Evidence of Increasing Antibiotic Resistance Gene Abundances in … 4b Evidence of Increasing... · 2011-08-11 · Evidence of Increasing Antibiotic Resistance Gene Abundances in

Evidence of Increasing AntibioticResistance Gene Abundances inArchived Soils since 1940C H A R L E S W . K N A P P , † , § J A N D O L F I N G , †

P H I L L I P A . I . E H L E R T , ‡ A N DD A V I D W . G R A H A M * , †

School of Civil Engineering & Geosciences, NewcastleUniversity, Newcastle upon Tyne, United Kingdom NE17RU,and Alterra, Wageningen University and Research Center, SoilScience Centre 6700 AA Wageningen, The Netherlands

Received April 24, 2009. Revised manuscript receivedOctober 6, 2009. Accepted December 8, 2009.

Mass production and use of antibiotics and antimicrobials inmedicine and agriculture have existed for over 60 years, and hassubstantially benefited public health and agricultural productivitythroughout the world. However, there is growing evidencethat resistance to antibiotics (AR) is increasing both in benignand pathogenic bacteria, posing an emerging threat topublic and environmental health in the future. Although evidencehas existed for years from clinical data of increasing AR,almost no quantitative environmental data exist that spanincreased industrial antibiotic production in the 1950s to thepresent; i.e., data that might delineate trends in AR potentiallyvaluable for epidemiological studies. To address this criticalknowledge gap, we speculated that AR levels might be apparentin historic soil archives as evidenced by antibiotic resistancegene (ARG) abundances over time. Accordingly, DNA wasextracted from five long-term soil-series from different locationsin The Netherlands that spanned 1940 to 2008, and 16S rRNAgene and 18 ARG abundances from different major antibioticclasses were quantified. Results show that ARG from allclasses of antibiotics tested have significantly increased since1940, but especially within the tetracyclines, with someindividual ARG being >15 times more abundant now than inthe 1970s. This is noteworthy because waste managementprocedures have broadly improved and stricter rules onnontherapeuticantibioticuse inagriculturearebeingpromulgated.Although these data are local to The Netherlands, theysuggest basal environmental levels of ARG still might beincreasing, which has implications to similar locations aroundthe world.

IntroductionOver the last few decades, there has been growing concernabout antibiotic resistance (AR). The emergence of methi-cillin-resistant Staphylococcus aureus (MRSA), vancomycin-resistant enterococci (VRE), resistant Clostridium difficile andmultiresistant pseudomonads, which are epidemic “Super

Bugs”, threatens health care systems worldwide. Interestingly,resistance to antibiotic compounds in environmental mi-croorganisms (i.e., species non-native to enteric systems) isa natural phenomenon, especially in soils; however, how theSuper Bug epidemic relates to resistance in environmentalspecies is not known. For example, although industrialproduction of antibiotic and antimicrobial agents hasincreased substantially since World War II (see Figure S1 inSupporting Information (SI); (1)), concurrent changes in basalAR levels in natural soil communities have not been assessed(2). This is relevant because exposure of environmentalorganisms to antimicrobial agents, heavy metals, and re-sistant commensal and pathogenic species might createreservoirs of resistance traits in soil organisms that may havebroader consequences to public health. Although the linkbetween resistant Super Bugs and environmental organismsis not defined, it would be of great value to contrast currentsoil AR levels with historic levels prior to large-scale industrialantibiotic production as a point of first investigation, whichis the basis of this study.

As a brief history, penicillin was one of the first widelyavailable antibiotics (discovered by Alexander Fleming in1928) and entered mass production in the early 1940s. It wassoon followed by streptomycin, tetracycline, and otherantibiotics (3, 4). Eventually, the increased productioncapacity of antibiotics and decreasing prices encouragedother applications of the drugs outside of medical settings.For example, low levels of antibiotic agents were morefrequently being added as a prophylactic to animal feedbecause they were found to promote growth in livestock (5).As additional applications were identified, antimicrobialproduction and use continued to increase. Although detailedestimates of annual use and production are broadly notavailable (6), trade data suggest an exponential increase inantibiotic production prior to 1990 (e.g., Figure S1) with >50%of that manufacture being for agricultural purposes (7, 8).

Although discoveries of resistant organisms and medicalscares have promoted more prudent use of antimicrobialagents in human medicine (9) and agriculture (10), antibioticresistance genes (ARG) are still readily detectable in sewage,surface water, oceans, sediments, and soils (11). Further,there is evidence that resistant bacteria in the environmentcan harbor AR traits, which might potentially confer resis-tance to pathogens via horizontal gene transfer and othermechanisms (12). Therefore, it is critical to assess whetherpast antibiotic and antimicrobial use has altered andcontinues to alter “background” resistance levels in natureto determine whether the chance of environmental organismsconferring resistance to pathogens is increasing.

Fortunately, it was recently shown that DNA extractedfrom dried archived soils provides a wealth of informationabout historical microbial communities (13, 14). Followingthis precept, we acquired archived soil samples from ruralsites across The Netherlands collected over five long-termexperiments of different durations between 1940 and 2008,and quantified 18 resistance determinants in the soils,including ARG for extended spectrum beta-lactamases,tetracyclines, erythromycins, and glycopeptides. AlthoughARG do not indicate expressed resistance, they reflect the“potential” for resistance, which is the only viable marker forcomparing historic samples over time to study long-termtrends. Here the archived soils varied according to soil type,irrigation water use, fertilizer applications, and heavy metalexposures, which permitted speculation on factors that mightaffect local ARG levels over time. Finally, bacterial 16S-rRNAgene levels were determined to compare bacterial levels in

* Corresponding author phone: (44)-0-191-222-7930; fax: (44)-0-191-222-6502; e-mail: [email protected].

