Scale-up of virus-like particles production: effects of sparging, agitation and bioreactor scale on cell growth, infection kinetics and productivity

Journal of Biotechnology 107 (2004) 55–64

Scale-up of virus-like particles production: effects of sparging,agitation and bioreactor scale on cell growth, infection kinetics

and productivity

Luis Marangaa, António Cunhaa, João Clementea, Pedro Cruza,b,c,Manuel J.T. Carrondoa,d,∗

a Instituto de Biologia Experimental e Tecnológica/Instituto de Tecnologia Qu´ımica e Biológica IBET/ITQB, Apartado 12,Oeiras P-2781-901, Portugal

b ECBio, Ed. IBET/ITQB, Lab. 4.11, Apartado 12, Oeiras P-2781-901, Portugalc Universidade Atlˆantica, Antiga Fábrica da Pólvora, Barcarena, Portugal

d Laboratório de Engenharia Bioqu´ımica, Faculdade de Ciˆencias e Tecnologia, Universidade Nova de Lisboa,Monte da Caparica P-2825-114, Portugal

Received 27 December 2002; received in revised form 4 September 2003; accepted 11 September 2003

Abstract

The baculovirus-insect cells expression system was used for the production of self-formingPorcine parvovirus(PPV) likeparticles (virus-like particles, VLPs) in serum-free medium. At 2 l bioreactor scale an efficient production was achieved byinfecting the culture at a concentration of 1.5 × 106 cells/ml using a low multiplicity of infection of 0.05 pfu per cell. In acontinuous bioreactor, it was shown that the uninfected insect cells were not sensitive to local shear stress values up to 2.25 N/m2

at high Reynolds numbers (1.5×104) in sparging conditions. Uninfected insect cells can be grown at scaled-up bioreactor at highagitation and sparging rates as long as vortex formation is avoided and bubble entrapment is minimized. An efficient processscale-up to 25 l bioreactor was made using constant shear stress criteria for scale-up. The kinetics of baculovirus infection at lowmultiplicity of infection, either at different cell concentration or at different scales, are very reproducible, despite the differentturbulence conditions present in the bioreactor milieu. The results suggest that the infection kinetics is controlled by the rate ofbaculovirus-cell receptor attachment and is independent of the bioreactor hydrodynamic conditions. Furthermore, the achievedspecific and volumetric productivities were higher at the 25 l scale when compared to the smaller scale bioreactor. Differentrates of cell lysis after infection were observed and seem to fully explain both the shift in optimal harvest time and the increasein cell specific productivity. The results emphasize the importance of integrated strategies and engineering concepts in processdevelopment at bioreactor stage with the baculovirus insect cell system.© 2003 Elsevier B.V. All rights reserved.

Keywords:Baculovirus; Insect cells; Virus-like particles; Bioreactor; Sparging; Scale-up

∗ Corresponding author. Tel.:+351-21-442-7787;fax: +351-21-442-1161.

E-mail address:[email protected] (M.J.T. Carrondo).

1. Introduction

The baculovirus-insect cell system has been estab-lished as a powerful technology for the production of

0168-1656/$ – see front matter © 2003 Elsevier B.V. All rights reserved.doi:10.1016/j.jbiotec.2003.09.012

56 L. Maranga et al. / Journal of Biotechnology 107 (2004) 55–64

high value heterologous proteins (Power et al., 1992).Since its introduction in the 1980s, proteins from alltypes of organisms (animals, plants, bacteria, viruses)have been expressed with this system (Agathos et al.,1990). Today it is widely used for its ability to achievea fast implementation at production scale and when-ever it is necessary to obtain mg quantities of proteinsfor structural studies (Maranga et al., 2002). The highproductivity derived from the late baculovirus promot-ers and the straightforward molecular biology of thevirus make it an attractive expression system (Kingand Possee, 1992; O’Reilly and Miller, 1989).

One of the most important technological break-throughs with the baculovirus insect cell system hasbeen the observation that the expression of viral struc-tural proteins could lead to the assembly of regularstructures known as virus-like particles, VLPs (forreviews seeCasal et al., 1999; Casal, 2001; Marangaet al., 2002; Roy, 1990). These VLPs often possessoutstanding immunogenic properties and their useas potential vaccines is currently being evaluated(Casal, 2001; Maranga et al., 2002). In addition, theVLPs have been claimed to be highly immunogenic(Martinez et al., 1992), and can be engineered to in-duce B-cell, CD4+ and CTL responses (Rueda et al.,1999).

Insect cells have been cultivated in almost all opera-tion modes ranging from batch to perfusion (Marangaet al., 2002). However, since the baculovirus infectioninduces cell lysis, for the production of VLPs the batchmode or eventually, fed-batch, is preferred. The devel-opment of production processes for VLPs, as for otherbioproducts, should be performed in scaleable andfully controlled bioreactors (Agathos, 1996; Marangaet al., 2002), in this way, parameters like pO2, pH andtemperature can be kept under tight control. Since theultimate goal of the bioprocess engineer is to developa transferable process to full industrial scale, scale-upconstraints should be taken into account early on thedevelopment timetable.

