Presence of Bacteroidales as a Predictor of Pathogens in Surface ...

APPLIED AND ENVIRONMENTAL MICROBIOLOGY, Sept. 2010, p. 5802–5814 0099-2240/10/$12.00 doi:10.1128/AEM.00635-10 Copyright © 2010, American Society for Microbiology. All Rights Reserved.

Vol. 76, No. 17

Presence of Bacteroidales as a Predictor of Pathogens in Surface Waters of the Central California Coast䌤 Alexander Schriewer,1 Woutrina A. Miller,2* Barbara A. Byrne,2 Melissa A. Miller,3 Stori Oates,3 Patricia A. Conrad,2 Dane Hardin,4 Hsuan-Hui Yang,2 Nadira Chouicha,2 Ann Melli,2 Dave Jessup,3 Clare Dominik,4 and Stefan Wuertz1 Department of Civil and Environmental Engineering, University of California, Davis, One Shields Avenue, Davis, California 956161; Department of Pathology, Microbiology & Immunology, School of Veterinary Medicine, University of California, Davis, One Shields Avenue, Davis, California 956162; Marine Wildlife Veterinary Care and Research Center, California Department of Fish and Game, 1451 Shaffer Road, Santa Cruz, California 950603; and Applied Marine Sciences and Central Coast Long-term Environmental Assessment Network, P.O. Box 8346, Santa Cruz, California 950614 Received 12 March 2010/Accepted 7 July 2010

The value of Bacteroidales genetic markers and fecal indicator bacteria (FIB) to predict the occurrence of waterborne pathogens was evaluated in ambient waters along the central California coast. Bacteroidales host-specific quantitative PCR (qPCR) was used to quantify fecal bacteria in water and provide insights into contributing host fecal sources. Over 140 surface water samples from 10 major rivers and estuaries within the Monterey Bay region were tested over 14 months with four Bacteroidales-specific assays (universal, human, dog, and cow), three FIB (total coliforms, fecal coliforms, and enterococci), two protozoal pathogens (Cryptosporidium and Giardia spp.), and four bacterial pathogens (Campylobacter spp., Escherichia coli O157:H7, Salmonella spp., and Vibrio spp.). Indicator and pathogen distribution was widespread, and detection was not highly seasonal. Vibrio cholerae was detected most frequently, followed by Giardia, Cryptosporidium, Salmonella, and Campylobacter spp. Bayesian conditional probability analysis was used to characterize the Bacteroidales performance assays, and the ratios of concentrations determined using host-specific and universal assays were used to show that fecal contamination from human sources was more common than livestock or dog sources in coastal study sites. Correlations were seen between some, but not all, indicator-pathogen combinations. The ability to predict pathogen occurrence in relation to indicator threshold cutoff levels was evaluated using a weighted measure that showed the universal Bacteroidales genetic marker to have a comparable or higher mean predictive potential than standard FIB. This predictive ability, in addition to the Bacteroidales assays providing information on contributing host fecal sources, supports using Bacteroidales assays in water quality monitoring programs.

ducing health risks in recreational waters, as required by the Beaches Environmental Assessment and Coastal Health Act (5), which amended the Clean Water Act (11). Groups of standard FIB monitored in water include total coliforms (TC), fecal coliforms (FC), Escherichia coli bacteria, and enterococci. These bacterial groups have been considered indicators of health risks in epidemiologic and quantitative microbial risk assessment (QMRA) studies (38, 42, 59, 66). To date, many monitoring programs have focused only on FIB measurements and do not test for pathogens. However, substantial evidence has been collected that challenges the usefulness of FIB data alone. A few limitations of using standard FIB to represent pathogens in water include the fact that FIB have been shown to multiply in the environment, that they are not host specific, and that the absence of FIB is not necessarily evidence of pathogen absence (21, 50, 51, 56). Consequently, alternative indicators of fecal pollution that address the weaknesses of standard FIB are needed. Ideally, these indicators would decay at rates similar to those of pathogens, be present at high concentrations in fecal sources, and be present at low concentrations in unpolluted environments. Proposed alternative indicators include (i) anaerobic bacteria such as bifidobacteria (46), Clostridium perfringens (22), and

Coastal waters worldwide have been influenced by human activities for centuries, as they are adjacent to densely populated areas, provide a means of transportation, and receive substantial recreational use. Consequently, impairments in nearshore water quality can result from enrichment of the coastal marine ecosystem with pollutants and nutrients that are transported down watersheds from land to sea. This poses health risks to humans and animals. Microbial pollution is caused by fecal contamination from a variety of sources, including humans, livestock, pets, and wildlife, and fecal pathogen pollution has been associated with numerous outbreaks of waterborne disease (14, 15, 27, 41, 49, 55). Fecal indicator bacteria (FIB) that normally reside in the gastrointestinal tracts of humans and animals are used throughout the world to assess the microbiological quality of drinking and recreational waters. In the United States, FIB are used to define bacterial water quality standards aimed at re* Corresponding author. Mailing address: Department of Pathology, Microbiology & Immunology, School of Veterinary Medicine, University of California, Davis, One Shields Avenue, Davis, CA 95616. Phone: (530) 219-1369. Fax: (530) 752-3349. E-mail: wamiller @ucdavis.edu. 䌤 Published ahead of print on 16 July 2010. 5802

VOL. 76, 2010

INDICATORS AND PATHOGENS IN COASTAL WATERS

Bacteroidales (20); (ii) viruses such as F-specific RNA (F-RNA)-specific coliphages (39), phages infecting Bacteroides fragilis (30), and host-specific viruses (25); and (iii) chemical compounds such as fecal sterols (29). An added benefit of using alternative indicators is that, in some cases, host sources of fecal contamination can be identified. Over a decade ago, PCR-based assays were developed to detect Bacteroides in an effort to monitor human fecal pollution in the environment (36, 37). This approach was adopted by others and further advanced to identify host-specific Bacteroidales 16S rRNA gene markers for different fecal sources. This has resulted in PCR and quantitative PCR (qPCR) assays for the detection of human, dog, pig, and cow Bacteroidales markers (6, 7, 16, 34, 57) as well as assays for the detection of general Bacteroidales markers (7, 34). The analysis of Bacteroidales markers has been incorporated in microbial source tracking (MST) studies, particularly in the United States, Japan, and Europe (24, 45, 52–54, 64). The objective of this study was to compare the abilities of Bacteroidales markers and FIB to predict the occurrence of waterborne pathogens in riverine and estuarine waters in California and to use several statistical approaches to better characterize the strengths and limitations of the assays. We hypothesized that Bacteroidales and FIB would correlate with bacterial and protozoal pathogen detection in surface waters. To test this hypothesis, four Bacteroidales-specific assays (universal, human, dog, and cow), three types of FIB (total coliforms, fecal coliforms, and enterococci), two protozoal pathogens (Cryptosporidium and Giardia spp.), and four bacterial pathogens (Campylobacter spp., E. coli O157, Salmonella spp., and Vibrio spp.) were monitored monthly for 14 months in 10 streams, rivers, and estuaries feeding into the Monterey Bay region of California. MATERIALS AND METHODS Site selection and sampling design. A total of 143 water samples were collected from 10 rivers and estuaries in the Monterey Bay region along the central California coast (Fig. 1). The sites were selected on the basis of the major freshwater contribution to the bay and/or historically high coliform counts detected under the Central Coast Long-term Environmental Assessment Network (CCLEAN) monitoring program (10). Water samples used in this study were collected monthly from August 2007 through September 2008. Seasons were categorized as spring (March to May), summer (June to August), fall (September to November), and winter (December to February). Salinity data to categorize sites as freshwater (mean salinity ⬍ 0.5 ppt) or marine influenced (mean salinity ⬎ 0.5 ppt) were obtained from the Central Coast Ambient Monitoring Program website (http://www.ccamp.org/ca300/3/3.htm). All samples were tested for bacterial and protozoal pathogens, FIB, and Bacteroidales markers, as described below. Sample collection and processing. Water samples were collected at field sites during daylight hours and when outgoing tides minimized the marine influence at stream mouths. Water grab samples (20 liter) were gathered just below the water surface after sterile 2-liter collection bottles were prerinsed three times in the river. Duplicate samples were randomly selected for quality assurance. After collection, water samples were placed on ice and initially processed at the field laboratory on the same day for FIB testing and pathogen detection. To determine the presence and concentration of FIB in water samples, serial dilutions were made and filtered through sterile 0.45-␮m membranes that were placed on transport media as described in EPA Method 9222C (1). These delayed incubation filters were kept chilled and further processed at UC Davis within 24 h. For protozoal detection, Envirochek HV filters were used to concentrate up to 10 liters of water per EPA Method 1623 (60) and were kept chilled until further processing at UC Davis within 36 h. Additionally, 2- to 4-liter water samples were kept chilled overnight for processing within 36 h for Bacteroidales markers at UC Davis. Bacteroidales present

