Probabilistic Analysis Showing That a Combination of Bacteroides and Methanobrevibacter Source Tracking Markers Is Effective for Identifying Waters Contaminated by Human Fecal Pollution

Probabilistic Analysis Showing That a Combination of Bacteroidesand Methanobrevibacter Source Tracking Markers Is Effective forIdentifying Waters Contaminated by Human Fecal PollutionChristopher Johnston,†,‡ Muruleedhara N. Byappanahalli,§ Jacqueline MacDonald Gibson,¶

Jennifer A. Ufnar,⊥ Richard L. Whitman,§ and Jill R. Stewart*,‡,¶

†Jardon and Howard Technologies Incorporated, Orlando, Florida 32826, United States‡Center for Coastal Environmental Health and Biomolecular Research and Hollings Marine Laboratory, U.S. National Oceanic andAtmospheric Administration, Charleston, South Carolina 29412, United States§U.S. Geological Survey, Great Lakes Science Center, Lake Michigan Ecological Research Station, Porter, Indiana 46304, UnitedStates⊥Division of Science and Technology, Southern Vermont College, Bennington, Vermont 05201, United States¶Department of Environmental Sciences and Engineering, Gillings School of Global Public Health, University of North Carolina,Campus Box 7431, Chapel Hill, North Carolina 27599, United States

ABSTRACT: Microbial source tracking assays to identify sources of waterbornecontamination typically target genetic markers of host-specific microorganisms.However, no bacterial marker has been shown to be 100% host-specific, andcross-reactivity has been noted in studies evaluating known source samples. Using485 challenge samples from 20 different human and animal fecal sources, thisstudy evaluated microbial source tracking markers including the BacteroidesHF183 16S rRNA, M. smithii nifH, and Enterococcus esp gene targets that havebeen proposed as potential indicators of human fecal contamination. Bayes’Theorem was used to calculate the conditional probability that these markers or acombination of markers can correctly identify human sources of fecal pollution.All three human-associated markers were detected in 100% of the sewage samplesanalyzed. Bacteroides HF183 was the most effective marker for determiningwhether contamination was specifically from a human source, and greater than 98% certainty that contamination was from ahuman source was shown when both Bacteroides HF183 andM. smithii nifH markers were present. A high degree of certainty wasattained even in cases where the prior probability of human fecal contamination was as low as 8.5%. The combination ofBacteroides HF183 and M. smithii nifH source tracking markers can help identify surface waters impacted by human fecalcontamination, information useful for prioritizing restoration activities or assessing health risks from exposure to contaminatedwaters.

■ INTRODUCTIONProtection of water resources, including waters used fordrinking, shellfish harvesting, or recreation, depends on ourability to detect and remediate fecal contamination andassociated human pathogens. Current methods used to monitorwater quality, however, cannot distinguish between sources ofpollution. Microbial source tracking (MST), an approach usedto discriminate among different sources of fecal pollution, mayhelp identify waters most likely to pose risks to human healthand could help ensure selection of appropriate and effectivemitigation strategies to restore impaired waters.Microbial source tracking is an approach to identify microbes

or genetic markers specifically associated with their host sources(e.g., human, dog, cow). Anaerobic bacteria common to gutmicroflora are among the most promising targets for MSTassays because there is evidence that strains of these bacteriahave coevolved with their hosts.1 Although detection ofanaerobic bacteria to assess fecal contamination has tradition-

ally been avoided due to difficulty associated with cultivation,modern molecular methods including the polymerase chainreaction (PCR) now make detection of these organisms rapidand practical, removing the need for cultivation.1,2

Source tracking methods in current use commonly target theanaerobic gut bacteria in the order Bacteroidales.3−5Bacteroidesspp. make up approximately one-third of the human fecalmicroflora,6 well outnumbering E. coli and Enterococcus spp.Bacteroides spp. are also obligate anaerobes, so there is littleconcern over prolonged persistence or regrowth in theenvironment.7,8Studies evaluating human-specific Bacteroidesmarkers in Australia reported that the HF183 marker wasdetected in 95% (n = 79) of tested sewage and individual

Received: August 22, 2013Revised: October 22, 2013Accepted: November 1, 2013Published: November 1, 2013

Article

pubs.acs.org/est

© 2013 American Chemical Society 13621 dx.doi.org/10.1021/es403753k | Environ. Sci. Technol. 2013, 47, 13621−13628

pubs.acs.org/est

human fecal samples and was absent in 94% (n = 201) ofanimal samples tested.9 Furthermore, a recent methodscomparison study involving 27 participating laboratoriesdemonstrated that HF183 is the most accurate MST targettested for discriminating human-source contamination.10 Thislatter study, however, was limited in the number andgeographic diversity of challenge samples, and it only includedchallenge samples prepared by mixing one or two sources offecal materials.11

Another promising anaerobic target proposed to tracksewage pollution is nifH of Methanobrevibacter smithii.12,13 M.smithii is the dominant Achaean in the human gut, occurring inconcentrations as high as 1010 per gram of dry weight.14−16

Moreover, M. smithii is the only species of Methanobrevibacterknown to specifically colonize the human large intestine.17,18

