Bacterial Pathogen Gene Abundance and Relation to Recreational Water Quality at Seven Great Lakes Beaches

Bacterial Pathogen Gene Abundance and Relation to RecreationalWater Quality at Seven Great Lakes BeachesRyan J. Oster,*,† Rasanthi U. Wijesinghe,† Sheridan K. Haack,† Lisa R. Fogarty,† Taaja R. Tucker,‡

and Stephen C. Riley§

†U.S. Geological Survey, Michigan Water Science Center, Lansing, Michigan 48911, United States‡CSS-Dynamic, 10301 Democracy Lane, Suite 300, Fairfax, Virginia 22030, United States§U.S. Geological Survey, Great Lakes Science Center, Ann Arbor, Michigan 48105, United States

*S Supporting Information

ABSTRACT: Quantitative assessment of bacterial pathogens,their geographic variability, and distribution in various matricesat Great Lakes beaches are limited. Quantitative PCR (qPCR)was used to test for genes from E. coli O157:H7 (eaeO157),shiga-toxin producing E. coli (stx2), Campylobacter jejuni(mapA), Shigella spp. (ipaH), and a Salmonella enterica-specific(SE) DNA sequence at seven Great Lakes beaches, in algae,water, and sediment. Overall, detection frequencies weremapA>stx2>ipaH>SE>eaeO157. Results were highly variableamong beaches and matrices; some correlations with environ-mental conditions were observed for mapA, stx2, and ipaHdetections. Beach seasonal mean mapA abundance in water wascorrelated with beach seasonal mean log10 E. coli concentration.At one beach, stx2 gene abundance was positively correlated with concurrent daily E. coli concentrations. Concentrationdistributions for stx2, ipaH, and mapA within algae, sediment, and water were statistically different (Non-Detect and DataAnalysis in R). Assuming 10, 50, or 100% of gene copies represented viable and presumably infective cells, a quantitativemicrobial risk assessment tool developed by Michigan State University indicated a moderate probability of illness forCampylobacter jejuni at the study beaches, especially where recreational water quality criteria were exceeded. Pathogen genequantification may be useful for beach water quality management.

■ INTRODUCTION

Fecal indicator bacteria (FIB) and pathogenic microorganismsaffect water quality at recreational beaches throughout theGreat Lakes.1−9 To prevent pathogen exposure, determiningthe influence of matrices like Cladophora glomerata (afilamentous green alga) and sediment is critical for under-standing bacterial pollution in water at recreational beaches.The U.S. Environmental Protection Agency’s (USEPA) recrea-tional water quality criteria (RWQC) rely on culturing FIB10

and concerns over culturing time and whether FIB accuratelyreflect contamination by specific pathogenic microbes are oftendiscussed.1,4,11,12 It is likely FIB-based beach advisories areposted after the highest risk for exposure has occurred.Rapid molecular methods like quantitative polymerase chain

reaction (qPCR) can significantly reduce time for processingand posting advisories.1 Quantifying specific genes frompathogenic microorganisms is an advantage to using qPCRover presence/absence (P/A) PCR, as P/A PCR can onlyestimate relative abundance. Comparing quantifiable abundan-ces of genes with land use, beach characteristics, andenvironmental variables is another benefit to using qPCR.Nevertheless, uncertainties also need to be considered with

qPCR, such as the inability to distinguish DNA from live asopposed to dead, or viable but unculturable, cells.13 Forexample, a qPCR method can be applied to monitorenterococci at recreational beaches according to USEPARWQC.10 Using this method, criteria based on qPCR-basedcalibrated cell equivalents of enterococci are 1−2 orders ofmagnitude greater than enterococci abundances based onculture.10 Determining qPCR-based abundances of bacterialpathogen genes is an important first step for understandingbeach water quality, interpreting pathogens relationships toFIB, and evaluating how to minimize the health risk forbeachgoers by using tools such as quantitative microbial riskassessment (QMRA), and for designing future studies.QMRA estimates infection probability using dose−response

models14 and may be useful for interpreting qPCR data toestimate health risk from exposure to specific pathogens. Thegoals of QMRA include investigating areas of concern,

Received: January 14, 2014Revised: November 13, 2014Accepted: November 14, 2014

Article

pubs.acs.org/est

© XXXX American Chemical Society A dx.doi.org/10.1021/es5038657 | Environ. Sci. Technol. XXXX, XXX, XXX−XXX

remediation, making regulatory decisions, or confirmingwhether regulations are being met.14 Limitations to using thismethod exist. Few studies have applied qPCR gene abundancedata to QMRA to characterize risk to bathers in recreationalareas, and as qPCR-generated gene abundances often exceedcultured cell counts,10 it remains unclear how qPCR-based geneabundances might be used in QMRA at beaches. Temporal andspatial variability demonstrated in FIB studies at beaches,environmental factors, and data gaps in hydrodynamics also canconfound QMRA interpretation.15 An online QMRA tooldeveloped by Michigan State University’s (MSU) Center forAdvancing Microbial Risk Assessment16 could potentially beused by beach managers to evaluate bacterial pathogen geneabundances from Great Lakes beach water, as an aid to beachmanagement and remediation.15

The occurrence of FIB in beach matrices throughout theGreat Lakes has been extensively studied.1,3,6,7,11,17−20 How-ever, quantification of bacterial pathogens has been limited incomparison.4,6,18,21 This study is one of the first extensivegeographic comparisons of bacterial pathogen gene abundancesin a variety of beach matrices in the Great Lakes. At sevenGreat Lakes beaches during the 2012 swimming season, and inthree environmental matrices (water, sediment, and algae), ourstudy tracked the occurrence of five genes: stx2, eaeO157, ipaH,mapA that are commonly associated with shiga-toxin 2producing E. coli, E. coli O157:H7, Shigella spp., Campylobacterjejuni, respectively, and an uncharacterized DNA sequencespecific for Salmonella enterica (hereafter SE). All of thesemicroorganisms are known to cause water-borne gastro-intestinal (GI) illness in humans.6,22 Gene abundances wererelated to RWQC at individual beaches and were evaluatedagainst a range of environmental parameters that mightinfluence results. Finally, we used qPCR gene abundance datafrom our analyses in a previously developed QMRA tool16 toestimate illness risk at beaches categorized by meeting or failingRWQC. Our study provides important information regardingthe geographic distribution of pathogen genes in the beachenvironment among various matrices, with respect to RWQC,and shows how gene abundances generated from qPCR canpotentially be applied to tools such as QMRA.

■ MATERIALS AND METHODSSampling Locations and Environmental Data. Seven

Great Lakes beaches on federal land, or in USEPA-designatedAreas of Concern (AOCs) with beach water quality impair-ments, were analyzed (Table 1). Land cover within the beachcatchment and other beach characteristics are reported in theSI. A total of 40 seasonal and 27 temporal parameters wereanalyzed with relation to pathogen gene abundance (SI TableS1). E. coli monitoring data were obtained from public beach-notification web sites for Indiana, Michigan, Ohio, and

Wisconsin. E. coli samples were collected at the same beachlocations and time range (8 am-12 pm), but not always on thesame dates, as the pathogen samples we collected. RWQC[seasonal geometric mean (GM) and statistical threshold value(STV)] were calculated using all available E. coli data for eachbeach, and were evaluated against the seasonal mean genecopies (GC) or total number of detections for each gene. Weused available E. coli data that matched our sampling dates forassessment of relationships between pathogen gene occurrenceand RWQC, except at Sleeping Bear-Esch Road (no matchingE. coli samples) and Portage Lakefront (only 2 matchingsamples).