5803

FIG. 1. Surface water sampling sites in the Monterey Bay region of California.

in the water samples were concentrated using hollow-fiber ultrafiltration (HFF) as described previously (43). The surrogate Acinetobacter baylyi ADP1 was added into all samples. Filtration recoveries were calculated by measuring concentrations of A. baylyi in subsamples of prefiltration and postfiltration samples. The water was pumped through an ultrafiltration unit with a 50,000-MW membrane cutoff (Microza AHP 2010; Pall Life Sciences, East Hills, NY), until the volume was reduced to approximately 50 ml. Two elution steps of the filter module and the system with 0.05 M glycine-NaOH and 0.1% Tween 80 were performed to increase recovery of organisms. The combined retentate and eluates yielded a volume of 100 to 200 ml and were subjected to DNA extraction and subsequent quantification using qPCR techniques. Enumeration of indicator bacteria. Detection of indicator bacteria was performed using a delayed incubation modification of the standard membrane filtration method, as described in Standard Methods for the Examination of Water and Wastewater (1). Biochemical, serologic, and/or PCR techniques were used for final bacterial identification as described below. Total coliforms (TC) were enumerated as red colonies with a metallic sheen on m-Endo LES agar (Hardy Diagnostics, Santa Maria, CA) after 24 h of incubation at 35.5°C. Fecal coliforms (FC) were identified as blue colonies on m-FC medium (Hardy Diagnostics) after 48 h of incubation at 44.5°C. Enterococcus spp. (ENT) were identified as small light and dark red colonies grown on m-Enterococcus agar after 48 h of incubation at 35.5°C. Suspect Enterococcus colonies were further characterized as esculin-positive colonies with biochemical testing on bile esculin agar (BEA) plates and growth in 6.5% NaCl broth. All counts were standardized to a 100-ml sample volume for statistical analyses. Enumeration of pathogenic bacteria. Isolation of pathogenic bacteria from riverine and estuarine samples targeted Campylobacter spp., E. coli O157:H7, Salmonella spp., and Vibrio spp. The general methodology for pathogen detection and enumeration utilized primarily serial dilutions with membrane filtration of water and delayed incubation of filters on selective media. Enumeration of Campylobacter spp. was done as outlined in Standard Methods for the Examination of Water and Wastewater, with slight modifications (1, 17). Briefly, a 100-ml volume of water was filtered through a 0.45-␮m-pore-size, 47-mm-diameter cellulose nitrate membrane filter. A second similar filter was placed in Campy thioglycolate broth (Hardy Diagnostics) for 24 h under

5804

SCHRIEWER ET AL.

APPL. ENVIRON. MICROBIOL. TABLE 1. Primers used in qPCR assays

Primer

Universal Bacteroidales BacUni-520f BacUni-690r1BacUni-690r2 BacUni-656p

Oligonucleotide sequence (5⬘–3⬘)a

34 CGTTATCCGGATTTATTGGGTTTA CAATCGGAGTTCTTCGTGATATCTA AATCGGAGTTCCTCGTGATATCTA FAM-TGGTGTAGCGGTGAAA-TAMRA-MGB

Human Bacteroidales BacHum-160f BacHum-241r BacHum-193p

TGAGTTCACATGTCCGCATGA CGTTACCCCGCCTACTATCTAATG FAM-TCCGGTAGACGATGGGGATGCGTT-TAMRA

Cow Bacteroidales BacCow-CF128f BacCow-305r BacCow-257p

CCAACYTTCCCGWTACTC GGACCGTGTCTCAGTTCCAGTG FAM-TAGGGGTTCTGAGAGGAAGGTCCCCC-TAMRA

Dog Bacteroidales BacCan-545f1 BacUni-690r1 BacUni-690r2 BacUni-656p

GGAGCGCAGACGGGTTTT CAATCGGAGTTCTTCGTGATATCTA AATCGGAGTTCCTCGTGATATCTA FAM-TGGTGTAGCGGTGAAA-TAMRA-MGB

Acinetobacter Acinet-137F Acinet-210R Acinet-159p

GATGCAACGCGAAGAACCTTA TTCCCGAAGGCACCAATC FAM-CTGGCCTTGACATAGTAGAAACTTTCC-TAMRA

a

Reference

34

6 34 34 34

This study

FAM, 6-carboxyfluorescein; TAMRA, 6-carboxytetramethylrhodamine, MGB, minor groove binder.

