at SciVerse ScienceDirect

Food Microbiology 36 (2013) 465e474

Contents lists available

Food Microbiology

journal homepage: www.elsevier .com/locate/ fm

Diversity, distribution and antibiotic resistance of Enterococcus spp.recovered from tomatoes, leaves, water and soil on U.S. Mid-Atlanticfarms

Shirley A. Micallef a,1, Rachel E. Rosenberg Goldstein a, Ashish George a, Laura Ewing b,Ben D. Tall b, Marc S. Boyer c, Sam W. Joseph a, Amy R. Sapkota a,*

aMaryland Institute for Applied Environmental Health, School of Public Health, University of Maryland, College Park, MD 20742, USAbVirulence Mechanisms Branch, Division of Virulence Assessment, OARSA, Center for Food Safety and Applied Nutrition (CFSAN),U. S. Food and Drug Administration (FDA), Laurel, MD 20708, USAcOffice of Food Defense, Communication and Emergency Response, Center for Food Safety and Applied Nutrition (CFSAN),U. S. Food and Drug Administration (FDA), College Park, MD 20740, USA

a r t i c l e i n f o

Article history:Received 3 July 2012Received in revised form17 April 2013Accepted 23 April 2013Available online 9 May 2013

Keywords:Pre-harvest tomatoesEnterococcus spp.Enterococcus faecalisAntibiotic resistanceFood safetySoilWaterAmplified fragment length polymorphism

* Corresponding author. University of Maryland Scland Institute for Applied Environmental Health, RoomPark, MD 20742, USA. Tel.: þ1 301 405 1772; fax: þ1

E-mail address: ars@umd.edu (A.R. Sapkota).1 Present address: Department of Plant Science and

Center for Food Safety and Security Systems, CollegResources, University of Maryland, College Park, MD

0740-0020/$ e see front matter � 2013 Elsevier Ltd.http://dx.doi.org/10.1016/j.fm.2013.04.016

a b s t r a c t

Antibiotic-resistant enterococci are important opportunistic pathogens and have been recovered fromretail tomatoes. However, it is unclear where and how tomatoes are contaminated along the farm-to-forkcontinuum. Specifically, the degree of pre-harvest contamination with enterococci is unknown. Weevaluated the prevalence, diversity and antimicrobial susceptibilities of enterococci collected from to-mato farms in the Mid-Atlantic United States. Tomatoes, leaves, groundwater, pond water, irrigation ditchwater, and soil were sampled and tested for enterococci using standard methods. Antimicrobial sus-ceptibility testing was performed using the Sensititre microbroth dilution system. Enterococcus faecalisisolates were characterized using amplified fragment length polymorphism to assess dispersal potential.Enterococci (n ¼ 307) occurred in all habitats and colonization of tomatoes was common. Seven specieswere identified: Enterococcus casseliflavus, E. faecalis, Enterococcus gallinarum, Enterococcus faecium,Enterococcus avis, Enterococcus hirae and Enterococcus raffinosus. E. casseliflavus predominated in soil andon tomatoes and leaves, and E. faecalis predominated in pond water. On plants, distance from the groundinfluenced presence of enterococci. E. faecalis from samples within a farmwere more closely related thanthose from samples between farms. Resistance to rifampicin, quinupristin/dalfopristin, ciprofloxacin andlevofloxacin was prevalent. Consumption of raw tomatoes as a potential exposure risk for antibiotic-resistant Enterococcus spp. deserves further attention.

� 2013 Elsevier Ltd. All rights reserved.

1. Introduction

Enterococci are enteric, commensal bacteria that colonize thedigestive tracts of a wide range of vertebrate hosts, and are there-fore, widespread in the environment and in agricultural settings(Fisher and Phillips, 2009). Some species, including Enterococcusfaecalis and Enterococcus faecium, are among the most importanthospital-acquired, multidrug-resistant microorganisms, causingsevere, life-threatening infections of the bloodstream, urinary tract,

hool of Public Health, Mary-2234P, SPH Building, College301 314 1012.

Landscape Architecture ande of Agriculture and Natural20742, USA.

All rights reserved.

skin and soft-tissue (Arias and Murray, 2012). Enterococci are alsoaccepted as suitable indicators of fecal contamination for recrea-tional waters (USEPA, 2002), and have been used as indicators ofmicrobiological quality of fresh produce (Ailes et al., 2008; Johnstonet al., 2006).

The pathogenicity of antibiotic-resistant enterococci in hospitalsettings and the possibility of community-acquired infectionsemphasize the potential importance of these microorganisms withregard to food safety (Arias and Murray, 2012; Franz et al., 2003;Giraffa, 2002). Although hospital-acquired infections are moreprevalent than community-acquired infections, cases ofcommunity-acquired urinary tract infections and other illnessesassociated with multidrug-resistant enterococci have been re-ported, with higher risks of infection being associated with antibi-otic therapy (Aznar et al., 2004; Tang et al., 2007; Fazal et al., 1995;Kwan and Onyett, 2008; Raja et al., 2005). Foodborne antibiotic-

S.A. Micallef et al. / Food Microbiology 36 (2013) 465e474466

resistant enterococci may colonize human digestive tracts, andcould become dominant gastrointestinal tract inhabitants in hos-pitalized patients being administered antibiotics, potentiallyserving as a source of hospital-acquired infections (Arias andMurray, 2012). In spite of this potential threat to public health, theprevalence of multidrug-resistant enterococci in the environmentand the community and the possibility of foodborne routes ofexposure remain under-researched.

Nevertheless, Enterococcus spp. have been isolated from variousvegetables, leafy greens and fruits obtained from retail markets(Johnston and Jaykus, 2004; Johnston et al., 2006; Ronconi et al.,2002) and specifically from tomatoes (Abriouel et al., 2008;McGowan et al., 2006; McGowan-Spicer et al., 2008). McGowanet al. (2006) found that 9 of 27 (33.3%) tomato samples harboredEnterococcus spp., the most predominant being Enterococcus cas-seliflavus, an organism that rarely causes human illness (Gasconet al., 2003; Iaria et al., 2005; Pappas et al., 2004). Since the fewstudies that have evaluated Enterococcus on tomatoes have testedretail tomatoes, it remains unclear whether colonization occursmainly pre-harvest during the production stage versus post-harvest during packing, handling and transport. Furthermore, in-formation regarding the antimicrobial susceptibilities of field-derived enterococci is scarce.