Sample Collection and Processing. Samples werecollected one or more times per week from May to September2012, along transects (Table 1) starting at the middle of thebeach with samples taken equidistant (100−400 m) from themiddle transect. Pathogen gene analysis was typically from themiddle transect, unless it did not offer all three matrices. Usinga sterile tongue depressor, algae (predominantly Cladophora)was scooped into a sterile Whirl-pak bag, until 3/4 of the bagwas filled. Water was retrieved at waist depth halfway betweenthe water surface and the lake bottom in sterile Whirl-pak bags.Lake bottom sediment was collected in knee depth water byforcing a sterile 50 mL tube 3/4 of the way down into thesediment. Samples were stored on ice in the field and frozen(−20 °C) the same day upon return to the laboratory. Waterwas mixed thoroughly, 100 mL was filtered through a 0.45 μmMillipore Isopore filter using Sterile Millipore Microfildisposable filter cups (EMD Millipore, Billerica, MA), thenthe filters were frozen (−20 °C) immediately. Dry weights wereobtained from each sediment sample and from 10% of algaesamples (representing a range of different beaches and dates,chosen randomly).

DNA Extraction. DNA from water filters was extracted byusing an UltraClean Soil DNA Isolation Kit (MoBio Inc.,Carlsbad, CA). The filters were placed into the bead tubes andthe tubes were placed in a mini bead beater for 1 min 30 s, inplace of the horizontal shaker recommended in the kitinstructions. DNA from sediment and Cladophora wasextracted using PowerMax Soil DNA Isolation Kits (MoBioInc., Carslbad, CA). For both sediment and Cladophorasamples, the horizontal bead beating time was increased to15 min. Sediment was first homogenized in a plastic weigh dish,and 9.5−10.0 g of material was weighed directly in a bead tubeprovided with the kit. The total amount of sample extractedwas recorded and used for determining the number of genecopies per gram of dry weight (copies gdw−1) from wet:dryratios. Cladophora was processed by transferring an ∼5.0 g (wetweight) subsample to a 50 mL plastic tube. A Cladophora slurrywas made by adding 5 mL of liquid from the Cladophorasample bag, or sterile PBS, to the 50 mL tube and the sample

Table 1. Beach Names, IDs, Locations, and Transect Lengths of Beaches in This Study

beach beach IDa latitude, longitude lake state transect length (m)

Deland Park WI949936 43.758461 N, −87.702118 W Michigan WI 829.3Jeorse Park IN319633 41.64936 N, −87.43324 W Michigan IN 788.1Portage Lakefront: Indiana Dunes National Lakeshore 41.630630 N, −87.181633 W Michigan IN 375.1Sleeping Bear Dunes National Lakeshore: Esch Rd/Otter Creek MI001811 44.76207 N, −86.07634 W Michigan MI 777.3Brimley State Park MI001552 46.417309 N, −84.555489 W Superior MI 625.5Bay City Recreation Area MI000290 43.67138 N, −83.90676 W Huron MI 550.0Maumee Bay State Park OH182884 41.6858 N, −83.3781 W Erie OH 238.6aBeach IDs are associated with state Beach Guard data.

Environmental Science & Technology Article

dx.doi.org/10.1021/es5038657 | Environ. Sci. Technol. XXXX, XXX, XXX−XXXB

was placed on a horizontal vortex mixer for 30 min. Threemilliliters of the homogenized Cladophora slurry was used forDNA extraction, and dry weight was determined from 10% ofthese aliquots. Extraction blanks for water were prepared byextracting a 0.45 μm membrane filter with 100 mL of sterilePBS buffer using the UltraClean Soil DNA isolation kit (MoBioInc., Carlsbad, CA). Cladophora and sediment extraction blankswere prepared using a PowerMax Soil DNA isolation kit(MoBio Inc., Carlsbad, CA) with nothing extra added to theextraction. All DNA concentrations were read on a NanoDrop2000 UV−Vis Spectrophotometer (Thermo Scientific, Wil-mington, DE).Quantitative Polymerase Chain Reaction (qPCR).

QPCR was performed using a StepOne Plus qPCR thermalcycler (Life Technologies, Grand Island, NY). Gene fragmentswere amplified from E. coli (ATCC 35150), Shigella sonnei(ATCC 9290), Campylobacter jejuni subsp. jejuni (ATCC29428), and Salmonella enterica subsp. enterica (ATCC 14028)genomic DNA. Plasmids used for standard curves weregenerated using a TOPO TA Cloning Kit (Life Technologies,Grand Island, NY) followed by a Promega PureYield MiniPrep(Promega, Madison, WI), both according to the manufacturer’sinstructions. Plasmid DNA concentration was measured with aNanoDrop ND-1000 UV spectrophotometer, and the genecopy numbers were determined using the formula: gene copyno. = 6.02 × 1023 × (plasmid DNA concentration) × (ng/μL)× template DNA (μL)/(molecular weight of plasmid DNA+molecular weight of gene insert) × (g/mol). Plasmid molecularweight was from the TOPO 2.1 cloning vector product manual.A 5-point standard curve was included on each qPCR plate.

All samples, standards, and controls were run in triplicate in a25 μL reaction volume. Inhibition tests were performed on 10%of the extracts from each matrix by running undiluted, 1:10, and1:25 diluted extracts and comparing Ct (threshold cycle) valuesbetween the dilution factors. To assess matrix inhibition, inmost runs, a 1000 copy plasmid standard (2.5 μL) was spikedinto 2.5 μL of extract that had previously tested negative for thegene of interest (all matrix types). To test the efficiency andaccuracy of the standard curve for each assay, a controlcontaining the 1000 copy standard was run as an unknownsample. A control DNA extraction blank was also tested onseveral qPCR runs.Three qPCR assays (stx2, eaeO157, SE) used SYBR Green

chemistry and were based on previously described meth-ods.23−25 Two assays (ipaH and mapA) used TaqManchemistry and were based on previously described meth-ods.25,26 Slight modifications to the original assays includedreduced cycle number for ipaH and SE (40 cycles), andreduction (ipaH) or increase (SE) by 2 degrees in annealingtemperature, as determined to be optimal using the VeriFlexblock option on the StepOne Plus qPCR thermal cycler.Additionally, an extra final step at 81 °C for the SE assay wasused as the data acquisition step to limit the influence of aminor second peak in the final fluorescence. The second peakwas not present as a DNA product on a 2.2% agarose gel,indicating proper amplification of the desired 261 bp product.Reaction conditions and primer sequences are in SupportingInformation (SI) Table S2.All assay master mixes using SYBR Green chemistry were

comprised (per reaction) of the following: 12.5 μL of SYBRGreen PCR Master Mix (Life Technologies, Foster City, CA),0.5 μL each of 10 μM forward and reverse primer, 0.5 μL 10μg/μL BSA, 6.0 μL nuclease free water, and 5 μL of DNA

template. A melt curve was produced in each SYBR Green assayto verify the specificity of the primer set, and 2.2% agarose gels(Lonza Group Ltd., Switzerland) were run on random samplesto determine that the correct product size was produced for therespective assays.Assay master mix for the TaqMan assays was comprised of

the following: 12.5 μL of TaqMan Environmental Master Mix(Life Technologies, Foster City, CA), 0.5 μL each of 10 μMforward and reverse primer, 0.5 μL of 10 μg/μL BSA, 5 μL ofDNA template, and nuclease-free water (5.75 μL for mapA; 5.0μL for ipaH). The final probe concentration for mapA was 100nM (0.25 μL of 10 μM probe) and 160 nM (1.0 μL of 4 μMprobe) for the ipaH assay. To verify expected amplicon basepair (bp) product size (SI Table S2), 2.2% agarose gels (LonzaGroup Ltd., Switzerland) were run using random amplifiedDNA from qPCR reactions.Limits of detection (LOD) and limits of quantification