The most important variables in batch process op-timization with the baculovirus-insect cell system arethe multiplicity of infection (MOI), cell concentrationat infection (CCI) and time of harvest (TOH), thesevariables are highly inter-related and are non-linearlyinterdependent. The use of a low MOI strategy ismandatory if there is a need to avoid the rapid de-pletion of the costly and certified master virus bank.

The use of a low MOI leads to an asynchronous in-fection process originating multiple cell populationstypes that coexist simultaneously, this further increasesthe complexity in the understanding and thus control-ling the system. When developing a process at differ-ent bioreactor scales it is also critical to ensure thatthe infection kinetics is reproducible between the dif-ferent scales, this is even more relevant when usinglow MOI.

No uniformity exists regarding the best scale-upcriteria. As reasonable confirmation that the scale-upprocess has been properly performed, volumetric pro-ductivity constancy works well as it is a good per-formance index in what concerns the optimization ofbioreactor operation, particularly for batch processeswith lytic systems (Maranga et al., 2002). The specificoxygen uptake rate (OURsp) normally plays a cen-tral role in process scale-up, leading to the specifica-tion of the bioreactor operational parameters (Marangaet al., 2002). Baculovirus infection generally imposesa higher metabolic burden on the infected cell withtransient increases in OURsp (Agathos, 1996; Cruzet al., 1998)and also in the glucose consumption rate(Raghunand and Dale, 1999). Through properly con-trolled bioreactor operations the increase in the OURof the culture is compensated for by sequentially vary-ing the agitation rate, the aeration rate and the oxygenpartial pressure in the gas inlet (Cruz et al., 1998).Nevertheless, as insect cells are more fragile than mi-croorganisms, their susceptibility to high shear stressimposes limitations on the maximum rate of oxygentransfer that can be attained. To avoid oxygen limita-tions, when producing HIV–VLPs,Cruz et al. (1998)infected the cells at half of the maximum cell yield thatwas attained in uninfected control cultures. Air/oxygensparging is also common in cultivation of insect cells atbioreactor scale, its importance increasing with scale,as headspace aeration, often used at lab scale, becomesirrelevant at large scale. Concerns with cell suscep-tibility to sparging, especially in serum-free media,were reduced after the beneficial effects of PluronicF-68 addition to insect cell cultures and its protec-tive effect upon bubble damage have been confirmed(Murhammer and Goochee, 1988, 1990a,b). The use ofPluronic F-68 is today ubiquitous in bioprocess devel-opment involving insect cell culture technology. Yet,little is still known regarding the susceptibility of in-fected cells, and infection time, to sparging conditions

L. Maranga et al. / Journal of Biotechnology 107 (2004) 55–64 57

since most of the previous studies were done withuninfected cells.

Porcine parvovirus(PPV) is an economically rel-evant veterinary ethiological agent and the majorcause of reproductive failure in pigs (Casal, 1996).Widespread vaccination has been proposed as acost-effective method to reduce the economical lossesdue to the endemic and worldwide prevalence ofthe virus (Gardner et al., 1996; Parke and Burgess,1993). Recombinant self-assembling VLPs composedof PPV VP2 capsid protein (PPV–VLPs) have beenexpressed with the baculovirus insect cell system andwere shown to be highly immunogenic (Martinezet al., 1992) and able to confer full protection againstthe disease (Casal, 1996).

In this study, PPV–VLPs were produced with Sf21cells in serum-free media at bioreactor stage using arecombinant baculovirus encoding the PPV VP2 cap-sid protein. A low MOI strategy was used and the ef-fects of scale, simultaneous use of sparging and highagitation rates were optimized. The process was alsoscaled-up to 25 l bioreactor using constant shear stresscriteria.

2. Materials and Methods

2.1. Cell line, virus and culture medium

A clonal isolate (Sf21) ofSpodoptera frugiperdaIPBL-SF21-AE adapted to grow in serum-free media(Hulst and Moormann, 1996) was used. Cells weremaintained in 500 ml (100 ml working volume) shakeflasks operated as repeated batch cultures at 27◦Cand 90 rpm. Inoculum of 2–4×105 cells/ml was used.When the cells reached 3–4× 106 cells/ml the cul-ture was diluted 10 times with fresh SF900II medium(Gibco, Glasgow, UK). For bioreactor experimentsPluronic F-68 (Sigma, St. Louis, MO) was added tothe medium (2 g/l, final concentration).

A recombinant baculovirus, AcAs3-PPV (Casal,1996) was used for viral infections. AcAs3-PPV isgenetically modified to express theLacZ gene underthe control of the Drosophila hsp70 promoter and toproduce aPorcine parvovirusVP2 protein under thecontrol of the very late promoter p10 and was kindlyprovided by Dr. J. Ignacio Casal (Ingenasa, Madrid,Spain). The titer of the baculovirus seed was deter-

mined by a modified end-point dilution assay (Deeand Shuler, 1997a).