This study aimed to evaluate the distribution, diversity andantimicrobial susceptibilities of Enterococcus spp. recovered fromtomato farms in the Mid-Atlantic region of the U.S. The relatednessamong environmental and tomato-associated E. faecalis was alsocompared by amplified fragment length polymorphism (AFLP) as ameans to evaluate relationships among isolates, and therefore,dispersal potential from one habitat to another within a farmsetting. In addition, pre-harvest tomatoes and leaves were dividedinto top-, middle- and bottom-portions of the plant to assesswhether tomato location on the vine is a risk factor for bacterialcontamination.

2. Materials and methods

2.1. Sampling sites

Eight tomato farms in the Mid-Atlantic Region of the U.S. weresampled during the 2009 tomato-harvesting season, from July toOctober. Six were large-scale industrial productions e coded TFL3,TFL9, TFL19, TFL25, TFL32 and TFL37 e that used plasticulture andchemical fertilization. On these large-scale farms, a single field wasrandomly selected and sampled in both July and October. The twoother farms included in the study were small-scale, family-ownedoperations e coded TFS1 and TFS2. On TFS1, two fields weresampled, TFS1-PC and TFS1-CC, where plasticulture or zero tillagewith a cover crop were used, respectively. On TFS2, one field wassampledwhere plasticulturewas used. The fields on the small-scalefarms were sampled in August and September. Small-scale opera-tions occasionally used composted poultry litter for fertilization. Allfields on both large-scale and small-scale farms had tomatoesplanted on raised beds in rows that were drip-irrigated using sand-filtered pond water.

2.2. Sample collection

Irrigation pond water, groundwater (water located under-ground) and water pooled in irrigation ditches between tomatorows were collected with gloved hands by filling sterile 1 L poly-ethylene Nalgene Wide Mouth Environmental Sample Bottles(Nalgene, Lima, OH, U.S.) as previously described (Micallef et al.,2012). Tomato and leaf samples as well as soil samples (200 g)from irrigation ditches, were collected from three randomly

selected locations in each sampled field with gloved hands usingsterile 798 ml whirl-pak bags (Nasco, Fort Atkinson, WI, U.S.) asdescribed previously (Micallef et al., 2012). Tomatoes and leaveswere sampled in triplicate (nine samples total) in a tiered fashionalong the vine: bottom (<30 cm from the ground), middle (30e60 cm from the ground) and top (>60 cm from the ground) plantportions as previously described (Micallef et al., 2012). All sampleswere transported to the lab on ice and stored at 4 �C.Water sampleswere analyzed within 24 h of collection.

2.3. Sample analysis

Standard membrane filtration was used to recover Enterococcusspp. from water samples (EPA, 2002). Briefly, ten-fold serial di-lutions in the range of 100e10�1 ml for pond water, 100e1 ml forgroundwater and 10e10�2 ml for irrigation ditch water werefiltered through 0.45 mm, 47 mm mixed cellulose ester filters(Millipore, Billerica, MA, U.S.). Filters were placed onmEI agar (EMDChemicals, Gibbstown, NJ, U.S.) and incubated at 42 �C for 24 h. Bluecolonies typical of enterococci on mEI were counted and thenumber of colony forming units in CFU/100 ml of water wasdetermined. One to six colonies were picked and purified on BrainHeart Infusion Agar (BD Diagnostic Systems, Franklin Lakes, NJ,U.S.). Confirmed catalase-negative isolates were tested for pyrroli-donyl peptidase (pyr) activity (Remel, Lenexa, KS, U.S.) beforearchiving in Brucella Broth (BD Diagnostic) with 15% glycerolat �80 �C.

For isolation of Enterococcus spp. from tomatoes, 100 ml ofBuffered Peptone Water (HiMedia Laboratories, Mumbai, India)was added to each bag and tomatoes were washed by hand rubbingthe bag for 2 min. A 5 ml aliquot of the rinsate was transferred to15 ml Enterococcosel Broth (EB), (BD Diagnostic Systems) for a 48 henrichment at 42 �C, and a 10 ml loopful of the enrichment wasstreaked onto Enterococcosel Agar (EA) (BD Diagnostic Systems)and incubated at 42 �C for 24 h. From each EA plate, up to six col-onies were purified, confirmed and archived as noted above.

For isolation of Enterococcus spp. from leaves, 100 ml of phos-phate buffered saline (PBS) was added to each bag and hand rubbedfor 30 s, vortexed for 2 min and the process repeated before trans-ferring a 2 ml aliquot to 18ml of EB in a culture tube and incubatingand processing as described above. For isolation of Enterococcus spp.from soil, 100 g of soil were re-suspended in 100 ml of EB for a 48 henrichment at 42 �C, and a 10 ml loopful of the enrichment wasstreaked on EA and incubated at 42 �C for 24 h. From each EA plate,up to 6 black colonies were purified on BHI agar plates, tested forcatalase and pyr activity and archived as described above.

2.4. Enterococcus spp. identification

Identification of presumptive Enterococcus spp. was performedon the Vitek 2.0 Compact 2.0 System (Biomeriuex, Marcy l’Etoile,France) using 24 h cultures grown on trypticase soy agar with 5%sheep’s blood (BD Diagnostic Systems) and GP cards with suspen-sions made up according to the manufacturer’s recommendations.For confirmation, a multiplex PCR assay described by Kariyamaet al. (2000) was modified as described below. Genomic DNAfrom Enterococcus was extracted by heat lysis as previouslydescribed (Micallef et al., 2012). Three microliters of this productwere used directly in a PCR reaction targeting the D-alanine:D-alanine ligase (ddl) genes of E. faecalis and E. faecium, the vanco-mycin resistance-encoding vanC1 and vanC2/3 genes of Entero-coccus gallinarum and E. casseliflavus, respectively, and an internalcontrol targeting a 350 base pair portion of the 16S rRNA genesusing primers as described in Micallef et al. (2012). Primers forEnterococcus avium and Enterococcus raffinosus targeted the

S.A. Micallef et al. / Food Microbiology 36 (2013) 465e474 467

superoxide dismutase gene (sodA) as described in Jackson et al.(2004). The PCR reaction contained 1� PCR buffer, 2 mM MgCl2,0.2 mMdNTPs, 0.05 mMeach of forward and reverse primers for 16SrDNA, 0.3 mM each for all ddl and vanC1 genes, 0.1 mM each forvanC2/3 genes and 0.25 mMeach for all sodA targets, and 0.6 Units ofTaq DNA Polymerase (New England Biolabs Inc., Ipswich, MA) in afinal volume of 30 ml. Amplification involved an initial denaturingstep of 95 �C for 3 min, followed by 35 cycles of denaturing at 94 �Cfor 30 s, annealing at 54 �C for 30 s and extension at 72 �C for 30 s,with a final extension of 5 min. Positive controls used for PCRamplification were E. faecalis ATCC 51299, E. faecium ATCC 51559,E. casseliflavus ATCC 25788, E. gallinarum ATCC 49573, E. aviumATCC 14025 and E. raffinosus ATCC 49427. PCR products wereresolved on 1% agarose gels.