(LOQ) were calculated for each assay. The limit of detection(LOD) was calculated based on the Ct values for the compilednontemplate controls (NTC). The 95% confidence value forthe NTC average when used in the compiled standard curveequation, generated LODs of less than the theoretical limit ofdetection11 of 3 copies μL−1. Thus, the LOD was conservativelyset at the theoretical limit, resulting in LODs of 15 copies per 5μL reaction. The lowest standard concentration where 95% ofthe compiled values were detectable was considered the LOQ95.The LOQ95 for each of the assays in this study was 50 copiesper 5 μL reaction. Individual standard curves were used tocalculate GC. Four to 10 copies of the ipaH gene occur onchromosomal and plasmid DNA,27 thus, for comparison withresults for other genes, and for use in QMRA, all ipaH copynumbers were adjusted assuming 5 GC per cell.28 For all watersamples the LOD and LOQ were, respectively, 150 and 500GC 100 mL−1 for algae, 3 × 103 and 1 × 104 GC mL−1 and forsediment, 1231 GC and 4105 GC gdw−1. Data were classifiedas quantifiable (Q), detectable but not quantifiable (DNQ),and nondetectable (ND). A sample was classified as Q if two ormore replicates were above the LOQ, DNQ if two or morereplicates were between the LOD and LOQ, and ND if two ormore replicates were below the LOD. Acceptable individualstandard curves had efficiencies from 85% to 115% and R2

above 0.95. Samples were rerun if standard curves did not meetthese criteria.

Data Reporting. We report GC per 100 mL beach waterbecause this is the convention for reporting RWQC based oncultured FIB or as numbers of calibrated cell equivalents forenterococci (as derived by qPCR using EPA Method 1611).10

FIB, pathogen, and gene densities in Great Lakes Cladophorasamples have been reported per mL, or per g wet or dry weight.Lacking a convention, we chose “per ml” so that Cladophoraresults could be compared to water results more directly. Theaverage dry weight of the algal slurry was 3.8 gdw 100 mL−1

thus our results can also be compared to those of prior studiesusing a dry weight convention.

Statistical Analyses. Due to the high percentage of leftcensored data (data below the LOD and LOQ) the Non-Detects and Data Analysis (NADA) statistical package in R29

was used to determine differences in gene distribution amongthe three matrices. This analysis uses nonparametric Kaplan−Meier methods to generate empirical cumulative distributionfunctions (ECDFs) and estimate summary statistics.30,31 Thecenstats function estimated means and standard deviations forthe matrices of each gene using three models: Kaplan−Meier,

Environmental Science & Technology Article

dx.doi.org/10.1021/es5038657 | Environ. Sci. Technol. XXXX, XXX, XXX−XXXC

regression on order statistics (ROS), and maximum likelihood.Boxplots were generated in NADA, using the default ROS,considered better at estimating summary statistics andmodeling distributions of censored data.30 The cendif f functiontested the null hypothesis that the ECDF distribution amongthe three matrices was identical and uses the Peto−Peto test,which is appropriate for left-censored log-normal data sets oftenobserved in environmental data.30

Statistical analyses were also performed using SigmaPlot 12.0(Systat Software Inc., San Jose, CA), using log10 transformedgene abundances. The Mann−Whitney U statistic was used totest for gene abundance differences among matrices inSigmaPlot. Pearson correlation and linear regression wereused to determine the relationship between seasonal mean ordaily pathogen gene abundances, seasonal mean E. coliabundance, and relation to seasonal or daily environmentalcharacteristics. The seasonal and daily environmental parame-ters evaluated are described in SI Table S1. In all cases, resultswere considered statistically significant when p < 0.05.Quantitative Microbial Risk Assessment (QMRA). A

QMRA tool developed by MSU’s Center for AdvancingMicrobial Risk Assessment16 was used to further assess qPCRgene abundance data to estimate infection risk based onestablished dose−response models for Shigella, Salmonella, andCampylobacter jejuni. The following model parameters weredescribed in the tool:16 Campylobacter jejuni was a beta-Poisson

best fit model with an LD50/ID50 of 8.9 × 102 required forinfection and the optimized parameters were α = 1.44 × 10−1

and N50 = 8.9 × 102,32 for Salmonella enterica a beta-Poissonmodel based on Salmonella Typhi with parameters α = 1.75 ×10−1 and LD50/ID50 = 1.11 × 106 was selected,33 a beta-Poisson model with parameters α = 2.65 × 10−1 and N50=1.48× 103 was selected for Shigella.34 Humans were the host forinfection in all three models. An assessment was not completedfor E. coli (stx2) or E. coli O157:H7 because the tool did nothave models for enterohemorrhagic E. coli specific to humanhosts. Assessments were based on the seasonal average GCmL−1 (using 1/2 LOD or LOQ for ND or DNQ samples) forwater samples at each beach location. Doses were calculated bymultiplying the GC ml−1 assuming 16 mL consumed by adultsand 37 mL by children in 45 min.35 We assumed that each genecopy represented one cell except for ipaH where 5 GC or morecan be present; however, as QPCR does not distinguishbetween viable and nonviable organisms or infectiousorganisms, dose response models were tested with dose rangesassuming 10%, 50%, or 100% of the mean GCs were fromviable and infectious organisms.After obtaining infection probability from the QMRA tool, an

illness probability was calculated by multiplying the infectionrisk by a morbidity factor (% of infections resulting in illness)for each organism.36 Shigella has an approximate morbidity of15%,37 Campylobacter jejuni 28%,36 and Salmonella enterica

Table 2. Detection Frequencies (%) of Pathogen Genes in Algae, Water, and Sediment at Individual Beachesa

beach location matrix n stx2 eaeO157 ipaH mapA Salmonella enterica

Deland Park algaeb 9 11.1 0.0 0.0 11.1 0.0water 14 57.1 0.0 0.0 14.3 0.0sediment 16 12.5 0.0 0.0 18.8 0.0

Jeorse Park algae 11 45.5 36.4 0.0 18.2 0.0water 13 0.0 15.4 0.0 100.0 0.0sediment 15/14ac 6.7 26.7 0.0 93.3 7.1a

Portage Lakefront algae 7 0.0 0.0 14.3 100.0 0.0water 10 0.0 0.0 90.0 100.0 10.0sediment 10 0.0 50.0 20.0 100.0 0.0

Sleeping Bear Dunes-Esch Rd./Otter Creek algae 7/6b 0.0 0.0 42.9 50.0b 0.0water 5 100.0 20.0 0.0 40.0 0.0sediment 10 0.0 0.0 0.0 50.0 0.0

Brimley State Park algae 3 0.0 0.0 0.0 0.0 0.0water 12 100.0 25.0 0.0 91.7 0.0sediment 14 14.3 21.4 0.0 0.0 7.1

Bay City Recreation Area algae 11/10c 90.9 0.0c 36.4 50.0c 63.6water 8 100.0 0.0 87.5 50.0 62.5sediment 9 0.0 0.0 66.7 88.9 0.0

Maumee Bay State Park algae 11 0.0 0.0 36.4 90.9 0.0water 13/12d 0.0 8.3d 92.3 83.3d 0.0sediment 12 0.0 8.3 41.7 16.7 8.3

total algae 59/58e/57f 27.1 6.9e 20.3 49.1f 11.9water 75/74g 44.0 9.5g 37.3 70.3g 8.0sediment 86/85h 5.8 15.1 15.1 48.8 35.h

aFor some genes in matrices from specific beaches, samples may not have been run due to the lack of DNA extract. bAlgae = Unidentified algae orCladophora. cSuperscript letters correspond to total samples (n) for matrices at beaches where total samples analyzed varied from others.

Environmental Science & Technology Article

dx.doi.org/10.1021/es5038657 | Environ. Sci. Technol. XXXX, XXX, XXX−XXXD

20%.36 An illness risk threshold of 0.03 was used to assess thesignificance of our results from the QMRA tool.15,36 Beacheswere categorized into three groups based on the RWQC: ingroup A both the statistical threshold value (STV) andgeometric mean (GM) RWQC were exceeded; in group Bbeaches either the STV or GM were exceeded; and in group Cbeaches the RWQC were not exceeded.