2.2. Bioreactor experiments

Cell growth and infection were performed respec-tively in 2 and 25 l fully controlled stirred biore-actors (B. Braun, Melsungen, Germany) with twosix-blade turbines (standard geometry), usually ref-ereed as Rushton turbines. The 2 l bioreactors wereinoculated at 3–5× 105 cells/ml with cells from theshake flasks. The 25 l bioreactors were inoculated at2–4× 105 cells/ml with cells from 2 l bioreactors.

For the continuous runs, once the cells reachedaproximately 2× 106 cells/ml the continuous modewas started. Fresh medium was continuously added tothe bioreactor at the same rate of medium removal. Adilution rate of 0.01 h−1 was used throughout the run.The pO2 was always controlled at 30% of air satura-tion by sparging a mixture of air and oxygen througha ring sparger. During the batch phase, varying se-quentially the agitation rate, the aeration rate and theoxygen partial pressure in the gas inlet controlled thepO2 level. In the continuous mode the agitation wasvaried as defined by the experimental design and thepO2 level was controlled by automatic adjustment ofthe oxygen concentration in the gas inlet using massflow controllers; the sparging rate during the contin-uous run was always kept at 0.04 vvm, a value oftenconsidered as optimal (Cruz et al., 1998).

2.3. Analytical methods

Cell concentration was determined by hemacytome-ter cell counts (Brandt, Wertheim/Main, Germany) andcell viability was evaluated by trypan blue exclusiondye (Merck, Darmstadt, Germany) in 0.4% of PBS.

Glucose, lactate, glutamine and glutamate were an-alyzed by a YSI Biochemical Analyzer Model 2700(Yellow Spring Instruments, Yellow Springs, OH).

The PPV–VLPs titer was determined by ELISA us-ing a quantitative protocol with the commercial kitINGEZIM PPV DAS (Ingenasa, Madrid, Spain) as de-scribed elsewhere (Rueda et al., 2001).

2.4. Computation of the scale-up parameters

The scale-up from 2 to 25 l was made using a criteriato keep constant the maximum impeller shear stress

58 L. Maranga et al. / Journal of Biotechnology 107 (2004) 55–64

(Cruz et al., 1998). The 2 and 25 l bioreactors haveRushton turbines of 5.1 and 10.4 cm of diameter (Di),respectively.

Several hydrodynamic parameters were computedto define the bioreactor operation conditions at thelarger scale; the impeller Reynolds number (Rei), therate of energy dissipation per unit of mass (ε), theKolmogoroff’s eddy size(η) and the shear stress (τ).The impeller Reynolds number measures the turbu-lence, also important for gas transfer and mixing

Rei = ρND2i

µ(1)

whereN is the agitation rate,ρ the medium densityandµ the medium viscosity. The Kolmogoroff eddysize was computed fromEq. (2).

η =(

ν3

ε

)1/4

(2)

whereν is the kinematic viscosity (µ/ρ) andε the rateof energy dissipation per unit of mass which can becalculated by

ε = NpN3D2i (3)

where Np is the power number.Eq. (3)assumes thatthe volume into which the energy is dissipated isD3

i

(Cruz et al., 1998; Placek and Tavlarides, 1985). Thepower number was obtained from a Power numbercurve as a function of the impeller Reynolds Num-ber and the impeller type (Tatterson, 1991). The shearstress (τ) was calculated as

τ =( ε

ν

)1/2µ (4)

Eq. (4)assumes as representative a shear gradient de-fined asγ = (ε/ν)1/2, where� is the shear gradi-ent. For further explanation seeHenzler (2000)andMoreira et al. (1995)and also the references on thosereports.

3. Results and discussion

3.1. Definition of scale-up parameters

Table 1summarizes the bioreactor operating con-ditions that were selected based upon the scale-upcriteria defined inSection 2. The supply of oxygen

Table 1Bioreactor operational conditions at 2 and 25 l scale for Sf21growth and baculovirus infection

Parameter 2 l bioreactor 25 l bioreactor

Total volume (l) 3 42Working volume (l) 2 25Total height (cm) 25.5 76.4Liquid height (cm) 17.5 48.6Internal diameter (cm) 12.9 25.8Impeller diameter (cm) 5.1 10.4Blade height (cm) 1.05 2.0Blade width (cm) 1.45 3Number of impellers 2 2pO2 (%) 30 30Stirrer speed (rpm) 70–270 60–170Airflow (vvm) 0.01–0.04 0.01–0.04Reynolds number 3033–11705 10034–27945Kolmogoroff eddy size (�m) 30–80 30–60Maximum shear stress (N/m2) 1.2 1.2CCI (106/ml) 1.5 1.5MOI (pfu per cell) 0.05 0.05Impeller type Six-blade-