† Newcastle University.‡ Wageningen University and Research Center.§ Current address: David Livingstone Centre for Sustainability,

Department of Civil Engineering, University of Strathclyde, Glasgow,United Kingdom G1 1XN.

Environ. Sci. Technol. 2010, 44, 580–587

580 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 44, NO. 2, 2010 10.1021/es901221x 2010 American Chemical SocietyPublished on Web 12/21/2009

the soils at the time of collection, assess the effect of storageon DNA preservation, and normalize ARG levels to totalbacteria within each soil for comparative analysis.

Materials and MethodsArchived Soils for Study. Soils were acquired from TAGA,the soil archive of Alterra at Wageningen University andResearch Centre, Wageningen, The Netherlands. TAGAincludes soils from many previous soil studies that have beenarchived since 1879; the specific soils used here are sum-marized in Table 1 (see Figure 1 for locations). It should bestrongly emphasized that original soil sampling was not forthe purpose of this study and limited background information

is available for the sites, especially the older soils. Therefore,we must limit our speculation to general trends relative togross characteristics.

For example, Site A was sampled between 1960 and 1986during an agricultural study on reclaimed marine silty loamsin Wieringerwerf (15). The site was reclaimed from theWaddenzee in 1930 and used for agricultural purposes fromabout 1940 onward. The soil series was collected during aninorganic fertilizer study on crop rotation among oats, wheat,and potatoes that commenced in 1955. This site primarilyreceived mineral fertilizers, including N, P, and K, althoughit also intermittently received manure for the oats, totallingabout 135 ton-manure/ha over the study, although the timing

FIGURE 1. Five locations in The Netherlands from which archived soils were originally collected.

TABLE 1. Summary of Archived Soil Series, Including Sampling Timeframe, Soil Type, Water Use, and Total Fertilizer Use for theFive Sites

site location years soil type water use fertilizer use

A Wieringerwerf 1960-1986marine silty loam,reclaimed fromWaddenzee.

intense irrigation fromIJsselmeer

mostly inorganic, some manure(135 ton/ha) on certain plots

B Heino 1950-1974 diluvial sand ditch management, somegroundwater (10-15%).

either manure (1250 ton/ha) orinorganic; history of manureapplication

C Slootdorp 1940-1975marine silty loam,reclaimed fromWaddenzee.

intense irrigationfrom IJsselmeer

inorganic; manure used in the1960s (80 ton-manure/ha)

D Marknesse 1976-2008 marine clay ditch irrigation from the IJsselmeer(Vollenhovermeer branch)

mineral phosphate fertilizers(80 kg/ha)

E Wijster 1978-2008 diluvial sand ditches: VAM kanaaland Oude Diep

mineral phosphatefertilizers (45 kg/ha)

VOL. 44, NO. 2, 2010 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 581

of manure addition was not recorded in detail. Site B sampleswere collected in Heino in the eastern Netherlands from ahumic sandy soil (1950-1974) exposed to either manure orinorganic fertilizer, which ultimately received about 1250ton-manure/ha over the experiment. It was assumed thissite would act as a “positive control” for high residual ARGlevels because of higher manure use; however, resultsindicated otherwise (see later).

Samples from Site C were collected from 1940 to 1975 inSlootdorp in the Wieringermeer polder from the top layer ofa marine silty loam (0-25 cm) from field experimentsassessing K fertilizers (although only non-K-fertilized soilswere available here). This experiment received about 80 ton-manure/ha over the experiment, primarily in the 1960s,although exact timing was not recorded. Site D consisted oflight marine clay soils similar to soils from Wieringerwerf(sites A and C) and irrigation water was provided from theinlet of the IJsselmeer. Site E had more sandy soils (76.8%sand, 16-2000 µm) with a variable groundwater table (<1.0m in winter; >1.5 m in summer). Both sites D and E wereexperimental tracts testing rock phosphate as fertilizers andhad no manure additions. Both had ditch irrigation systems;however, site E received water from VAM kanaal and OudeDiep, rather than the IJsselmeer.

Sample Processing and DNA Extraction. Soils were air-dried at 30-40 °C, ground, and then stored at roomtemperature in the archive (15). DNA was extracted from thesoils using the UltraClean Soil DNA Isolation Kit (MoBioLaboratories, Inc., Carlsbad, CA), with some modificationsto the manufacturer protocol. Soils were sterilely weighedinto prepared centrifuge tubes containing buffer and extrac-tion beads (by weighing tubes before and after soil addition);usually 150-300 mg (as dry weight) of soil was used. Thebuffer was prechilled to 4 °C and samples were incubated for30 min for rehydration. Cells were lysed in a Hybaid Ribolyzer(6.0 setting, 30 s; Hybaid Ltd., Middlesex, UK). The sampleswere then incubated at 70 °C for 10 min to aid the lysis ofGram-positive bacteria, and briefly reshaken in the ribolyzer.The samples were further purified following manufacturerprotocols. DNA eluants were temporarily stored at -20 °C(long-term storage in -80 °C). Current and previous DNAanalyses suggest that drying and storing has no major effecton the detectable part of the pro- and eukaryotic communities(15), although this was further verified in the current study.

qPCR Methods. Penicillins, tetracyclines, macrolides(such as erythromycins), and vancomycin have been usedfor up to 80 years, depending on the drug. Resistancedeterminants for these antibiotics and antimicrobials werechosen for comparison because they represent different drugclasses, targets, and modes of action, and have been useddifferently over time in agriculture and medicine.