2.5. Amplified fragment length polymorphism (AFLP) analysis

Following identification, all E. faecalis isolates were furthercharacterized by AFLP. DNA from each E. faecalis isolate wasextracted using the Ultraclean Microbial DNA Isolation kit (Mobio,Carlsbad, CA). DNA was quantified using a Nanodrop 1000 spec-trophotometer (Thermo Fisher Scientific, Willmington, DE) andapproximately 10 ng of DNAwere digested with 5 U of HindIII (NewEngland Biolabs Inc., Ipswich, MA) and 2 U of MboI (New EnglandBiolabs) for 2 h at 37 �C in a total volume of 10 ml. Enzymes weresubsequently inactivated at 65 �C. The HindIII and MboI adaptors(Invitrogen, Grand Island, NY) from Burtscher et al. (2006) werereconstituted to 1 mM and 10 mM, respectively in 10 mM Tris pH 7.5.HindIII adaptor was prepared by mixing equimolar amounts ofadaptors 1 (50-CTC GTA GAC TGC GTA CC-NH2) and 2 (50-AGC TGGTAC GCA GTC TAC), heating to 65 �C for 10 min, and then slowlycooling to 25 �C at a rate of �2 �C/min. The MboI adaptor wasprepared in the same way using adaptor 1 (50-GAC GAT GAG TCCTGA G) and adaptor 2 (50-GAT CCT CAG GAC TCA T-NH2). Ligation ofadaptors to DNAwas achieved by adding 10 ml of digested DNA to a10 ml mastermix containing 2 pmol of HindIII adaptor and 20 pmolof MboI adaptor, 0.2 U/ml of T4 ligase (New England Biolabs) and1 mM ATP (New England Biolabs), and incubating at 20 �C for 2 h.Aliquots of this product were subjected to PCR amplification withprimers specific to the adaptors. The forward primer HindIII-C-FAMwas labeled with 6-carboxyfluorescein (6-FAM) at the 50-end (50-FAM-GAC TGC GTA CCA GCT TC). The reverse primer used wasMboI-C (50-GAT GAG TCC TGA GGA TCC). For each sample, 2 ml ofDNA were mixed in a final volume of 25 ml for PCR in a reactionmixture containing 1� PCR buffer, 2.5 mM MgCl2, 0.8 mM dNTPs,0.25 mM of forward primer HindIII-C-FAM and 2.5 mM of reverseprimer MboI-C, 0.2 U of Taq DNA Polymerase (New England Bio-labs). PCR amplification involved an initial denaturing step of 94 �Cfor 3 min, followed by 35 cycles of denaturing at 94 �C for 30 s,annealing at 58 �C for 30 s and extension at 72 �C for 30 s, with afinal extension of 5 min. Negative controls contained water insteadof DNA. Sample aliquots of 2 ml were loaded onto 96 well platesmixed with 7.5 ml of Hi-Di formamide (ABI, Carlsbad, CA) and 0.5 mlGeneScan 500 ROX size standard (ABI) measuring 50e500 basepairs, in a final volume of 10 ml. Plates were sealed, heated to 95 �Cfor 3 min and placed on ice until loaded onto a 3730xl DNAAnalyzer (ABI) for fragment separation and analysis.

2.6. Antimicrobial susceptibility testing

Antimicrobial susceptibility testing was performed using theSensititre microbroth dilution system in accordance with the man-ufacturer’s instructions (Trek Diagnostic Systems Inc., Cleveland,OH). GPN3F cards were used (Trek Diagnostic Systems) to test forsusceptibility to the following antimicrobials, according to standards

published by the Clinical and Laboratory Standards Institute (CLSI)for Enterococcus spp. (CLSI, 2010): erythromycin, quinupristin/dal-fopristin (Q/D), vancomycin, daptomycin, tetracycline, ampicillin,rifampicin, levofloxacin, linezolid, penicillin, ciprofloxacin, gati-floxacin, and high level gentamicin and streptomycin. Minimalinhibitory concentrations (MICs) were recorded as the lowest con-centration of an antimicrobial that completely inhibited bacterialgrowth. Positive controls used forquality control and assurancewereE. faecalis ATCC 29212 and ATCC 51299 and Staphylococcus aureusATCC 29213 (American Type Culture Collection, Manassas VA).Multidrug resistancewasdefinedas resistance to twoormore classesof antibiotics.

2.7. Statistical analysis

Following identification and antimicrobial susceptibility testing,isolates from the same sample identified as the same species withantimicrobial susceptibility profiles that did not exhibit a four-folddifference for at least one antibiotic or a two-fold difference forthree or more antibiotics were considered to be clonal and removedfrom further analysis. The ManteleHaenszel row mean (ANOVA)statistic was used to perform a stratified statistical analysis of therelationship between presence or absence of enterococci, or therelationship between Enterococcus species, and the predictor vari-able, vine location (bottom, middle and top), after controlling forthe remaining predictor variable, farm size. In all cases, a p value<0.05 indicated a statistically significant effect of the predictorvariable on the outcome. To analyze which predictor variables, ifany, had a significant effect on multidrug resistance, a logisticregression model was fitted. All data analysis was done using SAS9.3 (SAS Institute Inc., Cary, NC).