Tests from the U.S. and U.K. using methane emissions from thebreath found that approximately 33% of humans harbormethanogens.14 These data suggest that the marker is a moresuitable indicator of sewage pollution rather than for identifyingsamples contaminated from individuals. Detection of M. smithiiby PCR has been demonstrated in treated sewage and surfacewaters,19,20 and environmental persistence has been demon-strated for up to 21 days.12 A recent validation study reportedthat the nifH marker was 96% specific and 81% sensitive forhuman contamination when tested against 272 fecal andwastewater samples in Southeast Queensland, Australia.21 Theauthors concluded the nifH marker is sewage-specific, but thatit may not be sensitive enough to detect sewage pollution in theenvironment unless it is combined with an additional marker.Using a combination of host-specific markers could help

increase confidence in source identifications. No proposedMST markers have proven to be 100% host-specific.10 Also,more rigorous statistical tests are needed to calculateprobabilities that a positive sample is indeed contaminatedwith the indicated host source, as well as to interpret negativeresults. Kildare et al.22 used Bayes’ Theorem to calculateconditional probabilities that four different Bacteroidales assays,targeting BacUni, BacHum, BacCow, and BacCan, couldcorrectly identify host-specific fecal pollution. Wang et al.23

built on that work by evaluating ratios of these host-specificmarkers to general Bacteroidales to allocate sources of fecalcontamination. No tests were conducted to evaluate whethercombinations of markers targeting the same host species (e.g.,humans) could increase confidence in source identifications.The goal of this study was to evaluate the sensitivity and

specificity of promising human-specific source tracking markersusing a large panel of fecal wastes from human and animalsources. A panel of DNA extracts consisting of 485 individualsamples from 20 different host fecal sources was employed. Toour knowledge, this study represents one of the largest panelsof reference and known-source samples against which thesemarkers have been tested. Using Bayes’ Theorem, we thencalculated the conditional probability that these markers orcombination of markers could correctly identify human sourcesof fecal pollution. Calculating these probabilities could makeMST data more useful in resolving legal disputes over watercontamination and could make MST more amenable foradoption in regulations aimed at protecting water resources.

■ MATERIALS AND METHODSBacterial Cultures, Standards, and Controls. Live

cultures of M. smithii were obtained from the American TypeCulture Collection (ATCC 35061) (ATCC, Manassas, VA)

and the German Collection of Microorganisms and CellCultures (DSM 861) (DSMZ, Braunschweig, Germany). DNAwas extracted directly from these cultures using the DNeasyTissue Kit following the manufacturer’s protocol for bacteriacells (Qiagen, Valencia, CA). Genomic DNA extracted from aculture of M. smithii was used as a positive control for the M.smithii nifH conventional PCR method and to createquantification standards for generating a standard curve forthe M. smithii nifH qPCR method. DNA concentration wasdetermined by spectrophotometry using a Nanodrop ND-1000spectrophotometer (Thermo Scientific, Waltham, MA). Enter-ococcus faecium C68 served as a positive control for theEnterococcus esp PCR method.24 DNA extracts from humansewage samples served as positive controls for the BacteroidesHF183 method.

Sample Collection and Processing. To test and comparethe specificity and sensitivity of different microbial sourcetracking gene targets, PCR methods were applied to a variety ofDNA extracts from different hosts. Two geographically distinctpanels of DNA extracts from animal and human source fecalsamples were created from locations in Northwestern Indianaand Mississippi. For the Indiana panel, fecal samples werecollected between June 2006 and December 2007 andprocessed as described by Whitman et al.24 Briefly, sewageinfluent samples were collected from sewage treatment plants atvarious locations, pit toilets were sampled from camping andrecreational areas, and samples were collected from differentseptic trucks at a local treatment facility. For domestic pets,livestock, or wildlife, samples were collected from animalboarding facilities, farms, or forested areas around the IndianaDunes National Lakeshore in Porter County. For all samples,swabs of the fecal material (∼0.3−0.5 g) or pellets weretransferred to sterile 15 mL centrifuge tubes containing 6 mL ofphosphate-buffered water (PBW) and vortexed vigorously tocreate fecal slurry. For composite animal fecal samples, equalvolumes of individual fecal slurries were pooled. The volume offecal slurry used for the composite was dependent on thenumber of samples collected from individual animals (i.e.,smaller volumes of fecal slurries were combined for a greaternumber of individual animal fecal samples collected).For the Mississippi panel, fecal samples were collected in

2005 and 2006 and processed as reported by Ufnar et al.12

Briefly, sewage samples (500 mL) were collected from sewersin Gulfport, MS. Individual human samples were collected andprocessed in Hattiesburg, MS. Individual animal fecal sampleswere collected from various farms, processing plants, or animalshelters in sterile 50 mL centrifuge tubes. Composite animalfecal samples consisted of bovine and swine waste lagoons andwere collected in either 500 or 50 mL volumes.

DNA Extractions. Aliquots (250 μL) of the fecal slurriescollected in Indiana were transferred to 2 mL centrifuge tubescontaining glass beads, and genomic DNA was extracted usingthe PowerSoil DNA Isolation Kit (MO BIO, Carlsbad, CA).The manufacturer’s protocol was followed with the followingmodifications: after adding solution C1 and vortexing, tubeswere incubated at 60 °C for 4 min; tubes were put in a beadbeater for 1 min rather than vortexing for 10 min; after addingsolution C2 and C3, tubes were incubated at 4 °C for 30 min;C6 solution was added in increments of 30, 30, and 40 μL withcentrifugation after each volume added.For the Mississippi DNA extract panel, individual fecal

samples were processed using the UltraClean Soil DNAExtraction Kit (MO BIO) and the PowerSoil DNA Kit