turbineSix-blade-turbine

Baffles No No

plays a critical importance in the bioreactor processdevelopment with the baculovirus insect cell system(Cruz et al., 1998; Hu and Bentley, 1999; Marangaet al., 2002; Palomares and Ramirez, 1996; Tsao et al.,1996). In this work, at both bioreactor conditions,the pO2 in the bioreactor milieu was kept constantthroughout the run at 30% of air saturation. At 2 lscale, to compensate for the increase in the oxygendemand of the cultures (OUR), due to cell growth,the agitation rate was increased automatically from70 to 270 rpm. After the culture reached an OUR thatwas no longer matched by the increase in the agita-tion rate, the sparging rate was also increased from0.01 to 0.04 vvm; afterwards, the oxygen concentra-tion in the gas inlet was also increased to compensatefor the increase in OUR. The conditions at 2 l scalewere based onCruz et al. (1998)since higher valuesthan those were reported to affect or even inhibit cellgrowth. Most of the variables shown inTable 1arealready normalized (i.e. scale independent) highlight-ing the importance of using such variables (e.g. MOI,CCI, air flowrate, etc.). To define the agitation rangeat 25 l Eqs. (1)–(4)were used to compute the maxi-mum value of agitation at that scale that originates ashear stress of 1.2 N/m2.

L. Maranga et al. / Journal of Biotechnology 107 (2004) 55–64 59

0.1

1

10

0 48 96 144 192 240

Via

ble

Cell

s (1

06/m

L)

Time (h)

Fig. 1. Growth curves of uninfected Sf21 cells cultivated in 100 mlshake flasks (�); 2 l (�) and 25 l (�) bioreactors. The cellswere inoculated at aproximately 2–4× 106 cells/ml in SF900IIsupplemented to 2 g/l of PF-68.

3.2. Uninfected growth of Sf21 cells

Two and 25 l runs were conducted to test if thebioreactor conditions had any effect on uninfectedcell growth.Fig. 1 depicts the growth of uninfectedSf21 insect cells in SF900II medium supplementedwith 2 g/l of PF-68. No detectable difference wasfound either between the different bioreactors scalesor the bioreactors and the shake flask; the averagegrowth rate was 0.024 h−1 at all scales. The shakeflask has an uncontrolled pO2 environment and thebioreactor have controlled pO2 throughout the run.The fact that no major differences were found is inaccordance with published reports that, at pO2 valueshigher than 10%, which is the Monod saturation con-stant (Maranga et al., 2002; Palomares and Ramirez,1996), the cell growth is generally independent of thepO2 value (Cruz et al., 1998; Palomares and Ramirez,1996); then, the maximum cell yield is actually lim-ited by the glutamine concentration in the media (datanot shown).

3.3. Continuous bioreactor experiments

The critical hydrodynamic conditions defined pre-viously were established in batch growth of a Sf9

clone at 2 l scale (Cruz et al., 1998). There wasstrong evidence that at agitation rates higher than270 rpm insect cell growth was severely affected.This is due to the formation of vortex and a higherlevel of bubble entrapment from the foam layerpresent at the liquid–gas surface leading to higherair flowrates than those expected from the spargingconditions. To separate the effect of varying spargingand agitation rates an experiment was conducted us-ing a continuous stirred tank bioreactor (chemostatmode).

The conditions are generally described inSection 2.After reaching a steady-state cell concentration, theagitation rate was increased stepwise. To avoid prob-lems with vortex formation the working volume waslowered to 1.2 l and only one Rushton turbine (stan-dard geometry) was used. The sparging rate was keptconstant at 0.04 vvm and the pO2 was controlled at30% in the continuous phase by adjusting the oxygenconcentration in the gas inlet; thus, as previously de-scribed, the combined effects of high agitation ratesand sparging could be rapidly assessed (Wang andBentley, 1994).

Fig. 2 shows the major results obtained in therun. The continuous mode was started at about 140 hpost-inoculation when the cell concentration reachedapproximately 2×106 cells/ml. The agitation rate wasthen fixed at 210 rpm between 145 and 500 h. Thecell concentration overshoot initially and returned toa value between 2.5 and 3.5 × 106 cells/ml. In thefollowing step changes in agitation rate no major ef-fect was observed for both the cell concentration andthe viability throughout the run. Cell concentrationoscillated between 2.5 and 3.5 × 106 cells/ml andcell viability was only slightly affected at the high-est agitation rates but always remained above 90%.Agitation rates up to 400 rpm were used with gassparging without any observable effect on the cellviability. This value of agitation rate corresponds to ashear stress of 2.2 N/m2, previously for the Sf9 clonethe cells were already affected at 1 N/m2 (Cruz et al.,1998).

It should be noted that the use of such high agita-tion rates and sparging led to a very high gas hold-upwith an abundant formation of foam that made vol-ume control of the bioreactor extremely difficult at theend of the run, leading to a run stop. The pO2 con-trol was also effective, as can be seen inFig. 2 the

60 L. Maranga et al. / Journal of Biotechnology 107 (2004) 55–64

0

100

200

300

400

0

10

20

30

40

50

60

Agit

ati

on

rate

(rp

m)

pO

2 and

oxy

gen

in th

e g

as in

let (%

)

Aera

tion ra

te (1

03 v

vm

)

0

1

2

0

20

40

60

80

0 200 400 600 800 1000

Sh

ear

stre

ss (

N/m

2)

Ko

lmo

go

roff e

dd

y siz

e ( µ

m)

Time (h)

0

1

2

3

4

5

6

7

8

0

20

40

60

80

100

Via

ble

Cell

s (1

06/m

L)