AFLP profiles of E. faecalis were analyzed by GeneMapper Soft-ware 3.7. (ABI) and inputted into PRIMER 6 (Plymouth Routines inMultivariate Ecological Research e ver. 6) (PRIMER-E Ltd., Ply-mouth, UK), a statistical software package for the analysis ofecological, multivariate data. After data clean-up, a similarity ma-trix for the isolate profiles was constructed by calculating similar-ities between each pair of isolates using the BrayeCurtis coefficient,as previously described (Micallef et al., 2009). To visualize the re-lationships among samples, the similarity matrix was analyzed byhierarchical cluster analysis and non-metric Multi-DimensionalScaling (MDS), an ordination method that seeks to reveal possiblerelationships on a continuous scale in reduced space. MDS plotswere generated from the best possible ordination following 100random restarts. For significance testing of sample data, the non-parametric permutation procedure ANOSIM (analysis of similar-ity), available in PRIMER, was employed. This test applies ranks tosimilarity matrices and combines this ranking similarity withMonte Carlo randomization to generate significance levels (pvalues). ANOSIM tests the null hypothesis, for which a test statisticR will have a value of 0, that all samples are the same. As R ap-proaches 1, the null hypothesis is rejected and this describes a casewhere replicates from one group are more similar to each otherthan to replicates from other groups.

3. Results

A total of 295 samples were collected from the tomato farms;228 samples from large-scale farms and 67 samples from small-scale farms. Out of the large-scale farm samples, 72 (32%) weretomato samples; 90 (39%) were leaf samples; 15 (7%) were pondsamples; 12 (5%) were groundwater and 9 (4%) were irrigationditch water samples; and 30 (13%) were soil samples. Out of thesmall-scale farm samples, 27 (40%) were tomato samples; 27 (40%)were leaf samples; 2 (3%) were pond samples; 2 (3%) were

S.A. Micallef et al. / Food Microbiology 36 (2013) 465e474468

groundwater or irrigation ditch water samples; and 9 (13%) weresoil samples.

3.1. Prevalence of Enterococcus spp. in environmental samples

Enterococcus spp. were detected on 34 of 99 (34.3%) tomatosamples tested: 21 of 72 (29.2%) from large-scale farms and 13 of 27(48.1%) from small-scale farms. Of the 117 leaf samples tested, 22(18.8%) were found to be positive for Enterococcus spp.: 9 out of 90(10%) leaf samples from large-scale farms and 13 out of 27 (48.1%)samples from small-scale farms. Enterococci were more likely to beisolated from bottom and middle tomatoes versus top tomatoes(p ¼ 0.132); and lower versus middle and top leaves (p < 0.0001)(Fig. 1). For all farms, enterococci were isolated from only 21.2% oftop tomato samples, compared to 42.4% of bottom tomato samples(Fig. 1). Likewise, only 5.1% of top leaf samples were positive forenterococci, in comparison to 38.5% of bottom leaf samples (Fig. 1).

Beyond tomatoes and leaves, enterococci were isolated from 7 of17 (41.2%) total irrigation pond water samples: 5 of 15 (33.3%)samples from large-scale farms and 2/2 (100%) samples from small-scale farms. Two out of 14 (14.3%) total groundwater samples werepositive for enterococci, the October sample from TFL37 and theAugust sample from TFS1. Five out of 9 (55.6%) water samplescollected from irrigation ditches were Enterococcus positive, allfrom large-scale farms. In addition, a total of 20/39 (51.3%) irriga-tion ditch soil samples were positive for enterococci: 11 of 30(36.7%) samples from large-scale farms and 9 of 9 (100%) samplesfrom small-farms.

3.2. Enterococcus spp. diversity

In total, 307 enterococci were isolated from tomatoes, leaves,groundwater, pond water, irrigation ditch water and irrigationditch soil, of which 255were determined to be unique and includedin subsequent data analyses. Of these, 157 (62%) were isolated fromlarge-scale farms and 98 (38%) from small-scale farms. Sevendifferent Enterococcus spp. were identified by Vitek analysis andconfirmed by PCR: E. casseliflavus, E. faecalis, E. gallinarum,E. faecium, E. avium, Enterococcus hirae and E. raffinosus.

The distribution of species by habitat is shown in Fig. 2A.E. casseliflavus was the predominant taxon followed by E. faecalisand E. gallinarum. These three species were isolated from all sampletypes, except for E. gallinarum, which was not detected ingroundwater. E. casseliflavuswas the dominant species on tomatoesand leaves and in irrigation ditch water and soil. E. faecalis was thepredominant species in pondwater. Most groundwater samples (12of 14 (86%)) were Enterococcus spp. negative. Only two isolates

Fig. 1. Percentage of leaf and tomato samples from which enterococci were recoveredfrom upper, middle and bottom portions of the plants. Letters in italics indicate sig-nificant differences.

were recovered from groundwater and identified as E. faecalis andE. casseliflavus, from the large-scale and small-scale farm, respec-tively (Fig. 2). Less common species were E. faecium detected on aleaf sample, E. avium from a tomato sample, E. hirae from a pondand an irrigation ditch soil sample from separate farms, andE. raffinosus from two tomato samples, from both a large- andsmall-scale farm.

In terms of the tomato plants, enterococci species distributionwas somewhat impacted by location on the plant on both small-and large-scale farms. Middle and top tomatoes supported themostdiversity (Fig. 2B); after controlling for farm size, the distribution ofEnterococcus spp. differed significantly on tomatoes along the vine(p ¼ 0.007). E. casseliflavus and E. faecalis were more likely to be onbottom tomatoes whereas E. gallinarum and E. raffinosuswere morelikely to be on top tomatoes. In terms of leaves, bottom leaves hadthe highest diversity. E. casseliflavuswasmore likely to be recoveredfrom bottom leaves than middle or top leaves (p < 0.0001).E. faecalis and E. gallinarum were only detected on bottom andmiddle leaves and E. faecium was only detected on a bottom leafsample (Fig. 2B).

