sterile disposable transfer pipettes• Sterile virus-free media for resuspending and diluting

virus• Polyethylene glycol, average molecular weight 6000-

8000—i.e., Fisher Scientific Carbowax PEG 8000, or SigmaAldrich Biochemika Ultra 8000


178

Steps:Clarify lysate

1. Centrifuge the lysate at 4000g for 30 min.2. Carefully decant and retain the supernatant.

Concentrate virus by PEG precipitation3. Dissolve 8% PEG (w/v) in clarified lysate and allow to

precipitate overnight at 4°C.4. Centrifuge the PEG solution at 10,000g for 20 min.5. Carefully decant the supernatant, retaining the pellet.6. Resuspend pellet in a small volume of residual PEG

solution and pool all pelleted material.7. Repeat steps 5-6 as needed to concentrate virus to < 1 mL.8. Resuspend virus in 10-50 volumes of culture media

to dilute PEG and allow virus pellet to disaggregateovernight at 4°C.

9. Concentrate sample to ~1 mL through a 30 kDa cut-off disposable centrifugal ultrafiltration device.

Prepare continuous, isopycnic, purifying gradients10. Prepare OptiPrep solutions using culture media as

the diluent. For many viruses a gradient from 25%-40% OptiPrep will provide a good range for separa-tion, but very dense or light viruses may requireadjustment. To achieve this range, prepare 25%,30%, 35%, and 40% v/v final-OptiPrep-concentra-tion solutions, remembering that OptiPrep is sold asa 60% solution. The actual densities these concen-trations achieve are dependent on the density of theculture media used as a diluent, and must be deter-mined for each system.

11. Using the underlayering technique with syringeand pipetting needle, pour 4-step gradients intoopen-toped ultracentrifuge tubes, beginning withthe least dense solution first. Be sure to leaveenough room at the top of the centrifuge tube toload the sample, with 2-3 mm of space at the top.Allow to blend for 2 h at room temperature. Makesure to prepare gradients to serve as balance tubeswhere appropriate.

12. Mark the top of the gradients with a fine-tippedmarker, and carefully overlay virus concentrate usinga transfer pipette. Overlay culture media on balancegradients to create balance tubes.

13. Balance the tubes by adding media to underweighttubes.

14. Load tubes into rotor and ultracentrifuge at maxi-mum permissible speed until density equilibrium isreached. As a guideline, a 4-mL gradient with 1-mLvirus sample in a Beckman Coulter MLS-50 should becentrifuged for at least 4 h 15 min at 200,620g(50,000 rpm); an 11-mL gradient with 1-mL virussample in a Beckman Coulter SW-41 should be cen-trifuged for at least 7 h 20 min at 207,570g (41,000rpm). These conditions should be determined empir-ically for different systems.

Collect viral fraction15. Using any fraction collection apparatus/technique,

carefully extract purified viral concentrate from eachtube. If bands are not visible, a starting point is to frac-tionate the gradient into 4 + fractions, and use a cou-ple of techniques to identify the virus-containing frac-tion (i.e., TEM, bioassay, absorbance at 260 nm, nucleicacid analysis, epifluorescence microscopy or flowcytometry). Note that each of these techniques hasdrawbacks and may lead to false results. For example,more than one virus-containing fraction may bedetected by TEM when analyzing non-axenic cultures,because contaminating bacteria are usually host tophage. Likewise, no virus-containing fractions may bedetected by epifluorescence microscopy if the virus inquestion contains a small ssRNA genome, since the flu-orescence yields of dyes are currently too low for visualdetection of small ssRNA genomes. A combination ofapproaches for identifying virus fractions may there-fore be required when working with novel viruses.

Discussion:OptiPrep must be removed from samples before examina-

tion of virus particles by negative staining and TEM. This canbe achieved using disposable Millipore centrifugal ultrafiltra-tion devices with a 30 kDa cutoff. For most other applicationsOptiPrep does not need to be removed prior to further analy-sis, although it should be assayed to determine effects on thegrowth of specific viral-hosts when re-infection assays are usedto confirm purification of the infectious agent.

Purification of viral assemblages from seawater in CsCl gradi-ents—OptiPrep gradients, as described above for purifying viralisolates, may also be used for purifying marine viral assem-blages. There are as yet, however, few descriptions of this appli-cation in the literature. CsCl gradients, on the other hand, havebeen used extensively for purifications of viral isolates and nat-ural viral assemblages. Given the continued popularity of CsClgradients, and their well-characterized performance, we presentthe following two-part purification and fractionation protocolfor viruses in CsCl gradients. The first part involves relativelyquick sedimentation through a step-gradient to remove thebulk of the contaminants. This is followed by a higher resolu-tion continuous gradient to separate the viruses from residualcontaminants and to separate viruses having differing buoyantdensities from one another. We assume the starting material forthe following procedure to be a concentrated suspension ofviruses (whether filtered to remove cells or not).Materials and Reagents:

• Ultracentrifuge• Swinging-bucket rotor (Beckman Coulter SW 41 Ti or Sor-

vall TH-641)• Polyallomer tubes (Beckman Coulter part No. 331372;

Sorvall part No. 03669)• CsCl (molecular biology grade)• SM Buffer (100 mM NaCl, 8 mM MgSO

4, 50 mM Tris pH 7.5)


179

Steps:Step gradient

1. Prepare CsCl solutions in SM buffer having densitiesof 1.2, 1.35, and 1.55 g mL–1 (see “Gradient prepara-tion” for helpful information)

2. Sequentially layer the CsCl solutions (2 mL of 1.55 gmL–1; 2 mL of 1.35 g mL–1; 1 mL of 1.2 g mL–1) fol-lowed by the sample, into an ultracentrifuge tube. Tominimize mixing at the interface where viruses willcollect, the bottom layer (1.55 g mL–1) can be chilledon ice prior to adding the next layer. The total volumeshould fill the tube to within 2-3 mm of the top. Thisprocedure will accommodate up to 6.5 mL sample, butsmaller sample sizes will ensure the most rapid sedi-mentation of viruses into the gradient. If necessary,top off the sample with mineral oil. NOTE: mineral oilis compatible with polyallomer tubes, but not Beck-man Coulter Ultra-Clear or equivalent tubes.

3. Prepare a balance tube (containing a second sampleor a blank) in the same manner as described above.

4. Verify that tubes that will be in opposing positions inthe rotor are well matched. Opposing tubes shouldhave the same layers that match in density and vol-ume. The total masses of opposing tubes can bematched by adjusting the volume of the top layer.The final total masses should be matched to wellwithin the tolerance of the rotor. Balancing to within1% of the total mass should be more than adequate.Modern centrifuges have much greater imbalancetolerance (as high as 10%), but consult the manualfor your rotor and centrifuge for recommendationsand always err on the side of caution.

5. Centrifuge the sample at 40,000 rpm for 2.5 h at20°C (4°C is also acceptable). Do this as soon as pos-sible after the gradient has been prepared. For olderrotors that have been permanently derated to 36,000rpm, spin at this lower speed, and increase centrifu-gation time to 3.5 h.

6. Unload the gradient immediately after centrifuga-tion by puncturing the side of the tube close to thebottom with a needle and stopcock assembly (see“Gradient Fractionation and Sample Collection”)and collecting 0.5 mL fractions. A low side punctureis recommended, because some unencapsidatednucleic acids may have pelleted and could contami-nate the virus fractions if the tube is punctured at thevery bottom. Measure the densities of the collectedfractions (see “Measuring fraction densities” forimportant considerations). Fractions having densi-ties in the appropriate range can be pooled and sub-jected to further fractionation and purification. Themajority of viruses in seawater have densities > 1.35and < 1.50, while most bacteria will have densities < 1.35 g mL–1. To ensure that the maximum numbers

of viruses are recovered with the minimum contam-ination, one may wish to count viruses (and bacteriaif appropriate) in each fraction by epifluorescence orelectron microscopy. One could also assay the frac-tions for nucleic acid content to determine the dis-tribution of RNA and DNA viruses, or by PCR todetermine the location of specific viruses of particu-lar interest.

7. If proceeding to the subsequent continuous gradientpurification, pool the virus-containing fractions ofthe appropriate density and purity, make sure theyare well mixed, then determine the initial density ofthe pooled sample (see “Measuring fraction densi-ties”) and proceed with the continuous gradient pro-tocol below. If further purification is not required,proceed to the postgradient cleanup step describedfollowing the continuous gradient protocol.

Continuous gradient1. Add CsCl and SM buffer as needed to the sample to

achieve a final density of 1.45 g mL–1 and a finalvolume of 4 mL. Make sure all of the CsCl is dis-solved, the sample is well mixed and near 20°C,and then verify its density (see “Measuring fraction densities”).

2. Prepare solutions of CsCl in SM having a density of 1.20g mL–1 (≥ 4.5 mL per sample) and 1.60 g mL–1 (≥ 3 mLper sample).

3. Layer 3 mL of the 1.60 solution, followed by 4 mLsample at 1.45 g mL–1, then 4.5 mL of the 1.20 g mL–1

solution into an ultracentrifuge tube.4. Prepare a balance tube (containing a second sample

or a blank) in the same manner as described above.5. Verify that tubes in opposing positions in the rotor

are well matched. Opposing tubes should have thesame density and volume and, therefore, total mass.The masses of opposing tubes should be matched towell within the tolerance of the rotor. Consult themanual for your rotor for recommendations, butwithin 1% of the total mass should be more thanadequate (guidelines for modern centrifuges are10%). Top off the gradients with CsCl (1.20 g mL–1)as needed to achieve balance and to ensure that eachtube is filled to within a few millimeters of the top.

6. Centrifuge the samples for ≥ 40 h at 30,000 rpm at20°C (4°C is also acceptable).

7. Unload the gradient immediately, top end first, bydirect unloading with an Auto DensiFlow, a pistonfractionator, or by displacement from below (see“Gradient Fractionation and Sample Collection”).

8. Screen the fractions for viruses (TEM, epifluorescence,nucleic acid assay, etc.) to determine which containthe viruses of interest. Alternatively, if the desiredbuoyant density range of the targeted viruses isknown, the appropriate fractions can be selected


180

based on measured densities of the fractions (see“Measuring fraction densities” for important consid-erations when measuring fraction density).

The CsCl in the relevant virus-containing gradient fractionscan be removed by buffer exchange using a centrifugal ultra-filtration unit with a nominal molecular weight cutoff of30,000 to 100,000 (Microcon, Millipore or Nanosep, Pall LifeSciences). In this case, the sample is repeatedly concentratedto a small volume, then resuspended in the desired buffer.After several such buffer exchanges, the sample is resuspendedand recovered in the final desired volume.Discussion:

The point of the step gradient in this two-step protocol isto achieve a quick initial purification of viruses, which canthen be purified to greater degree in the subsequent continu-ous gradient. The spin is kept short to minimize diffusionbetween layers, which would eventually result in a continuousgradient. Because the spin is short, not all material may reachits equilibrium position in the gradient. Viruses, having rela-tively high sedimentation coefficients, reach their equilibriumpositions more quickly than many other dissolved macromol-ecules. Since all the material starts at the top of the gradient,viruses can therefore be most efficiently separated from lessdense material (e.g., lipids, proteins, and most bacteria) as wellas molecules that are more dense, but which have low sedi-mentation rates (small pieces of DNA or RNA). The subsequentcontinuous gradient, when centrifuged to equilibrium, pro-vides good separation of viruses from contaminants and dif-ferent viruses from one another. Since the less dense contam-inants are mostly removed in the step gradient, which isunloaded from the bottom, we recommend unloading thesubsequent continuous gradient from the top, thereby mini-mizing contamination from free nucleic acids, which willeither band below the viruses (DNA) or form a pellet at thebottom of the tube (RNA). Viruses have been purified formetagenomic analyses using only a step gradient (Angly et al.2006; Breitbart et al. 2004; Breitbart et al. 2002; Vega Thurberet al. 2009), or no gradient at all (Bench et al. 2007; Heltonand Wommack 2009). In the latter cases 0.2-µm filtration andnuclease digestion are relied upon to remove nonviral nucleicacids. Since many virus genes recovered from the environ-ment may not be recognized as viral based only on theirsequence (Edwards and Rohwer 2005), having a highly puri-fied virus preparation increases one’s confidence that anynovel sequences recovered do indeed derive from virusesrather than from cellular life forms. For that reason, one mayfind the complete two-part gradient purification protocolespecially desirable for viral metagenomic studies. A rigorousgradient purification would be particularly important if thesample had not been first filtered to remove prokaryotes.

Fixed-angle rotors (including vertical, near-vertical, andothers) are commonly used for equilibrium buoyant densitygradients, especially self-forming gradients, because the cen-trifugation times required to approach the equilibrium gradi-

ent shape can be much shorter. However, the use of a swing-ing bucket rotor for the continuous gradient, as presentedhere, has several advantages: 1) only a single rotor is neededfor the both the step and the continuous gradients, 2) theopen-top tube simplifies gradient unloading from the top, and3) there is less chance of contaminating the viruses with dis-solved nucleic acids or cellular material. Any contaminatingnucleic acid, especially RNA, that pellets during the run will belocated at the very bottom of the tube where it will not be incontact with the virus bands. Material that is less dense thanthe least dense portion of the gradient will not pellet, but willfloat at the top of the gradient. In any type of fixed anglerotor, these potential contaminants will pellet to some degreeon the sides of the tube and come in contact with the viralbands during deceleration as the gradient reorients in thetube. This material could contaminate the recovered fractionsif it is dislodged or diffuses from the pellets during unloading.This is of particular concern for a sample that has not beenfirst purified through a step gradient.