■ RESULTSQuality Assurance. Data for individual and compiled

standard curves from all qPCR runs of acceptable quality are inSI Table S3. Quantitative data for all analyses can be found inSI Tables S4−S7. For all assays the compiled NTCs andextraction blanks had Ct values between 38 and 40. Sampleswere rerun if average Ct values for NTCs or blanks were <38for any run. All reactions were performed with extracts at 1:10dilution, which limited inhibition optimally (as determined by>2 Ct difference from undiluted samples) and all subsequentcalculations accounted for this dilution factor. Matrix spikesgenerally did not deviate more than 10% from the expectedvalue (500 GC) in any of the assays and GCs were neveradjusted. The deviation could be due to both pipetting errorand slight inhibition.Detection of Pathogen Genes in Cladophora, Sedi-

ment, and Water. The detection of pathogen genes(quantifiable and DNQ results) among the three matriceswas variable and highly beach specific (Table 2). Two to fourgenes were normally detected at each beach, but the gene typevaried and no beach had all genes present. Overall, there was atendency for greater detection frequencies of stx2, ipaH, andmapA in water samples than in Cladophora and sediment, butthis pattern did not necessarily hold at individual beaches, andthere was no trend for the eaeO157 or SE genes. The lower LODfor water samples than for sediment and algae could influence

detection frequencies, as could the amounts of materialextracted, and methods used. Stx2 was not as frequentlydetected within the sediments as in Cladophora, and lake waterhad the greatest detection frequency (Table 1). EaeO157 was theleast commonly detected gene among all of the beachesstudied. IpaH was detected at four of seven beaches, was onlydetected in the water at beaches where it was also detected inthe sediment and Cladophora, and detection frequency of ipaHin water and sediment was positively correlated (R = 0.89; p =0.007; n = 7; df = 6). Of the five pathogen genes, mapAoccurred most frequently in all matrices. While there was nostatistical correlation between sediment and water concen-trations for the mapA gene, the gene was often detected insediment and water samples at Jeorse Park, Portage Lakefront,and Bay City at the same time (SI Tables S4−S7).

Quantification of Pathogen Genes in Cladophora,Sediment and Water. Figure 1 shows NADA-determinedgene copy number distributions across all beaches for each geneand matrix. Except in a few cases, NADA-predicted means weresimilar to the means generated using only quantifiable samples(provided for reference in SI Figure S2). NADA analysis basedon ECDFs estimated the mean stx2 abundance to be 5.4 log10GC ml−1 in algae, 4.7 log10 GC ml−1 in sediment, and 4.3 log10GC 100 mL−1 in water. A Chi-square test comparing the ECDFdata distributions between the matrices showed a significantdifference (p < 0.001; Figure 1A). NADA statistics based onECDFs estimated the mean eaeO157 abundance to be 4.7 log10GC ml−1 algae, 4.1 log10 GC gdw−1 sediment, and 3.7 log10 GC100 mL−1 of water. However, only one water sample hadquantifiable GC of eaeO157 and no significant difference in thedata distribution among the matrices was observed (Chi-square; p = 0.4; Figure 1B). The NADA-predicted ipaH meanbased on the ECDFs was 3.9 log10 GC ml−1 for algae, 4.0 log10GC gdw−1 sediment, and 3.2 log10 GC 100 mL−1 water, and the

Figure 1. Comparison of log10 pathogen gene copy data distributions using NADA statistical analysis in R between Cladophora, sediment, and watermatrices. Water is expressed as log10 GC 100 mL−1, Cladophora as log10 GC mL−1, and sediment as log10 GC per gram dry weight (gdw−1). Panel Ashows stx2 abundance, panel B eaeO157 abundance, panel C ipaH abundance, and panel D mapA abundance. The horizontal black line in the box isthe median value. ALG = algae (predominantly Cladophora), SED = sediment, WAT = water.

Environmental Science & Technology Article

dx.doi.org/10.1021/es5038657 | Environ. Sci. Technol. XXXX, XXX, XXX−XXXE

Chi-square test showed a moderate significant differencebetween their data distributions (p = 0.07; Figure 1C). Themean mapA concentrations for each matrix determined byNADA were 5.4 log10 GC ml−1 algae, 4.6 log10 GC gdw−1

sediment, and 5.3 log10 GC 100 mL−1 water, and a significantdifference between their data distributions was observed (Chi-square; p = 0.03; Figure 1D). Bay City Recreational Area wasthe only beach that had quantifiable concentrations of SE in anyof the matrices; the detections at the other beaches were allDNQ. The NADA-predicted SE mean was 4.5 log10 GC mL−1

of algae and was 2.9 log10 GC 100 mL−1 in water. Mean SEabundance for sediment could not be determined because allvalues were censored data.Gene Concentrations and Temporal Variability at

Individual Beaches. There was a high degree of beach-specific temporal variability within each environmental matrix,especially in Cladophora and sediment (SI Tables S4−S7). Atbeaches where genes were commonly detected throughout thesampling season, their concentrations varied from non-detectable to quantifiable on a weekly basis. At Jeorse Park,where samples were collected daily from July 30 to August 2,2012 and on August 14 and 15, and 17, and 18, 2012, somegene abundances varied between ND and fully quantifiable overthe course of 24 h. However, other genes were consistentlydetected or nondetected over multiple successive days at this

beach. Between 15 and 100% of the total detections were DNQdepending on the gene and matrix (SI Tables S4−S7).

Associations with Environmental Conditions. Variableenvironmental conditions at these beaches likely play a role ingene detections or concentrations. We observed somecorrelations between environmental variables such as waterlevel, wind speed, and temperature, and detection frequenciesor concentrations of the mapA, stx2, and ipaH genes (Table 3).Notably, those are the genes that were detected mostfrequently, suggesting that the highly variable and relativelyinfrequent detection of genes in our study limited the ability tofully evaluate the influence of environmental variables, and thatmore intensive sampling regimes would be needed in futurestudies.

Fecal Indicator Bacteria and Gene Relationships inWater. At Brimley State Park, daily log10 E. coli (stx2)abundance in water was correlated to the concurrent daily log10E. coli most probable number (MPN) (R = 0.69; p = 0.059; n =8; df = 7). No other beaches in this study showed this sametemporal relationship with any other gene. At two of thebeaches (Portage Lakefront and Sleeping Bear-Esch Road) toofew daily E. coli values were collected on the same day as thepathogen samples to make this comparison.Seasonally, mapA occurred in higher abundance at beaches

where E. coli abundance was greater. The seasonal mean log10

Table 3. Correlations Among Gene Detection Frequencies or Concentrations and Environmental Variablesa

gene matrix variable Pearson R probability (p) number (n)

stx2seasonal detection frequency water no. point sources on beach 0.78 0.038 7daily log10 GC water mean air temp (24 h) −0.63 0.001 74daily log10 GC sediment depth averaged water velocity 0.36 0.002 72daily log10 GC Cladophora antecedent 24 h wind speed 0.35 0.01 51daily log10 GC Cladophora water level −0.33 0.01 57mapAseasonal detection frequency sediment seasonal mean wind speed at beach 0.86 0.014 7daily log10 GC sediment daily magnitude of water velocity −0.37 0.001 73daily log10 GC Cladophora water level 0.30 0.02 56ipaHdaily log10 GC sediment daily magnitude of water velocity −0.26 0.03 74daily log10 GC Cladophora antecedent 24 h wind speed 0.37 0.008 51

aDefinitions and sources of environmental data in SI Table S1.