3.3. AFLP analysis of E. faecalis

AFLP was employed on E. faecalis isolates to further characterizethese isolates genotypically, and assess the heterogeneity and dis-tribution of this species within the tomato farm environment. MDSplots of E. faecalis isolates grouped by farm size (Fig. 3A), revealedtwo distinct clusters that were about 55% divergent by completelinkage cluster analysis (data not shown), although some pondwater and tomato isolates from large-scale farms clustered closer tosmall-scale farm isolates. The Global R statistic generated by theANOSIM test was 0.34 (p < 0.001), therefore supporting the hy-pothesis that there are differences between isolates from large- andsmall-scale farms. Some degree of divergence was also apparentwhen data were grouped by farm (Fig. 3B), with most clustersexhibiting >80% similarity (data not shown). The Global R statisticfor the ANOSIM test for differences among isolates recovered fromthe various farms was 0.26 (p < 0.001). No clear clustering of datawas visible when isolates were grouped by substrate type (data notshown), however this observation was confounded by the fact thatseveral farms had E. faecalis isolates represented in only one sampletype. When only farms with E. faecalis isolates from multiple hab-itats were considered, namely TFL37, TFS1-PC, TFS1-CC and TFS2,the data indicated that isolates from samples within a farm weremore closely related than those recovered from samples betweenfarms, regardless of the substrate (Fig. 3C and D). The global Rstatistic for this comparison was 0.23 (p < 0.001), and may bepossibly inflated as a result of a large number of permutationspossible for certain pairwise comparisons, which tends tostrengthens the R statistic. A high confidence level however can begiven to the MDS plots since all had very low stress, indicating thatlittle distortion was introduced to fit the data in the 2-D plot.Superimposing Fig. 3C and D reveals that data points from the samefarm representing isolates from different environmental mediagroup together, although there was also one outlying cluster forpond water isolates from TFL37. These were isolates from July,branching away from all other TFL37 October isolates, implying thatseasonal effects are also at play.

3.4. Antimicrobial resistance

Antibiotic resistance profiles were obtained for 307 isolates. Ofthese, 255 were considered non-clonal and included in data ana-lyses. Only 28 of 255 (11.0%) isolates were pansusceptible to all ofthe antibiotics tested: 17 E. casseliflavus, 7 E. gallinarum, and two

Fig. 2. (A) Enterococcus spp. distribution across each habitat of Mid-Atlantic tomato farms (total number of isolates are given in parenthesis); and (B) throughout the tomato vines (B).

S.A. Micallef et al. / Food Microbiology 36 (2013) 465e474 469

each of E. avium and E. raffinosus isolates. Multidrug resistance wasobserved in 160 of 255 (62.7%) of the isolates. These included 112E. casseliflavus, 44 E. faecalis, two E. faecium and two E. gallinarumisolates. Multidrug-resistant isolates were recovered from allsamples types. Irrigation ditch water yielded the highest incidenceof multidrug-resistant isolates (87.5%), followed by leaves (71%),irrigation ditch soil (62.5%), tomatoes (55.4%), pond water (52.6%)and groundwater (50%). The percentage of multidrug-resistantisolates was higher on large- versus small-scale farms (66.9% and56.1%, respectively). Pansusceptibility, however, was more similarbetween large- and small-scale farms, 9.2% and 12.1%, respectively.Logistic regression for multidrug resistance of E. casseliflavus,E. faecalis, E. faecium and E. gallinarum recovered from tomato andleaves revealed that farm size and vine location did not have asignificant effect on multidrug resistance. On the other hand, spe-cies was significant (p < 0.003), with E. faecalis and E. casseliflavusbeing almost 36 and 13 times more likely than E. gallinarum,respectively, to be multidrug resistant.

A high percentage of E. faecalis isolates were resistant torifampicin (71.4%), and a low percentage of these isolates wereresistant to daptomycin (1.8%), levofloxacin (8.9%) and ciprofloxacin(7.1%) (Fig. 4). Resistance to Q/D is intrinsic in this species. A highpercentage of E. faecium isolates exhibited resistance to penicillin

(100%), ampicillin (66.7%) and rifampicin (66.7%). A high percent-age of E. casseliflavus isolates were resistant to Q/D (67.1%), rifam-picin (56.7%), ciprofloxacin (46.3%) and levofloxacin (18.3%) and afew of these isolates exhibited resistance to tetracycline (1.8%),erythromycin (1.2%), linezolid (0.6%), and penicillin (0.6%). Half ofthe E. gallinarum isolates expressed resistance to daptomycin, andsome also possessed resistance to rifampicin (19.2%), Q/D (11.5%),and levofloxacin (3.8%). One of the 2 E. hirae isolates was resistantto daptomycin and the other to tetracycline. E. avium andE. raffinosus were pansusceptible. None of the isolates exhibitedresistance to vancomycin, gatifloxacin, or high level gentamicin andstreptomycin.

Interestingly, isolates resistant to ciprofloxacin, levofloxacin,rifampicin, daptomycin, Q/D and rifampicin were widespreadamong the different types of environmental media tested: to-matoes, leaves, irrigation ditch soil, pond water and irrigation ditchwater (Fig. 5). One groundwater isolate also exhibited rifampicinresistance.

When looking at the resistance data by farm type, ciprofloxacin,Q/D and rifampicin resistance were ubiquitous and detected on allfarms. Levofloxacin resistance was detected on all farms exceptTFL3 and TFL32, and daptomycin resistance was detected on allfarms except TFS1-PC, TFS1-CC (same farm, separate fields), and

Habitat

TomatoLeafIrrigation Pond WaterGroundwaterIrrigation Ditch SoilE. faecalis ATCC 51299

2D Stress: 0.11

Farm ID

TFL37TFS1TFS1-PCTFS1-CCTFS2E. faecalis ATCC 51299

2D Stress: 0.11

Farm ID

TFL3TFL9TFL19TFL25TFL32TFL37TFS1TFS1-PCTFS1-CCTFS2E. faecalis ATCC 51299

2D Stress: 0.24

Farm Size

LargeSmallE. faecalis ATCC 51299

2D Stress: 0.24

Fig. 3. Multidimensional scaling (MDS) plots of amplified fragment length polymorphism (AFLP) data of E. faecalis from all tomato farms by farm size (A), and by specific farm (B);and for TFL37, TFS1 and TFS2 by environmental media (C) and by specific farm (D).

S.A. Micallef et al. / Food Microbiology 36 (2013) 465e474470

Fig. 4. Antibiotic resistance profiles of Enterococcus spp. recovered fromMid-Atlantic tomato farms. E. avium and E. raffinosus are not represented as all isolates were pansusceptible.No resistance to vancomycin, gatifloxacin, high level gentamicin or high level streptomycin was detected. Asterisk (*) denotes intrinsic resistance.

S.A. Micallef et al. / Food Microbiology 36 (2013) 465e474 471

TFL37 (Fig. 6). The rarer resistance patterns observed, namelyresistance to penicillin, linezolid, ampicillin, tetracycline anderythromycin, were only observed in isolates recovered from small-scale farms, with the exception of 1 isolate obtained from a large-scale farm that expressed resistance to erythromycin (Fig. 6).