Via

bility

(%)

Fig. 2. Viable cell concentration (�), cell viability (�), agitationrate (—), aeration rate (+), pO2 (�), oxygen in the gas inlet(�), shear stress (×) and Kolmogoroff eddy size (---) profilesfor a continuous culture of Sf21 cells in SF900II supplementedto 2 g/l of PF-68. The bioreactor was inoculated at aproximately3 × 105 cells/ml and the continuous mode was initiated when thecells reached 2× 106 cells/ml. A dilution rate of 0.01 h−1 wasused throughout the run. After the cells reached a steady-state, theagitation rate was varied at defined time intervals. The aerationrate was kept constant at 0.04 vvm and the pO2 was kept constantat 30% of air saturation.

oxygen concentration in the gas inlet decreased as theagitation rate increased, due to the higherkLa valueobtained in such conditions. The computed Kolmogo-roff eddy size was always higher than the cell size

(aproximately 15�m) and it can be concluded thatagitation rates up to 400 rpm do not cause any sig-nificant effect on the cells. These results are in accor-dance with previous reports showing that it is not theturbulent fluid flow which is responsible for the cellinjury at agitation rates normally shown to affect cellviability in culture, it is the combined effect of stir-ring, sparging and bubble entrapment by vortex for-mation that leads to cell death (Michaels et al., 1996).In addition, these results also highlight that when de-veloping bioreactor processes it is imperative to assesscombined conditions of agitation rate and sparging tofind the critical hydrodynamic conditions (Cruz et al.,1998).

3.4. Two liters culture infection at different CCI

MOI is, probably, the most widespread concept usedin research with virus cultivation. Infection of cellswith low MOI led ultimately to the overall culture be-ing infected at a much higher cell concentration byprogeny virus produced already in the system. Dueto the pO2 control strategy that was employed (varia-tions in both aeration rate and agitation rate at differenttime-points post-inoculation) the infection efficiencycan be potentially affected. If the same MOI is used butat different CCI, the bioreactor microenvironment willsurely be different, at the same times post-infection,with some conceivable impact in the infection kinet-ics. To test if the bioreactor operating conditions couldimpact the progression of the infection at different CCItwo bioreactors were run in parallel and were infectedat CCI of 0.6 and 1.5×106 cells/ml, respectively. Thenormalized viable cell concentration (i.e. the viablecell concentration divided by the cell concentration atinfection) at different times post-infection is shown inFig. 3. The profiles of the normalized viable cell con-centrations are very similar for both conditions sug-gesting a highly consistent infection kinetics betweenthe two CCI conditions used. The results also suggestthat for infections at different CCI, and at 2 L scale,the use of a low multiplicity of infection with the bac-ulovirus insect cell system is highly reproducible andit is not significantly affected by the hydrodynamicand sparging conditions, despite the increase in pro-cess complexity and its associated effects (Radfordet al., 1997).

L. Maranga et al. / Journal of Biotechnology 107 (2004) 55–64 61

0.2

0.4

0.6

0.8

1.0

3.0

0 24 48 72 96 120 144 168

Rela

tiv

e c

ell

co

ncen

trati

on

Time post-infection (h)

Fig. 3. Relative viable cell concentrations, with respect to the cellconcentration at infection, at different times post-infection for 2 lbioreactor cultures infected at different CCI. The cultures wereinoculated at 0.3 × 106 cells/ml and infected with AcAs3-PPV,at an MOI of 0.05 pfu per cell, when they reached the CCI of0.6 × 106 (�) and 1.5 × 106 cells/ml (�).

3.5. Two and 25 l bioreactor infection at constantCCI

Fig. 4 shows the viable cell concentration at dif-ferent times post-infection for 2 and 25 l cultures in-

0.2

0.4

0.6

0.8

1.0

3.0

0 24 48 72 96 120 144 168

Via

ble

Cell

s (1

06/m

L)

Time post-infection (h)

Fig. 4. Viable cell concentrations at different times post-infectionfor 2 l (�) and 25 l (�) bioreactor cultures infected withAcAs3-PPV. The cultures were inoculated at 0.3 × 106 cells/mland infected, at an MOI of 0.05 pfu per cell, when they reacheda CCI of aproximately 1.5 × 106 cells/ml.

fected with baculovirus at a CCI of 1.5 × 106 cells/mland a MOI of 0.05 pfu per cell (seeTable 1for theother operating conditions). As expected, due to thelow MOI, the cells still grow after infection reachinga maximum cell yield of 3.2 and 3.4×106 cells/ml, at2 and 25 l scale, respectively.

The profiles are identical highlighting that the rateof infection is also very similar between the twodifferent scales. Based solely on the hydrodynamicconditions of the system of study to characterize thebaculovirus attachment and applying the modifiedBrownian motion model of Valentine and Allison topredict the virus attachment rates, has always origi-nated values that are higher than the ones measuredexperimentally (Power et al., 1996). Furthermore,taking into account virus-receptor interactions, it hasbeen shown, in attached (Wickham et al., 1992) andsuspension cultures (Dee and Shuler, 1997b), that thebaculovirus infection kinetics is receptor mediated butalso attachment limited. Therefore, the fact that theturbulence level is higher at 25 l scale (Reof 3× 104)than at 2 l (Reof 104) also supports this hypothesis.These results are also in accordance withDee andShuler (1997b)claim that, in suspension cultures, theinfection rate is independent of the agitation rate.