Environmental Science & Technology Article

dx.doi.org/10.1021/es403753k | Environ. Sci. Technol. 2013, 47, 13621−1362813622

following the manufacturer instructions. Sewer samples wereprefiltered through a 3 μm cellulose acetate filter (PallCorporation, West Chester, PA) and concentrated onto a 0.2μm Supor-200 filter (Pall Corporation). Bound bacteria weredislodged from the filter by agitating with a magnetic stir bar for5 min in 5 mL of sterile phosphate-buffered saline (PBS). Cellswere pelleted by centrifugation for 15 min at 13 000g andresuspended in 2 mL, and DNA was extracted using theUltraclean Soil DNA Extraction kit (MO BIO). The resultingDNA extracts were stored at −80 °C and shipped to thelaboratory in Charleston, SC on dry ice.PCR Methods. M. smithii nifH qPCR. A TaqMan-based

qPCR assay was developed to detect the nifH gene of M.smithii.13 Thermal cycling and fluorescence detection werecarried out in the iQ5 Real-Time PCR detection system (Bio-Rad, Hercules, CA). The reactions were performed in a totalvolume of 25 μL containing 1× PCR buffer (50 mM KCl, 20mM Tris−HCl, pH 8.4), 5 mM MgCl2, 800 μM deoxynucleo-tide triphosphates (dNTPs), 800 nM primers, 240 nM of theMnif probe, 2.3 U Taq DNA polymerase, and 3 μL of the M.smithii quantification standard or DNA extract. Standards werediluted in nuclease-free water and stored in single use aliquotsat −80 °C. A five-point 10-fold serial dilution of the M. smithiigenomic DNA (10 to 100 000 fg) was run in triplicate witheach set of reactions to generate the standard curve, and thelower limit of detection was set using the lowest quantity ofgenomic DNA detected for the standard curve. All sampleswere run in triplicate with the standards acting as positivecontrols and no-template negative controls.The cycling conditions were an initial denaturation for 10

min at 95 °C and 50 cycles of denaturation for 10 s at 95 °Cand annealing/extension for 30 s at 57 °C. Real-timefluorescence measurements were collected by the iQ5 instru-ment beginning after the first 3 cycles to prevent any residualbubbles from causing any background fluorescence signal.Background well factors were collected from the experimentalplate, and the fluorescent thresholds were set automatically bythe iQ5 software in the PCR baseline subtracted curve fitanalysis mode. The cycle thresholds (CT) at which the samplefluorescence exceeds background fluorescence were recordedfor the extracted samples and quantitative standards. Thenumbers of M. smithii nifH gene targets are interpolated fromthe standard curve generated from the quantification standardsin relation to their CT.M. smithii nif H Conventional PCR. A conventional PCR

method for detecting the M. smithii nifH gene was firstdeveloped by Ufnar et al.12 PCR reactions were carried out ineither 10 or 20 μL volumes containing 1× PCR buffer, 0.1%BSA, 200 μM dNTPs, 0.5 U Taq polymerase, 0.5 μM eachprimer, and 1 μL of DNA template. Cycling conditions for thereaction consisted of initial denaturation for 2 min at 92 °C and30 cycles of denaturation for 1 min at 92 °C, annealing for 30 sat 55 °C, and elongation for 1 min at 72 °C. A final elongationwas performed for 6 min at 72 °C. PCR products wereseparated on a 1% agarose gel stained with ethidium bromideand viewed under UV light.Enterococcus esp. The Enterococcus esp gene has been

proposed as a marker of human pollution in environmentalwaters.25 PCR was performed as described by Whitman et al.24

and using E. faecium C68 as a positive control. Reactions wereperformed in 50 μL volumes containing 1× PCR buffer, 1.5mM MgCl2, 200 μM dNTPs, 0.3 μM each primer, 2.5 UAmpliTaq DNA polymerase, and 5 μL of DNA template. The

PCR amplification conditions consisted of initial denaturationfor 10 min at 95 °C and 35 cycles of denaturation for 1 min at94 °C, annealing for 1 min at 58 °C, and elongation for 1 minat 72 °C. PCR products were separated on a 1.5% agarose gelstained with ethidium bromide and viewed under UV light.

Bacteroides HF183. The Bacteroides HF183 assay, targetingthe 16S rRNA gene, was performed using a modified protocolfrom Bernhard and Field.3,26 PCR reactions were carried out in25 μL volumes containing 1× PCR buffer, 200 μM dNTPs, 1 UTaq polymerase, 0.5 μM each primer, and 2 μL of DNAtemplate. Cycling conditions for the reaction consisted of initialdenaturation for 5 min at 95 °C and 35 cycles of denaturationfor 30 s at 95 °C, annealing for 30 s at 58 °C, and elongationfor 1 min at 72 °C. A final elongation was performed for 6 minat 72 °C. PCR products were separated on a 1% agarose gelstained with ethidium bromide and viewed under UV light.

Data Analysis. Sensitivity is the ability to detect a sourcewhen it is present and is calculated by dividing the number oftrue-positive results by the number of samples that shouldcontain the target.27 Specificity is the ability to not detect asource when it is not present and is calculated by dividing thenumber of true-negative results by the number of samples thatshould not contain the target. For this study, sensitivity (r) andspecificity (s) were calculated as r = a/(a + c) and s = d/(b + d),where a is the number of DNA samples positive for the PCRmarker of its own species (true positive); b is the number ofDNA samples positive for a PCR marker of another species(false positive); c is the number of DNA samples negative for aPCR marker of its own species (false negative); d is the numberof DNA samples negative for a PCR marker of another species(true negative).27,28

To explore the potential use of the markers tested in thisresearch for identifying water bodies affected by human fecalcontamination, we employed Bayes’ Theorem29 to calculate theposterior probability that a human fecal source is present, givena prior assumption about the probability that human fecalcontamination is present and detection of each marker alone orin combination with the others. For individual markers, theposterior probability that human fecal contamination is present,once the marker is detected, can be derived from Bayes’Theorem, as follows:

| = | ×| × + | ×

= ×× + − ×

++

+ +P H MP M H P H

P M H P H P M H P Hr P H

r P H s P H

( )( ) ( )

( ) ( ) ( ) ( )( )

( ) (1 ) ( )

c c

c(1)

where H represents the event that human fecal contaminationis present, Hc represents the absence of human fecalcontamination, and M+ represents the event that the sampletests positive for the marker. Similarly, using Bayes’ Theorem,the probability that human fecal contamination is present giventhat two markers are detected in a sample can be computedfrom

| ∩

=× ×

× × + − × − ×

+ +P H M Mr r P H

r r P H s s P H

( )( )

( ) (1 ) (1 ) ( )

1 2

1 2

1 2 1 2c

(2)

where ri represents the sensitivity of marker i and si representsthe specificity of marker i.