microaerophilic conditions at 37°C. Each filter was placed face down on Campy agar containing cefoperazone, vancomycin, and amphotericin B (Campy-CVA) with 5% sheep blood (Hardy Diagnostics) and incubated for 24 h at 37°C under microaerophilic conditions (Pack-Micro Aero; Mitsubishi Gas Chemical Company). The filter was then placed face up on another CVA agar plate, and both plates were incubated again at 37°C. Plates were read for the presence of Campylobacter spp. after 48 h. Suspect Campylobacter sp. colonies were small, grayish brown, and smooth or flat, mucoid, and gray irregularly edged translucent colonies that Gram stained as Gram-negative curved rods. Biochemical tests, including oxidase, catalase, hippurate, and nalidixic acid (NA) sensitivity and cefoxitin (CF) sensitivity, as well as Campylobacter genus-specific PCR, were used to confirm isolates (4). Screening for Escherichia coli O157:H7 with immunomagnetic separation (IMS) began in September 2007. Briefly, a Dynabeads (Invitrogen Corporation, Carlsbad, CA) anti-E. coli O157:H7 IMS technique was used to concentrate suspect bacteria, followed by plating of separated bacteria on CHROMagar O157:H7 (Hardy Diagnostics) according to the manufacturer’s instructions. Suspect colonies were further characterized using biochemical testing and PCR with primers specific to the O157 and H7 genes (23, 28). Detection of Salmonella spp. used filter incubation on xylose lysine desoxycholate (XLD) medium and 24 h of incubation at 35.5°C. A second filter was preenriched in selenite broth for 24 h at 37°C prior to subculturing 100 ␮l onto XLD agar. Salmonella spp. were initially identified on XLD medium by the presence of red colonies with black centers. An additional modification of the EPA method (61) involved filtering 100 ml of water, enriching the filters in tryptic soy broth, and subculturing onto criterion-modified semisolid RappaportVassiliadis agar. Suspect Salmonella spp. were confirmed with triple sugar iron (TSI) agar (alkaline-acid-H2S production) and were o-nitrophenyl-␤-D-galactopyranoside (ONPG) negative, oxidase negative, urease negative, and indole negative. Salmonella was further characterized with Poly O (A-E) antiserum and group-specific antiserum (BD Difco). Vibrio spp. were isolated by incubation of filters on thiosulfate-citrate-bile salt-sucrose agar (Hardy Diagnostics) for 24 h at 35.5°C. Large and small yellow colonies were consistent with the isolation of Vibrio cholerae and Vibrio alginolyticus; small green colonies were consistent with the isolation of Vibrio parahaemolyticus. Upon further biochemical testing, V. cholerae strains showed an alkaline slant/acid butt or acid slant/acid butt on TSI agar after incubation at 37°C for 18 to 24 h, were oxidase and ONPG positive, and fermented sucrose. Vibrio cholerae confirmation on a subset of isolates utilized API 20E strips,

O1-specific antisera, and PCR assays targeting Vibrio spp., V. cholerae, and ctx toxin genes (35, 63). Vibrio parahaemolyticus was confirmed for production of oxidase, reaction in TSI agar (alkaline slant/acid butt), urease production, hemolysis on sheep blood agar, and species-specific PCR (63). All counts were standardized to CFU per 100 ml for statistical analysis. Enumeration of protozoa. Cryptosporidium oocysts and Giardia cysts were quantified according to EPA Method 1623 (60). This involved pumping up to 10 liters of water through the Envirochek HV filter (Pall Gelman Laboratory, Ann Arbor, Michigan), eluting the filter using a hand-wrist shaker, concentrating the parasites using Dynabeads (Invitrogen Corporation) immunomagnetic separation (IMS), and finally enumerating oocysts and cysts using a direct fluorescent antibody (DFA) technique with BTF kit reagents (BTF, Precise Microbiology, Sydney, Australia). Cryptosporidium oocysts were identified as ⬃5-␮m spheres outlined in apple green and often with a mid-line seam, whereas Giardia cysts were also apple green but oval and 9 to 14 ␮m long. Slides were viewed with fluorescein isothiocyanate (FITC) and DAPI (4⬘,6-diamidino-2-phenylindole) stains on an Axioskop epifluorescent microscope. Organisms were visualized at 20⫻ magnification, and identification was confirmed at 40⫻ magnification. All slides were read by the same microscopist. Nucleic acid extraction. Genomic DNA (gDNA) was extracted from 10 ml of HFF concentrate obtained from river and estuarine water samples. QIAamp DNA stool kits (Qiagen, Valencia, CA) were used, according to the manufacturer’s directions, with slight modifications by using a homemade guanidine isocyanate-based lysis buffer instead of buffer AL and centrifuging samples prior to application to spin filters (9, 34). Bacteroidales and Acinetobacter qPCR. Each 25-␮l PCR mixture contained 12.5 ␮l of commercially available TaqMan PCR mastermix (Eurogentec, San Diego, CA) with 400 nM (each) of forward and reverse primers and 80 nM of probe for the respective TaqMan system. For all TaqMan reactions, 10 ␮l of the diluted gDNA sample was assayed in a final reaction volume of 25 ␮l. In order to suppress inhibitors, bovine serum albumin (BSA) was added to each reaction in a final concentration of 50 ng/␮l, and four serial dilutions were performed to assess inhibition factors. Cycling conditions used were 2 min at 50°C and 10 min at 95°C, followed by 40 cycles of 15 s at 95°C and 60 s at 60°C, using an ABI Prism 7000 (Applied Biosystems, Carlsbad, CA). Bacteroidales assay primers (Table 1) and specificity were established previously (34). Concentrations and sample limits of detection (SLODs) were analyzed according to Rajal et al. (44). These SLODs are individual limits of detection for each sample and account for concentration factors, recoveries, and qPCR

VOL. 76, 2010


5805

TABLE 2. Prevalence of bacterial and protozoal pathogens in monthly surface water samples by site in the Monterey Bay region of California, 2007-2008 % of water samples positive for: Site

Marine influencea

Waddell Creek Scott Creek San Lorenzo River Soquel Creek Watsonville Slough Pajaro River Elkhorn Slough Salinas River Carmel River Big Sur River

Yes No Yes No No Yes Yes No No No

Overall % positive

V. parahaemolyticus

V. cholerae

V. alginolyticus

Salmonella spp.

Campylobacter spp.

E. coli O157:H7

Cryptosporidium spp.

Giardia spp.

—b — 14 — 21 7 14 — 7 —

50 43 64 29 79 79 21 64 14 —

— — — 21 14 — 43 — — —

21 13 21 7 — — 7 — — —

— 7 — 7 21 — 7 7 — —

— — — — — — — — — —

— 21 14 29 29 21 23 21 14 8

21 14 43 29 7 29 15 21 21 8

6

44

8

7

5

—

18

21

a

Considered marine influenced if site’s mean salinity was above 0.5 ppt. b —, not detected.

inhibition to help evaluate “nondetects.” The SLOD values in gene copies/ml (GC/ml) are calculated as follows: SLOD ⫽

1,000 䡠 ALOD 䡠 I Cfiltr 䡠 Cextr 䡠 R

where ALOD (in GC/␮l) is the assay limit of detection for the applied assay and specific conditions, C indicates concentration factors for filtration (Cfiltr) or nucleic acid extraction (Cextr), and I indicates the inhibition factor expressed as the inverse of the dilution factor. The overall recovery proportion, R, is assessed by measurement of known spike doses of a bacterial surrogate, Acinetobacter baylyi strain ADP1 (62), previously referenced as Acinetobacter sp. strain ADP1 (31). Quality assurance. A field blank and sample blank were run using membrane filtration techniques for each time point and site processed, respectively. For hollow-fiber ultrafiltration, laboratory blanks were processed with each sampling set. For each second qPCR plate, a negative control was measured. One duplicate field sample was run during each sampling period. Statistical methods. Data were compiled and analyzed using Excel 2007 (Microsoft Corp.). Box plots were created to show the medians, 25th and 75th percentiles, 10th and 90th percentiles, and data outliers using SigmaPlot version 10 (Systat Software Inc.). Kendall and Spearman correlations, as well as logistic regression analyses using Nagelkerke’s R square (26), were performed using SPSS version 14 (SPSS Inc.). Predictive qualifier. As a weighted measure of how well pathogen occurrence can be predicted by the presence of an indicator organism, a predictive qualifier (PQ) for different threshold cutoff levels of the indicator was calculated, using indicator and pathogen presence/absence: predictive qualifier 共%兲 ⫽