To partially compensate for the longer centrifugation timesrequired for gradient to reach equilibrium in a swingingbucket rotor, we present a protocol in which a continuous gra-dient forms by diffusion from an initial step gradient. Themodest amount of extra effort needed to prepare the step gra-dient is compensated by a significantly shorter run time. Ifone were to start with a homogeneous CsCl solution in an SW41 or TH-1641 rotor, it could take ≥ 80 h for the gradient, andthe constituents within it, to approach equilibrium. Anotheradvantage of the step gradient is that it provides a dense cush-ion at the bottom of the tube that prevents viruses from pel-leting early in the run before the gradient has fully formed.

WARNING! The protocol we present here is tailored to thespecified rotors and operating conditions. Many other rotorsand centrifugation conditions could be used instead. However,if you wish to change the conditions or adapt the methods toa different rotor, it is critical that you ensure that the new con-ditions are within the safe limits for centrifugation of CsClgradients. Centrifugation of CsCl solutions at certain combi-nations of concentration, temperature, and rotor speed canresult in CsCl crystallization at the bottom of the tube. Thehigh density of the crystals will exceed the tolerance of therotor and could result in catastrophic rotor failure.

Summary—Ultracentrifugation provides a highly flexiblemeans for purifying viruses. While it is undoubtedly one ofthe most effective approaches, there are many critical vari-ables to consider when developing and adapting methods foroptimal performance. As with any technique, researchers whoendeavor to apply ultracentrifugation should thoroughly con-sult the literature for previously developed methods intendedfor similar downstream applications, and be prepared to opti-mize for the particular material and equipment available. Itcannot be emphasized enough how important it is to monitoreach step in the purification process to ensure one is effec-tively purifying the target(s) of interest (Rowlands et al. 1971).


181

References

Anderson, N. G., W. W. Harris, A. A. Barber, C. T. Rankin, andE. L. Candler. 1966. Separation of subcellular componentsand viruses by combined rate- and isopycnic-zonal cen-trifugation. J. Nat. Cancer Inst. Monographs 21:253-269.

———, and G. B. Cline. 1967. New centrifugal methods forvirus isolation, p. 137-178. In M. K. Maramorosch and H.Koprowski [eds.], Methods in virology. V. 2. AcademicPress.

Anderson, T. F., C. Rappaport, and N. A. Muscatine. 1953. Onthe structure and osmotic properties of phage particles.Ann. Inst. Pasteur 84:5-15.

Angly, F. E., and others. 2006. The marine viromes of fouroceanic regions. PLoS Biol 4:2121-2131.

Bench, S. R., and others. 2007. Metagenomic characterizationof Chesapeake Bay virioplankton. Appl. Environ. Micro-biol. 73:7629-7641.

Brakke, M. K. 1967. Density-gradient centrifugation, p. 93-118. In M. K. Maramorosch and H. Koprowski [eds.], Meth-ods in virology. V. 2. Academic Press.

Bratbak, G., O. H. Haslund, M. Heldal, A. Nœss, and T. Røeggen.1992. Giant marine viruses? Mar. Ecol. Prog. Ser. 85:201-202.

Breitbart, M., and others. 2002. Genomic analysis of uncul-tured marine viral communities. Proc. Natl. Acad. Sci.U.S.A. 99:14250-14255.

———, and others. 2004. Diversity and population structureof a near-shore marine sediment viral community. Proc.Royal Soc. Biol. Sci. Ser. B 271:565-574.

Cline, G. B., and R. B. Ryel. 1971. Zonal centrifugation. Meth-ods Enzymol. 22:168-204.

Edwards, R. A., and F. Rohwer. 2005. Viral metagenomics.Nature Rev. Microbiol. 3:504-510.

Fauquet, C. M., M. A. Mayo, J. Maniloff, U. Desselberger, andL. A. Ball [eds.]. 2005. Virus taxonomy: VIIth Report of theInternational Committee on the Taxonomy of Viruses. Else-vier, Academic Press.

Garza, D. R., and C. A. Suttle.1995. Large double-strandedDNA viruses which cause the lysis of a marine het-erotrophic nanoflagellate (Bodo sp.) occur in natural marineviral communities. Aquat. Microb. Ecol. 9:203-210.

Griffith, O. M. 1994. Techniques of preparative, zonal, andcontinuous flow ultracentrifugation, 5th ed. BeckmanInstruments.

Helton, R. R., and K. E. Wommack. 2009. Seasonal dynamicsand metagenomic characterization of estuarine virioben-

thos assemblages by randomly amplified polymorphic DNAPCR. Appl. Environ. Microbiol. 75:2259-2265.

Hinton, R. H., B. M. Mullock, and C. C. Gilhuus-Moe. 1974.The use of metrizamide for the fractionation of ribonucleo-protein particles, p. 103-110. In E. Reid [ed.], Methodologi-cal developments in biochemistry. V. 4. Longman.

———, and ———. 1997. Isolation of subcellular fractions, p.31-69. In J. M. Graham and D. Rickwood [eds.], Subcellularfractionation: a practical approach. IRL Press at Oxford Uni-versity Press.

Lide, D. R. [ed.]. 2009. CRC handbook of chemistry andphysics, 90th ed. CRC Press.

Luttmann, W., K. Bratke, M. Kupper, and D. Mytrek. 2006. Cellseparation, p 44-54. In Immunology. Academic Press.

Rickwood, D. 1984. Centrifugation, a practical approach, 2nded. IRL Press.

——— [ed.]. 1992. Preparative centrifugation: a practicalapproach. IRL Press at Oxford Univ. Press.

Pertoft, H., L. Philipson, P. Oxelfelt, and S. Höglund. 1967.Gradient centrifugation of viruses in colloidal silica. Virol-ogy 33:185-196.

Rowlands, D. J., D. V. Sangar, and F. Brown. 1971. Buoyantdensity of picornaviruses in caesium salts. J. Gen. Virol.13:141-152.

Rüst, P. 1968. Cesium chloride density distributions in prepar-ative equilibrium ultracentrifugation. Biopolymers 6:1185-1200.

Sambrook, J., and D. W. Russell. 2001. Molecular cloning: A lab-oratory manual, 3rd ed. Cold Spring Harbor Laboratory Press.

Schmidt, E. R. 1973. Separation of isolated cell nuclei of tissuesof the meal moth, Ephestia kuehniella Z. by metrizamidedensity gradient centrifugation. Differentiation 7:107-112.


Vanden Berghe, D. A. 1983. Comparison of various density-gradient media for the isolation and characterization ofanimal viruses, p. 175-193. In D. Rickwood [ed.], Iodinateddensity gradient media: A practical approach. IRL Press.

Vega Thurber, R., M. Haynes, M. Breitbart, L. Wegley, and F.Rohwer. 2009. Laboratory procedures to generate viralmetagenomes. Nat. Protocols 4:470-483.

Velimirov, B. 2001. Nanobacteria, ultramicrobacteria and star-vation forms: a search for the smallest metabolizing bac-terium. Microbes Environ. 16:67-77.

182

The domain of aquatic viruses includes bacterial, archaeal,and algal viruses. Bacterial viruses, commonly known asphages or bacteriophages, include cyanophages or viruses ofcyanobacteria, formerly referred to as “viruses of blue-greenalgae.” Bacteriophages were discovered in 1915 by FrederickWilliam Twort and, independently, in 1917 by Felix d’Hérelle(Ackermann 2008). The first cyanophage, a “phycovirus”infecting Phormidium, was described in 1964 (Safferman andMorris), followed in 1974 by the first true algal virus, a virusof the brown alga Ectocarpus (Clitheroe and Evans) and thefirst archaeal viruses, two-tailed Halobacterium phages (Torsvikand Dundas). In 1991, 44 algal virus taxa were known to con-tain or produce viruslike particles (Van Etten et al.). Unfortu-nately, no recent review of this interesting subject is available.

Indigenous marine phages were found as early as 1960(Spencer), but it was realized rather belatedly that marine

phages, including cyanophages, were ubiquitous and occurredin enormous numbers in nature. It appears now that bacterio-phages are the most abundant life form on Earth, their globalpopulation being estimated at 1031 viruses (Hendrix 2003). Thefrequency of eukaryotic algal viruses can only be guessed, andtheir classification is still at its beginnings. So far, only algalviruses of the Phycodnaviridae and Marnaviridae families havebeen officially classified by the International Committee onTaxonomy of Viruses (ICTV) (Fauquet et al. 2005). It is clear thatmany more algal viruses occur in nature and await classifica-tion. In addition, seawater contains unusual and novel entitiessuch as mimiviruses (La Scola et al. 2003, Suttle 2005).

The ICTV classifies viruses primarily by nucleic acid andphysicochemical properties of the virion, but any viral prop-erty may be used for classification. Viruses contain dsDNA,ssDNA, dsRNA, or ssRNA and are isometric, helical, pleomor-phic, or tailed. They are presently categorized into threeorders, 67 families, and 282 genera (Fauquet et al. 2005). Newviruses are continuously described. Virus classification is along and stepwise procedure. Bacterial and archaeal viruses aregrouped together as prokaryote viruses. More than 5500prokaryote viruses have been examined in the electron micro-scope (Ackermann 2007), with less than 300 of them beingfully sequenced in 2007 (Ackermann and Kropinski). This sug-gests that we are still at the thresholds of viral ecology andthat new discoveries are waiting to be made anytime. Approx-imately 96% of prokaryote viruses are tailed. They belong tothe Myoviridae, Siphoviridae, and Podoviridae families (Acker-mann 2007; Ackermann and DuBow 1987; Fauquet et al.2005) and are easily identified in the electron microscope. Iso-

Basic electron microscopy of aquatic virusesHans-W. Ackermann1 and Mikal Heldal21Felix d’Hérelle Center for Bacterial Viruses, Department of Microbiology-Immunology-Infectiology, Faculty of Medicine, LavalUniversity, Quebec, Quebec, G1K 7P4, Canada2Department of Biology, University of Bergen, Jahnebakken 5, 5020 Bergen, Norway

AbstractFor identification and structural studies, viruses are concentrated and purified by differential or density gra-

dient centrifugation, stained with phosphotungstate (PT) or uranyl acetate (UA), and examined by transmissionelectron microscopy. Both PT and UA produce artifacts. PT yields negative staining only; UA produces both neg-ative and positive staining and often gives excellent contrast. Fixation is normally not necessary. UA-stainedpreparations can be stored for years. Virus particles may be sedimented directly from water samples onto gridsby means of special centrifuge tubes and subsequent staining. Positively stained particles are highly contrastedand easy to count at low magnification. Positive staining also provides information on bacteria containing viralparticles and rough estimates of burst sizes in individual bacteria. Isometric, filamentous, and pleomorphicviruses are identified after negative staining.

*Corresponding author: E-mail: [email protected],[email protected]



ISBN 978-0-9845591-0-7, DOI 10.4319/mave.2010.978-0-9845591-0-7.182Suggested citation format: Ackermann, H.-W., and M. Heldal. 2010. Basic electron microscopy ofaquatic viruses, p. 182–192. In S. W. Wilhelm, M. G. Weinbauer, and C. A. Suttle [eds.], Manualof Aquatic Viral Ecology. ASLO.


MANUALof


Ackermann and Heldal Electron microscopy

183

metric, filamentous, and pleomorphic viruses are much moredifficult to recognize. Algal viruses are isometric or filamen-tous. The major hosts of marine phages are proteobacteria ofthe genera Pseudomonas and Vibrio and cyanobacteria of thegenera Synechococcus and Prochlorococcus.

Electron microscopy was introduced in the 1930s. It is aresearch field on its own and a basic technique of microbiology.The first virus images were published in 1938 and showed themouse ectromelia virus, a poxvirus (von Borries et al.). The firstbacteriophages, mostly coliphages with short tails, wereobserved in 1940 (Pfankuch and Kausche; Ruska). As early as1943, a morphological classification of all viruses was proposed(Ruska). Viruses were first examined unstained and later aftershadowing with a metal to enhance contrast. Shadowing, acumbersome and aggressive procedure, was long ago replacedby negative and positive staining. Presently, the principal tech-nique in use is the examination of negatively stained particlesin transmission electron microscopes. The introduction of digi-tal electron microscopes and digital cameras has created a wholenew situation with occasionally unfortunate results (see below).

The challenges in the electron microscopy of aquaticviruses are still numerous. How do we identify a virus amongother particles in, for example, the marine environment? Whydo we hardly recognize filamentous viruses in these samples?Would more detailed studies of negatively stained prepara-tions add to our knowledge of viral impact on microbialecosystems? How to circumvent the use of uranyl acetate, as itis slightly radioactive and its use is restricted in many coun-tries? The objective of this article is to summarize the generalmethods used for morphological studies, total virus counts,and other applications of electron microscopy in an ecologicalcontext. Our ultimate aim is to improve the quality of envi-ronmental electron microscopy.

Materials and procedures1. General—Virological electron microscopy has essentially descriptive

purposes, namely the identification and characterization of freeviruses. Of the many biological applications of electronmicroscopy (EM), only transmission electron microscopy (TEM)is of significant importance in virology. Its other applications,e.g., scanning electron microscopy (SEM), cryo-EM with orwithout three-dimensional image reconstruction, immuno-EM,sectioning of infected cells, shadowing of individual particles,or enzymatic virus digestion on grids, have limited applicationsand are sometimes practiced in a few specialized laboratoriesonly. In aquatic virology, sectioning is still important because itpermits identification of intracellular assembly sites of novelviruses and the detection, in mixed host systems, of which hostcell produces viruses. An excellent overview of sectioning tech-niques may be found in a recent book (Kuo 2007). The directharvesting of particles onto grids for total viruslike counts wasand is of great importance for the understanding of viral activ-ity in aquatic ecosystems (Bergh et al. 1989).

Negative staining was introduced in 1959 (Brenner andHorne) and revolutionized virology in generating virus imagesof unprecedented clarity. It is indispensable for structural stud-ies and virus identification, but is also used for particle counts(below). The principle of negative staining is to mix the parti-cles to be examined with an electron-dense solution of a metalsalt of high molecular weight and small molecular size. On anelectron microscopic grid, the particles then appear white orgray on a dark background. Many variant procedures havebeen devised. A detailed description of stains and proceduresmay be found elsewhere (Ackermann and Dubow 1987; Hayatand Miller 1990; Kay 1965).