3.6. Expression of VLP at different bioreactorscales

Fig. 5 shows the VLPs expression profile at differ-ent times post-infection. Due to the low MOI used,the product is only detectable in the culture at 48 hpi.The profiles are similar up to 96 hpi but very differentat 120 hpi. The 2 l culture shows a clear maximum at96 hpi. The ELISA test for VLP quantification onlyrecognizes conformation-specific epitopes (Ruedaet al., 2001), thus, after cell lysis, the VLPs disas-semble (most probably due to the acidic pH of themedium) and are no longer detected by the ELISA,showing also decreased immunogenicity (lower qual-ity). At 25 l scale there is still an increase in producttiter between 96 and 120 hpi, the culture was har-vested at 120 hpi for downstream processing. It haspreviously been proposed that the harvest time has tobe carefully optimized for each product and for eachbioreactor size (Cruz et al., 1998), therefore theseresults are in agreement with that statement. In thiscase, a higher product titer was actually obtained upon

62 L. Maranga et al. / Journal of Biotechnology 107 (2004) 55–64

0.0

0.5

1.0

1.5

2.0

2.5

0 24 48 72 96 120 144

VL

Ps

(pg/c

ell

)

Time post-infection (h)

Fig. 5. Cell specific VLPs titers measured at different timespost-infection for 2 l (�) and 25 l (�) bioreactor cultures infectedwith AcAs3-PPV. The titers are the ratio between the VLP titer atdifferent times post-infection and the maximum cell yield obtainedin that culture. The cultures were inoculated at 0.3× 106 cells/mland infected, at an MOI of 0.05 pfu per cell, when they reacheda viable cell concentration of approximately 1.5 × 106 cells/ml.

scale-up reinforcing that the scale-up strategy that waschosen was not detrimental to the process efficiency.

The reason for the shift in the optimal harvest timebetween the two scales was assessed by analysis of thebioreactor hydrodynamic and aeration conditions atthe same times post-infection. This is shown inFig. 6for both runs. The 2 l run reaches the maximum aer-ation flowrate (0.04 vvm) but the 25 l only reaches avalue of 0.03 vvm. In addition, between 72 and 96 hpithe aeration flowrate is also much higher at 2 l scalethan at 25 l scale. Since both profiles of viable cellconcentration are very similar (seeFig. 4), this results,most probably, from the higher residence time of thebubbles at 25 l caused by the higherRe number, incomparison with the 2 l scale, and a more efficient sup-ply of oxygen at less harsh hydrodynamic conditions.Furthermore, analyzing the shear stress at the sametimes post-infection (Fig. 6) reveals a distinct profilebetween the two batches. The shear stress was muchhigher at 2 l scale between 72 and 120 hpi, than at 25 lscale. PPV–VLPs are accumulated intracellularly andinfected cell lysis leads to heavy losses in product titer.Fig. 7shows the profile of the total cell concentrationafter 72 hpi for the two batches. The lines inFig. 7

100

150

200

250

300

Ag

itati

on

rate

(rp

m)

0

10

20

30

40

Aera

tio

n r

ate

(1

03 v

vm

)

0.5

1.0

1.5

0 24 48 72 96 120

Sh

ear

stre

ss (

N/m

2)

Time post-infection (h)

Fig. 6. Agitation rate, aeration rate and shear stress profiles atdifferent times post-infection for 2 l (—) and 25 l (---) bioreactorcultures infected with AcAs3-PPV. The cultures were inoculatedat 0.3× 106 cells/ml and infected, at an MOI of 0.05 pfu per cell,when they reached a viable cell concentration of approximately1.5 × 106 cells/ml.

correspond to the best fit of the following equation (atypical cell death/lysis equation):

dXT

dt= −kLXT (5)

L. Maranga et al. / Journal of Biotechnology 107 (2004) 55–64 63

1.0

2.0

3.0

4.0

72 96 120 144 168

y = 7.4307 * e^(-0.0093223x) R= 0.99412

y = 4.9344 * e^(-0.0041879x) R= 0.98487 To

tal

Cell

s (1

06/m

L)

Time post-infection (h)

Fig. 7. Computation of the infected cell death constant at for 2 l(�) and 25 l (�) bioreactor cultures infected with AcAs3-PPV.The cultures were inoculated at 0.3×106 cells/ml and infected, atan MOI of 0.05 pfu per cell, when they reached a viable cellconcentration of aproximately 1.5 × 106 cells/ml. The lines werecomputed by a least-squares fit ofEq. (6)to the experimental data.