Table 4. Probability of Illness for Adults Assuming 10%, 50%, and 100% Bacterial Cell Viability and Infectivity for PathogenGenes

Shigella spp. (ipaH) Salmonella enterica Compylobacter jejuni (mapA)

beach group type 10%a 50%a 100%a 10% 50% 100% 10% 50% 100%

Ad 0.000b 0.002 0.004 0.000 0.000 0.000 0.163 0.187 0.196Be 0.000 0.002 0.004 0.000 0.000 0.000 0.055 0.097 0.114B 0.006 0.023 0.036c 0.000 0.000 0.000 0.059 0.100 0.116Cf 0.016 0.044 0.058 0.000 0.000 0.000 0.060 0.102 0.118C 0.030 0.062 0.076 0.000 0.000 0.001 0.014 0.038 0.055C 0.000 0.002 0.004 0.000 0.000 0.000 0.010 0.028 0.043C 0.000 0.002 0.004 0.000 0.000 0.000 0.010 0.037 0.052

aAssuming 10%, 50%, and 100% of the gene copies are from viable and infectious cells. bIllness probability. cBold numbers exceed a risk concernthreshold of 0.03, as recommended by Ashbolt et al. 2010.15 dBeach group type A; both seasonal GM and STV exceeded recreational water qualitycriteria. eBeach group type B; either seasonal GM or STV exceeded recreational water quality criteria. fBeach group type C; seasonal GM or STV didnot exceed recreational water quality criteria. gMean doses were calculated by multiplying the GC ml−1 by 16 mL (estimated volume of wateringested by adults),35 multiplied by the assumed doses of viable and infectious cells (10%, 50%, 100%) and these doses were evaluated with theirrespective dose−response models to give the probability of infection. Infection probabilities were then multiplied by morbidity factors to estimatethe probability of illness.

Environmental Science & Technology Article

dx.doi.org/10.1021/es5038657 | Environ. Sci. Technol. XXXX, XXX, XXX−XXXF

mapA GC 100 mL−1 in water was positively related to theseasonal mean E. coli log10 MPN 100 mL−1 (R = 0.88; p =0.009; n = 7; df = 6) across the beaches. MapA abundance wasalso related to the RWQC; the concentration in water waspositively correlated with the percentage of water samples thatexceeded the BAV (R = 0.81; p = 0.028; n = 7; df = 6) andthose that exceeded the STV (R = 0.78; p = 0.037; n = 7, df =6). However, BAV and STV relationships were driven mostlyby the Jeorse Park beach where seasonally, 69.8% of thesamples exceeded the BAV and 61.9% exceeded the STV. Noother pathogen gene in this study was related to the seasonal E.coli concentrations or RWQC in water.Application of qPCR Data to Risk Assessment Tools.

Using the online QMRA tool, we analyzed illness probabilitiesassuming first adult (16 mL) and then child (37 mL)consumption of water for Shigella spp., Salmonella enterica,and Campylobacter jejuni in the context of RWQC. Wecalculated illness probabilities assuming 10, 50, and 100% ofthe mean GC were from infective cells and assumed Shigellahad 5 ipaH GC per cell. Assuming adult water consumption,two beaches that had no exceedances of RWQC had illnessprobabilities for both Shigella and Campylobacter jejuni thatexceeded the recommended threshold of 0.03.15,36 Campylo-bacter illness probabilities were often above the 0.03 bench-mark, due to high GC and a low infectious dose for C. jejuni,even if only 10% of the GC were assumed to be from infectiveviable cells, but only at one beach where RWQC were met(Table 4). Surprisingly, the illness probability for Shigella wasabove the benchmark at two beaches that met RWQC when10% or 50% of the GC were assumed to represent infectivecells. QMRA results for water consumption by children (37mL) were similar to those of adults. However, illnessprobabilities were slightly greater for children, and threebeaches had illness probabilities for Shigella greater than 0.03,assuming 10% of the GC came from infectious cells (SI TableS8). SE was rarely detectable in this study and QMRA resultsindicated low probability of illness induced by this organism(Table 4). Cabral et al. (2010)22 showed a 50% concentrationreduction for E. coli in 1.5−3 days, 0.1−0.67 days forSalmonella, and 1 day for Shigella in surface water, thus thedoses calculated are plausible for potentially viable organismsdetected at the beaches.

■ DISCUSSIONThe detection frequency and abundance of the pathogen genesdiffered in the beaches studied and detection frequency in algaeand sediment was typically less than in water, but higherabundances were often observed in the algae and sediment. ThemapA gene was widespread throughout the Great Lakes, stx2and ipaH were moderately detected, and eaeO157 and Salmonellawere infrequently detected. While the sources of these genesare unknown, source likely plays a key role for pathogendetection in the beach environment. For example, there areseveral sources for stx2,23,38−40 and sources other than thebeach matrices studied likely have an influence on theoccurrence of this gene in water. While the average geneabundances indicated that water typically harbored fewer of thegenes studied, there were three beaches where the average stx2abundance in water was greater than in sediment (Bay CityRecreation Area, Brimley State Park, and Deland Park Beach).The Deland Park Beach catchment was 98.1% urban, but theBay City catchment was dominantly agricultural (36.4%) andforest (47%), and the Brimley State Park catchment was

primarily wetland (34%) land use. All three of these beaches arelocated in close proximity to major rivers or ditches and couldbe influenced by land use in the beach catchment. With arelatively high percentage of urbanization and the variety ofsources for stx2 in urban, agricultural or wetland environments,it may not be surprising that the water had greaterconcentrations at these locations.A previous study at Portage Lakefront, showed 100%

detections of Campylobacter spp. in Cladophora in August and60% detection in September of 2006 (using MPN-PCR with16SrDNA gene for the genus), but none in August using qPCRof the same gene (attributed to inhibition of the assay).6 Ourstudy showed 100% detection of mapA in all matrices tested atPortage Lakefront. The relatively widespread occurrence ofmapA may suggest the sources of this gene are common amongmost beaches. The detection of SE was highly beach specificwith most of the detections occurring at Bay City RecreationalArea. Similarly, the ipaH gene, which is suggestive of a humansource, was frequently detected in the water, sediment, andCladophora at Maumee Bay, Bay City, and Portage Lakefront.All three of these beaches have a large percentage of urban and/or agricultural land in their watersheds.Similar to the pathogen genes we studied, others have shown

that FIB are significantly more abundant in Cladophora andsediment than water.9,17,41−49 The ultimate source(s) of FIBand pathogens in Cladophora, sediment, and water remainsunknown, however, a dynamic relationship may exist betweenthem. Cladophora and sediment may act as a sink and sourcefor not only FIB, but bacteria containing pathogen-specificgenes in the Great Lakes6,17,21,41,43,44,48,49 and contact withthese materials may increase exposure risk.50 Because of thegreater abundance of genes in Cladophora compared to thewater, Cladophora or algae could be trapping,21 protecting,6

and/or supporting the growth of attached enteric bacte-ria.9,21,42,47 In a laboratory environment Salmonella persistedin Cladophora mats for up to 10 days, Shigella 2 days, and E. coliup to 45 days.42 A study at a harbor in Lake Superior showedthat both sediment and sand can act as a genetic reservoir forFIB with the ability to load the surrounding water column.41

Environmental factors like wind speed, wave height, andrunoff could be responsible for suspension of sediment orCladophora and may cause temporary gene concentrationincreases in the water.41,51 Lower abundance in the watercolumn could be due to dilution and/or water could beinoculating the algae and sediment. One might expect decreasesin sediment gene concentrations and increases in the watercolumn when wind and water velocities are greater. Wind andwave action can result in free floating algae, suspendedsediment, and bacterial loading to surrounding water.9,42,44

Temperatures during the summer can result in higherconcentrations of E. coli in algae.9,52 It is unclear from ourstudy whether microbes containing these genes are growing inthe sediment and Cladophora or if DNA from nonviableorganisms is being detected. Either way, any detection indicatesan organism carrying the gene was growing in or delivered tothe system. We observed some correlations between environ-mental variables and gene detection frequencies or abundances.Although our results are constrained by relatively fewdetections of most genes, our data suggest that influences ofenvironmental variables on pathogen gene occurrence andabundance warrants further study. More intensive studies onmaterials transport and the association of pathogens on