2. Specific techniques—a. Purification:For high-resolution images, samples should be sterile and

free from bacteria and other large particles. This is achieved byfiltration through membrane filters of 0.22–0.45 µm pore size.Except for particle counts on EM grids (below), purification isa must because proteins and slime interfere with staining andresolution and their amounts must be reduced to acceptablelevels. The most convenient procedure is washing in a buffer,e.g., 0.1 M ammonium acetate (pH 7), in an ultracentrifugewith a fixed-angle rotor (25,000g or 25,000 rpm for 1 h). Cen-trifugation in swinging-bucket rotors requires higher g forcesand centrifugation times (e.g., 70,000–80,000g for 60–90 min).Clearly, fixed-angle rotors are advantageous as they permit theuse of relatively inexpensive medium-size centrifuges. Phos-phate buffer is acceptable, but physiological saline is notbecause it leads to the formation of salt crystals on grids. Alter-natively, viruses may be purified by CsCl or sucrose gradientcentrifugation followed by dialysis.

b. Particle counts on EM grids:Viruses in lysates are counted by TEM after depositing or

vaporizing a fixed volume of virus suspension on the grid (Figs. 1and 2). Viruslike particles (VLPs) in water can be counted byTEM, flow cytometry, or epifluorescence microscopy (EFM). Thedirect harvesting of particles on TEM grids ensures that bacteria,algae, and other particles can be observed simultaneously. Thesepreparations can also be used for estimating bacterial size distri-bution and the frequency of infected cells and dividing bacteria,and to visualize lysed bacteria.

The technique requires an ultracentrifuge with a swinging-bucket rotor and centrifugation tubes (Polyallomer®) with flatbottoms, created by introducing Epoxy resin into a conicaltube. The resin is left to harden at 60°C overnight, and an elec-tron microscopic grid is deposited on the flat resin surface.Samples are centrifuged at 80,000g for 90 min (Borsheim et al.1990). Tubes fitting Beckman SW-41 rotors allow particle har-vesting from a 60-mm water column. Use cellulose nitrate fil-ters fitting the diameter of the tubes, add a small piece of dou-ble-sided tape, and position the grids at the edge of the tape inthe middle of the tube. Fill the tubes with sample and thenadd the filter paper with the grids attached to avoid air bub-bles being trapped under the grids. The tubes should always be


184

filled up to 2–3 mm from the top. If particle density is high(107–108 cells mL–1), the sample volume can be reduced. Thiscan be done by reducing the height of tubes. The calculationof total virus contents is described by Bratbak and Heldal(1993) (see Appendix).

Montanié et al. (2002) presented a different approach tototal virus counts by TEM. A purification step is followed bypelleting viruses from the supernatant by means of an ultra-centrifuge. The particles are then transferred by capillarity togrids with carbon–collodium membranes. The recovery ofviruses is reported to be 71% to 79%. The filtration techniqueuses membrane filters of 0.2 µm pore size for virus concentra-tion. Virus particles on the filter are then resuspended in waterand the suspension is examined in the EM (Torrella andMorita 1979). Again, this technique is inexact because parti-cles smaller than 0.2 µm may be lost and also because ofuncertainties introduced by resuspension. The technique ofEwert and Paynter (1980) involves sedimentation of viruses bymeans of a special rotor onto agar blocks and the preparationof pseudoreplicas, which are then stained and examined inthe EM. The technique is quite complicated and has, to ourknowledge, not been used by anybody else. We have no per-sonal experience with it.

c. Special applications:TEM has been used to visualize viruses within infected,

unstained cells (Fig. 1). The aim is to determine burst sizes,

lytic cycle lengths, percentages of bacteria in a stage of lyticinfection, and mortality rates. The subject has been aptlyreviewed by Weinbauer (2004).

d. Stains:The stains used are tungstates (potassium, sodium, lithium,

silicotungstates), uranyl salts (acetate, formate, magnesiumacetate, nitrate, oxalate), molybdic acid, and ammonium orvanadium molybdates. The most common stains are aqueoussolutions of sodium or potassium phosphotungstate (PT; usu-ally 2% and pH 7.2) and uranyl acetate (UA; 1%–2%, pH4–4.5). Stains are prepared by dissolving phosphotungstic acidor uranyl acetate in distilled water and adjusting the pH withKOH or NaOH. They can be kept in stoppered bottles at 4°Cfor 2 years. Uranyl salts have a strong affinity to double-stranded DNA (Huxley and Zubay 1961), are toxic andradioactive, and should be handled with care. PT producesnegative staining only. UA produces negative staining and,because of its affinity to dsDNA, will stain positively in blackany particles with sufficient dsDNA content, for example,phage heads or herpesvirus capsids.

e. Staining:The required supplies include pointed pipettes and tweez-

ers, strips of filter paper, and common electron microscopicgrids (Athene-type, copper or steel, 100–400 mesh or squareholes). Grids must carry a Formvar or collodium film stabilizedby a carbon layer. They may be prepared by the electronmicroscopist or purchased ready-made. After 1 month, gridstend to be hydrophobic and reject particles. Grids can bemade hydrophilic again by glow discharge in a carbon evapo-rator or by rinsing with a wetting agent (Alcian blue, baci-tracin, poly-L-lysine, serum albumin, sucrose) (Gentile andGelderblom 2005; Gregory and Pirie 1973). One of the authors(H.-W. Ackermann) sometimes uses UA or PT supplementedwith 2 drops/mL of a 0.5% bacitracin solution.

Figs. 1 and 2. Particles harvested by centrifugation onto electronmicroscopic grids. Samples from Raunefjord near Bergen were sedi-mented onto 400-mesh Ni grids carrying carbon-coated Formvar films,using a Beckman Coulter ultracentrifuge, an SW-41 rotor, and Polyallomertubes with molded flat bottoms. Grids were stained for 1 min with 2%UA, air-dried, rinsed in distilled water, and examined in a JEOL TEM 100.

Fig. 1. Marine bacterium with mature phages before lysis and freephages in the vicinity. Note the size variation in viruslike particles. Thesquares and crownlike structures are scales of algae, e.g., Pyramimonasorientalis (Chlorophyta).

Fig. 2. Viral cluster outside a lysed cell.


185

The staining procedure is extremely simple and near-instantaneous. A drop of virus suspension is deposited on agrid, left 30 s for adsorption of viruses, and then a drop ofstain is added, and the liquid is withdrawn with filter paper.Positive staining takes a little longer, but is complete after 1min. Preparations prepared for TEM by centrifugation arestained with 1%–2% UA in water for 2–3 min, gently washedin distilled water, and air-dried. This will give a positive stain-ing which facilitates total virus counts and an estimation ofthe number of infected cells and burst sizes (see Weinbauer2004 and references therein). Positively stained viruses appearas electron-dense and deep black. The presence of impuritiesmay create a high background noise.

f. Electron microscopes and imaging:Transmission electron microscopes (TEMs) are operated at

60–80 kV. To limit contamination by hydrocarbon moleculespresent in the column and hence loss of resolution, the objectstage may be cooled with liquid nitrogen. This is not necessaryfor experienced users. Three types of instruments are presentlyin use.

(1) Conventional TEMs. Lens columns, astigmatism, focus-ing, and magnification are adjusted by the user. Images arerecorded on films or plates and are as much a product of thedarkroom as of the EM. They require a darkroom equippedwith a enlarger of professional quality. Mercury lamps andexhausts are not necessary, but small table-top enlargers asused by amateur photographers are inadequate. Suppliesinclude fine-grain films, high-speed developers, polycontrastpaper, and graded gelatin filters for enhancing contrast. Imagecontrast generally depends on developers, and papers and canbe dramatically improved by filters. It is unfortunate thatsome photographical supplies, especially papers, have becomedifficult to find or have even disappeared from the market fol-lowing the introduction of digital cameras.

(2) Automated TEMs. Images are recorded as above on filmor plates, but focusing is automated, columns are largely self-contained, and resolution and magnification can no longer(or only with great difficulty) be adjusted by the electronmicroscopist. Their operation has thus been considerably sim-plified, but the operator has little control over the instrumentand depends largely on the installer for alignment, resolution,and magnification control. Most of these instruments are nowequipped with digital cameras and are commonly called “dig-ital electron microscopes.”

(3) Digital (charged-couple device [CCD]) cameras. The cam-eras are not made by EM manufacturers and must be pur-chased separately. Images are recorded electronically and areeasy to store, retrieve, copy, and exchange. A darkroom is nolonger necessary, but images must be printed on special paperand by means of special printers.

If no TEM is available, it is possible to use a scanning elec-tron microscope (SEM) equipped with a field emission gun(FESEM) and scanning–transmission electron microscopy(STEM) detector. Most of these instruments can be run in

STEM mode and used as ordinary TEMs. Preparation of sam-ples and staining are as described for TEM. FESEM instrumentsare normally operated at 10–40 kV, but for magnifications upto ×150,000, reasonably good total counts of viruslike particlesare obtained. One example is the JEOL instrument JSM-7400F(cold emitter), with which unfixed and unstained marineviruses were observed in STEM mode at a magnification of×100,000 (M. Heldal, unpubl. data).

The other advantage of FESEMs is the possibility of obtain-ing high-resolution images in scanning mode, even at lowaccelerating voltages (1.5–2.5 kV). In such cases, the evapora-tion of gold/palladium on the material may give a coarsemetal surface, and evaporation of iridium should be preferred.As an example, a Zeiss SUPRA 55 VP instrument (FESEM)equipped with an in-lens detector can be run at enlargementsof up to ×500,000. Altogether, the new-generation SEMs (FES-EMs) may be a good choice for providing good resolutionmicroscopy for both TEM and SEM preparations. A drawbackis the time-consuming setup of the STEM detector.

g. Magnification control:High magnification (×250,000 or more) is controlled by

means of grids with beef liver catalase crystals (Luftig1967) or T4 bacteriophage tails. Diffraction grating replicasare suitable for low magnification only (below ×30,000).Calibration grids can be prepared in advance and kept forat least 1 month. Catalase crystals have parallel lines with8.8-nm periodicity, can be used after PT and UA staining,and are commercially available. At a magnification of×300,000, 20 lines correspond to 5.2 cm. T4 phage tails are114 nm long, which corresponds to 34 mm at the samemagnification. In conventional electron microscopes,magnification should be monitored often because of possi-ble fluctuations in the local electricity supply. This is eas-ily done by taking a picture of a calibration grid every20–50 photographs and adjusting the darkroom enlarger.Digital electron microscopes and cameras are calibrated bythe supplier or installer. This is a difficult procedure; it isdoubtful whether these adjustments are all done by com-petent operators and it is so far unclear how stable digitalcameras are over long periods.

3. Assessment—a. Purification:Centrifugation does not eliminate bacterial debris (cell

wall fragments, flagella, pili, capsule material), but this isnot a real problem as these particles do not interfere withresolution and it is always possible to find areas withoutthem. In addition, almost all centrifuged preparations con-tain phage debris (e.g., isolated heads or tails) and abnormalor defective particles (polyheads or polytails). Using a fixed-angle rotor, a force of 25,000g is amply sufficient for tailedphages, the largest of which sediment at 6000g for 1 h. It isjust sufficient for very small viruses, e.g., microviruses or fil-amentous phages, which require 30,000g for 1 h for com-plete sedimentation (Ackermann and Dubow 1987). The


186

technique described above is akin to centrifugation in theBeckman Coulter Airfuge which, however, uses extremelysmall volumes (180–240 µL) and is thus unsuitable for envi-ronmental studies (see above, 2a). Density gradient purifica-tion is not mandatory for environmental studies. Suchpreparations can be very rich and pure, but some are full ofcrystals and impurities. The technique critically depends onadequate dialysis, demands large amounts of material, andcan be recommended only if the performing scientist is sureof his or her skills.

b. Particle counts and special applications:Counting of viruses in lysates can be difficult because

viruses often do not disperse well or adhere poorly to the grid.For example, phages may aggregate with each other or cellulardebris and even separate into areas of intact and damaged par-ticles. Both problems can be overcome by washing grids with asolution of poly-L-lysine (Müller et al. 1980). Visualizing viruseswithin unstained infected bacteria is problematic because ofbacterial inclusion bodies, notably carboxysomes (Shively et al.1973) and polyhydroxybutyrate (Tian et al. 2005). Car-boxysomes, which occur among others in the importantcyanobacterium Synechococcus, resemble phycodnaviruses indiameter and hexagonal shape, but are recognizable by theirvariable size and distinctive capsomerlike structures.

c. Stains and staining:Phosphotungstate (Table 1) is a neutral stain and produces

negative staining only. Artifacts are few and results are pre-dictable and constant. Full phage heads often appear roundedin PT; empty heads appear as membranes or are filled by stainand then resemble black vesicles. Ammonium molybdate pro-duces the same type of staining as PT and has no advantageover it.

UA is an acid stain and causes both negative and positivestaining. It is unpredictable and there is no way to producepositive or negative staining consistently and reliably. Bothtypes of staining are usually present on the same grid andeven on adjacent parts of the same grid hole (Fig. 3). How-ever, proteinic impurities seem to facilitate negative staining.Negatively stained phage heads are white or grayish whenfull, but appear as thick membranes when empty. Positivelystained viruses (or phage heads) are deep black, indicating thepresence of dsDNA. They can easily be detected at low mag-nification, which is advantageous in environmental studies(see Borsheim et al. 1990). Positively stained phage heads areinvariably shrunken by 10% to 15% (Figs. 4 and 5), and theirdimensions are useless for virus identification and descrip-tion. UA produces positive staining with large dsDNA-con-taining particles and not with ssRNA-containing phages, fila-mentous phages, or dsDNA viruses with low DNA contentsuch as marine phage PM2. Consequently, these viruses can-not be detected by positive staining. In addition, UA causesswelling of purely proteinic structures such as phage tails andempty capsids. UA produces excellent contrast and is particu-larly suitable for resolution of tail striations. Negatively

stained heads are more angular and presumably better pre-served than in PT. UA spreads more evenly than PT, but tendsto crystallize on grids.