When one marker is detected and the second is not, then theposterior probability of human fecal contamination is given by

| ∩

=× − ×

× − × + − × ×

+ −P H M Mr r P H

r r P H s s P H

( )(1 ) ( )

(1 ) ( ) (1 ) ( )

1 2

1 2

1 2 1 2c

(3)

where Mi− represents the event that marker i is absent. Bayes’

Theorem was used to derive similar equations for the posteriorprobability that human fecal contamination is present givenvarious other combinations of results for the genetic markerstested in this research (for example, two markers positive andthe third marker negative, and so on).

■ RESULTSThe panel of DNA extracts, representing 485 individualsamples from 20 different host fecal sources, was tested usingPCR methods proposed for microbial source tracking.12,24,30

The M. smithii nifH qPCR method was tested on the completepanel of DNA extracts from both Mississippi and Indiana(Tables 1 and 2). This method detected the marker in 100% ofthe sewage samples (both influent and effluent) from bothMississippi and Indiana. The marker was also detected in 47%

and 45% of samples collected from pit toilet and septic trucksin Indiana, respectively, as well as 78% of individual humanfeces sampled in Mississippi. The conventional PCR methodfor M. smithii nifH was only tested on the Mississippi DNAextract panel (Table 2) because there was not enough genomicDNA from the Indiana panel, and this gene was detected in93% of the sewage samples and 29% of the individual fecalsamples. While the conventional PCR method did not detectthe marker in any of the nonhuman sources, the qPCR methoddid display cross reactivity, notably in fecal samples from cow,goats, sheep, deer, and pigs from both Mississippi and Indiana.This marker was also detected in 48% of the pig samples fromthe Mississippi panel, which were not part of the host sourcesfor the Indiana panel. Fecal waste tested from birds, horses, andhousehold pets were all negative (Table 2). Quantitative resultsfor the nifH qPCR marker varied depending on the host sourceas well as among individuals from the same source (data notshown). As a general trend, sheep and goat fecal samples werefound to have higher levels of the nifH marker (at levelscomparable to humans) than pigs or cows. The nifH qPCRmarker was detected at the lowest levels in deer samples.The Bacteroides HF183 gene marker was tested on the

Indiana DNA extract panel (Table 1). It was detected in 100%of the sewage samples, 47% of pit toilet samples, and 5% of

Table 1. PCR Results from the M. smithii nifH, Bacteroides HF183, and Enterococcus esp Gene Markers against a Panel ofHuman and Nonhuman Fecal Sources Collected in Indianaa

sample type source M. smithii nifH qPCR Bacteroides HF183 PCR Enterococcus espfmb PCR

human pit toilet 47 (7/15) 47 (7/15) 0 (0/15)septic truck 45 (9/20) 5 (1/20) 30 (6/20)sewage influent 100 (24/24) 100 (24/24) 100 (24/24)

individual animals cow 46 (6/13) 0 (0/13) N/Abird 0 (0/5) 0 (0/5) 0 (0/4)cat 0 (0/5) 0 (0/5) N/Adog 0 (0/5) 0 (0/5) 35 (9/26)goose 0 (0/10) 0 (0/10) N/Agull 0 (0/5) 0 (0/5) 10 (2/20)mouse 0 (0/5) 0 (0/5) N/Araccoon 0 (0/5) 0 (0/5) N/Achicken 0 (0/5) 0 (0/5) N/Agoat 100 (2/2) 0 (0/2) N/Asheep 50 (1/2) 0 (0/2) N/Adeer 0 (0/13) 0 (0/13) 0 (0/1)chipmunk 0 (0/1) 0 (0/1) 0 (0/1)rabbit 0 (0/1) 0 (0/1) 0 (0/1)

composite animals bird 0 (0/8) 0 (0/8) 0 (0/5)cat 0 (0/7) 0 (0/7) 0 (0/7)dog 0 (0/10) 0 (0/10) 60 (6/10)goose 0 (0/3) 0 (0/3) 0 (0/2)gull 0 (0/4) 0 (0/4) 25 (1/4)mouse 0 (0/3) 0 (0/3) 0 (0/3)raccoon 0 (0/3) 0 (0/3) 0 (0/3)chicken 0 (0/1) 0 (0/1) N/Ahorse 0 (0/1) 0 (0/1) N/Asquirrel 0 (0/1) 100 (1/1) N/Aturkey 0 (0/1) 0 (0/1) N/Arooster 0 (0/1) 0 (0/1) N/Adeer N/A N/A 0 (0/1)

aResults for each gene marker shown as percent positive detection (positive PCR results/total number of samples tested). bData from Whitman etal.24



septic truck samples. The only nonhuman source that theBacteroides HF183 marker was detected in from this study was asquirrel composite sample. Not enough genetic material wasavailable to test this marker against the samples collected inMississippi.The Enterococcus esp gene frequency was similarly tested on a

portion of the samples in the Indiana panel (Table 1). The

marker was detected in 100% of sewage samples and 30% ofseptic truck samples. It was not detected in any of the pit toiletsamples. The marker was also detected in both individual andcomposite samples from dogs and gulls and was absent in allother animal samples. A subset of the positive samples wasconfirmed by sequencing as described in Byappanahalli et al.31