Environmental Science & Technology Article

dx.doi.org/10.1021/es5038657 | Environ. Sci. Technol. XXXX, XXX, XXX−XXXG

sediment and Cladophora are needed to better understand theinterconnectedness between algae, sediment, and water.We observed a high degree of temporal and geographic

variability in pathogen gene detections. Variability has beendemonstrated previously in water and Cladophora matrices.4,6,21

Temporal and spatial variability of stx2 was exhibited in beachand tributary water at Presque Isle State Park on Lake Eriewhere beach water was an unstable reservoir for stx2.4

Variability in gene detection could also be due to weather orenvironmental factors such as runoff, temperature fluctuations,and solar radiation.6 A recent study showed FIB concentrations,microbial source tracking (MST) markers, and bacterialpathogen genes varied regionally and certain beach catchmentvariables (e.g., urban land cover, impervious surface, nearbyrivers, and the presence of drains or creeks, etc.) influencedtheir occurrence at local Great Lakes beaches.20 Environmentalconditions at beaches vary widely by day over the summer andcould affect the detection and abundance of FIB and pathogengenes within the matrices on individual days.Few relationships existed between FIB and seasonal or

temporal pathogen gene abundances. While our resultsindicated a relationship between daily E. coli and stx2abundances at one beach, another study on Lake Eriedetermined that the stx2 concentration was not correlatedwith the total E. coli abundance.4 We also determined thatseasonal E. coli abundances were positively related to seasonalmapA abundances at the seven beaches. Average mapA GCwere greater at locations where the GM, STV, and the BAVwere exceeded more often. The source and mode of delivery ofE. coli to a beach may explain why genes at individual beacheshave a relationship to FIB. Previous studies have shown thatstorm drains, combined sewage overflows, and stormwaterdischarge contribute to the presence of FIB and entericpathogens in water that influence beaches,7,20,53,54 and BrimleyState Park had the greatest number of storm drains and creeksthat flowed directly over the beach. Birds, specifically gulls, arelikely an important ecological link for the relationship betweenCampylobacter and E. coli. Gulls are common among thebeaches and are a significant source of FIB.55 FIB abundanceand Campylobacter occurrence in beach waters increases whengulls are present.56−58 Microbial source tracking (MST) wouldbe useful at beaches where the dominant sources are unknownto help determine where pathogen genes are originating.Our quantitative approach has shown that sediments and

Cladophora likely have a negative effect on water quality atrecreational beaches around the Great Lakes. Concern whetherFIB monitoring is adequately protecting the public frompathogens1,4,11,12 is one reason why QMRA may be beneficialfor beach management. C. jejuni is a leading cause of diarrhea inhumans and C. jejuni infections can be 2−7 times morecommon than infections due to Shigella or E. coli O157:H7.59

Our analysis shows Campylobacter jejuni is more likely tonegatively influence water quality and human health at ourstudy beaches compared to the other modeled pathogens.Beach managers could monitor for both FIB and pathogensusing qPCR to see if any correlations exist; this informationwould make managers more confident about what risk is posedwhen FIB are high. Without epidemiological information in thisstudy, it is difficult to assess the accuracy of this QMRA tool.Regardless, this tool is useful from a management standpointbecause pathogens such as Shigella and Campylobacter havespecific sources: humans are the common source for Shigella6,22

and birds and ruminants for Campylobacter.6,56,60 Knowing this

information may lead to better beach management strategiesthat aim to reduce sources of bacterial pathogens at recreationalbeaches.15 Management strategies might include bird control,sewage infrastructure improvements, reduction of storm drainoutfalls on beaches, understanding agricultural impacts fromnonpoint sources, or classifying beaches as nonrecreational incases where improvements cannot be feasibly made.QMRA tools are useful to make generalized assumptions

about the risk microorganisms may pose to humans. However,large sources of variability and other confounding factors mustbe considered when interpreting the results. Temporal andspatial variation of pathogen genes we observed would beimportant to consider when using QMRA to make manage-ment decisions.14,15 Due to the variation in gene detectionamong matrices, risk to human health at specific beachlocations and times is only an estimate.36 In addition,standardized methods for amounts of materials to be analyzed,extraction methods, and reporting conventions will be requiredto move ahead with qPCR data. QMRA suffers from the lack ofepidemiological studies that describe the fraction of infectiouspathogens in ambient microbial populations.15,36 QPCR alsoposes constraints, as analysis does not distinguish DNA fromviable or nonviable cells.6,11,12,41,61 PMA DNA binding dyecould be used in PMA-qPCR to better characterize the genesthat are from live cells.13,18,41,62 Even if the cells are viable, theymay not be infective,15 and PCR inhibition can reduceestimated GC. Our study used 10, 50, and 100% viabilityestimates, but it is difficult to know at this time if these areappropriate ranges. Ultimately, more needs to be learned aboutthe types (genotypic and phenotypic), viability, and infectivityof bacterial pathogens in recreational water, and the relation-ships between qPCR-based pathogen gene assessments and riskfrom actual pathogens.Additionally, filling research gaps in the hydrodynamics of

the beach system and the source and fate of fecal matter wouldassist in obtaining accurate risk assessments.15,36 As our studydetected pathogen genes intermittently at most beaches,intensive sampling (collecting samples multiple times perweek) throughout the recreational year at specific beaches maybe needed to establish stronger associations between geneabundance and environmental parameters. Concurrentlycollecting FIB and pathogen-specific gene abundances wouldbe an important step in fully assessing water quality degradationat beaches and the factors contributing to their occurrence.Quantitative bacterial pathogen data has been lacking in manybeach studies, and our study only begins to address pathogengene abundance at beaches. In future studies, assessingquantitative pathogen data along with physical and environ-mental factors will help researchers, beach managers, and publichealth officials better understand what influences human healthin the beach environment.

■ ASSOCIATED CONTENT

*S Supporting InformationAdditional text and tables as mentioned in the text. Thismaterial is available free of charge via the Internet at http://pubs.acs.org.

■ AUTHOR INFORMATION

Corresponding Author*Phone: 218-290-0944; e-mail: [email protected].

Environmental Science & Technology Article

dx.doi.org/10.1021/es5038657 | Environ. Sci. Technol. XXXX, XXX, XXX−XXXH

NotesThe authors declare no competing financial interest.

■ ACKNOWLEDGMENTS

This work was funded by the Great Lakes Restoration Initiative.The authors would like to acknowledge the beach monitorswho collected samples. We thank Joe Duris and Erin Stelzer fordata and manuscript reviews. We express gratitude to LoriFuller for assisting with GIS and to Alex Totten and HeatherJohnson for their assistance in the laboratory. Any use of trade,firm, or product names is for descriptive purposes only anddoes not imply endorsement by the U.S. Government.