The abundances of 18 ARG and 16S rRNA genes (16) werequantified here using real-time qPCR (iCycler; BioRad,Hercules, CA). Although it was not possible to quantify allpossible genes, the genes chosen crossed drug classes andreflected different mechanisms of resistance. Suites of primersand probes were either selected or designed to specificallytarget resistance genes, which are summarized Table 2. Assaysfor tetracycline-resistance determinants (tet(B), tet(L), tet(M),tet(O), tet(Q) and tet(W)) were based on previously publishedmethods (17, 18). The assays for the detection of erythro-mycin-resistant-methylase determinants (erm(B), erm(C),erm(E) and erm(F)) were designed de novo (see below) withsome previously published primers (19), whereas other assayswere adapted from previously published methods: vanA (20),vanB (21), mecA (22, 23), and ampC (20), blaTEM-1 (24), blaCTX-M

(25), blaSHV-1 (26, 27), and blaOXA-1 (28, 29).Probe and Primer Designs. The 16S rRNA gene assay

was based on universal eubacterial probes and primers (16),and is commonly used in many qPCR reactions as a measure

of the total bacteria. The suites of primers and probes for thetet determinants (17, 18), ampC (ampicillin resistance, agroup-1, class-C ESBL) (20), vanA (acquired vancomycin-resistance gene) (20), and blaCTX-M (a group 2, class D ESBL)(25) were adapted from established methods and usedTaqMan probes. A second vancomycin-resistance assay(vanB) was modified from Palladino et al. (21). Minor changeswere made to increase the universality of the assay via BLASTnsearches on NCBI GenBank Web site (30); in particular, theoriginal forward primer was replaced with the VanB-FL probesequence, and the VanB-640 was incorporated as the TaqManprobe. The reverse primer (VanBR) remained the same. Themethicillin-resistance (mecA) assay included the forwardprimer by Rieschl et al. (22) and TaqMan and reverse primerby Tan et al. (23).

Erm determinants were modified to accommodate theneed of qPCR to have shorter amplicons (usually <200 bp),although our assays still used either forward or reverseprimers published previously (19). However, additionalprimers were designed based on suggested results fromPrimer3 software (31) and aligned GenBank sequences(erm(B): AF109075, K00551, M11180, M19270, U86375,U48430, U18931, X52632, and Y00116; erm(C): AF019140,J01755, M12730, M13761, M17990, M19652, U35228, U82697,V0128, X82668, and Y09001-Y09003; erm(E): M11200; anderm(F): M14730, M17124, M17808, and M62487). The speci-ficity of the primers was tested using BLASTn software andthen all were experimentally tested for targeted reactions atdifferent annealing temperatures.

blaTEM was developed using two previously publisheduniversal primers: blaTEM-F (24) and 5′-GGAATAAGGGC-GACA as the reverse primer. Primer selection was based onalignments for blaTEM-1 and blaTEM-2 (TEM-1: AY538698-AY538702, AY729027, U48775, X54607, AB194682, AB201242,AB282997, AY293072, DQ873693, EF125012, AF309824, X54604,AY392531, AY271827, AY302260, Y18200; TEM-2: X54606,AJ251946) and Primer3 software. According to a recentBLASTn search (22/07/2008), the universality of the selectedblaTEM primer pair was determined to include TEM-variants:1-3, 6, 10, 12, 21, 22, 26, 54, 60, 67, 70-72, 77-84, 89, 104-107,109, 128-129, 131, 138, 144, 150-152, 158, 162, and 163.Primers for blaSHV (blaSHV-F (26) and blaSHV-R (27)) weresimilarly developed using alignments (EU280315, X98098-X98100, DQ478720, AY826416, EF035565, EF035567); uni-versality was also tested, and includes SHV-variants: 1-2, 5,7, 11-14, 18, 25-26, 28, 30-31, 36-37, 42, 55-56, 59-66,69-71, 73-83, 89, 93-96, and 104. blaOXA-1 is designed totarget the first cluster of blaOXA; the forward primer is thereverse-complement of a reverse primer by Moland et al.(29), and an unchanged reverse primer (28). GenBanksequences used include AY008291, J02967, EF415651,EU117233, and AF227505, and the assay also targeted blaOXA-4

and blaOXA-30.qPCR Methods: Reaction Conditions. DNA template (2

µL), appropriate primers, and TaqMan probe (if available)were combined with iQ supermix PCR reagent (BioRad,Hercules, CA) and molecular-grade water to create 25 µLreaction volumes. Reactions were then performed using aBioRad iCycler with an iCycler iQ fluorescence detector andsoftware v. 2.3 (BioRad). Temperature cycle was 95 °C (10min), and 40-45 cycles of 94 °C (20 s), and then followed byannealing and elongation conditions as required (see Table2). All samples were analyzed in duplicate; any samples witha major discrepancy (high analytical variability) were rerun.Typical duplicate values ranged within ( 0.3 (log scale). Inmany cases, a TaqMan probe, a dual-labeled oligonucleotidewith a 5′-fluorophore and 3′-quencher, was used for detec-tion. Otherwise SYBR-green I, a nonspecific fluorescent dye,was used and followed with a postanalytical temperature

582 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 44, NO. 2, 2010

TABLE 2. Primer and Probe Sequences Used in the Study, and PCR Reaction Conditions

target primer/ probe (concentration) sequence (5′-3′)aannealingconditions

elongationconditions reference

totaleubacteria BAC338-F (500 nM) ACTCCTACGGGAGGCAG 60 °C/30 s (16)