Both PT and UA form glassy films around particles and actas fixatives. PT-stained preparations are stable for a monthwithout loss of detail. UA-stained grids keep easily for 10years in grid boxes at room temperature. One of the authors(H.-W. Ackermann) has several grids that are 30 years old andstill observable. Because PT and UA are complementary,investigators should always use both of them. Fixation, forexample with 2.5% glutaraldehyde, may be necessary formarine phages that cannot be processed immediately andhave to travel weeks before examination (Cochlan et al.1993). The stability of glutaraldehyde- or formaldehyde-fixedmaterial seems to be highly variable. Bacteria (Gundersen etal. 1996) and some viruses have a variable loss rate in fixedsamples, whereas the formaldehyde-fixed Emiliania huxleyivirus showed reasonably high stability for 2–3 years understorage in the dark (Bratbak et al. 1993). As a rule, we recom-mend harvesting and preparation on grids of fresh materialwithout any fixation.

d. Electron microscopes and imaging:The electron-optical resolution of EMs has not improved

since the 1960s; all technical improvements made since con-cern ease of operation, automated focusing, digitalization, andgadgetry. The introduction of digital cameras (and hence dig-ital electron microscopes) has not improved image quality,speed or ease of manipulations, or image recording. On thecontrary, it has generated a flurry of gray, contrastless picturesof poor resolution which may irreverently be called “graypulp.” Even company prospects show similar gray pictures.Neither prospects nor instruction manuals provide guidancefor contrast improvement. It appears that the various EM andcamera companies marketed their products in haste and with-out knowing how to operate them. The relative failure of dig-ital electron microscopes is largely due to lack of information.Fortunately, image contrast can be greatly improved bymanipulation of the camera software, particularly of the his-togram governing luminescence (gamma), brightness, andintensity values (Tiekotter and Ackermann 2009). These

Table 1. Comparison of PT and UA

Parameter PT UA

pH Neutral Acidic

Crystallizes on grids Generally not Yes, often in blades

Acts as a fixative No Yes

Type of staining Negative Negative or positive

Grid life 1 month 10 years or more

Contrast Poor to good Poor to excellent

Negatively stained capsids Often rounded Angular

Positively stained capsids Black, shrunken

Protein structures Swollen


187

parameters can be adjusted separately for selected areas, andexcellent, high-contrast pictures can be so obtained. A majordrawback of CCD cameras is the high price of cameras($50,000 to $300,000) and printers ($5,000–10,000). Thiscompares poorly with $2,000 for a Philips EM300 conven-tional camera and $10 for photographic trays and puts digitalEM out of reach of most laboratories and institutes.

e. Magnification control:The specifications of manufacturers cannot be trusted, as

they depend on lens setting and local electricity supply. Con-ventional and digital EMs suffer from the same problem andboth must be controlled by means of test specimens. Unfortu-nately, both catalase crystals and T4 tails have their problems.Catalase crystals may be deformed and it can be difficult tofind enough parallel lines of subunits for accurate mea-surements. In addition, the crystals are soluble in water andtend to dissolve into subunits. T4 tails stained with UA (butnot PT) may be stretched beyond their normal length. On theother hand, T4 phages are easy to prepare in a microbiologicallaboratory and much easier to use than catalase crystals. Itremains to see how stable and trustworthy digital cameras areover long periods of operation. Unfortunately, to our knowl-edge, the adjustment of digital EMs and cameras is lengthyand difficult and may take 2–3 h. It is essentially out of reachfor investigators, who depend for adjustments on manufac-turers and their technicians.

f. Measurements:Virus particles on conventional photographic images must

be measured after enlargement on prints and never on films,plates, or screens. The reason is that measurements of smallparticles at low magnification are difficult per se and cannotbe accurately extrapolated to high magnification. Digitizedparticles can be measured directly on the monitor screen.Long flexible phage tails may be measured with special flexi-ble rulers. Investigators should be aware of artifacts, shouldnot measure positively stained particles, and should measureat least 10 particles to obtain statistically meaningful data atsufficiently high magnifications (×250,000 or higher) to mini-mize measurement errors. Isometric viruses or phage headsshould be measured between opposite apices. The literaturesometimes includes mentions of phage heads of 30 nm indiameter. This puts tailed phages into the vicinity ofpolioviruses and is clearly an error. The smallest tailed phagesknown, podoviruses of the φ29 type (Fauquet et al. 2005),have head diameters of 40 nm. All others have heads of 50–60nm in diameter or larger.

g. Virus detection and identification:Preparations enriched by incubation with host bacteria

may reveal a wide variety of phages or phage-like particles(Rachel et al. 2002). Virus detection is both simple and diffi-cult. It is easy because tailed phages, the mainstay of watersamples, are instantly recognizable and differentiated in theEM. They belong to three clear-cut families, Myoviridae,Siphoviridae, and Podoviridae (Ackermann 2007; Ackermann

and DuBow 1987; Fauquet et al. 2005). Members of theMyoviridae family are characterized by contractile tails likephage T4. Tails are generally thick, rigid, and constituted by atail tube and a sheath separated from the head by a constric-tion or “neck.” Tail sheaths may be contracted. Siphoviridaehave long, noncontractile, often very flexible tails like phagelambda. Podoviridae, similar to coliphage T7, have short tails ofabout 10 nm in length. Positively stained podoviruses mayappear as headless tails and be difficult or impossible to iden-tify. All tailed phage families have members with isometric orelongated heads. Empty capsids and headless tails will notelicit positive staining and remain invisible.

The detection of isometric, filamentous, or pleomorphicviruses is intrinsically difficult. These viruses seem to be rare innature, constituting only 4% of prokaryote virus isolates (Ack-ermann 2007). Further, ssDNA, ssRNA, and small dsRNAviruses do not stain positively. In crude, nonenriched samples,they can be diagnosed (a) after negative staining only and ifthey (b) occur in large numbers of identical particles, (c) pres-ent regular structures such as capsomers, and/or (d) havedimensions corresponding to those of known viruses. Theidentification of enveloped viruses without inner capsids,such as paramyxoviruses, is difficult or even impossible.Indeed, no pleomorphic virus has ever been identified by sim-ple electron microscopic examination of environmental sam-ples. Even the visualization of spikes on membranous particlesis no help, as they also occur on mammalian cell membranes.The detection of filamentous phages is particularly difficultbecause these viruses are thin, break easily, and may be con-founded with pili, flagella, or slime filaments. Algal viruses areisometric or filamentous (Van Etten et al. 1991). Their pres-ence may be suspected if particles are isometric, at least 100nm in diameter (Fauquet at al. 2005), and show capsomersand transverse edges. Filamentous algal viruses are known inthe brown alga, Chara corallina (Skotnicki et al. 1976), butshould be diagnosed only if they show transverse striations.Pitfalls are many; for example, one of the authors (H.W. Ack-ermann) once misidentified the mastigoneme filaments ofOchromonas algae (Bouck 1971) as possible filamentousviruses. In conclusion, isometric, filamentous, or pleomorphicviruses should be diagnosed only after their concentration ina density gradient or their propagation as pure cultures.

For a definitive diagnosis and identification with knownviruses, high-magnification (×250,000 to ×300,000) imagesand exact dimensions of negatively stained viruses arerequired. Images of positively stained viruses are unsuitablefor this purpose and their dimensions are always inexact (seeabove, 3c; Figs. 4, 5).

h. Artifacts:Damaged particles and bacterial debris are frequent in

phage lysates and environmental samples. For example, tailedphages are quite frequently decapitated, simulating taillessviruses or short-tailed phages. The frequency of tail lossdepends on the phage and, possibly, storage conditions. Some


188

Fig. 6. (a), Cell wall and cytoplasma membrane debris of lysedPseudomonas aeruginosa cells; UA, ×92,400. (b), Vesicles budding fromthe cell membrane of a phage-infected Salmonella bacterium; thin sec-tion, ×59,400. (c), False herpesvirus simulated by cell wall debris sur-rounding an unknown particle; Vibrio phage lysate, UA, ×297,000.

Fig. 4. Negatively and positively (right) stained T4-like Vibrio para-haemolyticus phage; UA, ×297,000.

Fig. 5. Negatively and positively (right) stained phage 71A-6 of Vibriovulnificus; UA, ×297,000. The phage is a podovirus with a very long cap-sid and a barely visible short tail.

Figs. 3–8. Various bacteriophages and artifacts. Philips EM 300 electronmicroscope; uranyl acetate (2%, UA) or phosphotungstate (2%, PT)except Fig. 5b, which is a section. Bars indicate 100 nm.

Fig. 3. Negatively and positively (center) stained Salmonella phageswithin the same field; UA, ×92,400.

Fig. 7. Slime filament and various bacterial debris. Bacillus cereus phageBL1, PT, ×148.500.


189

phages are stable over years and some, such as coliphage Mu,fall into pieces within weeks. The frequency of damaged par-ticles increases with repeated freezing and thawing. Thisshould be considered if samples are taken during a maritimeexpedition, frozen, and processed weeks later. Bacterial debrisinclude round particles derived from cell walls or cytoplasmicmembranes, pili, flagella, and capsule-related mucus. Roundparticles (Fig. 6), often of the size of normal phage heads andthick-shelled, are particularly frequent in lysates ofPseudomonas phages, but also occur in those of Vibrio andSynechococcus. They may be misdiagnosed by inexperiencedobservers as enveloped viruses. We have even seen round celldebris lodged within other spherical cellular vesicles, simulat-ing enveloped viruses (Fig. 6c) or Tectiviridae. All these debrisare likely to adsorb DNA. Unless treated with DNase, they arebound to mimic DNA phages in epifluorescence microscopy.Slime, produced for example by xanthan-producing Xan-thomonas bacteria, is likely to occur in freshwater. When cen-trifuged, slime blobs may become elongated and then mas-querade as filamentous viruses (Fig. 7) and even tailed phages.Short flagella fragments often resemble contractile phage tails,and pili are almost impossible to distinguish from filamentousviruses of the Inoviridae family. Further, salts may obscure

phage structures (Fig. 8), and UA sometimes simulatesenvelopes by producing haloes around positively stainedviruses (Fig. 9). In conclusion, observers must be wary andconstantly on the outlook for artifacts (Table 2).

Discussion and conclusionsThe investigation of waterborne viruses suffers from the

absence of atlases of cyanophages and algal viruses. Theyshould include photographs with scale markers and tableswith reliable dimensions for virus identification and compar-ison. This is a condition for further progress. Uranyl acetate,despite its shortcomings and problems, is irreplaceable sim-ply because it produces good contrast in negative stainingand stains viruses positively. At present, it can be replacedonly by other uranyl salts. Would more detailed studies ofnegatively stained preparations improve our knowledge ofviral impact on microbial ecosystems? This is certain, but onlyso if micrographs are of sufficient quality that environmentalviruses can be unambiguously linked to (or differentiatedfrom) viruses already known. This is far from being the case.

Fig. 8. Phage T4 with a NaCl crystal, UA, ×297,000. Fig. 9. False envelope in a phage of Rhizobium sp.; UA positive staining,×297,000. Haloes of this type are sometimes seen after UA-positive stain-ing. They expand with prolonged irradiation. Their nature is unclear.


190

Positive staining misses ssDNA viruses, ssRNA viruses, andempty virus capsids entirely and therefore does not give acomplete picture of virus populations.

Electron microscopy is time-consuming and still an art.Electron microscopes are as good as their operators, and noth-ing can replace the eyes and brains of a skilled scientist. Elec-tron microscopy depends critically on the knowledge and ded-ication of the investigator. It ensues that EM examinationscannot normally be entrusted to a technician or farmed out toanother institution. Unfortunately, electron microscopy is tak-ing a bad turn. In former times, using an old-style conven-tional EM, an electron microscopist could control all parts ofa column and adjust astigmatism, resolution, and magnifica-tion. The quality and contrast of pictures depended entirelyon him or her. This is now less (or not at all) possible withautomated and digitized electron microscopes. For any prob-lems of image sharpness and magnification with these instru-ments, one has to resort to company technicians and to paythem. This can be very expensive, especially if the techniciansare located in another city, and success is not guaranteed. Thequality of virological, especially bacteriophage, electronmicroscopy had already sharply declined since 1990 and ismuch worse than 50 years ago (Ackermann 2004). Presently,with the advent of CCD cameras, dependence on companytechnicians has become almost total.

This situation is unlikely to reverse itself because (1) digitalEMs are popular with inexperienced users, (b) conventionalphotographical supplies may be difficult to find since theintroduction of digital technology, and (c) manufacturers areoffering digital EMs only. These instruments may be twice asexpensive as conventional EMs and can no longer be adjustedand corrected by users. Their maintenance depends on costlyservice contracts and company technicians, so underfundedinstitutions may be tempted to skip maintenance. The finaland general consequence is that electron microscopes become

concentrated in a few institutions and are accessible for a feeonly and sporadically. We fear that environmental scientistswill be particularly affected by this trend and we foresee areduced activity in environmental electron microscopy.

AppendixCalculation of total virus contents—The water volume (V) from which particles are harvested

during centrifugation may be calculated as follows:(1) V = A * (R2 – r2)/2R

where A is the area in cm2 of the view fields observed on agrid; R and r are the maximum and minimum radius, respec-tively, in cm of the water in the centrifuge tube and are calcu-lated from the maximum radius of the rotor used (see themanufacturer’s data sheet), the thickness of the molded flatbottoms, and the height of the water column in the centrifugetube. The area of the view fields may be calculated as

(2) A = π * (rview field/mag)2

where π = 3.14, rview field is the radius of view field of the micro-scope in cm, i.e., the fluorescent screen, and mag is the mag-nification. The concentration of particles in the sample (unitis mL–1) is calculated by multiplying the average particlecounts per view field with V–1.