This equation can be solved, with appropriate bound-ary conditions, to

XT = XT0 e(−kL t) (6)

whereXT is the total cell concentration andkL the celllysis constant.The computed values for 2 and 25 l scaleare kL,2L = 0.0093± 0.0006 h−1 (P < 0.001) andkL,25L = 0.0042± 0.0004 h−1 (P < 0.005), respec-tively. This represents a reduction of more than 50%in the cell lysis rate for the 25 l scale compared withthe 2 l. This result can be explained by the lower sheardamage (both from agitation and from the bubbles) atthe 25 l scale. This result seems a likely explanationfor the higher product titer that was obtained at thisscale. A lower infected cell lysis rate means that theintracellular accumulation of VLPs is extended longerreaching an higher concentration. This level of shearstress has been shown above (Fig. 4) not to affect un-infected cells but infected cells can actually be moresensitive to this type of stresses. It is known that in-sect cell average size increases after infection (Zeiseret al., 1999, 2000) also, as a response to baculovirusinfection, insect cells shut down the synthesis of theirown proteins (Du and Thiem, 1997). This might leadto a decrease in cell membrane resistance and thus a

faster lytic destruction of the cells at the higher shearrates prevailing in the 2 l environment.

4. Conclusions

The major conclusions of this study are as follows:

• Uninfected Sf21 insect cells can be grown atscaled-up bioreactor at high agitation and spargingrates as long as vortex formation is avoided andbubble entrapment is minimized.

• The kinetics of baculovirus infection at low MOIseither at different CCI or at different scales are veryreproducible despite the different turbulence condi-tions present in the bioreactor milieu.

• Efficient scale-up was achieved using shear stressas a scale-up criteria and long-standing engineeringconcepts for insect cells, yielding consistent pro-cesses with improved productivities even throughthe low MOI strategy that was implemented.

• Different kinetics of infected cell lysis observed atdifferent scales seem to result from a higher level ofshear stress and can be responsible for the shift inthe optimal harvest time between production scalesbased on maximum volumetric productivity.

Acknowledgements

The authors acknowledge and appreciate the finan-cial support received for the project from the Euro-pean Commission (BIO4-CT98-0215) and for a grantto L.M. from Fundação para a Ciencia e Tecnologia,Portugal (PRAXIS XXI/BD/16136/98).

References

Agathos, S.N., Jeong, Y.H., Venkat, K., 1990. Growth kinetics offree and immobilized insect cell cultures. Ann. N. Y. Acad.Sci. 589, 372–398.

Agathos, S.N., 1996. Insect cell bioreactors. Cytotechnology 20,173–189.

Casal, J.I., 1996. Parvovirus diagnostic and vaccine production ininsect cells. Cytotechnology 20, 261–270.

Casal, J.I., Rueda, P., Hurtado, A., 1999. Parvovirus-like particlesas vaccine vectors. Methods 19, 174–186.

Casal, J.I., 2001. Use of the baculovirus expression system forthe generation of virus-like particles. Biotechnol. Genet. Eng.Rev. 18, 73–87.

64 L. Maranga et al. / Journal of Biotechnology 107 (2004) 55–64

Cruz, P.E., Cunha, A., Peixoto, C.C., Clemente, J., Moreira,J.L., Carrondo, M.J., 1998. Optimization of the productionof virus-like particles in insect cells. Biotechnol. Bioeng. 60,408–418.

Dee, K.U., Shuler, M.L., 1997a. Optimization of an assay forbaculovirus titer and design of regimens for the synchronousinfection of insect cells. Biotechnol. Prog. 13, 14–24.

Dee, K.U., Shuler, M.L., 1997b. A mathematical model of thetrafficking of acid-dependent enveloped viruses: application tothe binding, uptake, and nuclear accumulation of baculovirus.Biotechnol. Bioeng. 54, 468–490.

Du, X., Thiem, S.M., 1997. Responses of insect cells to baculovirusinfection: protein synthesis shutdown and apoptosis. J. Virol.71, 7866–7872.

Gardner, I.A., Carpenter, T.E., Leontides, L., Parsons, T.D., 1996.Financial evaluation of vaccination and testing alternatives forcontrol of parvovirus-induced reproductive failure in swine. J.Am. Vet. Med. Assoc. 208, 863–869.

Henzler, H.J., 2000. Particle stress in bioreactors. Adv. Biochem.Eng. Biotechnol. 67, 35–82.

Hu, Y.-C., Bentley, W.E., 1999. Enhancing yield of infectiousBursal disease virus structural proteins in baculovirusexpression systems: focus on media, protease inhibitors, anddissolved oxygen. Biotechnol. Prog. 15, 1065–1071.

Hulst, M.M., Moormann, R.J., 1996. Classical swine fevervirus diagnostics and vaccine production in insect cells.Cytotechnology 20, 271–277.

King, L.A., Possee R.D., 1992. The Baculovirus ExpressionSystem: A Laboratory Guide. Chapman Hall, London.

Maranga, L., Cruz, P.E., Aunins, J.G., Carrondo, M.J.T., 2002.Production of core and virus-like particles with baculovirusinfected insect cells. Adv. Biochem. Eng. Biotechnol. 74, 183–206.