BAC516-F (300 nM) Hex-TGCCAGCAGCCGCGGTAATAC-TAMRABAC805-R (500 nM) GACTACCAGGGTATCTAATCC

Tet(B) TetB-F (900 nM) ACACTCAGTATTCCAAGCCTTTG 60 °C/30 s (17)TetB-Taq (300 nM) FAM-AAAGCGATCCCACCACCAGCCAAT-TAMRATetB-R (900 nM) GATAGACATCACTCCCTGTAATGC

Tet(L) TetL-F (900 nM) GGTTTTGAACGTCTCATTACCTGAT 60 °C/30 s (17)

TetL-Taq (300 nM)FAM-CCACCTGCGAGTACAAACTGGGTGAAC-TAMRA

TetL-R (900 nM) CCAATGGAAAAGGTTAACATAAAGGTet(M) TetM-F (900 nM) GGTTTCTCTTGGATACTTAAATCAATCR 60 °C/30 s (17)

TetM-Taq (300 nM)FAM-ATGCAGTTATGGARGGGATACGCTATGGY-TAMRA

TetM-R (900 nM) CCAACCATAYAATCCTTGTTCRCTet(O) TetO-F (900 nM) AAGAAAACAGGAGATTCCAAAACG 60 °C/30 s (18)

TetO-Taq (300 nM) FAM-ACGTTATTTCCCGTTTATCACGG-TAMRATetO-R (900 nM) CGAGTCCCCAGATTGTTTTTAGC

Tet(Q) TetQ-F (900 nM) AGGTGCTGAACCTTGTTTGATTC 60 °C/30 s (18)TetQ-Taq (300 nM) FAM-TCGCATCAGCATCCCGCTC-TAMRATetQ-R (900 nM) GGCCGGACGGAGGATTT

Tet(W) TetW-F (900 nM) GCAGAGCGTGGTTCAGTCT 60 °C/30 s (18)TetW-Taq (300 nM) TTCGGGATAAGCTCTCCGCCGATetW-R (900 nM) GACACCGTCTGCTTGATGATAAT

Erm(B) ErmB-F (500 nM) AAAACTTACCCGCCATACCA 60 °C/30 sthisstudy

ErmB-R (500 nM) TTTGGCGTGTTTCATTGCTT

Erm(C) ErmC-F (500 nM) GAAATCGGCTCAGGAAAAGG 60 °C/30 sthisstudy

ErmC-R (500 nM) TAGCAAACCCGTATTCCACG

Erm(E) ErmE-F (500 nM) TGTTCGAGTGGGAGTTCGT 60 °C/30 sthisstudy

ErmE-R (500 nM) GGTACTTGCGCAGAAGCGA (19)Erm(F) ErmF-F (500 nM) TCGTTTTACGGGTCAGCACTT 60 °C/30 s (19)

ErmF-R (500 nM) CAACCAAAGCTGTGTCGTTTthisstudy

AmpC Lak2FP (300 nM) GGGAATGCTGGATGCACAA 60 °C/60 s (20)P-Lak1 (200 nM) FAM-CCTATGGCGTGAAAACCAACGTGCA-TAMRALak1RP (300 nM) CATGACCCAGTTCGCCATATC

VanA VanA3FP (300 nM) CTGTGAGGTCGGTTGTGCG 60 °C/60 s (20)VanA3FAM (200 nM) HEX-CAACTAACGCGGCACTGTTTCCCAAT-TAMRAVanA3RP (300 nM) TTTGGTCCACCTCGCCA

VanB VanBFl (400 nM) ATGGGAAGCCGATAGTC 56 °C/60 s 74 °C/60 s (21)VanB-640 (200 nM) FAM-CACGGTCAGGTTCGTCCT-TAMRAVanBR (400 nM) GATTTCGTTCCTCGACC

MecA Mec-F (400 nM) CAAGATATGAAGTGGTAAATGGT 57 °C/60 s 74 °C/60 s (22)MecA-P (200 nM) AAACAAGCAATAGAATCATCAGAT (23)MecA-R (400 nM) ACTGCCTAATTCGAGTGCTAC (23)

blaTEM BlaTEM-F (400 nM) TCGGGGAAATGTGCG 50 °C/60 s 72 °C/60 s (24)BlaTEM-R (400 nM) GGAATAAGGGCGACA

blaSHV-1 blaSHV-F (400 nM) TGATTTATCTGCGGGATACG 55 °C/60 s 76 °C/30 s (26)BlaSHV-R (400 nM) TTAGCGTTGCCAGTGCTCG (27)

blaCTX-M CTX-M-F (200 nM) ATGTGCAGYACCAGTAARGTKATGGC 58 °C/60 s 72 °C/30 s (25)CTX-M-1-groupprobe (100 nM) HEX-CCCGACAGCTGGGAGACGAAACGT-TAMRA

CTX-M-probe (100 nM) FAM-CGACAATACNGCCATGAAMGB-TAMRA

CTX-M-R (200 nM) ATCACKCGGRTCGCCNGGRAT

blaOXA-1

OXA1B14(as forwardprimer) (400 nM) CACTTACAGGAAACTTGGGGTCG 55 °C/60 s 72 °C/30 s (29)

OXA-probe (200 nM) HEX-ATCAAGCATAAAAGCCAAGAAAATGC-TAMRAthisstudy

blaOXA1-R (400 nM) AGTGTGTTTAGAATGGTGATC (28)a Sequence modification added: 5′-FAM (6-carboxyfluorescein; fluorophore); 5′-HEX (hexachlorofluorescein; fluorophore);

3′-TAMRA (carboxytetramethylrhodamine; quencher). b Degenerate IUB nucleotide symbols: R ) A/G, W ) A/T, S ) C/G, Y) C/T, K ) G/T, B ) C/G/T, and N ) A/T/C/G.