Expendable supplies and small equipment (North Americaonly)—

Ernest F. Fullam, 900 Albany Shaker Road, Latham, NY121110-1491, USA; (800) 833-4024, [email protected]

Ladd Research Industries, 83 Holly Ct., Williston, VT05495, USA; (800) 451-3406, [email protected]

Ted Pella, 4595 Mountain Lakes Blvd., Redding, CA 96003-1443, USA; (800) 237-3526, [email protected]

SPI Supplies, P.O. Box 656, 569 East Gay St., West Chester,PA 19381-0656, USA; (800) 242-4774, [email protected]

Soquelec, 5757 Cavendish Blvd. #101, Montreal, QC H4W2W8, Canada; (514) 482-6427, [email protected]

Table 2. Frequently asked questions.

Can I use grids without a carbon layer? No. The support film is unstable.

What is better, PT or UA? Neither. They are different.

What is negative staining for? Virus descriptions.

What is positive staining for? Virus counts on grids.

Which viruses can be positively stained? Those with high dsDNA contents.

What happens to the others? They are stained negatively or not at all.

Aspect of empty phage heads in PT? Membranes or black stain-filled vesicles.

Aspect of empty phage heads in UA? Membranes or shadows.

How to improve virus adhesion to grids? By glow discharge or wetting agents.

Why should samples be washed? To remove proteins, sugars, and salts.

Shall I examine unwashed lysates? Never.

Is fixation necessary? Generally not.

What is the average diameter of phage heads? 55 to 60 nm.

Do phage heads of 30 nm in diameter exist? Probably not.

When to suspect large algal viruses (Phycodnaviridae)? In icosahedral, nonenveloped, tailless particles of more than 120 nm in diameter.

How to correct faulty magnification in digital EMs? By calculation after photographing and measuring phage T4 tails.


191

References

Ackermann, H.-W. 2004. Declining electron microscopy. Lab.News (12):25. No volume.

———. 2007. 5550 Phages examined in the electron micro-scope. Arch. Virol. 152:277-243.

———. 2008. History of virology: Bacteriophages, p. 442-450.In B. Mahy and M.H.V. Van Regenmortel [eds.], Encyclope-dia of virology, 3rd ed. V. 2. Elsevier.

———, and M. S. DuBow. 1987. Description and identificationof new phages, p. 103-142. In Viruses of prokaryotes. CRCPress. V. I, General properties of bacteriophages.

———, and A. M. Kropinski. 2007. Curated list of prokaryoteviruses with fully sequenced genomes. Res. Microbiol.158:555-566.

Bergh, O., K. Y. Borsheim, G. Bratbak, and M. Heldal. 1989.High abundance of viruses found in aquatic environments.Nature 340:467–468.

Borsheim, K. Y., G. Bratbak, and M. Heldal. 1990. Enumerationand biomass estimation of planktonic bacteria and virusesby transmission electron microscopy. Appl. Environ. Micro-biol. 56:352-356.

Bouck, G. B. 1971. The structure, isolation, and compositionof the tubular mastigonemes of Ochromonas flagella. J. CellBiol. 50:362-384.

Bratbak, G., J. K. Egge, and M. Heldal. 1993. Viral mortality ofthe marine alga Emiliania huxleyi (Haptophyceae) and ter-mination of algal blooms. Mar. Ecol. Prog. Ser. 93:39-48.

———, and M. Heldal. 1993. Total counts of viruses in aquaticenvironments. In P. F. Kemp, B. F. Sherr, E. B. Sherr, and J. J.Cole [eds.], Handbook of methods in aquatic microbialecology. Lewis Publishers.

Brenner, S., and R. W. Horne. 1959. A negative stainingmethod for high resolution electron microscopy of viruses.Biochim. Biophys. Acta 34:103-110.

Clitheroe, S. B., and L.V. Evans. 1974. Viruslike particles in thebrown alga Ectocarpus. J. Ultrastruct. Res. 49:211-217.

Cochlan, W. P., J. Wikner, G. F. Steward, D. C. Smith, and F.Azam. 1993. Spatial distribution of viruses, bacteria andchlorophyll a in neritic, oceanic and estuarine environ-ments. Mar. Ecol. Progr. Ser. 92:77-87.

Ewert, D. L., and M. J. B. Paynter. 1980. Technique for deter-mining total bacterial virus counts in complex aqueous sys-tems. Appl. Environ. Microbiol. 39:253-260.

Fauquet, C. M., M. A. Mayo, J. Maniloff, U. Desselberger, andL. A. Ball [eds.]. 2005. Virus taxonomy: VIIIth report of theInternational Committee on Taxonomy of Viruses. ElsevierAcademic.

Gentile, M., and H. R. Gelderblom. 2005. Rapid viral diagno-sis: Role of electron microscopy. New Microbiol. 28:1-2.

Gregory, D. W., and B. J. S. Pirie. 1973. Wetting agents for bio-logical electron microscopy. I. General considerations andnegative staining. J. Microsc. 99:251-255.

Gundersen, K., G. Bratbak, and M. Heldal. 1996. Minimizing

the loss of bacteria in preserved seawater samples. Mar.Ecol. Progr. Ser. 137:305-310.

Hayat, M. A., and S. E. Miller. 1990. Negative staining.McGraw-Hill.

Hendrix, R. W. 2003. Bacteriophage genomics. Curr. Opin.Microbiol. 6:506-511.

Huxley, H. E., and Zubay, G. 1961. Preferential staining ofnucleic acid-containing structures for electron microscopy.J. Biophys. Biochem. Cytol. 11:273-296.

Kay, D. H. [ed.] 1965. Techniques for electron microscopy, 2nded. Blackwell Scientific. See especially pp. 58 and 328.

Kuo, J. [ed.] 2007. Electron microscopy: Methods and proto-cols, 2nd ed. Humana Press. See chaps. 1, 2, and 4–6. Meth-ods in molecular biology 369.

La Scola, B., and others. 2003. A giant virus in amoebae. Sci-ence 299:2033.

Luftig, R. B. 1967. An accurate measurement of the catalasecrystal period and its use as an internal marker for electronmicroscopy. J. Ultrastruct. Res. 20:91-102.

Montanié, H., H. J. Hartmann, C. Crottereau, and C. Trichet.2002. Virus-like particle analysis in a seston-rich coastalpond using transmission electron microscopy. Aquat.Microb. Ecol. 28:105-115.

Müller, G., G. Nielsen, and C. L. Baigent. 1980. Electronmicroscopical particle counting in virology. III. Parametersgoverning the efficiency of sedimentation techniques andits importance with the application of poly-L-lysine. Arch.Virol. 64:311-318.

Pfankuch, E., and G. A. Kausche. 1940. Isolierung und über-mikroskopische Abbildung eines Bakteriophagen. Natur-wissenschaften 28:46.

Rachel, R., M. Bettstetter, B. P. Hedlund, M. Häring, A. Kessler,K. O. Stetter, and D. Prangishvili. 2002. Remarkable diver-sity of viruses and virus-like particles in hot terrestrial envi-ronments. Arch. Virol. 147:2419-2429.

Ruska, H. 1940. Über die Sichtbarmachung der bakteriophagenLyse im Übermikroskop. Naturwissenschaften 28:45-46.

———. 1943. Versuch zu einer Ordnung der Virusarten. Arch.Ges. Virusforsch. 2:480-498.

Safferman, R. S., and M. E. Morris. 1964. Growth characteris-tics of the blue-green algal virus LPP-1. J. Bacteriol. 88:771-775.

Shively, J. M., F. S. Ball, and B. W. Kline. 1973. Electronmicroscopy of the carboxysomes (polyhedral bodies) ofThiobacillus neapolitanus. J. Bacteriol. 116:1405-1411.

Skotnicki, A., A. Gibbs, and N. G. Wrigley. 1976. Further stud-ies on Chara corallina virus. Virology 75:457-468.

Spencer, R. 1960. Indigenous marine bacteriophages. J. Bacte-riol. 79:614.

Suttle, C. A. 2005. Viruses in the sea. Nature 437:356-361.Tian, J., A. J. Sinskey, and J. Stubbe. 2005. Kinetic studies of poly-

butyrate granule formation in Wautersia eutropha H16 bytransmission electron microscopy J. Bacteriol. 187:3814-3824.

Tiekotter, K. L., and H.-W. Ackermann. 2009. High-quality

virus images obtained by TEM and CCD technology. J.Virol. Methods 159:87-92.

Torrella, F., and R. Y. Morita. 1979. Evidence by electron micro-graphs for a high incidence of bacteriophage particles in thewaters of Yaquina bay, Oregon: Ecological and taxonomicalimplications. Appl. Environ. Microbiol. 37:774-778.

Torsvik, T., and I. D. Dundas. 1974. Bacteriophage of Halobac-terium salinarium. Nature 248:680–681.

Van Etten, J., L. C. Lane, and R. H. Meints. 1991. Viruses andviruslike particles in eukaryotic algae. Microbiol. Rev.55:586-620.

Von Borries, B., E. Ruska, and H. Ruska. 1938. Bakterien andVirus in übermikroskopischer Aufnahme. Klin. Wochen-schr. 17:921-925.

Weinbauer, M. G. 2004. Ecology of prokaryotic viruses. FEMSMicrobiol. Rev. 28:127-181.

192


193

Ribonucleic acid (RNA) viruses that infect marine planktonhave been poorly studied compared with their deoxyribonu-cleic acid (DNA)-containing counterparts, but evidence oftheir importance has been accumulating. RNA-containingviruses have been isolated that infect marine protists, includ-ing a noxious dinoflagellate

(Tomaru et al. 2004), a common fungoid protist (Takao etal. 2005), a red tide–forming raphidophyte (Tai et al. 2003),and three types of diatom (Nagasaki et al. 2004; Shirai et al.2008; Tomaru et al. 2009). There is also at least one report ofa marine RNA-containing bacteriophage (Hidaka and Ichida1976). More detailed descriptions of these viral isolates can befound in reviews by Munn (2006), Brussaard and Martinez(2008), Nagasaki (2008) and Lang et al. (2009).

The isolation and cultivation of a virus from the environ-ment is essential if the virus is to be fully characterized. Nev-

ertheless, cultivation is an unrealistic approach to describingthe full range of diversity in natural marine viral assemblages.Cultivation-independent methods have thus been adopted,which include categorization based on morphology (Frankand Moebus 1987), genome size distributions (Steward et al.2000), the phylogeny or finger print of a specific viral gene(Chen et al. 1996; Fuller et al. 1998; Short and Suttle 1999;Filée et al. 2005; Larsen et al. 2008), or construction of metage-nomic libraries (Breitbart et al. 2002, 2004; Angly et al. 2006;Bench et al. 2007; Helton and Wommack 2009).

Characterizing RNA virus diversity based on morphology isthe least informative of the above approaches because it islaborious and provides little resolution among viruses. To thebest of our knowledge, most RNA viruses are relatively smalland untailed with little morphological detail to distinguishthem from small, untailed DNA-containing viruses. Moreover,there appears to be no relationship between the morphology ofan RNA virus and the type of host that it is able to infect. Char-acterizing the distribution of genomes sizes within a sample isa more tractable approach to determining RNA virus diversity,because these results are relatively quick and easy to obtain andcan provide quantitative data on diversity. Our preliminaryresults suggest that RNA sizing by denaturing agarose gels elec-trophoresis can be used to generate “fingerprints” of RNA virusgenotypes from marine assemblages. However, this method islimited because relatively large amounts of viral RNA arerequired, and only significant differences in genome size arediscriminated and not differences in sequence.

The superior sensitivity and resolution achievable bysequence analysis has made sequence-based methods the mostwidely used approach to categorizing microbial diversity, andmolecular surveys have made it clear that much of the diver-sity among RNA viruses has yet to be discovered. Many of thenovel gene sequences recovered so far appear to derive fromviruses of marine protists. Their high diversity suggests that

Characterization of the diversity of marine RNA virusesAlexander I. Culley1*, Curtis A. Suttle2, and Grieg F. Steward1

1Department of Oceanography, University of Hawai’i at Manoa, Honolulu, HI 968222University of British Columbia, Department of Earth and Ocean Sciences, Botany, and Microbiology & Immunology, 1461Biosciences-6270 University Blvd., Vancouver, BC, V6T 1Z4

AbstractThe diversity of ribonucleic acid (RNA) viruses in the ocean and the ongoing isolation and characterization

of RNA viruses that infect important primary producers suggests that RNA viruses are active members of themarine microbial assemblage. At this point, little is known about the composition, dynamics, and ecology ofthe RNA virioplankton. In this chapter, we describe two methods to assess RNA virus diversity from seawater.

*Corresponding author: E-mail: [email protected], phone: (808) 956-8629



The writing of this chapter was supported in part by grants fromNSF to the authors (OCE-04-42664 and OCE-0826650) and to theCenter for Microbial Ecology Research and Education (EF-0424599). Theauthors acknowledge the efforts of two anonymous peer reviewers andtheir suggestions to improve the manuscript.

ISBN 978-0-9845591-0-7, DOI 10.4319/mave.2010.978-0-9845591-0-7.193Suggested citation format: Culley, A. I., C. A. Suttle, and G. F. Steward. 2010. Characterization ofthe diversity of marine RNA viruses, p. 193–201. In S. W. Wilhelm, M. G. Weinbauer, and C. A.Suttle [eds.], Manual of Aquatic Viral Ecology. ASLO.


MANUALof


viral infections are a persistent force-shaping protistan com-munity composition in the sea.