Martinez, C., Dalsgaard, K., Lopez de Turiso, J.A., Cortes, E.,Vela, C., Casal, J.I., 1992. Production of porcine parvovirusempty capsids with high immunogenic activity. Vaccine 10,684–690.

Michaels, J.D., Mallik, A.K., Papoutsakis, E.T., 1996. Spargingand agitation-induced injury of cultured animals cells: docell-to-bubble interactions in the bulk liquid injure cells?Biotechnol. Bioeng. 51, 399–409.

Moreira, J.L., Cruz, P.E., Santana, P.C., Aunins, J.G., Carrondo,M.J.T., 1995. Formation and disruption of animal cellaggregates in stirred vessels: mechanisms and kinetic studies.Chem. Eng. Sci. 50, 2747–2764.

Murhammer, D.W., Goochee, C.F., 1988. Scale-up of insect cellcultures: protective effects of Pluronic F-68. Biotechnol. 6,1411–1418.

Murhammer, D.W., Goochee, C.F., 1990a. Sparged animal cellbioreactors: mechanism of cell damage and Pluronic F-68protection. Biotechnol. Prog. 6, 391–397.

Murhammer, D.W., Goochee, C.F., 1990b. Structural features ofnonionic polyglycol polymer molecules responsible for theprotective effect in sparged animal cell bioreactors. Biotechnol.Prog. 6, 142–148.

O’Reilly, D.R., Miller L.K., 1989. Baculovirus Expression Vectors,A Laboratory Manual. Freeman, New York.

Palomares, L.A., Ramirez, O.T., 1996. The effect of dissolvedoxygen tension and the utility of oxygen uptake rate in theinsect cell culture. Cytotechnology 22, 225–237.

Parke, C.R., Burgess, G.W., 1993. An economic assessment ofporcine parvovirus vaccination. Aust. Vet. J. 70, 177–180.

Placek, J., Tavlarides, L.L., 1985. Turbulent flow in stirred tanks.Part I. Turbulent flow in the turbine impeller region. AIChEJ. 31, 1113–1120.

Power, J., Greenfield, P.F., Nielsen, L., Reid, S., 1992. Modellingthe growth and protein production by insect cells followinginfection by a recombinant baculovirus in suspension culture.Cytotechnology 9, 149–155.

Power, J.F., Reid, S., Greenfield, P.F., Nielsen, L.K., 1996. Thekinetics of baculo virus adsorption to insect cells in suspensionculture. Cytotechnology 21, 155–163.

Radford, K.M., Cavegn, C., Bertrand, M., Bernard, A.R., Reid,S., Greenfield, P.F., 1997. The indirect effects of multiplicityof infection on baculovirus expressed proteins in insect cells:secreted and non-secreted products. Cytotechnology 24, 73–81.

Raghunand, N., Dale, B.E., 1999. Alteration of glucoseconsumption kinetics with progression of baculovirus infectionin Spodoptera frugiperda cells. Appl. Biochem. Biotechnol. 80,231–242.

Roy, P., 1990. Use of baculovirus expression vectors: developmentof diagnostic reagents, vaccines and morphologicalcounterparts of bluetongue virus. FEMS Microbiol. Immunol.2, 223–234.

Rueda, P., Martinez-Torrecuadrada, J.L., Sarraseca, J., Sedlik,C., del Barrio, M., Hurtado, A., Leclerc, C., Casal, J.I.,1999. Engineering parvovirus-like particles for the inductionof B-cell, CD4(+) and CTL responses. Vaccine 18, 325–332.

Rueda, P., Fominaya, J., Langeveld, J.P., Bruschke, C., Vela, C.,Casal, J.I., 2001. Effect of different baculovirus inactivationprocedures on the integrity and immunogenicity of porcineparvovirus-like particles. Vaccine 19, 726–734.

Tatterson, G.B., 1991. Fluid Mixing and Gas Dispersion in AgitatedTanks. MacGraw-Hill, New York.

Tsao, E.I., Mason, M.R., Cacciuttolo, M.A., Bowen, S.H.,Folena-Wasserman, G., 1996. Production of parvovirus B19vaccine in insect cells co-infected with double baculoviruses.Biotechnol. Bioeng. 49, 130–138.

Wang, M.Y., Bentley, W.E., 1994. Continuous insect cell (Sf-9)culture with aeration through sparging. Appl. Microbiol.Biotechnol. 41, 317–323.

Wickham, T.J., Shuler, M.L., Hammer, D.A., Granados, R.R.,Wood, H.A., 1992. Equilibrium and kinetic analysis ofAutographa californica nuclear polyhedrosis virus attachmentto different insect cell lines. J. Gen. Virol. 73, 3185–3194.

Zeiser, A., Bedard, C., Voyer, R., Jardin, B., Tom, R., Kamen, A.A.,1999. On-line monitoring of the progress of infection in Sf-9insect cell cultures using relative permittivity measurements.Biotechnol. Bioeng. 63, 122–126.

Zeiser, A., Elias, C.B., Voyer, R., Jardin, B., Kamen, A.A., 2000.On-line monitoring of physiological parameters of insect cellcultures during the growth and infection process. Biotechnol.Prog. 16, 803–808.