VOL. 44, NO. 2, 2010 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 583

melt curve to verify reaction quality (50-95 °C, ∆T )0.1 °C/second).

All reactions were run with serially diluted plasmid-DNAstandards of known quantity (similarly to previous methods:17, 18), created using gene-positive bacteria. The presenceof inhibitory substances in the sample matrix was checkedby spiking the samples with known amounts of templateand comparing differences in concentration threshold values(CT) between the matrix and controls (always less than onecycle difference). Either a 1:100 (sites A-B) or 1:1000 (sitesC-E) dilution with molecular-grade water worked well tominimize inhibitory effects. Further, to test for inhibition,PCR efficiencies (typically between 75 and 110%) wereexamined by comparing serial dilutions of selected samples(those with high concentrations of DNA) and plasmidcontrols. Correlation coefficients were more than 0.990 forall calibration curves, and log gene abundance values (exceptthose below detection limits) were within the linear rangeof the calibration curve.

Data Analysis. ARG abundances were first normalized to16S rRNA gene abundances to minimize variance caused bydifferential extraction and analytical efficiencies, and dif-ferences in background bacterial abundances. This createda time-series of relative ARG for each site that was normalizedto the background bacterial 16S rRNA signal. Using thesedata, trends over time for each individual gene were evaluatedusing different regression models (e.g., linear, exponentialand power), and patterns of residuals and coefficients ofdetermination (R2) were used to chose the most appropriatemodel for comparing all genes and sites. The exponentialfunction consistently best-fit resistance gene patterns in thetime series and was used for all comparisons.

Various approaches had been considered for integratingthe different time-series for longer-term trend assessment(i.e., from 1940 to 2008) and it was concluded the least-biasedmethod was to unitize all “ARG per 16S-rRNA values” to thevalues observed in the one window of time when data existedfor all series; i.e., the mid 1970s. Therefore, the mean valuefor each ARG, averaged between 1970 and 1979, was assigneda value of 1.0, and unit ARG levels in the decades before andafter 1970-1979 were calculated on a proportional basis.This unitization step created common units among time-series, and also removed site-specific variations in back-ground ARG and 16S-rRNA gene levels due to differing soiland other conditions at each site. These unitized values werethen used to determine longer-term trends of individual genesand also collective trends of all detected genes within eachantibiotic class. This latter step was valuable because itpermitted the pooling of data across time-series and increasedthe statistical significance of the different trend estimations.

Results and Discussion

The five soil-series tested were selected because theybracketed increased antibiotic production after World WarII to the present day, including periods when waste-treatmentprocedures were improved in The Netherlands in the late1970s and also when nontherapeutic antibiotic use inagriculture became more restricted over the last 10 years.Figure 1 and Table 1 summarize the soil locations and types,water sources, reported levels of manure exposure, anddurations of available samples, which varied among soil-series. Sites from other soil archives had been considered forthis study, possibly representing longer individual time-windows; however, we chose to focus only on soils fromTAGA because handling and storage methods were knownand identical.

Between 108 and 1010 16S-rRNA genes per gram-soil (dryweight) were detected in all samples, typical of many upper-horizon soils. To test the possibility of sample deterioration

(or degradation), linear regression was performed on log-transformed 16S-rRNA values over time for each series.However, no significant changes over time in bacterial 16S-rRNA levels per gram of dry soil were observed in any storedsoil (see Table S1; r2 < 0.48, p > 0.13). Therefore, contaminationor differential DNA degradation was not apparent duringsample storage, which has been noted elsewhere in carefullyarchived soils (15).

ARG per 16S-rRNA relative abundances (normalized) werecalculated for both individual genes and genes associatedwith each class of antibiotics for the five sampling sites (seeTable S2 for the actual data employed, including absoluteand relative gene abundances for each soil and site). As afirst assessment of ARG trends over time, normalizedindividual ARG were curve-fit to an exponential model overtime to determine general trends among genes at each site(Table 3). Seventy-eight percent (50 of 64) of normalizedindividual ARGs displayed increasing levels over time,although only 31% of the rate coefficients were statisticallysignificant (p < 0.10). However, four of the five sites hadconsistently increasing normalized ARG levels; i.e., all coef-ficients at site C were positive, whereas sites A, E, and D allhad less than three negative coefficients. In fact, only site Bdid not show broadly increasing normalized relative ARGlevels. Further, the major classes of antibiotic resistance testedhere showed broad increase over time, although increasesin tetracycline resistance gene levels were statisticallysignificant most frequently.

Although these results are informative and show generallyincreasing levels of ARG over time, they do not describelonger-term trends in ARG levels because each soil-seriesrepresented a different window of time. As such, thesenormalized values were unitized to average the valuesobserved between 1970 and 1979, which is when data existfor all five time-series. This approach provides a unitizedtime-series for each ARG and class of ARG that spans 1940to 2008, and allows comparison of longer-term trends fromsites with different background soil and other conditions.Figure 2 (and Table S3, which provides trend data for theindividual genes) summarizes trends in unitized data (basedon ARG detected at all sites, including site B), and showssignificantly increasing trends in ARG abundance since 1940.These trends exist among all tested classes of antibiotics,and are very consistent with trends of normalized genes ateach site (Table 3) and trends for individual genes over thelonger-term based on the unitized ARG data (Table S3).

Figure 2 confirms that the relative proportion of residentbacteria containing ARG is broadly increasing over time,ranging from 2 to 15 times higher in 2008 compared with1970-1979 levels. It should be noted that trends in Figure2 display higher levels of statistical significance than Table3 (p < 0.05), probably because rate coefficients are beingestimated based on larger data sets. Among individual ARGs,the highest rates of increase in normalized and unitized ARGabundance were the tetracyclines, particularly tet(Q), tet(O),and tet(M), and also the beta-lactamase, blaTEM-1. Interest-ingly, we see evidence of elevated extended spectrum beta-lactamases of the CTX-M family (i.e., blaCTX-M-1; see Table S3)that predate current clinical problems (32), which suggeststhat this resistance determinant may have originated fromsoils as has been speculated (33).