TOTtrue 共%兲 ⫹ POStrue 共%兲 2

In the equation, TOTtrue is the total percentage of matches that represent either simultaneous indicator and pathogen presence or simultaneous indicator and pathogen absence, and POStrue is the percentage of matches when pathogens were positive and matched correctly by indicators. For example, at a low-level threshold cutoff concentration of 1 GC/ml for universal Bacteroidales, the prediction would be that 100% of the samples are positive for pathogens, because universal Bacteroidales markers were detected in 100% of the samples above 1 GC/ml. At a certain threshold, most of the indicator samples will predict a pathogen nondetect, and therefore, the percentage of total matches will be rather high when most of the samples are also nondetects for pathogens. At some optimal threshold, both pathogen presence and absence will be matched by indicator occurrence at or above this threshold. This simple equation helps to qualify the predictive value for thresholds of an indicator, with emphasis on pathogen occurrence. Conditional probability analysis. To determine whether the host-specific assays correctly measure the organism for which they were designed, Bayes’ Theorem was used as previously reported (34) for the calculation of the probability of a given source of contamination existing in a water sample, given that a

positive test result is obtained; likewise, the probability that a given source of contamination in a water sample exists when a negative test result is obtained was also calculated. To apply Bayes’ Theorem, the assays were assumed to be independent discrete random variables, as is appropriate for presence/absence data. Conditional probabilities for each of the assays, predictive values, and prevailing (background) rates for each set of assay results from this study were calculated using validation results obtained by Kildare et al. (34). The conditional probability gives the probability that the signal of the tested assay actually results from the targeted host source. The diagnostic sensitivity is the proportion of samples that correctly tested positive over all the samples that actually experienced fecal contamination from the targeted host source. Diagnostic specificity is the proportion of samples that correctly tested negative over the total number of samples that actually did not experience fecal contamination from the targeted host source. The positive predictive value of the test is the proportion of samples that correctly tested positive over the total number of samples that tested positive. Lastly, the negative predictive value of the test is the proportion of samples that are correctly negative over the total number of samples that test negative.

RESULTS Pathogens in surface water. Pathogens were detected at freshwater sites and at sites with a tidal marine influence (Table 2). Out of all surface water samples tested, Salmonella spp. (7%) and Campylobacter spp. (5%) were detected in less than 10% samples, while E. coli O157:H7 was not detected in any samples. Vibrio spp. were commonly detected and included V. cholerae (tested positive in 44% of all surface water samples), V. parahaemolyticus (6%), and V. alginolyticus (8%). Vibrio cholerae was detected at all study sites except the Big Sur River and had a median concentration of 23 CFU/100 ml in positive samples, with all isolates negative for the ctx toxin gene. Vibrio parahaemolyticus was detected at half of the study sites, with a median concentration of 55.5 CFU/100 ml. Overall, the concentrations ranged from 0.2 to 1,980 CFU/100 ml for V. cholerae, from 0.2 to 720 CFU/100 ml for V. parahaemolyticus, and from 0.4 to 720 CFU/100 ml for V. alginolyticus. None of the target bacteria were detected at the Big Sur River, the southernmost site and one of the least developed sites. Target bacteria distribution was not highly seasonal, with detection occurring sporadically throughout the year, other than for Vibrio, which was most prevalent in the summer. The fecal protozoal pathogens Cryptosporidium spp. and Giardia spp. were detected in 18% and 21% of all surface

5806

SCHRIEWER ET AL.

water samples, respectively (Table 2). Cryptosporidium oocysts were detected at 9 of 10 study sites, and Giardia cysts were detected at all 10 sites. Cryptosporidium parasites were detected most often (in 29% of samples) at Soquel Creek and Watsonville Slough, two sites in the central Monterey Bay region that have urban and rural uses upstream of sampling sites. Giardia cysts were detected most often (in 43% of samples) in water samples from the San Lorenzo River, another highly developed tributary to the bay. For Cryptosporidium spp., the range of detected concentrations in positive samples was low, from 1 to 9 oocysts/10 liters, with a median value within positive samples of 2 oocysts/10 liters. The concentrations of Giardia cysts in positive samples ranged from 1 to 58 oocycts/10 liters, with a median value of 1 cyst/10 liters. Cryptosporidium spp. were detected throughout the year, with the most positive samples detected in November and February (40%), while no seasonal trends were observed for Giardia spp., with cysts detected in 10 to 40% of samples throughout the year. FIB in surface waters. Indicator bacteria measured in riverine and estuarine samples included total coliforms (TC), fecal coliforms (FC), and Enterococcus spp. (ENT). As expected for watersheds with heavy usage, TC could be detected in all samples, FC in 99% of samples, and ENT in 75% of samples. The highest levels were observed in surface waters from the high-use central bay tributaries, including Watsonville Slough and the San Lorenzo River. Overall, indicator bacterial concentrations showed no clear seasonal trends (Fig. 2). The seasonal medians for TC ranged from 50 CFU/100 ml in winter (December to February) to 198 CFU/100 ml in fall (September to November); for FC, medians ranged from 29.5 CFU/100 ml in spring (March to May) to 84 CFU/100 ml in fall; and for ENT, medians ranged from 10 CFU/100 ml in summer (June to August) to 20 CFU/100 ml in winter. The occurrence of lower FIB counts in winter coincided with Vibrio sp. detection, which was also lowest in winter. However, FIB results were seen to diverge from Vibrio results at Elkhorn Slough, an estuarine site with the largest marine influence, where FIB concentrations were low, and 75% of samples were positive for Vibrio spp. The highest concentrations of both FIB and Vibrio were observed at Watsonville Slough, a sampling site with minimal marine tidal influence. Bacteroidales and microbial source tracking in surface waters. Like FIB, the universal Bacteroidales (100%) marker was detected in all samples. The human Bacteroidales marker was detected in 37% of samples, and the cow (8%) and dog (6%) Bacteroidales markers were detected in less than 10% of samples. Median concentrations of universal Bacteroidales markers were highest at the Salinas River (34,805 GC/ml) and Watsonville Slough (19,037 GC/ml) sites. Universal Bacteroidales marker concentrations were lowest at the Big Sur River (937 GC/ml) and Scott Creek (1,126 GC/ml) sites located at the less-developed outer edges of the Monterey Bay region (Fig. 2). Overall, detected concentrations ranged from 87 to 1.3 million GC/ml for universal Bacteroidales markers, 45 to 17,268 GC/ml for human markers, 3 to 92 GC/ml for cow markers, and 12 to 575 GC/ml for dog markers. As with FIB, no marked seasonality for Bacteroidales markers was observed, with respect to prevalence or concentration (Fig. 2). Human Bacteroidales markers were detected most fre-

APPL. ENVIRON. MICROBIOL.