The overall specificity of the M. smithii qPCR marker was0.93 and 0.58 for the Indiana and Mississippi panels,respectively (Tables 3 and 4). In comparison, the M. smithiiconventional PCR method specificity was 1.0 for theMississippi panel; however, the sensitivity of the conventionalPCR method was 0.46, while the sensitivity of the qPCRmethod was 0.68 and 0.82 for the Indiana and Mississippipanels. The Enterococcus esp marker specificity and sensitivitywere 0.80 and 0.51 for the Indiana panel (Table 3). TheBacteroides HF183 marker specificity and sensitivity were 0.99and 0.54, respectively for the Indiana panel (Table 3).In order to explore the potential usefulness of the results

presented here for analysis of fecal source contributions to agiven water body, we employed Bayes’ Theorem to estimate theposterior probability that a fecal organism isolated from a watersample originates from a human source, given variouscombinations of the presence or absence of the markers.Figure 1 shows the results for differing prior probabilities thatcontamination originates from human sources. The priorprobability is defined in the typical manner employed inBayesian statistics (see, for example, DeGroot and Schervish32):it is the decision-maker’s advance prediction (before collectingmicrobial source-tracking data) of the likelihood that the waterbody is impacted by a human source. Such prior probabilitiesmight be based on characteristics of the watershed, for example,or on previous test results using traditional fecal indicator

Table 2. Frequency ofM. smithii nifH Detection by PCR andqPCR Methods against a Panel of Human and NonhumanFecal Sources Collected in Mississippia

sample type sourceM. smithii nifH

qPCRM. smithii nifH

PCRb

human individuals 78 (38/49) 29 (20/70)sewageinfluent

100 (12/12) 93 (25/27)

treated sewage 100 (1/1) ND

individualanimals

cow 41 (18/44) 0 (0/46)horse 0 (0/40) 0 (0/23)goat 91 (21/23) 0 (0/2)sheep 100 (23/23) 0 (0/2)deer 38 (9/24) 0 (0/20)rat 0 (0/20) NDpig 48 (12/25) 0 (0/24)chicken 0 (0/24) 0 (0/23)

compositeanimals

cow lagoon 89 (8/9) 0 (0/2)pig lagoon 91 (10/11) ND

aResults for each gene marker shown as percent positive detection(positive PCR results/total number of samples tested). bData fromUfnar et al.12

Table 3. Determination of the Specificity (s) and Sensitivity (r) of the M. smithii nifH, Bacteroides HF183, and Enterococcus espGene Markers against a Panel of Human and Nonhuman Fecal Sources Collected in Indiana

M. smithii nifH Bacteroides HF183 Enterococcus espfma

source s (%) r (%) n s (%) r (%) n s (%) r (%) n

human 67.8 59 54.2 59 50.9 59nonhuman 92.5 120 99.2 120 79.6 88pit toilet 46.7 15 46.7 15 0 15septic truck 45 20 5 20 30 20sewage influent 100 24 100 24 100 24cow 53.9 13 100 13 N/A 0bird 100 13 100 13 100 9cat 100 12 100 12 100 7dog 100 15 100 15 58.3 36goose 100 13 100 13 100 2gull 100 9 100 9 87.5 24mouse 100 8 100 8 100 3raccoon 100 8 100 8 100 3chicken 100 6 100 6 N/A 0goat 0 2 100 2 N/A 0sheep 50 2 100 2 N/A 0deer 100 13 100 13 100 2chipmunk 100 1 100 1 100 1rabbit 100 1 100 1 100 1horse 100 1 100 1 N/A 0squirrel 100 1 0 1 N/A 0turkey 100 1 100 1 N/A 0rooster 100 1 100 1 N/A 0

aData from Whitman et al.24



bacteria. Where no prior knowledge is available, the priorprobability can be assumed to be 50%. That is, the decision-maker or analyst has no information about whethercontamination is more likely than not to be present. Theresults in Figure 1 assume the sensitivity and specificity of themarkers are as observed in the data set from Indiana becausethe sensitivity and specificity data needed for similarcomputations using the Mississippi data were not available.These data show that the Bacteroides HF183 marker is the mosteffective marker for determining whether a sample may havebeen impacted by a human source of fecal contamination: even

with low prior probabilities that human fecal contamination ispresent, the presence of this marker results in a very highposterior probability that the contamination is, in fact, ofhuman origin. Near certainty (probability close to 100%) that asample is impacted by contamination of human origin can beobtained if both the M. smithii nifH and Bacteroides HF183markers are present, even in cases where the prior probability ofhuman fecal contamination is very low. Both of these markersare more powerful in confirming human fecal contaminationthan is the Enterococcus esp marker.

■ DISCUSSION

This study serves to establish the probability that detecting aparticular MST marker or combination of markers is accurate inidentifying contaminant sources in impacted waters, althoughthe approach still needs to best tested for environmentalsamples. The current results show that Bacteroides HF183 is themost effective single marker for determining fecal inputs fromhuman sources. Greater than 98% certainty that a sample isimpacted by fecal contamination of human origin can beobtained if a combination of markers, HF183 and M. smithiinifH, is present, even in cases where the prior probability ofhuman fecal contamination is as low as 8.5%.Using one of the largest panels of reference samples available,

results of this study show high but not absolute sensitivity ofthe markers for their intended targets. Both HF183 and M.smithii nifH markers were consistently detected in human-source fecal samples, indicating their widespread occurrence inthese sources. Exceptions were noted in the ability toconsistently detect these markers from pit toilets. The toiletswere in use at the time of sampling, so it is not clear whetherthe bacteria or target DNA were highly degraded in negativesamples or if individuals contributing wastes to the toilets didnot actually harbor the bacterial targets.Among the animal samples tested, M. smithii was detected in

ruminants (cows, goats, and sheep) using the qPCR assay andHF183 was detected in a composite scat sample from squirrels.