While the majority of the detected ARG increased inabundance over time, variations definitely existed amonggenes and among sites, which can be informative of factorsthat might influence longer term retention of resistance genesin specific locations. Accordingly, various “local” factors wereassessed, including ambient soil heavy metal levels; adsorp-tion and drainage characteristics of each soil; differing levelsof applied manure or other fertilizers; and varying irrigationpractices among sites (see Table 1), and some limited

584 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 44, NO. 2, 2010

speculation is possible. For example, soil heavy metal levelswere measured to determine whether elevated antibioticresistance in specific soils might be due to previous heavymetal coselection (34). However, soil heavy metal levelsshowed no significant increases over the 68-year samplingrecord (see Figure S2 and Table S3 for data). In reality, FigureS2 shows increases in metal levels that parallel increasingARG levels between 1940 and the late 1960s, but metal levelsdeclined thereafter when unitized ARGs increased the most.

Local differences in soil type, irrigation water source andpatterns, and fertilization might also elucidate possible factorsthat influence ARG over time and space, although ourobservations are only qualitative due to very limited historicdocumentation. The soil series were not collected for ourpurposes, and important data, such as other chemicals usedat the sites, patterns of manure application, and the qualityand amounts of irrigation water used, were neither recordednor are known. Regardless, some speculation is possible.For example, if one compares ARG trends at the three sitessampled prior to 1986, sites A and C generally displayincreases in relative ARG levels, whereas site B does not.Although the three sites received both irrigation water andmanure, only site B had well-drained sandy soils and did notreceive previously contaminated irrigation water. As back-ground, both sites A and C have chronic problems withsaltwater intrusion, therefore they have always requiredextensive freshwater irrigation from IJsselmeer, which, priorto the 1980s, was heavily contaminated by domestic andother waste inputs (35). In contrast, site B received 10 timesmore manure than sites A and C, although applications werenot well recorded. Therefore, based on gene data, it wouldappear that irrigation water quality (either irrigation waterand background saltwater) and soil drainage may be moreimportant than manure application, although this is contraryto our original hypotheses about the effects of manure.Considerable evidence suggests that manure placement caninput ARG into soils (36); however, within the context onlong-term impacts, irrigation water quality and drainage seemto be more important, although this needs to be verified byfurther studies.

Given that environment protection seems to be improving,the question is why might ARG levels be still increasing now?In our case, the answer may lie in specific practices in TheNetherlands, although such local activities are not uncom-

mon around the world. For example, despite greater emphasison conservative antibiotic use in agriculture, a surprisingincrease in use has occurred in The Netherlands since 1997(see Figure S3; (37)). The actual reason for this increase underthe more stringent European rules is under debate; however,this increase mimics increases in ARG levels at sites D andE, especially associated with tetracyclines (Table 3 and Figure2). Regardless of cause, this recent upturn in ARG levels isjust another step in apparently increasing resistance in thesoils. If one looks back over time, one can see that differentfactors may have been more or less important during differenthistoric periods, including heavy metal pollution, contami-nated surface waters for irrigation, saltwater intrusion, andincreasing use of antibiotics in agriculture. In total, ourobservations imply that single remediation approaches willnot likely be successful in reducing overall AR proliferation,and more comprehensive strategies are needed that simul-taneously address all causes, especially if antibiotic massproduction continues.

Although this study shows trends of increasing resis-tance in The Netherlands, the results argue for more andsimilar studies around the world because of the work’simplications. For example, our results imply there may bea progressively increasing chance of encountering organ-isms in nature that are resistant to antimicrobial therapy.With horizontal exchange of genetic material and increaseddiversity of bacterial hosts, each potential extrinsic sourceof resistance genes, either in the environment or amongcommensal organisms, increases the chance of acquiredresistance in a pathogen (38). Further, given that soils mightact as harbors of beneficial determinants to residentorganisms, past antibiotic and antimicrobial exposure mayhave a lingering effect on the resistance gene pool, evenwhen more prudent antibacterial use is common, althoughthis must still be proven. Overall, our data imply thatenvironmental factors promoting AR proliferation may nothave yet been fully defined, which argues for furtherdirected research into links between AR, and environmentaland anthropogenic causes. The Standing Medical AdvisoryCommittee’s report, The Path of Least Resistance (39),stated “.. .there is an uneasy sense that micro-organismsare ‘getting ahead’ and that therapeutic options arenarrowing”, and our results hint that increasing resistance

TABLE 3. Estimated First-Order Rate Coefficients for Individual Antibiotic Resistance Genes (ARG; Normalized to Bacterial16S-rRNA Gene Levels) for the Five Study Sites over the Defined Sampling Windows (Positive Rate Coefficients IndicateIncreasing ARG Levels over Time)

VOL. 44, NO. 2, 2010 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 585

in soil bacteria may be another factor impacting our battleagainst global AR.

AcknowledgmentsWe thank T. Schwartz and H. J. Monstein for providing DNAstandards, and William Sloan for comments on the manu-script. C.W.K., J.D., and D.W.G. were funded by ECOSERV,an EU Marie Curie Excellence Programme (MEXT-CT-2006-023469).