■ REFERENCES(1) Nevers, M. B.; Byappanahalli, M. N.; Whitman, R. L. Choices inrecreational water quality monitoring: New opportunities and healthrisk trade-offs. Environ. Sci. Technol. 2013, 47, 3073−3081.(2) Nevers, M. B.; Whitman, R. L. Beach monitoring criteria:Reading the fine print. Environ. Sci. Technol. 2011, 45, 10315−10321.(3) Vanden Heuvel, A.; McDermott, C.; Pillsbury, R.; Sandrin, T.;Kinzelman, J.; Ferguson, J.; Sadowsky, M.; Byappanahalli, M.;Whitman, R.; Kleinheinz, G. T. The green alga, Cladophora, promotesEscherichia coli growth and contamination of recreational waters inLake Michigan. J. Environ. Qual. 2010, 39, 333−344.(4) Smith, C. J.; Olszewski, A. M.; Mauro, S. A. Correlation of shigatoxin gene frequency with commonly used microbial indicators ofrecreational water quality. Appl. Environ. Microbiol. 2009, 75 (2), 316−321.(5) Wade, T. J.; Calderon, R. L.; Sams, E.; Beach, M.; Brenner, K. P.;Williams, A. H.; Dufour, A. P. Rapidly measured indicators ofrecreational water quality are predictive of swimming-associatedgastrointestinal illness. Environ. Health Perspectives 2006, 114 (1),24−28.(6) Ishii, S.; Yan, T.; Shively, D. A.; Byappanahalli, M. N.; Whitman,R. L.; Sadowsky, M. J. Cladophora (Chlorophyta) spp. harbor humanbacterial pathogens in nearshore water of Lake Michigan. Appl.Environ. Mircobiol. 2006, 72 (7), 4545−4553.(7) Haack, S. K.; Fogarty, L. R.; Wright, C. Escherichia coli andEnterococci at beaches in the Grand Traverse Bay, Lake Michigan:Sources, characteristics, and environmental pathways. Environ. Sci.Technol. 2003, 37, 3275−3282.(8) Lauber, C. L.; Glatzer, L.; Sinsabaugh, R. L. Prevalence ofpathogenic Escherichia coli in recreational waters. J. Great Lakes Res.2003, 29 (2), 301−306.(9) Whitman, R. L.; Shively, D. A.; Pawlik, H.; Nevers, M. B.;Byappanahalli, M. N. Occurrence of Escherichia coli and enterococci inCladophora (Chlorophyta) in nearshore water and beach sand of LakeMichigan. Appl. Environ. Microbiol. 2003, 69 (8), 4714−4719.(10) U.S. EPA. Recreational Water Quality Criteria, 820-F-12- 058;U.S. EPA Office of Water: Washington, DC, 2012.(11) Whitman, R. L.; Ge, Z.; Nevers, M. B.; Boehm, A. B.; Chern, E.C.; Haugland, R. A.; Lukasik, A. M.; Molina, M.; Przybyla-Kelley, K.;Shively, D. A.; White, E. M.; Zepp, R. G.; Byappanahalli, M. N.Relationship and variation of qPCR and culturable Enterococciestimates in ambient surface waters are predictable. Environ. Sci.Technol. 2010, 44, 5049−5054.(12) Francy, D. S.; Bushon, R. N.; Brady, A. M. G.; Kephhart, C. M.Comparing rapid and culture indicator bacteria methods at inland lakebeaches. Lake Reservoir Manage. 2013, 29, 99−102.(13) Bae, S.; Wuertz, S. Rapid decay of host-specific fecalBacteroidales cells in seawater as measured by quantitative PCR withpropidium monoazide. Water Res. 2009, 43, 4850−4859.(14) Interagency Microbiological Risk Assessment Guideline Work-group. Microbial Risk Assessment Guideline: Pathogenic Microorganismswith Focus on Food and Water, 100-J-12-001; U.S. EPA, 2012.(15) Ashbolt, N. J.; Schoen, M. E.; Soller, J. A.; Roser, D. J. Predictingpathogen risks to aid beach management: The real value of

quantitative microbial risk assessment (QMRA). Water Res. 2010,44, 4692−4703.(16) MSU. Michigan State University, Center for AdvancingMicrobial Risk Assessment, QMRAwiki, 2013. http://qmrawiki.msu.edu/index.php?title=Dose-Response_Calculator.(17) Ishii, S.; Hansen, D. L.; Hicks, R. E.; Sadowsky, M. J. Beach sandand sediments are temporal sinks and sources of Escherichia coli inLake Superior. Environ. Sci. Technol. 2007, 41, 2203−2209.(18) Eichmiller, J. J. The distribution and persistence of geneticmarkers on fecal pollution on Lake Superior Beaches. Ph.D.Dissertation, University of Minnesota: Minneapolis, MN, 2013.(19) Bauer, L.; Alm, E. Escherichia coli toxin and attachment genes insand at Great Lakes recreational beaches. J. Great Lakes Res. 2012, 38,129−133.(20) Haack, S. K.; Fogarty, L. R.; Stelzer, E. A.; Fuller, L. M.;Brennan, A. K.; Isaacs, N. M.; Johnson, H. E. Geographic settinginfluences Great Lakes beach microbiological water quality. Environ.Sci. Technol. 2013, 47, 12054−12063.(21) Byappanahalli, M. N.; Sawdey, R.; Ishii, S.; Shively, D. A.;Ferguson, J. A.; Whitman, R. L.; Sadowsky, M. J. Seasonal stability ofCladophora-associated Salmonella in Lake Michigan watersheds. WaterRes. 2009, 43, 806−814.(22) Cabral, J. P. S. Water microbiology. Bacterial pathogens andwater. Int. J. Environ. Res. Public Health 2010, 7, 3657−3703.(23) Beutin, L.; Kruger, U.; Krause, G.; Miko, A.; Martin, A.; Strauch,E. Evaluation of major types of shiga toxin 2e-producing Escherichiacoli bacteria present in food, pigs, and the environment as potentialpathogens for humans. Appl. Environ. Microbiol. 2008, 74 (15), 4806−4816.(24) Ibekwe, A. M.; Watt, P. M.; Grieve, C. M.; Sharma, V. K.; Lyons,S. R. Multiplex fluorogenic real-time PCR for detection andquantification of Escherichia coli O157:H7 in dairy wastewaterwetlands. Appl. Environ. Microbiol. 2002, 68 (10), 4853−4862.(25) Wang, L.; Li, Y.; Mustapha, A. Rapid and simultaneousquantitation of Escherichia coli O157:H7, Salmonella, and Shigella inground beef by multiplex real-time PCR and immunomagneticseparation. J. Food Protection 2007, 70 (6), 1366−1372.(26) Best, E. L.; Powell, E. J.; Swift, C.; Grant, K. A.; Frost, J. A.Applicability of a rapid duplex real-time PCR assay for speciation ofCampylobacter jejuni and Campylobacter coli directly from cultureplates. FEMS Microbiol. Letters 2003, 229, 237−241.(27) Thong, K. L.; Hoe, S. L. L.; Puthucheary, S. D.; Yasin, R. M.Detection of virulence genes in Malaysian Shigella species by multiplexPCR assay. BMC Infect. Dis. 2005, 5 (8), 1−7.(28) Hartman, A. B.; Venkatesan, M.; Oaks, E. V.; Buysse, J. M.Sequence and molecular characterization of a multicopy invasionplasmid antigen gene, ipaH, of Shigella f lexneri. J. Bacteriol. 1990, 172(4), 1905−1915.(29) R Core Team. R: A language and environment for statisticalcomputing. R Foundation for Statistical Computing. Vienna, Austria,2014. http//www.R-project.org.(30) Lee, L.; Helsel, D. Statistical analysis of water-quality datacontaining multiple detection limits II: S-language software fornonparametric distribution modeling and hypothesis testing. Comput.Geosci. 2007, 33, 696−704.(31) Helsel, D. R. More than obvious: Better methods forinterpreting nondetect data. Environ. Sci. Technol. 2005, 419A−423A.(32) Black, R. E.; Levine, M. M.; Clements, M. L.; Hughes, T. P.;Blaser, M. J. Experimental Campylobacter jejuni infection in humans. J.Infect. Dis. 1988, 157 (3), 472−479.(33) Hornick, R. B.; Woodward, T. E.; McCrumb, F. R.; Snyder, M.J.; Dawkins, A. T.; Bulkeley, J. T.; De la Macorra, F.; Corozza, F. A.Study of induced typhoid fever in man. I. Evaluation of vaccineeffectiveness. Trans. Assoc. Am. Physicians 1996, 79, 361−367.(34) Dupont, H. L.; Hornick, R. B.; Snyder, M. J.; Libonati, J. P.;Formal, S. B.; Gangarosa, E. J. Immunity in Shigellosis. I. Response ofman to attenuated strains of Shigella. J. Infect. Diseases 1972, 125 (1),5−11.