Table 4. Determination of the Specificity (s) and Sensitivity(r) of the M. smithii nifH Gene Markers against a Panel ofHuman and Nonhuman Fecal Sources Collected inMississippi

M. smithii nifH qPCR M. smithii nifH PCRa

source s (%) r (%) n s (%) r (%) n

human 82.3 62 46.4 97nonhuman 58.4 243 100 142

individual human 77.5 49 28.6 70sewage influent 100 12 92.6 27treated sewage 100 1 N/A 0

cow 59.1 44 100 46horse 100 40 100 23goat 8.7 23 100 2sheep 0 23 100 2deer 62.5 24 100 20rat 100 20 N/A 0pig 52 25 100 24chicken 100 24 100 23cow lagoon 11.1 9 100 2pig lagoon 9.1 11 N/A 0

aData from Ufnar et al.42

Figure 1. Predicted probability that a sample may be impacted by fecal contamination of human origin, given various combinations of the presenceor absence of the genetic markers tested in this study and different prior assumptions about the likelihood of human contamination.



These results are consistent with other reports in the literaturethat show exceptions to the host-specificity of source trackingmarkers. For example, cross-reactivity has been reportedbetween the HF183 marker and samples from chickens, dogs,and birds.9,33,34 Ahmed et al.21 reported 81% sensitivity of thenifH marker, with cross-reactivity noted for one bird (n = 30)and six composite pig (n = 20) samples. Such nonspecificreactions in domesticated animals (chickens, dogs, pigs) mayemphasize the need for additional confirmatory tests, especiallyin watersheds impacted by intense animal husbandryoperations, such as dairy, poultry, and swine farms.In contrast to the cross-reactivity noted for the nifH qPCR

assay, the conventional PCR assay tested against the Mississippipanel showed 100% specificity. That is, this marker was notdetected in any of the animal fecal samples tested. Discordantresults between the PCR and qPCR method may be due togreater sensitivity of the qPCR assay. The qPCR method hasbeen shown to be more sensitive with probe-based detection(limit of detection 10 fg of genomic DNA) compared to theconventional PCR method with gel-based detection (1 pg ofgenomic DNA).12,13

We recognize that our study did not fully address potentialgeographic or temporal variability for the tested markers.Geographic and temporal variation was determined to be anissue with library-based source-tracking methods, wherecharacteristics of organisms were compared to databases todetermine their sources.35−38 However, it is not clear the extentto which this will be an issue for library-independent sourcetracking markers, such as the ones tested in this study. Acollaborative study among scientists in the European Unionreported consistently high specificity of the HF183 markeramong all countries that participated, although regionalvariations in specificity of the ruminant marker CF128 werenoted.34 The panel of challenge samples used in this studyconsisted of feces and wastewaters obtained from twogeographically different locations in the United States: Indianaand Mississippi. Although the sensitivity and specificity of thehuman markers differed slightly between the Indiana andMississippi panels (Tables 3 and 4), the results were generallyconsistent with each other and with similar validation studiesconducted in Australia9,21,39 and California.40 Other locationsmay differ in host distributions of the tested markers, andvalidation of MST markers for a particular geographic areashould be conducted prior to using the markers in field studies.Application of probabilistic analysis to MST data, as

demonstrated in this study, could offer a powerful tool totranslate results from field studies into information needed fordecision-making. This approach could also help with selectionof the most appropriate tools available in the MST “tool-box”.27,41 Identifying waters with the highest probability ofhuman-source contamination could help prioritize those mostlikely to pose a threat to human health.20 Identifying particularanimal sources of contamination could help identify effectiveand practical strategies for mitigating impacted waters. Thisapproach could also allow managers and regulators to linkcontaminant sources with impacts. Apportioning impacts isimportant to water quality managers, utilities, and others whoare tasked with bringing water bodies into compliance withstandards, but who lack appropriate means to assess theimpacts of improved treatment or reduced discharges.11 Themove toward tools that identify the probability of watercontamination from specific sources could help confirm

linkages between discharges, nonpoint source pollution, andpotential ecological impacts such as receiving water quality.

■ AUTHOR INFORMATIONCorresponding Author*Phone: 919-966-7553; fax: 919-966-7911; e-mail: [email protected].

NotesThis publication does not constitute an endorsement of anycommercial product or intend to be an opinion beyondscientific or other results obtained by the National Oceanic andAtmospheric Administration (NOAA). No reference shall bemade to NOAA, or this publication furnished by NOAA, to anyadvertising or sales promotion which would indicate or implythat NOAA recommends or endorses any proprietary productmentioned herein, or which has as its purpose an interest tocause the advertised product to be used or purchased becauseof this publication. Any use of trade, product, or firm names isfor descriptive purposes only and does not imply endorsementby the U.S. Government.The authors declare no competing financial interest.

■ ACKNOWLEDGMENTSThe authors would like to thank Dawn Shively for help withsample collection and analyses. This work was funded in partby the U.S. Geological Survey Ocean Research Priorities Plan.This article is contribution 1802 of the U.S. Geological SurveyGreat Lakes Science Center.