Supporting Information AvailableAdditional information on the regression analysis of log-transformed 16S-rRNA gene abundances versus time (TableS1); the abundance of 16S-rRNA genes and ARG for eachsample-year and sample-site (Table S2); exponential ratecoefficients for ARG genes and heavy metals based on unitized

data from 1940 to 2008 (Table S3); antibiotic production levelsin the United States (kton per year) from 1950 to 1986 (FigureS1); heavy metals over time among soils based on unitizeddata from 1940 to 2008 (Figure S2); and utilization of selectedantibiotics in agriculture in The Netherlands between 1997and 2007 (Figure S3). This information is available free ofcharge via the Internet at http://pubs.acs.org/.

Literature Cited(1) Committee on Human Health Risk Assessment of using Sub-

therapeutic Antibiotics in Animal Feeds. Human Health Riskswith the Subtherapeutic Use of Penicillin or Tetracyclines inAnimal Feed; Institute of Medicine, Division of Health Promotionand Disease Prevention: National Academy Press, Washington,DC, 1989.

(2) Martinez, J. L. Antibiotics and antibiotic resistance genes innatural environments. Science 2008, 321, 365–367.

(3) Bergstrom, C. T.; Feldgarden, M., The Ecology and Evolutionof Antibiotic-Resistant Bacteria In Evolution in Health andDisease; Stearns, S., Koella, J., Eds.; Karen Bush/J&J, 2008.

(4) Taubes, G. The bacteria fight back. Science 2008, 321, 356–361.(5) Frost, A. J., Antibiotics and animal production In Microbiology

of Animals and Animal Products; Woolcock, J. B., Ed. Elsevier:Amsterdam, 1991; pp 181-194.

(6) Sarmah, A. K.; Meyer, M. T.; Boxall, A. B. A. A global perspectiveon the use, sales, exposure pathways, occurrence, fate and effectsof veterinary antibiotics (VAs) in the environment. Chemosphere2006, 65, 725–759.

(7) Levy, S. B. Factors impacting on the problem of antibioticresistance. J. Antimicrob. Chemother. 2002, 49 (1), 25–30.

(8) Ungemach, F. R. Figures on quantities of antibacterials usedfor different purposes in the EU countries and interpretation.Acta Vet. Scand. 2000, 89–98.

(9) Commission of the European Communities. Report from thecommission to the council on the basis of member states’ reportson the implementation of the council recommendation (2002/77/EC) on the prudent use of antimicrobial agents in humanmedicine; Council of the European Union: Brussels, Belgium,2002.

(10) Swan Report. Report of the joint committee on the use ofantibiotics in animal husbandry and veterinary medicine;H.M.S.O.: London, UK, 1969.

(11) Kummerer, K. Resistance in the environment. J. Antimicrob.Chemother. 2004, 54 (2), 311–320.

(12) Davies, J. Inactivation of antibiotics and the dissemination ofresistance genes. Science 1994, 264 (5157), 375–382.

(13) Clark, I. M.; Hirsch, P. R. Survival of bacterial DNA and culturablebacteria in archived soils from the Rothamsted Broadbalkexperiment. Soil Biol. Biochem. 2008, 40 (5), 1090–1102.

(14) Dolfing, J.; Vos, A.; Bloem, J.; Ehlert, P. A. I.; Naumova, N. B.;Kuikman, P. J. Microbial diversity in archived soils. Science 2004,306 (5697), 813–813.

(15) Tzeneva, V. A.; Salles, J. F.; Naumova, N.; de Vos, W. M.; Kuikman,P. J.; Dolfing, J.; Smidt, H. Effect of soil sample preservation,compared to the effect of other environmental variables, onbacterial and eukaryotic diversity. Res. Microbiol. 2009, 160 (2),89–98.

(16) Yu, Y.; Lee, C.; Kim, J.; Hwang, S. Group-specific primer andprobe sets to detect methanogenic communities using quan-titative real-time polymerase chain reaction. Biotechnol. Bioeng.2005, 89 (6), 670–679.

(17) Peak, N.; Knapp, C. W.; Yang, R. K.; Hanfelt, M. M.; Smith, M. S.;Aga, D. S.; Graham, D. W. Abundance of six tetracyclineresistance genes in wastewater lagoons at cattle feedlots withdifferent antibiotic use strategies. Environ. Microbiol. 2007, 9(1), 143–151.

(18) Smith, M. S.; Yang, R. K.; Knapp, C. W.; Niu, Y. F.; Peak, N.;Hanfelt, M. M.; Galland, J. C.; Graham, D. W. Quantification oftetracycline resistance genes in feedlot lagoons by real-timePCR. Appl. Environ. Microbiol. 2004, 70 (12), 7372–7377.

(19) Patterson, A. J.; Colangeli, R.; Spigaglia, P.; Scott, K. P. Distributionof specific tetracycline and erythromycin resistance genes inenvironmental samples assessed by macroarray detection.Environ. Microbiol. 2007, 9 (3), 703–715.

(20) Volkmann, H.; Schwartz, T.; Bischoff, P.; Kirchen, S.; Obst, U.Detection of clinically relevant antibiotic-resistance genes inmunicipal wastewater using real-time PCR (TaqMan). J. Mi-crobiol. Meth. 2004, 56, 277–286.

FIGURE 2. Relative increase of antibiotic resistance genesamong soils collected at five sites in The Netherlands from1940 to 2008. All values have been normalized to 16S rRNAgene abundances. Normalized values were then groupedaccording to decade and unitized relative to mean observedvalues from 1970 to 1979 for each site. Normalization andunitization were required to account for differences inbacterial abundances among sites and place data from eachsite into a common unit of measure. Each time seriesrepresents the unbiased sum of standardized values from allfive sites. Table S2 provides detailed data for each site.mecA, blaOXA-1, vanA and ampC were analyzed, but werebelow detection limits. Shaded areas are the best-fit curvesfor each class of detected antibiotics assuming a first-ordermodel, which represents the basal level of resistance geneswithin the soils. Inset rate coefficients are for each class ofantibiotic. Rate coefficients for each individual detectedgene are provided in Table S3.