Because this is a relatively new area of research, the numberof methods described in the literature for investigating thediversity of marine RNA viruses is still limited. We presenthere two protocols that have been used successfully in pub-lished articles (Culley et al. 2006; Culley and Steward 2007;Djikeng et al. 2008, 2009), but note that this is a nascent fieldand there are no routine methods. We hope the protocolsdescribed will at least serve as starting point for furtherimprovements.

Protocol 1. A degenerate primer reverse transcription-polymerase chain reaction–based protocol to deter-mine the diversity of picorna-like viruses

This protocol is designed to detect picorna-like viruses frommarine samples. It is based on a strategy first reported by Cul-ley et al. (2003) and refined by Culley and Steward (2007). Allpicorna-like viruses have single-stranded positive-sense RNAgenomes and are classified in the order Picornavirales (Le Gallet al. 2008). Picorna-like viruses are responsible for several sig-nificant human and animal diseases and infect a diversity ofmarine protists including a diatom (Nagasaki 2008; Lang et al.2009). The procedure outlined below uses direct filtration ofrelatively small volumes of water and reverse transcription-polymerase chain reaction (RT-PCR) amplification of an RNA-dependent RNA polymerase (RdRP) gene fragment usingdegenerate primer sets.

With this method, viral polymerase sequences were ampli-fied in several distinct aquatic environments, including anestuarine urban canal (Ala Wai canal, Waikiki, HI, USA), a trop-ical bay (Kaneohe Bay, HI, USA), and a temperate bay (Mon-terey Bay, CA, USA). Amplification occurred in samples fromthe same site in different seasons and at different depths show-ing that RNA viruses are widespread and consistently present.Sequencing of the amplified gene fragments revealed novelsequences that are highly divergent from any known isolates.

Material and equipmentEquipment—Peristaltic pump and pump heads, thermocy-

cler, heating block, incubator, gel electrophoresis unit, elec-troporator, gel documentation system

Supplies—Sterile, 1.7 mL nucleic acid-free microcentrifugetubes; 0.2 mL sterile, nucleic acid-free PCR tubes; 0.02 µm alu-minum oxide filters (Anotop, Whatman); 10 mL sterile syringes;peristaltic pump tubing; sterile razor blades; MinElute GelExtraction kit (Qiagen); MinElute PCR Purification kit (Qiagen);Masterpure Complete DNA and RNA Purification kit (EpicenterBiotechnologies); PCRTerminator End Repair kit (Lucigen);CloneSmart HCKan Blunt Cloning kit with Ecloni Supreme cells(Lucigen); Turbo DNA-free Kit (Applied Biosystems)

Solutions, reagents, and media—0.02-filtered, sterile, nucleicacid-free TE buffer (10 mM Tris, 1 mM EDTA; pH 8); 0.1 Mdithiothreitol (DTT); Superscript III Reverse Transcriptase and

buffers (Invitrogen); RNase Out (Invitrogen); RNase H (Invitro-gen); Platinum Taq DNA polymerase and buffers (Invitrogen)

Primers in concentrations discussed as follows: 0.5 × TBE(45 mM Tris-borate, 1 mM EDTA; pH 8.0) electrophoresisbuffer; nucleic acid–free, sterile water; 10 mM dNTP mix

StepsCollection—Whole seawater (if intracellular viruses are to be

included in the analysis) or 0.22-filtered seawater (if only thefree virus community is targeted) is gently filtered (7 mmHg)through a 0.02 µm aluminum oxide filter (Anotop, What-man). Filtration should continue until the rate slows dramati-cally or drops to zero (note that it is important to measure thetotal volume of sample that passes through the filter). Usepumped air to remove as much residual sample as possible.Once dry, seal the filter inlets and outlets with parafilm, labelthe filter, and then flash-freeze the filter in liquid nitrogen andstore it at –80°C until extraction.

Extraction—Total nucleic acids are extracted from the alu-minum oxide filters using a Masterpure complete DNA andRNA Purification kit (Epicenter, Biotechnologies) with a pro-tocol slightly modified from the manufacturer’s instructions.The detailed protocol of this method is presented elsewhere inthis book (Steward and Culley 2010, this volume).

DNase treatment—To reduce the likelihood of misprimingand increase the sensitivity of the subsequent cDNA synthesisand PCR reactions, we remove contaminating DNA from ourRNA preparations with the Turbo DNA-free kit (AppliedBiosystems) as described in the protocol provided with the kit.Although the manufacturers suggest that treated RNA tem-plate should not exceed 40% of the total cDNA synthesis reac-tion, using the reaction discussed below, we have used 55%template without inhibition.

cDNA synthesis—We have synthesized complementary DNA(cDNA) using reverse transcriptase (Superscript III, Invitrogen)in reactions primed with random hexamers (N6) or the spe-cific primers RdRp2, Mpl.sc1R, Mpl.sc2R, Mpl.sc3R, orMpl.cdhR (details for these primers are listed in Table 1), andhave found no detectable difference in sensitivity between thetwo approaches. Priming with random hexamers is preferablebecause the resultant cDNA can be used in a PCR reaction withany of the five primer sets listed in Table 1, eliminating thenecessity for a different cDNA reaction for each primer set.

Each cDNA reaction contains 11 µL of extracted TurboDNA-free–treated RNA template, 0.2 mM of each dNTP, and100 ng of N6 primer in a total volume of 13 µL. Denaturationand annealing of the primers to the RNA template occurs byheating the sample to 65°C for 5 min and then cooling it onice. While still on ice, DTT (0.5 mM final conc.) is added to thereaction as an enzyme stabilization reagent with 40 U RNaseOUT (Invitrogen) to protect the sample from RNAse activity.The complementary DNA strand is synthesized with 200 USuperscript III (Invitrogen) Reverse Transcriptase. The finalreaction volume should be 20 µL.

Culley et al. Characterization of marine RNA viruses

194

The reaction is incubated initially at 25°C for 5 min so thatthe relatively unstable hexamer primers remain annealed tothe template while cDNA synthesis commences. The tempera-ture is then increased to 50°C, the temperature at whichSuperscript III’s processivity is highest, for 60 min. The activ-ity of the enzyme is terminated by incubating the reaction at85°C for 5 min. After cDNA synthesis, add 1 µL (2 U) of RNaseH (Invitrogen) to the reaction and incubate at 37°C for 20 minto digest the RNA template from the cDNA:RNA molecule.

PCR—PCR can be performed with the RdRp, Mpl.sc1,Mpl.sc2, Mplsc3, and Mpl.cdh primer sets listed in Table 1 andthe cDNA synthesized in the previous step. In a 0.2 mL nucle-ase-free PCR tube, add reaction components to achieve finalconcentrations of 1× Platinum Taq buffer, 3 mM MgCl2, 0.2mM of each dNTP, 1 µM of each primer, and 1 unit of PlatinumTaq DNA polymerase (Invitrogen). Incubate the reactions withthe following thermal cycling conditions: 94°C for 75 s (this isnecessary to activate the enzyme and ensures complete initialdenaturation of template), followed by 35 cycles of denatura-tion at 94°C for 45 s, annealing at a primer-specific tempera-ture (Table 1) for 45 s, and extension at 72°C for 45 s (note thatthe target product size is approximately 500 bp). Complete thecycle with a single final extension step of 9 min 15 s at 72°C.

Before gel separation, we purify and concentrate the PCRreactions with a MinElute PCR cleanup column (Qiagen) asdescribed by the manufacturer. A smaller volume of productallows one to load the product on a thinner gel, ultimatelyresulting in a more efficient recovery of target DNA from theexcised band. Purified PCR products are loaded onto a 1%agarose gel containing 1× SYBR safe stain (Invitrogen) and0.5× TBE buffer. Bands of DNA of the appropriate size (approx-imately 500 bp) are excised and purified with a MinElute GelExtraction kit (Qiagen) according to the manufacturer’sinstructions. In the final step of this procedure, we recom-

mend eluting DNA from the column with three washes of 10µL nuclease-free water in preparation for the end repair reac-tion described below.

Cloning and sequencing—Several commercial kits are avail-able for the efficient cloning of PCR products. We use theCloneSmart Blunt Cloning kit from Lucigen. Some advantagesof this cloning kit are the high efficiency of recombinants thateliminate the need for screening (e.g., XGAL/IPTG) and theprevention of transcription of inserts in the vector reducescloning bias. However, the pSMART vector requires blunt endswith 5′ phosphate groups (unlike TOPO-TA [Invitrogen]cloning kits for example). The additional step adds time,increases cost, and increases risk of sample loss, but we havefound that it works well.

In preparation for ligation, PCR products are polished andphosphorylated with the PCRTerminator End Repair kit (Luci-gen) as described by the manufacturer. After a 15-min incuba-tion at room temperature the reaction is purified and concen-trated with a MinElute Reaction Cleanup column (Qiagen).The eluted PCR products are subsequently ligated into thepSMART-HCKan vector (Lucigen) according to the manufac-turer. In this reaction, we typically use the maximum amountof template (6.5 µL) and the maximum incubation time (2 h).After terminating the ligation reaction with by incubating for15 min at 70°C, 2 µL of the ligation reaction is transformedinto Ecloni 10G Supreme cells (Lucigen) via electroporation.We do not recommend using greater than 2 µL of the ligationmixture in the transformation reaction because the chances ofarcing are greatly increased during electroporation. Before ini-tiating a large-scale sequencing effort, we recommend screen-ing 10 to 20 colonies for inserts by PCR amplification with theprimers SL1 and SR2 (Table 1) to assess the quality of thelibrary. Colony PCR products can be visualized on a gel andthe products in the correct size range purified and sequenced.


195

Table 1. RT and PCR primer details.

Name Sequence (5′-3′) °C annealing ~Product (bp)

Mpl.sc1F TIGCIGGWGAYTWYARM 50 500

Mpl.sc1R YTCCTTWTCRGSCATKGTA

Mpl.sc2F ITWGCIGGIGATTWCA 43 500

Mpl.sc2R CKYTTCARRAAWTCAGCATC

Mpl.sc3F TIATIGMKGGIGAYTA 49 500

Mpl.sc3R TTMARGAAIKMAGCATCTT

Mpl.cdhF GMIGGTGAYTAYAGCGCTTWYGAY 44 500

Mpl.cdhR ATACCCAATGCCTYTTIARRAA

RdRp1 GGRGAYTACASCIRWTTTGAT 50 450

RdRp2 MACCCAACKMCKCTTSARRAA

SL1 CAGTCCAGTTACGCTGGAGTC 50 NA

SR2 GGTCAGGTATGATTTAAATGGTCAGT

FR26RV-N GCCGGAGCTCTGCAGATATCNNNNNN NA NA

FR40RV-T GCCGGAGCTCTGCAGATATC(T)20 NA NA

FR20RV GCCGGAGCTCTGCAGATATC 65 NA

NA, not applicable.

Assessment

The method described above is based on reverse transcrip-tion (RT) of RNA into cDNA and PCR amplification of mixedtemplates. PCR in combination with cloning, sequencing andbioinformatic analysis, has resulted in the identification andcultivation-independent classification of thousands ofmicrobes (Rappé and Giovannoni 2003). However, bias associ-ated with every step of a PCR-based assay, including samplecollection, nucleic acid extraction, PCR amplification, andsequence analysis of PCR amplicons, can contribute to aninaccurate portrayal of the community under examination(von Wintzingerode et al. 1997).

Reverse transcriptase synthesizes DNA from an RNA tem-plate. The enzyme lacks 3′ to 5′ proofreading exonucleaseactivity, contributing to its relatively high error rate in com-parison with other DNA polymerases (Roberts et al. 1988). Forexample, Superscript II reverse transcriptase (Invitrogen) isapproximately 13 times more error prone than Platinum Taqpolymerase (Invitrogen) (Roberts et al. 1988). Factors such asthe concentration of target RNA, the amount of template sec-ondary structure and priming conditions including annealingtemperature can significantly affect the precision, efficiencyand production of the RT reaction (Stahlberg et al. 2004).

Amplification of environmental targets with PCR can resultin differential amplification, the formation of chimeras andheteroduplexes, and artifacts from DNA polymerase error,among other biases (von Wintzingerode et al. 1997; Kanagwa2003). Polz and Cavanaugh (1998) found that PCR with degen-erate primers did not maintain the original ratio of templateafter 25 cycles and that templates with GC-rich priming siteswere preferentially amplified. Moreover, Suzuki and Giovan-noni (1996) observed that in reactions with greater than 35cycles, a 1:1 ratio of products occurred regardless of the initialratio of target sequences. PCR can also produce artifacts.Chimeras, molecules formed from parts of two differentsequences, can comprise approximately 10% of PCR amplifiedproducts and appear to increase in frequency with cycle num-ber (von Wintzingerode et al. 1997). The DNA polymerase inPCR can mis-incorporate bases during amplification resultingin sequencing artifacts. Eckert and Kunkel (1990) calculatedTaq error rates as high as 3 × 10–3, and Acinas et al. (2005) iden-tified Taq DNA polymerase error as the primary cause of arti-facts during the construction of rRNA clone libraries.

The cloning of amplified products can be another signifi-cant source of bias. Factors that may lead to cloning biasinclude the expression of deleterious genes (a significant con-cern when cloning phage genes), a decrease in cloning effi-ciency with increasing insert size, the formation of heterodu-plexes, and inappropriate antibiotic resistance (Kanagwa2003). For example, Rainey et al. (1994) observed significantdifferences in community composition between clonelibraries constructed with different cloning systems from thesame sample. Although PCR-based methods can be effective in

characterizing the richness of a community as well as theidentity and phylogeny of its members, estimates of evennessare dubious, because of the biases discussed above.