■ REFERENCES(1) Bernhard, A. E.; Field, K. G. Identification of nonpoint sources offecal pollution in coastal waters by using host-specific 16S ribosomalDNA genetic markers from fecal anaerobes. Appl. Environ. Microbiol.2000, 66 (4), 1587−1594.(2) Field, K. G.; Bernhard, A. E.; Brodeur, T. J. Molecular approachesto microbiological monitoring: Fecal source detection. Environ. Monit.Assess. 2003, 81 (1−3), 313−326.(3) Bernhard, A. E.; Field, K. G. A PCR assay to discriminate humanand ruminant feces on the basis of host differences in Bacteroides-Prevotella genes encoding 16S rRNA. Appl. Environ. Microbiol. 2000,66 (10), 4571−4574.(4) Dick, L. K.; Bernhard, A. E.; Brodeur, T. J.; Santo Domingo, J.W.; Simpson, J. M.; Walters, S. P.; Field, K. G. Host distributions ofuncultivated fecal Bacteroidales bacteria reveal genetic markers for fecalsource identification. Appl. Environ. Microbiol. 2005, 71 (6), 3184−3191.(5) Shanks, O. C.; Domingo, J. W.; Lu, J.; Kelty, C. A.; Graham, J. E.Identification of bacterial DNA markers for the detection of humanfecal pollution in water. Appl. Environ. Microbiol. 2007, 73 (8), 2416−2422.(6) Madigan, M. M.; Martinko, J. M.; Parker, J. Brock biology ofmicroorganisms, 10th ed.; Prentice Hall: Upper Saddle River, N.J.,2003.(7) Fiksdal, L.; Maki, J. S.; LaCroix, S. J.; Staley, J. T. Survival anddetection of Bacteroides spp., prospective indicator bacteria. Appl.Environ. Microbiol. 1985, 49 (1), 148−150.(8) Kreader, C. A. Persistence of PCR-detectable Bacteroidesdistasonis from human feces in river water. Appl. Environ. Microbiol.1998, 64 (10), 4103−4105.(9) Ahmed, W.; Masters, N.; Toze, S. Consistency in the hostspecificity and host sensitivity of the Bacteroides HF183 marker forsewage pollution tracking. Lett. Appl. Microbiol. 2012, 55 (4), 283−289.(10) Boehm, A. B.; Van De Werfhorst, L. C.; Griffith, J. F.; Holden,P. A.; Jay, J. A.; Shanks, O. C.; Wang, D.; Weisberg, S. B. Performance



mailto:[email protected]

mailto:[email protected]

of forty-one microbial source tracking methods: A twenty-seven labevaluation study. Water Res. 2013, 47 (18), 6812−6828.(11) Stewart, J. R.; Boehm, A. B.; Dubinsky, E. A.; Fong, T. T.;Goodwin, K. D.; Griffith, J. F.; Noble, R. T.; Shanks, O. C.; Vijayavel,K.; Weisberg, S. B. Recommendations following a multi-laboratorycomparison of microbial source tracking methods. Water Res. 2013, 47(18), 6829−6838.(12) Ufnar, J. A.; Wang, S. Y.; Christiansen, J. M.; Yampara-Iquise,H.; Carson, C. A.; Ellender, R. D. Detection of the nifH gene ofMethanobrevibacter smithii: A potential tool to identify sewagepollution in recreational waters. J. Appl. Microbiol. 2006, 101 (1),44−52.(13) Johnston, C.; Ufnar, J. A.; Griffith, J. F.; Gooch, J. A.; Stewart, J.R. A real-time qPCR assay for the detection of the nifH gene ofMethanobrevibacter smithii, a potential indicator of sewage pollution. J.Appl. Microbiol. 2010, 109 (6), 1946−1956.(14) Bond, J. H., Jr.; Engel, R. R.; Levitt, M. D. Factors influencingpulmonary methane excretion in man. An indirect method of studyingthe in situ metabolism of the methane-producing colonic bacteria. J.Exp. Med. 1971, 133 (3), 572−588.(15) Lin, C.; Miller, T. L. Phylogenetic analysis of Methanobrevibacterisolated from feces of humans and other animals. Arch. Microbiol.1998, 169, 397−403.(16) Miller, T. L.; Wolin, M. J. Enumeration of Methanobrevibactersmithii in human feces. Arch. Microbiol. 1982, 131 (1), 14−18.(17) Eckburg, P. B.; Bik, E. M.; Bernstein, C. N.; Purdom, E.;Dethlefsen, L.; Sargent, M.; Gill, S. R.; Nelson, K. E.; Relman, D. A.Diversity of the human intestinal microbial flora. Science 2005, 308(5728), 1635−1638.(18) Miller, T. Genus II. Methanobrevibacter. In Bergey’s Manual ofSystematic Bacteriology; Williams & Wilkins: Baltimore, MD, 1984; Vol.1, pp 2178−2183.(19) Rosario, K.; Symonds, E. M.; Sinigalliano, C.; Stewart, J.;Breitbart, M. Pepper mild mottle virus as an indicator of fecalpollution. Appl. Environ. Microbiol. 2009, 75 (22), 7261−7267.(20) Gentry-Shields, J.; Rowny, J. G.; Stewart, J. R. HuBac and nifHsource tracking markers display a relationship to land use but notrainfall. Water Res. 2012, 46 (18), 6163−6174.(21) Ahmed, W.; Sidhu, J. P.; Toze, S. Evaluation of the nifH genemarker of Methanobrevibacter smithii for the detection of sewagepollution in environmental waters in Southeast Queensland, Australia.Environ. Sci. Technol. 2012, 46 (1), 543−550.(22) Kildare, B. J.; Leutenegger, C. M.; McSwain, B. S.; Bambic, D.G.; Rajal, V. B.; Wuertz, S. 16S rRNA-based assays for quantitativedetection of universal, human-, cow-, and dog-specific fecalBacteroidales: A Bayesian approach. Water Res. 2007, 41 (16),3701−3715.(23) Wang, D.; Silkie, S. S.; Nelson, K. L.; Wuertz, S. Estimating truehuman and animal host source contribution in quantitative microbialsource tracking using the Monte Carlo method. Water Res. 2010, 44(16), 4760−4775.(24) Whitman, R. L.; Przybyla-Kelly, K.; Shively, D. A.;Byappanahalli, M. N. Incidence of the enterococcal surface protein(esp) gene in human and animal fecal sources. Environ. Sci. Technol.2007, 41 (17), 6090−6095.(25) Scott, T. M.; Jenkins, T. M.; Lukasik, J.; Rose, J. B. Potential useof a host associated molecular marker in Enterococcus faecium as anindex of human fecal pollution. Environ. Sci. Technol. 2005, 39 (1),283−287.(26) Seurinck, S.; Defoirdt, T.; Verstraete, W.; Siciliano, S. D.Detection and quantification of the human-specific HF183 Bacteroides16S rRNA genetic marker with real-time PCR for assessment ofhuman faecal pollution in freshwater. Environ. Microbiol. 2005, 7 (2),249−259.(27) Stoeckel, D. M.; Harwood, V. J. Performance, design, andanalysis in microbial source tracking studies. Appl. Environ. Microbiol.2007, 73 (8), 2405−2415.