Environmental Science & Technology Article

dx.doi.org/10.1021/es5038657 | Environ. Sci. Technol. XXXX, XXX, XXX−XXXI

(35) Dufour, A. P.; Evans, O.; Behymer, T. D.; Cantu, R. Wateringestion during swimming activities in a pool: A pilot study. J. WaterHealth 2006, 4, 425−430.(36) Soller, J. A.; Schoen, M. E.; Bartrand, T.; Ravenscroft, J. E.;Ashbolt, N. J. Estimated human health risks from exposure torecreational waters impacted by human and non-human sources offaecal contamination. Water Res. 2010, 44, 4674−4691.(37) Kotloff, K. L.; Winickoff, J. P.; Ivanoff, B.; Clemens, J. D.;Swerdlow, D. L.; Sansonetti, P. J.; Adak, G. K.; Levine, M. M. Globalburden of Shigella infections: Implications for vaccine developmentand implementation of control strategies. Bull. World Health Org.1999, 77 (8), 651−666.(38) Mauro, S. A.; Koudelka, G. B. Shiga toxin: Expression,distribution, and its role in the environment. Toxins 2011, 3, 608−625.(39) Casas, V.; Sobrepena, G.; Rodrigues-Mueller, B.; AhTye, J.;Maloy, S. R. Bacteriophage-encoded shiga toxin gene in atypicalbacterial host. Gut Pathogens 2011, 3 (10), 1−7.(40) Muniesa, M.; Jofre, J.; Garcia-Aljaro, C.; Blanch, A. R.Occurrence of Escherichia coli O157:H7 and other enterohemorrhagicEscherichia coli in the environment. Environ. Sci. Technol. 2006, 40,7141−7149.(41) Eichmiller, J. J.; Hicks, R. E.; Sadowsky, M. J. Distribution ofgenetic markers of fecal pollution on a freshwater sandy shoreline inproximity to wastewater effluent. Environ. Sci. Technol. 2013, 47,3395−3402.(42) Englebert, E. T.; McDermott, C. M.; Kleinheinz, G. T. Impact ofthe alga Cladophora on the survival of E.coli, Salmonella, and Shigella inlaboratory microcosm. J. Great Lakes Res. 2008, 34 (2), 377−382.(43) Halliday, E.; Gast, R. J. Bacteria in beach sands: An emergingchallenge in protecting coastal water quality and bather health. Environ.Sci. Technol. 2011, 45 (2), 370−379.(44) Verhougstraete, M. P.; Byappanahalli, M. N.; Rose, J. B.;Whitman, R. L. Cladophora in the Great Lakes: Impacts on beachwater quality and human health. Water Sci. Technol. 2010, 62, 68−76.(45) Skalbeck, J. D.; Kinzelman, J. L.; Mayer, G. C. Fecal indicatororganism density in beach sands: Impact of sediment grain size,uniformity, and hydrologic factors on surface water loading. J. GreatLakes Res. 2010, 36 (4), 707−714.(46) Olapade, O. A.; Depas, M. M.; Jensen, E. T.; McLellan, S. L.Microbial communities and fecal indicator bacteria associated withCladophora mats on beach sites along Lake Michigan shores. Appl.Environ. Microbiol. 2006, 72 (3), 1932−1938.(47) Byappanahalli, M. N.; Shively, D. A.; Nevers, M. B.; Sadowsky,M. J.; Whitman, R. L. Growth and survival of Escherichia coli andenterococci populations in the macro-alga Cladophora (Chlorophyta).FEMS Microbiol. Ecol. 2003, 46, 203−211.(48) Alm, E. W.; Burke, J.; Hagan, E. Persistence and potentialgrowth of the fecal indicator bacteria, Escherichia coli, in shoreline sandat Lake Huron. J. Great Lakes Res. 2006, 32, 401−405.(49) Alm, E. W.; Burke, J.; Spain, A. Fecal indicator bacteria areabundant in wet sand at freshwater beaches. Water Res. 2003, 37,3978−3982.(50) Heaney, C. D.; Sams, E.; Wing, S.; Marshall, S.; Brenner, K.;Dufour, A. P.; Wade, T. J. Contact with beach sand among beachgoersand risk of illness. Am. J. Epidem. 2009, 170 (2), 164−172.(51) Pachepsky, Y. A.; Shelton, D. R. Escherichia coli and fecalcoliforms in freshwater and estuarine sediments. Crit. Rev. Environ. Sci.Technol. 2011, 41, 1067−1110.(52) Ksoll, W. B.; Ishii, S.; Sadowsky, M. J.; Hicks, R. E. Presence andsources of fecal coliform bacteria in epilithic periphyton communitiesof Lake Superior. Appl. Environ. Microbiol. 2007, 73 (12), 3771−3778.(53) Sauer, E. P.; VandeWalle, J. L.; Bootsma, M. J.; McLellan, S. L.Detection of the human specific Bacteroides genetic marker providesevidence of widespread sewage contamination of stormwater in theurban environment. Water Res. 2011, 45, 4081−4091.(54) Newton, R. J.; Bootsma, M. J.; Morrison, H. G.; Sogin, M. L.;McLellan, S. L. A microbial signature approach to identify fecalpollution in the waters off an urbanized coast of Lake Michigan.Microb. Ecol. 2013, 65, 1011−1023.

(55) Hansen, D. L.; Ishii, S.; Sadowsky, M. J.; Hicks, R. E. Waterfowlabundance does not predict the dominant avian source of beachEscherichia coli. J. Environ. Qual. 2011, 40, 1924−1931.(56) Converse, R. R.; Kinzelman, J. L.; Sams, E. A.; Hudgens, E.;Dufour, A. P.; Ryu, H.; Santo-Domingo, J. W.; Kelty, C. A.; Shanks, O.C.; Siefring, S. D.; Haugland, R. A.; Wade, T. J. Dramaticimprovements in beach water quality following gull removal. Environ.Sci. Technol. 2012, 46, 10206−10213.(57) Leevesque, B.; Brousseau, P.; Simard, P.; Dewailly, E.; Meisels,M.; Ramsay, D.; Joly, J. Impact of the ring-billed gull (Larusdelawarensis) on the microbiological quality of recreational water. Appl.Environ. Microbiol. 1993, 59 (4), 1228−1230.(58) Kleinheinz, G. T.; McDermott, C. M.; Chomeau, V. Evaluationof avian waste and bird counts as predictors of Escherichia colicontamination at Door County, Wisconsin beaches. J. Great Lakes Res.2006, 32, 117−123.(59) Allos, B. M. Campylobacter jejuni infections: Update onemerging issues and trends. Food Safety 2001, 32, 1201−1206.(60) Stanley, K.; Jones, K. Cattle and sheep farms as reservoirs ofCampylobacter. J. Appl. Microbiol. 2003, 104S−113S.(61) Lavender, J. S.; Kinzelman, J. L. A cross comparison of QPCR toagar-based or defined substrate test methods for the determination ofEscherichia coli and enterococci in municipal water quality monitoringprograms. Water Res. 2009, 43, 4967−4979.(62) Bae, S.; Wuertz, S. Survival of host-associated Bacteriodales cellsand their relationship with Enterococcus spp., Campylobacter jejuni,Salmonella enterica serovar Typhimurium, and adenovirus in freshwatermicrocosms as measured by propidium monoazide-quantitative PCR.Appl. Environ. Microbiol. 2012, 78 (4), 922−932.

Environmental Science & Technology Article

dx.doi.org/10.1021/es5038657 | Environ. Sci. Technol. XXXX, XXX, XXX−XXXJ