4. Discussion

4.1. Prevalence of Enterococcus spp.

In this study,we found that tomato leaves and fruit are frequentlycolonized pre-harvest with a diverse group of enterococci. Preva-lence of enterococci on tomatoes from large-scale farms concurredwith published data regarding that of retail tomatoes (McGowan

Fig. 5. Distribution of antibiotic-resistant Enterococcus spp. among ecological niches sampleIntrinsic resistance of E. faecalis to Q/D is omitted.

et al., 2006) but higher rates were obtained on tomatoes recov-ered fromsmall-scale farms. Thedifferences in agricultural practicesbetween large- and small-scale production may account for thesedisparities. Some of these differences included fertilizationmethods, with exclusive chemical fertilization on large-scale farms,and a mixture of chemical and organic fertilization on small farms.The farms may also differ with regard to their accessibility to wild-life. Smaller farms tended to have wooded and riparian areas incloser proximity to production areas, as a factor of their size,increasing the likelihood of wildlife populations entering fields.However, evidence ofwildlife intrusion, including animal tracks andfeces, were observed on both small and large-scale farms.

Our findings also show that position on the vine appears to be adeterminant for Enterococcus spp. colonization. The probability of

d. n ¼ number of isolates across all sample types that were resistant to that antibiotic.

Fig. 6. Distribution of antibiotic-resistant Enterococcus spp. by farm. Large-scale farms and small-scale farms are depicted in patterned and solid bars, respectively. n ¼ number ofisolates that were resistant to that antibiotic.

S.A. Micallef et al. / Food Microbiology 36 (2013) 465e474472

Enterococcus spp. colonization decreased with increasing distancefrom the ground. This has implications for food safety. One possiblemeans of transmission of human pathogens onto tomatoes issplashingof contaminated soil onto tomato crops. A recent study thatsimulated rain events to assess dispersal of gfp-tagged SalmonellaTyphimurium showed that Salmonella could be transferred fromplastic mulch or soil to tomato leaflets and survived for 3 days(Cevallos-Cevallos et al., 2012). Splashing is also a known route ofdispersal for multiple species of the plant pathogen Colletotrichumspp. onto strawberries and is greatly enhanced by rainfall(Ntahimpera et al.,1999). Intermittent splashing by rain or tramplingbyworkers in tomatofields could effectively act asmultiple episodesof spot inoculation of tomato surfaces, not just with plant pathogensbut also with microorganisms that can compromise human health.Further investigation could assess if pathogen colonization risk var-ies with plant location, and whether selective harvesting mightreduce the likelihood of picking potentially contaminated produce.

4.2. Diversity and distribution of Enterococcus spp.

The diversity of enterococci detected on tomatoes and leaves isnotable. The high prevalence of E. casseliflavus, a species that isoften naturally associated with plants (Müller et al., 2001) andother aquatic environments (Badgley et al., 2010), potentially di-minishes the suitability of Enterococcus spp. as a fecal indicator fortomatoes and their production environments. E. faecalis andE. faecium are the two most common human infection-causingspecies and the two species that may be the most reliable as in-dicators of fecal contamination. Identification of enterococci to thespecies level to identify these taxa would provide a more accuratemetric of fecal contamination. Testing for the presence of E. faecalisand E. faecium in irrigation water and on produce should beconsidered as a potentially useful tool in assessing the microbio-logical quality of tomatoes and the food safety risk of a tomatoproduction area.

Genotyping by AFLP was performed on the E. faecalis isolates asa means to assess genetic relatedness among isolates and henceidentify clusters of closely related isolates within the tomato farmenvironment. By comparing fingerprints of the E. faecalis isolatesretrieved from various ecological niches, it appears that geneticheterogeneity among E. faecalis may be more dependent on

geography, agricultural practices and seasonal effects than habitat.The large-scale farms are located in a different area than the small-scale farms, suggesting that the E. faecalis heterogeneity observedmay be a factor of the geographical separation between large- andsmall-scale farms. However, since no E. faecalis isolates from othergeographical areas or regions were available, the impact of crop-ping practices cannot be discounted.

Differences between E. faecalis fingerprints from isolatesrecovered in July and October were evident for one farm, TFL37,suggesting that the populations may be dynamic over time. How-ever, more isolates collected over a longer period of time areneeded to assess the significance of temporal influences onenterococci contamination. Hierarchical clustering (data notshown) and MDS suggest that habitat is not a barrier for E. faecalistransmission and that E. faecalis are able to disperse efficientlywithin a farm setting, transferring from one habitat to another suchas from water and soil to tomato leaves and fruit. But it is possiblethat the homogeneity among E. faecalis isolates within a single farmmight have been overestimated by the ordination method used,since ANOSIM testing supported differences among isolatesrecovered from the various environmental media. However, ifenteric pathogens such as Salmonella or Listeria mono-cytogenesdtwo organisms that are often prevalent in the envi-ronmentdalso disperse easily between habitats, occurrence ofthese pathogens anywhere on a farm, such as previously reported(Micallef et al., 2012), could potentially pose a food safety risk. Theproximity of cultivated fields to ponds, and the location of compostheaps and windrows may be important risk factors to considerwhen cultivating high risk crops, even if these resources are notbeing used directly in production areas. Monitoring enteric path-ogens in various environmental media, not just irrigation water asis currently recommended in Good Agricultural Practices (GAPs)guidelines in the U.S. (FDA, CFSA, 1998), might be a better practicefor enhancing food safety of fresh produce.

4.3. Antibiotic resistance of Enterococcus spp.

Resistance against a number of antibiotics was observed amongthe environmental enterococci recovered in this study in spite ofthe absence of a direct selective pressure, since no antibiotics areapplied to tomato fields. In particular, among isolates tested,

S.A. Micallef et al. / Food Microbiology 36 (2013) 465e474 473

pansusceptibility was low (11%), compared to multidrug resistance(60%). The susceptibility of environmental and tomato isolates tovancomycin in this study corroborates published reports (Abriouelet al., 2008; Johnston and Jaykus, 2004; McGowan et al., 2006).However, in this study, resistance to rifampicin was noted for alarge proportion of E. casseliflavus, E. faecalis, E. faecium and less sofor E. gallinarum. A high incidence of rifampicin resistance was alsoobserved for E. faecalis isolated from a variety of retail fruits andvegetables in Spain (Abriouel et al., 2008), but contrasted with thelow resistance previously noted for E. faecalis isolated from U.S.produce (Johnston and Jaykus, 2004). Resistance to Q/D, dapto-mycin and linezolid observed in our study is notable, since theseantibiotics are now used as alternative treatments for multidrugresistant Enterococcus strains (Arias and Murray, 2012).