The advantages of the method described above are that it isrelatively simple and inexpensive (compared with a metage-nomic approach for example). Using this approach, novelmarine RNA viruses are being discovered and characterized atan increasing rate (Lang et al. 2009; Nagasaki 2008). Theprimers used in the assay can be redesigned to incorporate themost current available sequence information. Moreover, thisassay is not limited to picorna-viruses, and can be targetedtoward any group of RNA viruses that have an appropriatemolecular marker (e.g., RdRP, helicase, or capsid genes).

Protocol 2. Construction of shotgun libraries fromRNA virus assemblages

The following protocol is used to construct shotgunlibraries from RNA virus assemblages. This method is designedto detect all RNA viruses, regardless of their genome orienta-tion, and therefore provides a broader assessment of RNA virusdiversity compared to the single-gene–based methoddescribed in Protocol 1.

Metagenomics is loosely defined as the analysis of genomesequences from a community of organisms. In the field ofmarine virology, metagenomic approaches have been pre-dominantly used to characterize natural assemblages ofmarine DNA phages (see Edwards and Rohwer 2005 for areview). These studies have generated an immense quantity ofsequence data that has resulted in the discovery of a signifi-cant number of novel viral genes and has revealed the extraor-dinary diversity of DNA viruses in the ocean.

The diversity of RNA virus assemblages in human feces(Zhang et al. 2006b; Victoria et al. 2009), reclaimed water(Rosario et al. 2009), and a freshwater lake (Djikeng et al.2009) have been characterized with metagenomic methods.These studies detected unexpected viral genomes from eachrespective sampling environment as well as sundry novel RNAvirus genes. The only metagenomic analysis of marine RNAvirus assemblages to date characterized two marine RNA viruscommunities from coastal British Columbia (Culley et al.2006). In this study, most of the sequences were unrelated toknown sequences. For the sequences that were related toknown sequences, positive-sense ssRNA viruses that are dis-tant relatives of known RNA viruses dominated the samples.One RNA virus library was characterized by diverse picorna-like viruses, while the second library was dominated by virusesrelated to members of the family Tombusviridae and genusUmbravirus (Culley et al. 2006). In the latter library, 59% of thesequence fragments that formed overlapping contiguous sec-tions fell into one segment. Similarly, 66% of JP sequence frag-ments contributed to only four contigs (Culley et al. 2006).These sequences presumably represent the most abundantviruses in the two marine RNA virus communities sampled.This first “snap shot” suggests that marine RNA viruses have


196

primarily eukaryotic hosts and are orders of magnitude lessdiverse than DNA viruses.

There are two primary challenges to constructing shotgunlibraries of marine RNA virus communities: producing a puri-fied fraction of viral RNA from seawater and the unbiased pro-duction of DNA appropriate for sequencing from a relativelysmall amount of RNA template. We have constructed librarieswith a low percentage of non-viral sequences from concentratedRNA virus assemblages that have simply been filtered and enzy-matically treated (Culley et al. 2006). However, we have subse-quently found that filtration-only is ineffective in removingcontaminating nucleic acid from some samples. To consistentlyproduce the necessary purity of viral RNA for library construc-tion, we recommend that filtration is followed by two rounds ofdensity gradient purification of the viruses (Lawrence and Stew-ard 2010, this volume, Protocol B), followed by DNase treat-ment in order to remove contaminating DNA. There are severalmeans of producing microgram amounts of double-strandedDNA from sub-nanogram amounts of RNA beyond the methoddescribed here. Examples include 1) linear amplification of RNA(Frias-Lopez et al. 2008) and ds cDNA synthesis, 2) RNA-tailing(Botero et al. 2005) and PCR, 3) cDNA tailing and PCR (Iscoveet al. 2002), 4) ds cDNA synthesis, primer adapter addition andPCR (Culley et al. 2006; Zhang et al. 2006b), and 5) random dscDNA synthesis and multiple displacement amplification(Berthet et al. 2008; Rosario et al. 2009). These alternatives mayprove to be less biased and more sensitive and merit furtherinvestigation. However, to the best of our knowledge, the accu-racy, reproducibility, and limitations of these methods have notbeen evaluated in a systematic manner and/or the protocol hasnot been used with RNA viruses.

The method described below, based on the work of Reyesand Kim (1991), Froussard (1992), and Allander et al. (2005), isreferred to as random priming-mediated sequence-independ-ent single-primer amplification or RP-SISPA, and was devel-oped by Djikeng et al. (2008) to generate whole genome shot-gun libraries of virus communities. Victoria et al. (2008)described a similar approach to characterize virus diversityfrom tissue samples. In RP-SISPA, cDNA is synthesized from aDNA-free viral RNA template with a primer mixture that con-tains two types of primers. The first primer type, FR26RV-N, hasa 5′, 20-bp primer sequence (FR20RV), and a 3′ random hexa-mer (N6) segment. The second primer type (FR40RV-T) has thesame 5′ 20 bp primer (FR20RV) sequence coupled with a 3′ polyT tail (Figure 1A). Second strand synthesis is primed by rean-nealing the cDNA primers (FR26RV-N) remaining from the RTreaction to the newly synthesized cDNA strand and catalyzedby Klenow exo – DNA polymerase (Figure 1B). Single-strandPCR is performed with the double stranded cDNA template anda single 20 bp primer (FR20RV) that is the same sequence as the5′ cDNA primers (Figure 1B). PCR products are visualized on agel and DNA of the target size can be excised and either clonedor processed directly for pyrosequencing. An evaluation of theaccuracy of RP-SISPA by Djikeng et al. (2008) demonstrated

that the method consistently produced libraries with > 90%genome coverage and 15-fold depth of coverage of positive andnegative sense RNA virus genomes. Subsequently, RP-SISPA wasused to characterize seasonal differences in the diversity ofRNA viruses in a freshwater lake (Djikeng et al. 2009). Just as inthe marine environment, a majority of sequences did not showsignificant similarity to known sequences. The sequences thatwere characterized suggest that the lake RNA virus assemblageincludes pathogens of protists, plants, insects, fish, andhumans (Djikeng et al. 2009).


197

Figure 1. RP-SISPA Schematic. This figure is based on Djikeng et al.(2008). Figure 1A shows the three primers used in RP-SISPA. FR26RV-N iscomposed of the 20 bp 5′ primer sequence, FR20RV (black bar), and arandom hexamer at the 3′ end (light blue bar). Primer FR40RV-T is a com-posite of the FR20RV primer (black bar) and a 20 bp poly T tail (red bar).FR20RV (black bar) is used in PCR. The reverse complement of FR26RV-N(green and orange bar) and FR40RV-T (green and yellow bar) are shown.In Figure 1B, cDNA is synthesized from viral RNA (gray segmented line)with reverse transcriptase primed with FR26RV-N and FR40RV-T. To syn-thesize the second strand, FR26RV-N is annealed to the newly synthesizedcDNA strand where it primes Klenow exo – DNA polymerase “gap filling”activity, resulting in an FR20RV site on both 5′ and 3′ ends. The DNA tem-plates, which represent random stretches of the initial viral RNA genomes,are then amplified with the FR20RV primer to generate more material forcloning and sequencing.

Material and equipment

Equipment—Thermocycler, heating block, incubator, gelelectrophoresis unit, gel documentation system

Supplies—1.7 mL sterile, nucleic acid–free microcentrifugetubes; 0.2 mL sterile, nucleic acid-free PCR tubes; sterile razorblades; QIAamp Viral RNA Mini kit (Qiagen); MinElute GelExtraction kit (Qiagen); MinElute PCR Purification kit (Qia-gen); MinElute Reaction Cleanup kit (Qiagen); PCRTerminatorEnd Repair kit (Lucigen); CloneSmart HCKan Blunt Cloningkit with Ecloni Supreme cells (Lucigen); Turbo DNA-free Kit(Applied Biosystems); electroporation cuvettes

Solutions, reagents, and media—Superscript III Reverse Tran-scriptase and buffers (Invitrogen), 10 µM dNTP mix, RNaseOut (Invitrogen), Klenow Fragment, 3′-5′ exo – (New EnglandBiolabs); primers (Table 1): 0.5 × TBE (45 mM Tris-borate, 1mM EDTA [pH 8.0]) electrophoresis buffer; nucleic acid-free,sterile water; Ampligold Taq Polymerase and buffers (AppliedBiosystems); 0.02-filtered SM Buffer (100 mM NaCl, 8 mMMgSO4, 50 mM Tris pH 7.5)

StepsSample collection, purification, and extraction—The ideal start-

ing amount for this method is on the order of nanograms orgreater of purified viral RNA, however the sensitivity of RT-PCR suggests that sub-nanogram starting template will be suc-cessful. Because of how little RNA is present in a viral genome,the virus community in tens to thousands of liters of seawatermust be isolated and concentrated before one can proceedwith shotgun library construction. Other chapters in this book(Wommack et al. 2010, this volume; Steward and Culley 2010,this volume) discuss approaches to concentrating viral com-munities from seawater.

RP-SISPA will not discriminate between viral and non-viralnucleic acids, and it is therefore critical that only pure viralRNA is present as template. We have found that purificationthrough two sequential cesium chloride gradients (Protocol Bin Lawrence and Steward 2010, this volume) is effective forremoving contaminating cellular and exogenous nucleic acidsprior to library construction.

Once the cesium chloride fractions have been collectedand concentrated with a centrifugal ultrafiltration unit asdescribed, viruses are recovered by eluting in 3 × 50 µL of0.02-filtered SM buffer. Each density fraction is thenextracted with the QIAamp Viral RNA Mini kit (Qiagen) asdirected by the manufacturer with the following exception.We do not add carrier RNA as suggested to the “AL” lysisbuffer to avoid introducing non-target RNA to the sample.Note that there is no RNase treatment before extractionbecause exogenous RNA will pellet and is thus removed dur-ing the CsCl fractionation procedure. We recommend quan-tifying the RNA from each density fraction to identify inwhich fraction the concentration of viral RNA is highest. Thetemplate for RP-SISPA can be RNA extracted from mutiple

pooled fractions or from a single fraction depending on yourresearch objectives.

Enzymatic treatment—As in Protocol 1, we remove contami-nating DNA from our RNA preparations with the Turbo DNA-free kit (Applied Biosystems) as described in the protocol pro-vided with the kit.

RP-SISPAcDNA synthesis: In preparation for cDNA synthesis, 10 µL

purified RNA viral template is mixed with a final dNTP con-centration of 0.2 mM and 1 µM and 5 nM final concentra-tions of FR26RV-N and FR40RV-T primer, respectively (seeTable 1 for the sequence of each primer). FR40RV-T is addedto take advantage of the fact that a majority of characterizedRNA virus genomes have poly(A) tails. The addition of apoly(T) primer may increase the likelihood of the 3′ endsbeing sequenced (Figure 1B). The reaction is heated to 65°Cthen cooled on ice to allow the primers to anneal. While stillon ice, DTT (0.5 mM final conc.) is added to the reaction asan enzyme stabilization reagent with 40 U RNase OUT (Invit-rogen) to protect the sample form RNAse activity. The com-plementary DNA strand is synthesized with 200 U of Super-script III (Invitrogen) reverse transcriptase. The final reactionvolume should be 20 µL. The reaction is incubated initially at25°C for 10 min so that the hexamer 3′ end of primerFR26RV-N and the poly(T)20 3′ end of primer FR40RV-Tremain annealed to the template while cDNA synthesis com-mences. The temperature is then increased to 50°C, the tem-perature at which Superscript III’s processivity is highest, for60 min.

Second strand synthesis: In a simple and elegant step, thesecond strand synthesis reaction results in a ds cDNA templatewith a primer site added to both the 5′ and 3′ end (Figure 1B).After the hour-long incubation at 50°C, the first strand syn-thesis reaction is heated immediately to 94°C for 3 min andthen rapidly cooled on ice. This step results in the reannealingof excess FR26RV-N primer to the nascent cDNA strand. Acomplementary second strand is subsequently synthesized at37°C for 60 min with the addition of 2.5 U of Klenow Frag-ment, 3′-5′ exo – (New England Biolabs). The Klenow reactionis terminated with a final incubation at 75°C for 10 min.

PCR: PCR is used to produce a sufficient quantity of DNAfrom the ds cDNA template from the second strand reactionfor sequencing (Figure 1B). One PCR reaction contains 5 µL oftemplate taken directly from the second strand synthesis reac-tion, 40 pM of FR20RV primer (see Table 1), a final dNTP con-centration of 0.2 mM, 1 × Gold buffer, 2.5 mM MgCl2, and 2.5U of Ampligold DNA polymerase (Applied Biosystems) in afinal volume of 50 µL. The reaction is incubated at 94°C for 10min to fully denature the template and activate the hot startenzyme (we have found that a hot start is absolutely essential),followed by 35 cycles of denaturation at 94°C for 1 min,annealing at 65°C for 1 min, and extension at 72°C for 2 minand a final extension for 13 min that permits the completionof complementary strand synthesis.


198

Before gel separation, we purify and concentrate the PCRreactions with a MinElute PCR cleanup column (Qiagen) asdescribed by the manufacturer. This reduces the thickness ofthe gel, ultimately resulting in a more efficient recovery of tar-get DNA from the excised band. Purified PCR products areloaded onto a 1% agarose gel containing 1 × SYBR safe stain(Invitrogen) and 0.5 × TBE buffer. Bands of DNA of the appro-priate size range (the size range excised should be based onwhat type of sequencing method is being used; for example,we target the 800-2000 bp size range for Sanger sequencing)are excised and purified with a MinElute Gel Extraction kit(Qiagen) according to the manufacturer’s instructions. Whenvisualizing the gel, prolonged exposure to UV irradiation candamage the DNA and greatly reduce downstream cloning effi-ciency. An illuminator with blue light-emitting bulbs is idealfor gel visualization. If this is not available, take steps toreduce the exposure of the gel to ultraviolet irradiation duringexcision as much as possible. To mitigate the biases associatedwith high cycle number PCR (see the “Assessment” section ofProtocol 1), we recommend pooling the products from multi-ple PCR reactions using the same ds cDNA template.