FIG. 2. Box plots of site-specific variation of universal Bacteroidales concentrations from north to south (top) and seasonal variation for universal Bacteroidales (light gray) and fecal coliforms (dark gray) (bottom). Upper and lower bounds of boxes denote the 75th and 25th percentiles. Upper and lower bars show the 90th and 10th percentiles, with outliers represented by filled circles. Note the logarithmic vertical axis.

quently at the San Lorenzo River study site (57% of water samples were positive), and that site also had the highest ratio of human to universal Bacteroidales concentrations (0.22), as shown in Table 3. Watsonville Slough was the site with the highest pathogen and FIB concentrations, and it had the second highest median concentration of universal Bacteroidales markers among the sites. All three Bacteroidales host markers were detected at this site, which is consistent with known fecal inputs from human, cow, and dog sources in the area. All three Bacteroidales markers were also detected in water samples from Soquel Creek, a highly developed suburban site where Cryptosporidium oocysts were most frequently detected. In this study, less than 1% of the universal Bacteroidales signal was made up of cow- or dog-specific marker signals. On average, the human-specific assay signals for each site made up 1 to 22% of the universal signals, which is still a minority of the total Bacteroidales load in the surface water samples and suggests that noncharacterized fecal sources such as wildlife or

VOL. 76, 2010


5807

TABLE 3. Site characteristics with respect to land uses, main fecal sources upstream, and detection of Bacteroidales markers in surface waters

Site ID

Major land uses

Waddell Creek Scott Creek San Lorenzo River Soquel Creek Watsonville Slough Pajaro River Elkhorn Slough Salinas River Carmel River Big Sur River

Major host fecal source(s)b

a

RR, REC RR, AG, REC URB, RR, REC URB, RR, REC URB, RR, AG, REC URB, RR, AG, REC RR, AG, REC URB, RR, AG, REC URB, RR, AG, REC RR, AG, REC

W L, W H, P, W H, P, L, W H, P, L, W H, P, L, W H, P, L, W H, P, L, W H, P, L, W H, L, W

Overall % positive

No. of samples

14 16 14 14 14 14 15 14 16 12

% of water samples positive for markersc

Mean host Bacteroidales marker/ universal Bacteroidales marker concentration ratio

Human

Cow

Dog

Human

Cow

Dog

36 25 57 21 29 21 47 36 44 42

— — 14 7 7 — — — — —

7 — — 21 7 7 — 7 — 8

0.12 0.10 0.22 0.11 0.09 0.10 0.06 0.08 0.16 0.15

⬍0.01 ⬍0.01 ⬍0.01 ⬍0.01 ⬍0.01 ⬍0.01 ⬍0.01 ⬍0.01 ⬍0.01 ⬍0.01

⬍0.01 ⬍0.01 ⬍0.01 ⬍0.01 ⬍0.01 ⬍0.01 ⬍0.01 ⬍0.01 ⬍0.01 ⬍0.01

36

3

6

URB ⫽ urban development; RR ⫽ rural residential development; AG ⫽ agriculture, including crops, rangeland, and livestock; REC ⫽ recreational areas, including forest and trails. b H ⫽ humans; P ⫽ pets, such as dogs and cats; L ⫽ livestock; W ⫽ wildlife. c —, not detected. a

other domestic animals predominate. The median sample-specific detection limits (SLODs) were relatively low, as follows: 24 GC/ml (universal, cow, and dog markers) and 84 GC/ml (human markers). The SLOD ranges were relatively wide, with values from 3 to 3,155 GC/ml for universal, cow, and dog markers and from 9 to 10,842 GC/ml for human markers. Recoveries of spiked surrogates were relatively good for environmental samples (median of 66%), and inhibition factors were relatively low (median of 3) (44), but for some samples, recoveries below 10% were observed, which are reflected in high SLOD values. Conditional probabilities. The ability to use host-specific fecal assays for microbial source tracking is strongly linked to the specificity of the assays. Based on validation of host-specific assays using known fecal samples, the conditional probability, diagnostic sensitivity and specificity, and positive and negative predictive values that fecal contamination from a given source is present or absent in a water sample can be calculated (Table 4). By combining validation data that we have previously obtained (34) with our study-specific results, the conditional probability for the universal marker is determined to be 100%, which means that it is virtually certain that samples testing positive for the universal Bacteroidales assay truly have the host feces present in the sample. The probabilities for human- and dog-specific assays were 75% and 15%, respectively. This can be interpreted as a relatively high (75%) probability that water

TABLE 4. Results from four 16S rRNA-based qPCR assays in surface waters Assay results Bacteroidales marker assay Universal Human Cow Dog

Predictive value

Conditional probability

Sensitivity

1.00 0.75 0.19 0.15

1.00 0.84 0.73 0.74

Positive

Negative

Prevailing rate

1.00 0.75 0.19 0.15

1.00 0.92 0.95 0.98

1.00 0.32 0.11 0.05

Specificity 1.00 0.87 0.63 0.78

samples testing positive with the human-specific assay truly have human feces present. In contrast, the dog-specific assay reliability is lower, with only a 15% probability of dog feces truly being present in samples testing positive, based on previous specificity data. Likewise, for the cow-specific assay, the conditional probability is only 19% due to the crossspecificity with 37% of horse samples reported previously (34) and the relatively low prevailing rate of detection in this study (Table 4). Negative Bacteroidales assay results are estimated to be 100% reliable for the universal marker, 92% reliable for the human marker, 98% reliable for the dog marker, and 95% reliable for the cow Bacteroidales marker detection. Thus, in this study, the likelihood of human, dog, or cow fecal contamination being present in a surface water sample when a negative test result was obtained is 8%, 2%, and 5%, respectively. As additional specificity and prevailing host data are obtained, updated conditional probabilities can be calculated. When considering the actual presence/absence observations for Bacteroidales-specific markers in this study, we have a very high likelihood that all of the universal assay results are true positives. For samples testing positive with the human-specific assay, it is possible that there could be some false-positive test results, and we estimate that at least 39 of the 53 positive test results are truly positive for human feces. We have reduced confidence in the results from the dog- and cow-specific assays because these assays may also detect human- and horse-derived fecal pollution, respectively. However, both the dog and cow assay have excellent negative predictive values of 0.98 and 0.95, respectively, which means that the absence of detection gives us high confidence that no dog or cow fecal pollution was present. An external demographic or chemical measure of host distribution at our study sites would allow for additional insights into contributing host sources. Correlation of indicators and pathogens. Many of the data sets were nonnormally distributed, partially due to the fact that the majority of samples tested negative for the selected pathogens. Thus, nonparametric Kendall and Spearman correlation

5808

SCHRIEWER ET AL.


TABLE 5. Significant Spearman correlations between Bacteroidales markers, indicator bacteria, and pathogens for all sites combined Predictora

All data Universal Bact

Comparatorb

No. of samples

Spearman values R

P

Dog Bact Cryptosporidium spp.

143 137

0.227 0.213

0.006 0.013

Human Bact

Total coliforms Fecal coliforms Enterococci

137 137 137

0.264 0.310 0.224

0.002 ⬍0.001 0.008

Cow Bact

Dog Bact

143

0.240

0.004

Total coliforms

Fecal coliforms Enterococci V. cholerae Vibrio spp.