Unlike large-scale farms that fertilized exclusively with chemi-cal nutrients, small-scale farms occasionally used poultry litter forfertilization (although the specific litter source was unknown).Current recommended composting times for poultry litter may beinsufficiently long to eliminate all enterococci (Graham et al., 2009),resulting in the risk of introducing MDR organisms onto crops viacontaminated manure application. Antibiotic resistance rate dis-crepancies between small and large farms were apparent, althoughnot significant by logistic regression, and the rarer antibioticresistance phenotypes all originated from small farms. If organicfertilization is in fact a potential source of antibiotic-resistantbacteria that might be carried over from animal operations, thenpoultry litter composting and application might require stricterregulations.

A number of Enterococcus spp. identified on tomatoes in thisstudy are known to colonize the human gut. Aside from thecommonly found E. faecalis and E. faecium, other species includingE. casseliflavus, E. avium and E. raffinosus have also been described inthe gut microbiota (Kubota et al., 2010). Yet, the efficiency ofgastrointestinal tract colonization by enterococci via consumptionof contaminated produce is unknown. Colonization by strainspossessing resistance to clinically important antibiotics such asQ/D, rifampicin, ciprofloxacin, daptomycin, erythromycin, tetracy-cline, ampicillin and linezolid, could result in later infections,especially under instances of antibiotic exposure or surgery. Theunclear distinction between clinical and environmental strains ofE. faecalis has been noted (Willems et al., 2011), and the dissemi-nation and sharing of strains between nosocomial settings and thecommunity could be high for this species, highlighting the impor-tance of further investigating foodborne routes of exposure.Considering the antibiotic resistance reservoir potential of entero-cocci, any risks associated with the consumption of food carryingantibiotic-resistant enterococci that could colonize the human gutcalls for further study.

5. Conclusions

This study shows that colonization of tomatoes by enterococcican occur during the pre-harvest stage of tomato production. Po-sition of tomatoes and leaves on the vine appears to influenceenterococcal colonization. The likelihood of Enterococcus spp.colonization was higher closer to the ground, suggesting thatirrigation water or soil could be sources for most enterococci onproduce. Vine position could also impact enterococcal speciesdistribution, which could be a result of different environmentalreservoirs for each species. E. faecalis isolates from within a farmwere more genotypically related to each other than to isolatesfrom other farms, regardless of the ecological niche from whichthey originated, suggesting that dispersal occurs within tomatofarms. Antibiotic resistance to a wide range of clinically relevantantibiotics was detected, including rifampicin, Q/D, ciprofloxacin,

levofloxacin and daptomycin. The public health risks associatedwith the consumption of tomatoes colonized by antibiotic-resistant Enterococcus spp. remains unknown but deservesfurther investigation.

Acknowledgments

The authors would like to thank Steve Rideout, Andrew J. Estrinand Cristina R. McClaughlin for their help in the planning stages ofthis project and the Virginia Institute for Marine Sciences for lab-oratory space during sampling. This research was supported by theUniversity of Maryland, Joint Institute for Food Safety and AppliedNutrition (JIFSAN) and Grant/Cooperative Agreement Number5U01CI000310 from the Centers for Disease Control and Prevention(CDC). Its contents are solely the responsibility of the authors anddo not necessarily represent the official views of CDC.

References

Abriouel, H., Omar, N.B., Molinos, A.C., López, R.L., Grande, M.J., Martínez-Viedma, P.,Ortega, E., Cañamero, M.M., Galvez, A., 2008. Comparative analysis of geneticdiversity and incidence of virulence factors and antibiotic resistance amongenterococcal populations from raw fruit and vegetable foods, water and soil,and clinical samples. International Journal of Food Microbiology 123, 38e49.

Ailes, E.C., Leon, J.S., Jaykus, L.-A., Johnston, L.M., Clayton, H.A., Blanding, S.,Kleinbaum, D.G., Backer, L.C., Moe, C.L., 2008. Microbial concentrations on freshproduce are affected by postharvest processing, importation, and season.Journal of Food Protection 71, 2389e2397.

Arias, C.A., Murray, B.E., 2012. The rise of the Enterococcus: beyond vancomycinresistance. Nature Reviews Microbiology 10, 266e278.

Aznar, E., Buendía, B., García-Peñuela, E., Escudero, E., Alarcón, T., López-Brea, M.,2004. Community-acquired urinary tract infection caused by vancomycin-resistant Enterococcus faecalis clinical isolate. Revista Española de Quimioter-apia 17, 263e265.

Badgley, B.D., Thomas, F.I.M., Harwood, V.J., 2010. The effects of submerged aquaticvegetation on the persistence of environmental populations of Enterococcusspp. Environmental Microbiology 12, 1271e1281.

Burtscher, M.M., Köllner, K.E., Sommer, R., Keiblinger, K., Farnleitner, A.H.,Mach, R.L., 2006. Development of a novel amplified fragment length poly-morphism (AFLP) typing method for enterococci isolates from cattle faeces andevaluation of the single versus pooled faecal sampling approach. Journal ofMicrobiological Methods 67, 281e293.

Cevallos-Cevallos, J.M., Danyluk, M.D., Gu, G., Vallad, G.E., van Bruggen, A.H.C., 2012.Dispersal of Salmonella Typhimurium by rain splash onto tomato plants. Journalof Food Protection 75, 472e479.

CLSI, 2010. In: Clinical and Laboratory Standards Institute (Ed.), PerformanceStandards for Antimicrobial Susceptibility Testing; Twentieth InformationalSupplement. Clinical and Laboratory Standards Institute, Wayne, PA. CLSIDocument M100-S20.

EPA, 2002. Method 1600: Enterococci in Water by Membrane Filtration UsingMembrane-Enterococcus Indoxyl-B-D-Glucoside Agar (mEI). Office of Water.U.S. Environmental Protection Agency, Washington, DC.