If the sample is to be cloned for Sanger sequencing, we rec-ommend eluting DNA from the column with three washes of10 µL nuclease-free water in preparation for the PCRTermina-tor (Lucigen) end repair reaction. For the cloning protocol,please refer to the “Cloning and sequencing” section of Proto-col 1 in this chapter. The procedure from this point forward isthe same. The material purified from the gel may instead beprocessed for pyrosequencing.

AssessmentThe primary strengths of the RP-SISPA are that it requires

no previous knowledge of the composition of the target RNAvirus assemblage, it has a low detection limit (nanograms orless) and that it is a relatively straightforward, inexpensiveprocedure that uses equipment standard to any molecular lab-oratory. A systematic comparison of metagenomic approachesto characterizing marine RNA virus diversity is not available.However, several different methods have been developed toaccurately interrogate the transcriptome of single cells (Peanoet al. 2006). These methods share several of the same chal-lenges as the examination of natural communities of RNAviruses, starting with very low concentrations of starting RNAtemplate. Surprisingly, those methods that include an expo-nential amplification step after the addition of primer sites tothe target template (such as RP-SISPA) generally introducedless bias than other approaches such as linear RNA amplifica-tion (Iscove et al. 2002; Subkhankulova and Livesey 2006).Presumably those protocols that rely on multiple displace-ment amplification during RNA virus shotgun library con-struction are subject to the same high incidence of chimeraformation as found in previous studies (Zhang et al. 2006a;Lasken and Stockwell 2007), however this has yet to bedemonstrated.

The limitations of the RP-SISPA method included the gen-eral biases associated with RT-PCR discussed in the assessmentsection of Protocol 1. RP-SISPA appears to achieve full cover-age of RNA viral genomes less than or equal to 15 kb in size,but only recovered an average of 50% of a 50 kb ds DNA bac-teriophage genome. Furthermore sequences from the 3′ end ofall RNA viral genomes tested were consistently underrepre-sented, as expected with a method based on randomly primedRT. Gaps in the sequence occurred and may have been due toregions of secondary structure, extreme codon bias, or shear-ing of the RNA template (Djikeng et al. 2009). Finally, therequirement of having viral RNA free from contamination canbe a significant challenge with some samples.

ConclusionsStudy of the ecology of marine RNA viruses is still in its

infancy and thus the number of methods available are limited.In this chapter, we have presented two molecular approachesto determining RNA viral diversity from seawater samples. Thefirst is a relatively rapid and inexpensive approach based onthe amplification of a single-gene, wheras the second method,based on the sequencing of whole genome shotgun libraries,requires more investment but produces a more comprehensivepicture of the RNA virus community. The choice of method isdependent on the research objective and the resources avail-able. We hope that this chapter will kindle interest in marineRNA viruses and serve as an informative starting point for fur-ther research and development on this understudied compo-nent of the plankton.

ReferencesAcinas, S. G., R. Sarma-Rupavtarm, V. Klepac-Ceraj, and M. F.

Polz. 2005. PCR-induced sequence artifacts and bias:Insights from comparison of two 16S rRNA clone librariesconstructed from the same sample. Appl. Environ. Micro-biol. 71:8966-8969.

Allander, T., M. T. Tammi, M. Eriksson, A. Bjerkner, A.Tiveljung-Lindell, and B. Andersson. 2005. Cloning of ahuman parvovirus by molecular screening of respiratorytract samples. Proc. Natl. Acad. Sci. U.S.A. 102:12891-12896.

Angly, F. E., and others. 2006. The marine viromes of fouroceanic regions. PLoS Biol. 4:2121-2131.

Bench, S. R., and others. 2007. Metagenomic characterizationof Chesapeake bay virioplankton. Appl. Environ. Microbiol.73:7629-7641.

Berthet, N., and others. 2008. Phi29 polymerase based randomamplification of viral RNA as an alternative to random RT-PCR. BMC Mol. Biol. 9:77.

Botero, L. M., S. D’imperio, M. Burr, T. R. Mcdermott, M.Young, and D. J. Hassett. 2005. Poly(A) polymerase modifi-cation and reverse transcriptase PCR amplification of envi-ronmental RNA. Appl. Environ. Microbiol. 71:1267-1275.

Breitbart, M., and others. 2004. Diversity and population


199

structure of a near-shore marine sediment viral community.Proc. Royal Soc. Biol. Sci. B 271:565-574.

———, and others. 2002. Genomic analysis of unculturedmarine viral communities. Proc. Natl. Acad. Sci. U.S.A.99:14250-14255.

Brussaard, C. P. D., and J. M. Martinez. 2008. Algal bloomviruses. Plant Viruses 2:1-13.

Chen, F., C. A. Suttle, and S. M. Short. 1996. Genetic diversityin marine algal virus communities as revealed by sequenceanalysis of DNA polymerase genes. Appl. Environ. Micro-biol. 62:2869-2874.

Culley, A. I., A. S. Lang, and C. A. Suttle. 2003. High diversityof unknown picorna-like viruses in the sea. Nature424:1054-1057.

———, ———, and ———. 2006. Metagenomic analysis ofcoastal RNA virus communities. Science 312:1795-1798.

———, and G. F. Steward. 2007. New genera of RNA viruses insubtropical seawater, inferred from polymerase genesequences. Appl. Environ. Microbiol. 73:5937-5944.

Djikeng, A., and others. 2008. Viral genome sequencing byrandom priming methods. BMC Genomics 9:5.

———, R. Kuzmickas, N. G. Anderson, and D. J. Spiro. 2009.Metagenomic analysis of RNA viruses in a fresh water lake.PLoS ONE 4:e7264.

Eckert, K. A., and T. A. Kunkel. 1990. High fidelity DNA-syn-thesis by the Thermus-aquaticus DNA-polymerase. NucleicAcids Res. 18:3739-3744.

Edwards, R. A., and F. Rohwer. 2005. Viral metagenomics. Nat.Rev. Microbiol. 3:504-510.

Filée, J., F. Tétart, C. A. Suttle, and H. M. Krisch. 2005. MarineT4-type bacteriophages, a ubiquitous component of thedark matter of the biosphere. Proc. Natl. Acad. Sci. U.S.A.102:12471-12476.

Frank, H., and K. Moebus. 1987. An electron microscopicstudy of bacteriophages from marine waters. Helgol. Mar.Res. 41:385-414.

Frias-Lopez, J., and others. 2008. Microbial community geneexpression in ocean surface waters. Proc. Natl. Acad. Sci.U.S.A. 105:3805-3810.

Froussard, P. 1992. A random-PCR method (RPCR) to con-struct whole cDNA library from low amounts of RNA.Nucleic Acids Res. 20:2900.


Helton, R. R., and K. E. Wommack. 2009. Seasonal dynamicsand metagenomic characterization of estuarine virioben-thos assemblages by randomly amplified polymorphic DNAPCR. Appl. Environ. Microbiol. 75:2259-2265.

Hidaka, T., and K. Ichida. 1976. Properties of a marine RNA-containing bacteriophage. Mem. Fac. Fish. KagoshimaUniv. 25:77-89.

Iscove, N. N., M. Barbara, M. Gu, M. Gibson, C. Modi, and N.Winegarden. 2002. Representation is faithfully preserved inglobal cDNA amplified exponentially from sub-picogramquantities of mRNA. Nat. Biotechnol. 20:940-943.

Kanagwa, T. 2003. Bias and artifacts in multi-template poly-merase chain reaction (PCR). J. Biosci. Bioeng. 96:317-323.

Lang, A. S., M. L. Rise, A. I. Culley, and G. F. Steward. 2009.RNA viruses in the sea. FEMS Microbiol. Rev. 33:295-323.

Larsen, J. B., A. Larsen, G. Bratbak, and R. A. Sandaa. 2008.Phylogenetic analysis of members of the Phycodnaviridaevirus family, using amplified fragments of the major capsidprotein gene. Appl. Environ. Microbiol. 74:3048-3057.

Lasken, R. S., and T. B. Stockwell. 2007. Mechanism of chimeraformation during the multiple displacement amplificationreaction. BMC Biotechnol. 7:19.

Lawrence, J. E., and G. F. Steward. 2010. Purification of virusesby centrifugation, p. 166-181. In S. W. Wilhelm, M. G.Weinbauer, and C. A. Suttle [eds.], Manual of Aquatic ViralEcology. ASLO.

Le Gall, O., and others. 2008. Picornavirales, a proposed orderof positive-sense single-stranded RNA viruses with apseudo-T=3 virion architecture. Arch. Virol. 153:715-727.

Munn, C. B. 2006. Viruses as pathogens of marine organisms- from bacteria to whales. J. Mar. Biolog. Assoc. U.K. 86:453-467.

Nagasaki, K. 2008. Dinoflagellates, diatoms, and their viruses.J. Microbiol. 46:235-243.

———, and others. 2004. Isolation and characterization of anovel single-stranded RNA virus infecting the bloom-form-ing diatom Rhizosolenia setigera. Appl. Environ. Microbiol.70:704-711.

Peano, C., M. Severgnini, I. Cifola, G. De Bellis, and C.Battaglia. 2006. Transcriptome amplification methods ingene expression profiling. Expert Rev. Mol. Diagn. 6:465-480.

Polz, M. F., and C. M. Cavanaugh. 1998. Bias in template-to-product ratios in multi-template PCR. Appl. Environ.Microbiol. 64:3724-3730.

Rainey, F. A., N. Ward, L.I. Sly, and E. Stackerbrandt. 1994.Dependence on the taxon composition of clone librariesfor PCR amplified, naturally occurring 16S rDNA, on theprimer pair and the cloning system used. Experientia50:796-797.

Rappé, M. S., and S. J. Giovannoni. 2003. The unculturedmicrobial majority. Ann. Rev. Microbiol. 57:369-394.

Reyes, G. R., and J. P. Kim. 1991. Sequence-independent, sin-gle-primer amplification (sispa) of complex DNA popula-tions. Mol. Cell. Probes 5:473-481.

Roberts, J. D., K. Bebenek, and T. A. Kunkel. 1988. The accu-racy of reverse-transcriptase from HIV-1. Science 242:1171-1173.

Rosario, K., C. Nilsson, Y. Lim, Y. Ruan, and M. Breitbart. 2009.Metagenomic analysis of viruses in reclaimed water. Envi-ron. Microbiol. 11:2806-2820.


200

Shirai, Y., Y. Tomaru, Y. Takao, H. Suzuki, T. Nagumo, and K.Nagasaki. 2008. Isolation and characterization of a single-stranded RNA virus infecting the marine planktonic diatomChaetoceros tenuissimus Meunier. Appl. Environ. Microbiol.74:4022-4027.

Short, S. M., and C. A. Suttle. 2002. Sequence analysis ofmarine virus communities reveals that groups of relatedalgal viruses are widely distributed in nature. Appl. Envi-ron. Microbiol.68:1290-1296.

Stahlberg, A., J. Hakansson, X. J. Xian, H. Semb, and M.Kubista. 2004. Properties of the reverse transcription reac-tion in mRNA quantification. Clin. Chem. 50:509-515.


———, and A. I. Culley. 2010. Extraction and purification ofnucleic acids from viruses, p. 154-165. In S. W. Wilhelm, M.G. Weinbauer, and C. A. Suttle [eds.], Manual of AquaticViral Ecology. ASLO.

Subkhankulova, T., and F. J. Livesey. 2006. Comparative eval-uation of linear and exponential amplification techniquesfor expression profiling at the single-cell level. GenomeBiol. 7: R18.1-R18.16.

Suzuki, M. T., and S. J. Giovannoni. 1996. Bias caused by tem-plate annealing in the amplification of mixtures of 16SrRNA genes by PCR. Appl. Environ. Microbiol. 62:625-630.

Tai, V., J. E. Lawrence, A. S. Lang, A. M. Chan, A. I. Culley, andC. A. Suttle. 2003. Characterization of HaRNAV, a single-stranded RNA virus causing lysis of Heterosigma akashiwo(Raphidophyceae). J. Phycol. 39:343-352.

Takao, Y., K. Nagasaki, K. Mise, T. Okuno, and T. Honda. 2005.

Isolation and characterization of a novel single-strandedRNA virus infectious to a marine fungoid protist,Schizochytrium sp. (Thraustochytriaceae, Labyrinthulea). Appl.Environ. Microbiol. 71:4516-4522.

Tomaru, Y. and others. 2004. Isolation and characterization oftwo distinct types of HcRNAV, a single-stranded RNA virusinfecting the bivalve-killing microalga Heterocapsa circular-isquama. Aquat. Microb. Ecol. 34:207-218.

———, Y. Takao, H. Suzuki, T. Nagumo, and K. Nagasaki. 2009.Isolation and characterization of a single-stranded RNAvirus infecting the bloom-forming diatom Chaetocerossocialis. Appl. Environ. Microbiol. 75:2375-2381.

Victoria, J. G., A. Kapoor, K. Dupuis, D. P. Schnurr, and E. L.Delwart. 2008. Rapid identification of known and new RNAviruses from animal tissues. PLoS Pathog. 4:e1000163.

———, and others. 2009. Metagenomic analyses of viruses instool samples from children with acute flaccid paralysis. J.Virol. 83:4642-4651.

Von Wintzingerode, F., U. B. Gobel, and E. Stackebrandt. 1997.Determination of microbial diversity in environmentalsamples: pitfalls of PCR-based rRNA analysis. FEMS Micro-biol. Rev. 21:213-229.


Zhang, K., and others. 2006a. Sequencing genomes from singlecells by polymerase cloning. Nat. Biotechnol. 24:680-686.

Zhang, T., and others. 2006b. RNA viral community in humanfeces: Prevalence of plant pathogenic viruses. PLoS Biol.4:108-118.


201

Manual of Aquatic Viral Ecology

Documents