138 138 138 138

0.719 0.331 0.245 0.254

⬍0.001 ⬍0.001 0.004 0.003

Fecal coliforms

Enterococci Cryptosporidium spp. V. cholerae Vibrio spp.

138 138 138 138

0.434 0.180 0.240 0.225

⬍0.001 0.035 0.005 0.008

Human Bact Fecal coliforms

53 142

0.412 0.200

0.002 0.017

Total coliforms

Fecal coliforms Enterococci

142 112

0.725 0.626

⬍0.001 ⬍0.001

Fecal coliforms

Enterococci Cryptosporidium spp.

112 25

0.602 0.453

⬍0.001 0.023

Detects only Universal Bact

a

Universal, human, or cow Bacteroidales marker (Bact). Nondetects (test-negative samples) were included as zero values (all data) or omitted (detects only). b

analyses were performed and showed the same trends when calculated by (i) using the complete data sets with nondetects (negative test results) replaced by zeros and (ii) using only positive test result data. For the former case, detection of the universal Bacteroidales marker correlated significantly with confirmation of Cryptosporidium spp., while total coliforms and fecal coliform detection correlated significantly with that of Cryptosporidium spp. and V. cholerae (Table 5). Correlation of enterococcal detection with that of any other pathogen was not observed. When organisms are detected in a minority of samples, then the many nondetects match with each other during a correlation analysis between the organisms. Given the fact that no detection of a certain parameter does not mean that the parameter is absent, all nondetects for the latter case described above were removed to see if there was a relationship between positives test results alone. We consider this a valid approach under the assumption that parameters present but not detected might also correlate. In this case, there was no significant correlation between detection of universal Bacteroidales markers and isolation of any pathogen. For the FIB, the only correlation that remains is between fecal coliforms and Cryptosporidium spp. (Spearman’s r ⫽ 0.453; P ⫽ 0.023), while levels of enterococci and host-specific Bacteroidales markers were not significantly correlated with specific pathogen detec-

tion. The only significant correlation between Bacteroidales and pathogen levels was the concentration ratio of human to universal Bacteroidales markers (using any ratio ⬎ 0) and Vibrio cholerae detection, with an R value of 0.529 (P ⫽ 0.017). Sparse data precluded correlation analysis for V. parahaemolyticus and V. alginolyticus (n ⱕ 5). Using binary logistic regression with presence/absence data, no association between detection of the human Bacteroidales marker and isolation of any Vibrio spp. (maximum Nagelkerke’s R square ⫽ 0.057) was found. Predictive qualifier. Pathogens were categorized as positive or negative in surface water samples in order to calculate predictive qualifier (PQ) values at various thresholds of indicator bacterial concentrations. This PQ approach provided insights into how threshold cutoffs for categorizing indicator (FIB and Bacteroidales) data are related to specific pathogen detection (presence/absence). Categorization of indicator data can be useful, for example, when TC or universal Bacteroidales signals were detected in almost every sample, and so presence/ absence data alone would not provide informative data. The PQ gives a pathogen presence-weighted percentage of how often indicator presence cooccurs with pathogen detection when certain indicator threshold concentrations are applied. For environmental samples, the question of what percentage exceeds randomness is not trivial to answer, as different factors can cooccur. For example, cooccurrence of indicators and pathogens at a certain sampling point in a river might simply be the result of a confluence of two streams with separate origins of indicators and pathogens further upstream. In this study, we considered predictive qualifier values of ⱖ66% (PQ66) noteworthy because they are above what might be considered random. If a specific pathogen from a given location was present only in a few samples, and the indicator was never detected at that site, then total matches would be close to 100% (TOTtrue), but pathogen occurrence in relation to indicators would be lowered (POStrue ⫽ 0%), leading to a PQ below 50%. To achieve a PQ above 66% with averaging, both TOTtrue and POStrue must be close to or above 66%, which means that two of three cases are predicted correctly. The PQ66 target level (or greater) was reached for specific threshold concentration ranges of universal Bacteroidales and FIB but not for human-specific Bacteroidales with several pathogens (Fig. 3 and 4). PQ66 ranges for all study sites combined were observed for universal Bacteroidales with V. alginolyticus (1,400 to 1,499 GC/ml), V. cholerae (100 to 800 GC/ml), and Vibrio spp. (100 to 1,599 GC/ml). When separating sites by their average salinity, the most apparent change was that prediction of the presence of Cryptosporidium spp. in marineinfluenced sites increased from a PQ of 66% to up to ⬎77% (between 1,100 and 6,299 GC/ml). Given the small number of dog- and cow-positive Bacteroidales samples, their PQ performance was not evaluated. Total coliforms showed the highest PQ values, with overlapping ranges for all three Vibrio species analyzed individually as well as for their combined signals in freshwater. The PQ66 range was from 10 to 399 CFU/100 ml in freshwater and from 10 to 199 CFU/100 ml with all sites pooled. Both ranges span beyond the median value of 117 CFU/100 ml for total coliformpositive samples. A similar range for Salmonella spp. in marine-influenced sites (20 to 239 CFU/100 ml) was observed.

VOL. 76, 2010


5809

FIG. 3. Threshold concentration ranges for universal Bacteroidales (top) and total coliforms (bottom), with resulting PQ values greater than 66% for data from all sites combined and separated by average site salinity (freshwater salinity of ⬍0.5 ppt). Maximum PQ values are shown in parentheses. Vibrio parahaemolyticus found at marine-influenced sites is not displayed because only one positive sample was detected. The vertical dotted line indicates the median value of all measured samples.

5810

SCHRIEWER ET AL.


FIG. 4. Threshold concentration ranges for fecal coliforms (top) and enterococci (bottom), with resulting PQ values greater than 66% for data from all sites combined and separated by average salinity (freshwater salinity of ⬍0.5 ppt). Maximum PQ values are shown in parentheses. Vibrio parahaemolyticus found at marine-influenced sites is not displayed because only one positive sample was detected. The vertical dotted line indicates the median value of all measured samples.