(28) Van Belle, G.; Fisher, L. D.; Heagerty, P. J.; Lumley, T.Biostatistics: a methodology for the health sciences, 2nd ed.; Wiley-interscience: Hoboken, NJ, 2004.(29) Wasserman, L. All of statistics: a concise course in statisticalinference; Springer Verlag: New York, 2004.(30) Ufnar, J. A.; Wang, S. Y.; Ufnar, D. F.; Ellender, R. D.Methanobrevibacter ruminantium as an indicator of domesticated-ruminant fecal pollution in surface waters. Appl. Environ. Microbiol.2007, 73 (21), 7118−7121.(31) Byappanahalli, M. N.; Przybyla-Kelly, K.; Shively, D. A.;Whitman, R. L. Environmental occurrence of the enterococcal surfaceprotein (esp) gene is an unreliable indicator of human fecalcontamination. Environ. Sci. Technol. 2008, 42 (21), 8014−8020.(32) DeGroot, M. H.; Schervish, M. J. Probability and Statistics, 3rded.; Addison Wesley: Boston, 2002.(33) Haugland, R. A.; Varma, M.; Sivaganesan, M.; Kelty, C.; Peed,L.; Shanks, O. C. Evaluation of genetic markers from the 16S rRNAgene V2 region for use in quantitative detection of selectedBacteroidales species and human fecal waste by qPCR. Syst. Appl.Microbiol. 2012, 33 (6), 348−357.(34) Gawler, A. H.; Beecher, J. E.; Brandao, J.; Carroll, N. M.; Falcao,L.; Gourmelon, M.; Masterson, B.; Nunes, B.; Porter, J.; Rince, A.;Rodrigues, R.; Thorp, M.; Walters, J. M.; Meijer, W. G. Validation ofhost-specific Bacteriodales 16S rRNA genes as markers to determinethe origin of faecal pollution in Atlantic Rim countries of the EuropeanUnion. Water Res. 2007, 41 (16), 3780−3784.(35) Jenkins, M. B.; Hartel, P. G.; Olexa, T. J.; Stuedemann, J. A.Putative temporal variability of Escherichia coli ribotypes from yearlingsteers. J. Environ. Qual. 2003, 32 (1), 305−309.(36) Scott, T. M.; Parveen, S.; Portier, K. M.; Rose, J. B.; Tamplin, M.L.; Farrah, S. R.; Koo, A.; Lukasik, J. Geographical variation in ribotypeprofiles of Escherichia coli isolates from humans, swine, poultry, beef,and dairy cattle in Florida. Appl. Environ. Microbiol. 2003, 69 (2),1089−1092.(37) Blyton, M. D. J.; Banks, S. C.; Peakall, R.; Gordon, D. M. Hightemporal variability in commensal Escherichia coli strain communitiesof a herbivorous marsupial. Environ. Microbiol. 2013, 15 (8), 2162−2172.(38) Parveen, S.; Lukasik, J.; Scott, T. M.; Tamplin, M. L.; Portier, K.M.; Sheperd, S.; Braun, K.; Farrah, S. R. Geographical variation inantibiotic resistance profiles of Escherichia coli isolated from swine,poultry, beef and dairy cattle farm water retention ponds in Florida. J.Appl. Microbiol. 2006, 100 (1), 50−57.(39) Ahmed, W.; Stewart, J.; Powell, D.; Gardner, T. Evaluation ofBacteroides markers for the detection of human faecal pollution. Lett.Appl. Microbiol. 2008, 46 (2), 237−242.(40) Layton, B. A.; Cao, Y.; Ebentier, D. L.; Hanley, K. T.; Balleste,E.; Brandao, J.; Byappanahalli, M. N.; Converse, R. R.; Farnleitner, A.;Gentry-Shields, J.; Gidley, M. L.; Gourmelon, M.; Lee, C. S.; Lee, J.;Lozach, S.; Madi, T.; Meijier, W.; Noble, R. T.; Peed, L.; Reischer, G.;Rodrigues, R.; Rose, J. B.; Schriewer, A.; Sinagalliano, C. D.;Srinivasan, S.; Stewart, J. R.; Van De Werfhorst, L.; Wang, D.;Whitman, R. L.; Wuertz, S.; Jay, J.; Holden, P.; Boehm, A. B.; Shanks,O. C.; Griffith, J. F. Performance of human fecal-associated PCR-basedassays: An international source identification method evaluation.WaterRes. 2013, 47 (18), 6897−6908.(41) Scott, T. M.; Rose, J. B.; Jenkins, T. M.; Farrah, S. R.; Lukasik, J.Microbial source tracking: Current methodology and future directions.Appl. Environ. Microbiol. 2002, 68 (12), 5796−5803.(42) Ufnar, J. A.; Ufnar, D. F.; Wang, S. Y.; Ellender, R. D.Development of a swine-specific fecal pollution marker based on hostdifferences in methanogen mcrA genes. Appl. Environ. Microbiol. 2007,73 (16), 5209−5217.



Documents

Probabilistic Analysis Showing That a Combination of Bacteroides and Methanobrevibacter Source Tracking Markers Is Effective for Identifying Waters Contaminated by Human Fecal Pollution