586 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 44, NO. 2, 2010

(21) Palladino, S.; Kay, I. D.; Costa, A. M.; Lambert, E. J.; Flexman,J. P. Real-time PCR for the rapid detection of vanA and vanBgenes. Diagn. Microbiol. Infect. Dis. 2003, 45 (1), 81–84.

(22) Reischl, U.; Linde, H. J.; Metz, M.; Leppmeier, B.; Lehn, N. Rapididentification of methicillin-resistant Staphylococcus aureus andsimultaneous species confirmation using real-time fluorescencePCR. J. Clin. Microbiol. 2000, 38 (6), 2429–2433.

(23) Tan, T. Y.; Corden, S.; Barnes, R.; Cookson, B. Rapid identificationof methicillin-resistant Staphylococcus aureus from positiveblood cultures by real-time fluorescence PCR. J. Clin. Microbiol.2001, 39 (12), 4529–4531.

(24) De Gheldre, Y.; Avesani, V.; Berhin, C.; Delmee, M.; Glupczynski,Y. Evaluation of Oxoid combination discs for detection ofextended-spectrum beta-lactamases. J. Antimicrob. Chemother.2003, 52 (4), 591–597.

(25) Birkett, C. I.; Ludlam, H. A.; Woodford, N.; Brown, D. F. J.; Brown,N. M.; Roberts, M. T. M.; Milner, N.; Curran, M. D. Real-timeTaqMan PCR for rapid detection and typing of genes encodingCTX-M extended-spectrum beta-lactamases. J. Med. Microbiol.2007, 56 (1), 52–55.

(26) Haeggman, S.; Lofdahl, S.; Paauw, A.; Verhoef, J.; Brisse, S.Diversity and evolution of the class a chromosomal beta-lactamase gene in Klebsiella pneumoniae. Antimicrob. AgentsChemother. 2004, 48 (7), 2400–2408.

(27) Hanson, N. D.; Thomson, K. S.; Moland, E. S.; Sanders, C. C.;Berthold, G.; Penn, R. G. Molecular characterization of a multiplyresistant Klebsiella pneumoniae encoding ESBLs and a plasmid-mediated AmpC. J. Antimicrob. Chemother. 1999, 44 (3), 377–380.

(28) Brinas, L.; Moreno, M. A.; Zarazaga, M.; Porrero, C.; Saenz, Y.;Garcia, M.; Dominguez, L.; Torres, C. Detection of CMY-2, CTX-M-14, and SHV-12 beta-lactamases in Escherichia coli fecal-sample isolates from healthy chickens. Antimicrob. AgentsChemother. 2003, 47 (6), 2056–2058.

(29) Moland, E. S.; Hanson, N. D.; Black, J. A.; Hossain, A.; Song,W. K.; Thomson, K. S. Prevalence of newer beta-lactamasesin gram-negative clinical isolates collected in the United States

from 2001 to 2002. J. Clin. Microbiol. 2006, 44 (9), 3318-3324.

(30) Altschul, S. F.; Madden, T. L.; Schaffer, A. A.; Zhang, J. H.; Zhang,Z.; Miller, W.; Lipman, D. J. Gapped BLAST and PSI-BLAST: anew generation of protein database search programs. Nucl. AcidRes. 1997, 25 (17), 3389–3402.

(31) Rozen, S.; Skaletsky, H., Primer3 on the WWW for general usersand for biologist programmers. In Bioinformatics Methods andProtocols: Methods in Molecular Biology; Krawetz, S., Misener,S., Eds.; Humana Press: Totowa, NJ, 2000; pp 365-386.

(32) Canton, R.; Coque, T. M. The CTX-M beta-lactamase pandemic.Curr. Opin. Microbiol. 2006, 9 (5), 466–475.

(33) Wright, G. D. The antibiotic resistome: the nexus of chemicaland genetic diversity. Nat. Rev. Microbiol. 2007, 5 (3), 175–186.

(34) Baker-Austin, C.; Wright, M. S.; Stepanauskas, R.; McArthur,J. V. Co-selection of antibiotic and metal resistance. TrendMicrobiol. 2006, 14 (4), 176–182.

(35) Cioc, M. The Rhine - An Eco-biography 1815-2000; Universityof Washington Press: Seattle, WA, 2006.

(36) Chee-Sanford, J. C.; Mackie, R. I.; Koike, S.; Krapac, I. G.; Lin,Y.-F.; Yannarell, A. C.; Maxwell, S.; Aminov, R. I. Fate andtransport of antibiotic residues and antibiotic resistance genesfollowing land application of manure wastes. J. Environ. Qual.2009, 38, 1086–1108.

(37) FIDIN Werkgroep Antibioticabeleid Antibioticarapportages;FIDIN (Vereniging van Fabrikanten en Importeurs van Dier-geneesmiddelen in Nederland): The Hague, 2007.

(38) Smith, D. L.; Harris, A. D.; Johnson, J. A.; Silbergeld, E. K.; Morris,J. G., Jr. Animal antibiotic use has an early but important impacton the emergence of antibiotic resistance in human commensalbacteria. Proc. Natl. Acad. Sci. 2002, 99 (9), 6434–6439.

(39) Standing Medical Advisory Committee The Path of LeastResistance; Department of Health, UK; 1997.

ES901221X

VOL. 44, NO. 2, 2010 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 587