VOL. 76, 2010


Fecal coliforms were the only indicators that showed a PQ66 for the same pathogens in salinity-separated sites and pooled sites, specifically for Cryptosporidium spp., Salmonella spp., and among Vibrio spp., for V. parahaemolyticus and V. cholerae. While for Salmonella the highest PQ value (88.4%) and the largest PQ66 range (10 to 239 CFU fecal coliforms per 100 ml) were reached at marine influenced sites, similar trends were observed at freshwater sites for V. cholerae (10 to 179 CFU/100 ml; maximum PQ value [PQmax] ⫽ 76.4%), V. parahaemolyticus (20 to 149 CFU/100 ml; PQmax ⫽ 83.5%), and Vibrio spp. overall (10 to 169 CFU/100 ml; PQmax ⫽ 79.1%). When all sites were pooled, enterococci reached PQ66 for Salmonella spp. (10 to 39 CFU/100 ml; PQmax ⫽ 74.2%) and for V. parahaemolyticus (40 to 79 CFU/100 ml; PQmax ⫽ 73.6%). To compare the predictive abilities of indicators, mean PQ values for V. cholerae, Cryptosporidium spp., and Giardia spp. were chosen for evaluation because these pathogens were detected most commonly. For the elimination of outliers, PQ values for each indicator within a threshold concentration range from zero (nondetect) to the 75th percentile of the measured concentrations were considered. Fecal coliforms exhibited the highest mean PQ value for this comparison overall (55.6%), followed closely by enterococci (54.5%), total coliforms (54.0%), and universal Bacteroidales (51.1%). When the same criteria were applied to marine-influenced sites only, universal Bacteroidales showed the highest mean PQ value (52.3%), followed by enterococci (50.8%), fecal coliforms (48.9%), and total coliforms (47.0%). For single pathogens, the mean PQ values can also vary significantly. For example, the highest mean PQ value for Cryptosporidium spp. was 66.7% when predicted by universal Bacteroidales markers in samples from marine-influenced sites compared to 55.2% in freshwater samples. DISCUSSION This study was the first in California to compare detection of Bacteroidales markers and fecal indicator bacteria in ambient water samples, with concurrent isolation of bacterial and protozoal waterborne pathogens. A suite of laboratory screening tests were utilized to detect and quantify the presence of standard fecal indicator bacteria, Bacteroidales as alternative indicator bacteria useful for microbial source tracking, and selected pathogens in impaired and more pristine surface waters of the central coast over 14 months. Significant correlations were observed for some, but not all, indicator and pathogen combinations. The calculation of a new parameter, the predictive qualifier (PQ), allowed for comparison and evaluation of both conventional and alternative indicator bacteria as predictors for selected pathogens. This was possible because the PQ gives a pathogen presence-weighted percentage of how often indicator presence cooccurs with pathogen detection when certain indicator threshold concentrations are applied. There was widely distributed pathogen occurrence across the study sites. The protozoa Cryptosporidium and Giardia were detected at low levels at most study sites, similar to the findings of other environmental studies by Miller et al. (40) and those reviewed by Fayer (19). Contributing to the seemingly ubiquitous nature of these fecal protozoa in surface waters is the fact that protozoal oocysts and cysts are known to survive

5811

well in aquatic environments (32, 48) and are shed in the feces of a wide range of human and animal hosts (19). With respect to bacterial pathogens, Salmonella and Campylobacter spp. were detected infrequently, while Vibrio species were commonly detected and varied seasonally, consistent with previous studies reporting variation in relative bacterial abundance attributable to local environmental conditions (8, 33, 47). The fact that protozoa were detected more frequently than bacterial pathogens could be due to different host loading dynamics and survival in the environment or due to differences in detection methods, such as sample volume analyzed and competitive growth of other microorganisms. While universal Bacteroidales markers, total coliforms, and fecal coliforms were detected in at least 99% of samples, enterococci were detected in only 75% of the samples, human Bacteroidales markers in 37%, and dog- or cow-specific markers in less than 10%. These findings are consistent with uncharacterized, nonhuman sources such as wildlife or other domestic animals making a large contribution to fecal loading of the various water bodies evaluated during this study. Such a scenario is plausible, given the large populations of wild birds and small mammals that reside along the waterways studied, and the significant number of backyard animals, including small ruminants, poultry, and cats, that are known to be present in these areas. The concentration ratio of host to universal Bacteroidales markers is only a rough estimate of the proportion of the total Bacteroidales signal attributable to each host source because ratios can vary in individual fecal samples within a host species, and ratios are assumed to be constant over time. There is supporting evidence for the latter based on controlled decay studies of marine waters using the same Bacteroidales assays as those used in the present study (2). The change in a ratio of a specific marker can be useful for comparisons between sites and time points. As development and validation of Bacteroidales assays progress, our ability to characterize contributing host sources will improve, and additional host-specific assays may be used to help facilitate source-tracking efforts. All tested rivers had some degree of human activity; detection of both human and universal Bacteroidales markers at all sites suggests that current management of human fecal waste is not preventing contamination of ambient waterways. By using the prevailing (or background) detection rates for each marker in this study, we were able to calculate the conditional probabilities of correctly identifying contributing host fecal sources for each assay. For samples where the human Bacteroidales assay was positive, they are estimated to correctly identify human fecal sources 75% of the time, which is lower than the 98% probability calculated in a previous study (34). However, when samples test negative for human Bacteroidales markers, we can expect the samples to be truly negative 92% of the time. As additional studies are performed and background rates for detection are better understood, more accurate probability calculations will be possible. The traditional concept of employing fecal indicators to estimate health risks, the presence of FIB, and even the presence of pathogen in surface waters is being increasingly called into question. The presence and concentration of FIB may not be well correlated with specific pathogen exposure and actual health outcomes from water contact recreation, given the wide

5812

SCHRIEWER ET AL.

variation in infectivity, exposure dose, pathogen virulence, and differential immunity that may affect disease expression. Different microbial species and strains may vary in virulence and ability to cause disease in human and animal hosts (3, 12, 13, 58), thus no conclusions should be drawn regarding their pathogenic potential without further characterization. Vibrio was deliberately included in the suite of target bacteria in order to explore relationships with traditional FIB and Bacteroidales, even though its niches in coastal ecosystems are more complex than those of some other potentially pathogenic organisms. However, the majority of environmental V. cholerae strains are nonpathogenic unless the cholera toxin gene has been acquired via phage transduction (18), and the same may be true for other pathogens evaluated in this study. The proportion of virulent strains can also vary by water type, water temperature, and season, and different risk factors may exist for persons exposed to freshwater versus marine environments (56). Epidemiologic and QMRA studies that evaluate associations between environmental risk factors and health outcomes in a variety of settings are needed. The findings from this study can help prioritize selection of microbes and risk factors that are most important to consider for subsequent research efforts. Significant Spearman and Kendall correlations were seen for some, but not all, indicator and pathogen combinations. Analyses were performed both as an all-inclusive data set, with nondetect samples set to 0, and by excluding nondetects from the data set. The results from the two approaches differed slightly, but overall correlations ranged from 0.14 to 0.73 (perfect correlation would equal 1). The significant indicatorpathogen correlations are shown in Table 5, and the general pattern of values of ⬍0.5 may reflect the dynamic environments and complex niches of indicators and pathogens that have differing abilities to survive over time. Similarly, in the Field and Samadpour review (21), enterococcus enumeration was not highly correlated with Campylobacter spp., Cryptosporidium spp., Giardia spp., or Salmonella spp. detection in surface waters. For most environmental studies, pathogen detection occurs in the minority of samples, but these are the ones that are most important for the assessment of relationships between indicators and pathogens. A limitation of using the complete data set for correlation analysis is that these important positive samples can be underrepresented, although if only positive samples are included for analysis, information from nondetect samples is lost. Additionally, correlation analysis does not allow for different indicator threshold concentrations to be used for categorizing results, as can be done in PQ analysis. The calculation of predictive qualifier levels used a weighted average to evaluate how often pathogen occurrence was correctly predicted or missed when selecting different threshold cutoff concentrations for indicator bacteria. We extended the approach of Harwood et al. (26), who expressed the relationship of indicator and pathogen detection by percentages of true/false, positive/negative combinations, and added a component of weighting different indicator concentrations by the actual pathogen occurrence in paired samples. Use of binary logistic regression is also applicable for a stepwise analysis of signal thresholds but does not allow for a weighted evaluation of pathogen occurrence (52, 65). We did not find correlations with detection of any of the pathogens and the presence of the