Fazal, B., Turett, G., Chilimuri, S., Mendoza, C., Telzak, E., 1995. Community-acquiredenterococcal meningitis in an adult. Clinical Infectious Diseases 20, 725e726.

FDA, Center for Food Safety and Applied Nutrition (CFSAN), 1998. Guidance forIndustry: Guide to Minimize Microbial Food Safety Hazards for Fresh Fruits andVegetables.

Fisher, K., Phillips, C., 2009. The ecology, epidemiology and virulence of Entero-coccus. Microbiology 155, 1749e1757.

Franz, C.M.A.P., Stiles, M.E., Schleifer, K.H., Holzapfel, W.H., 2003. Enterococci infoodseA conundrum for food safety. International Journal of Food Microbiology88, 105e122.

Gascon, F., Castano, M., Gonzalez, A., Cordon, M., 2003. Endocarditis due toEnterococcus casseliflavus. Enfermedades Infecciosas y Microbiología Clínica 21,275e276.

Giraffa, G., 2002. Enterococci from foods. FEMS Microbiology Reviews 26, 163e171.Graham, J.P., Evans, S.L., Price, L.B., Silbergeld, E.K., 2009. Fate of antimicrobial-

resistant enterococci and staphylococci and resistance determinants in storedpoultry litter. Environmental Research 109, 682e689.

Iaria, C., Stassi, G., Costa, G., Di Leo, R., Toscano, A., Cascio, A., 2005. Enterococcalmeningitis caused by Enterococcus casseliflavus. First case report. BMC Infec-tious Diseases 5, 3.

Jackson, C.R., Fedorka-Cray, P.J., Barrett, J.B., 2004. Use of a genus- and species-specific multiplex PCR for identification of enterococci. Journal of ClinicalMicrobiology 42, 3558e3565.

Johnston, L.M., Jaykus, L.-A., 2004. Antimicrobial resistance of Enterococcus speciesisolated from produce. Applied and Environmental Microbiology 70, 3133e3137.

S.A. Micallef et al. / Food Microbiology 36 (2013) 465e474474

Johnston, L.M., Jaykus, L.-A., Moll, D., Anciso, J., Mora, B., Moe, C.L., 2006. A fieldstudy of the microbiological quality of fresh produce of domestic and Mexicanorigin. International Journal of Food Microbiology 112, 83e95.

Kariyama, R., Mitsuhata, R., Chow, J.W., Clewell, D.B., Kumon, H., 2000. Simple andreliable multiplex PCR assay for surveillance isolates of vancomycin-resistantenterococci. Journal of Clinical Microbiology 38, 3092e3095.

Kubota, H., Tsuji, H., Matsuda, K., Kurakawa, T., Asahara, T., Nomoto, K., 2010.Detection of human intestinal catalase-negative, Gram-positive cocci by rRNA-targeted reverse transcription-PCR. Applied and Environmental Microbiology76, 5440e5451.

Kwan, C.W., Onyett, H., 2008. Community-acquired urinary tract pathogens andtheir resistance patterns in hospitalized children in Southeastern Ontario be-tween 2002 and 2006. Paediatrics & Child Health 13, 759e762.

McGowan, L.L., Jackson, C.R., Barrett, J.B., Hiott, L.M., Fedorka-Cray, P.J., 2006.Prevalence and antimicrobial resistance of enterococci isolated from retailfruits, vegetables, and meats. Journal of Food Protection 69, 2976e2982.

McGowan-Spicer, L.L., Fedorka-Cray, P.J., Frye, J.G., Meinersmann, R.J., Barrett, J.B.,Jackson, C.R., 2008. Antimicrobial resistance and virulence of Enterococcusfaecalis isolated from retail food. Journal of Food Protection 71, 760e769.

Micallef, S.A., Shiaris, M.P., Colón-Carmona, A., 2009. Influence of Arabidopsisthaliana accessions on rhizobacterial communities and natural variation in rootexudates. Journal of Experimental Botany 60, 1729e1742.

Micallef, S.A., Rosenberg Goldstein, R.E., George, A., Kleinfelter, L., Boyer, M.S.,McLaughlin, C.R., Estrin, A., Ewing, L., Jean-Gilles Beaubrun, J., Hanes, D.E.,Kothary, M.H., Tall, B.D., Razeq, J.H., Joseph, S.W., Sapkota, A.R., 2012. Occurrence

and antibiotic resistance of multiple Salmonella serotypes recovered fromwater,sediment and soil on Mid-Atlantic tomato farms. Environmental Research 114,31e39.

Müller, T., Ulrich, A., Ott, E.M., Müller, M., 2001. Identification of plant-associatedenterococci. Journal of Applied Microbiology 91, 268e278.

Ntahimpera, N., Wilson, L.L., Ellis, M.A., Madden, L.V., 1999. Comparison of raineffects on splash dispersal of three Colletotrichum species infecting strawberry.Phytopathology 89, 555e563.

Pappas, G., Liberopoulos, E., Tsianos, E., Elisaf, M., 2004. Enterococcus casseliflavusbacteremia. Case report and literature review. Journal of Infection 48, 206e208.

Raja, N.S., Karunakaran, R., Ngeow, Y.F., Awang, R., 2005. Community-acquiredvancomycin-resistant Enterococcus faecium: a case report from Malaysia. Jour-nal of Medical Microbiology 54, 901e903.

Ronconi, M.C., Merino, L.A., Fernández, G., 2002. Detection of Enterococcus withhigh-level aminoglycoside and glycopeptide resistance in Lactuca sativa (let-tuce). Enfermedades Infecciosas y Microbiología Clínica 20, 380e383.

Tang, Chen-Wei, Cheng, Cheng-Kuo, Lee, T.-S., 2007. Community-acquired bleb-related endophthalmitis caused by vancomycin-resistant enterococci. Cana-dian Journal of Ophthalmology 42, 477e478.

USEPA, 2002. Method 1106.1: Enterococci in Water by Membrane Filtration UsingMembrane-Enterococcus-Esculin Iron Agar (mE-EIA). EPA 821-R-02e021. U.S.Environmental Protection Agency Office of Water, Washington, DC.

Willems, R.J.L., Hanage, W.P., Bessen, D.E., Feil, E.J., 2011. Population biology of gram-positive pathogens: high-risk clones for dissemination of antibiotic resistance.FEMS Microbiology Reviews 35, 872e900.