human-specific Bacteroidales marker in paired water samples; this finding is consistent with the low PQ percentages observed for threshold concentrations of the human marker with all pathogens. The fact that 100% of samples were positive for universal markers demonstrates why threshold cutoffs are necessary for the estimation of predictive values, because presence/absence is often uninformative. Threshold cutoffs could be varied depending on the pathogen of interest, so for example, with respect to Vibrio cholerae detection, the highest predictive qualifier percentage was associated with a signal threshold of 500 GC/ml. In a study of wastewater and river samples in the Sapporo City area of Japan, good predictive values for total and human Bacteroidales markers were found in association with E. coli O157:H7 and Salmonella detection; both of these pathogens were either not detected or scarce in our study (52). In a Canadian study, ruminant Bacteroidales assays were predictive of E. coli O157:H7 detection (65). We also demonstrated that salinity plays an important role for the predictive ability of indicators. While Cryptosporidium spp. were distributed equally at freshwater- and marine-influenced sites for universal Bacteroidales detection, a spread of more than 10 percent between both water types could be observed for mean PQ values. Additional studies are needed to fill data gaps and provide guidance as to the most promising approaches for predicting health risks. Our results demonstrate how, in combination with traditional fecal indicator bacteria, Bacteroidales markers can provide complementary information to facilitate microbial source tracking efforts. In this study, the detection frequencies of the host-specific Bacteroidales markers were relatively low, and so a more detailed discussion of these concentrations was not considered worthwhile (though concentrations are presented indirectly in Table 3 in the form of concentration ratios of universal to host-specific Bacteroidales markers). However, some of our other ongoing studies, especially those in urbanized areas, show a much different picture, and concentration data will be very important. It is with the assessment of relative concentrations that one is able to identify hot spots, prioritize treatment measures, and identify whether natural or anthropogenic sources have higher contributions. Lastly, Bacteroidales cannot be cultured, and so molecular methods such as qPCR are required to provide the most useful suite of data. One benefit of such molecular approaches is that, in contrast with culture-based methods, the samples are all treated the same for Bacteroidales assays. Another benefit is that nucleic acid extracts can be stored long term (under appropriate conditions to minimize degradation), thus facilitating additional or improved assays. Future studies that analyze assay performance in other settings and development of complementary assays to characterize loading sources (e.g., aging fecal inputs) will further augment our understanding of potential uses for Bacteroidales in microbial source tracking and mitigation efforts. The fact that Bacteroidales are anaerobic and less likely to multiply in the environment than traditional FIB may also make Bacteroidales a more useful indicator of recent fecal pollution, which is more likely to contain viable fecal pathogens than pollution events that occurred longer ago. Given that the probability of a viable pathogen presence increases with the amount of recent fecal pollution found in a watershed

VOL. 76, 2010


when numbers of contributing individuals are high, the ability to quantify recent fecal pollution in an indicator assay is desirable. This study has added to our understanding of the associations, or lack thereof, that exist between traditional indicator bacteria, Bacteroidales host markers, and the presence of specific waterborne pathogens in surface waters. We found that traditional indicator bacteria, including coliforms and enterococci, correlated with some, but not all, bacterial and protozoal pathogens in this study, suggesting that monitoring for indicator bacteria alone may not accurately represent the presence of fecal pathogens in surface waters. In addition, we observed that both Bacteroidales and FIB were detected in most surface water samples, which demonstrates the high sensitivity of Bacteroidales as an alternative fecal indicator assay. This conclusion is supported by prior specificity testing that demonstrated high discrimination between human and animal fecal sources. The value of using Bacteroidales markers rather than traditional FIB testing has long been acknowledged to come from the information obtained about contributing host fecal sources. We have shown here that Bacteroidales perform similarly to traditional FIB when predicting the presence of pathogens in surface waters, which will give resource managers more flexibility in choosing their monitoring approaches to evaluate water quality, remediate pollution problems, and thereby reduce health risks. We conclude that utilizing a combination of indicator and pathogen assays can provide useful information regarding the presence, relative abundance, and contributing sources of fecal contamination in environmental water samples. This approach may be increasingly important for future monitoring and mitigation efforts in shared watersheds. ACKNOWLEDGMENTS We thank B. S. McSwain Sturm for designing the Acinetobacter baylyi qPCR assay. This study was made possible by the contributions of numerous collaborators, staff, students, and volunteers at the University of California at Davis, the California Department of Fish and Game, and Applied Marine Sciences. Funding for this project was provided by the city of Watsonville, CA, and the Central Coast Long-term Environmental Assessment Network through a grant from the California State Water Resources Control Board. REFERENCES 1. APHA. 2001. Standard methods for the examination of water and wastewater, 21st ed. American Public Health Association, Washington, DC. 2. Bae, S., and S. Wuertz. 2009. Rapid decay of host-specific fecal Bacteroidales cells in seawater as measured by quantitative PCR with propidium monoazide. Water Res. 43:4850–4859. 3. Bag, P. K., P. Bhowmik, T. K. Hajra, T. Ramamurthy, P. Sarkar, M. Majumder, G. Chowdhury, and S. C. Das. 2008. Putative virulence traits and pathogenicity of Vibrio cholerae non-O1, non-O139 isolates from surface waters in Kolkata, India. Appl. Environ. Microbiol. 74:5635–5644. 4. Bang, D. D., F. Scheutz, P. Ahrens, K. Pedersen, J. Blom, and M. Madsen. 2001. Prevalence of cytolethal distending toxin (cdt) genes and CDT production in Campylobacter spp. isolated from Danish broilers. J. Med. Microbiol. 50:1087–1094. 5. Beaches Environmental Assessment and Coastal Health Act. 2000. Beaches Environmental Assessment and Coastal Health Act. Stat. 870–877, Public Law 106–284. 106th U.S. Congress, Washington, DC. 6. Bernhard, A. E., and K. G. Field. 2000. Identification of nonpoint sources of fecal pollution in coastal waters by using host-specific 16S ribosomal DNA genetic markers from fecal anaerobes. Appl. Environ. Microbiol. 66:1587– 1594. 7. Bernhard, A. E., and K. G. Field. 2000. A PCR assay to discriminate human and ruminant feces on the basis of host differences in Bacteroides-Prevotella genes encoding 16S rRNA. Appl. Environ. Microbiol. 66:4571–4574. 8. Blackwell, K. D., and J. D. Oliver. 2008. The ecology of Vibrio vulnificus,

9.

10. 11. 12. 13.

14.

15.

16.

17.

18.

19. 20.

21. 22.

23.

24.

25.

26.

27.

28.

29.

30.

31. 32.

33.

34.

5813

Vibrio cholerae, and Vibrio parahaemolyticus in North Carolina estuaries. J. Microbiol. 46:146–153. Boom, R., C. J. A. Sol, M. M. M. Salimans, C. L. Jansen, P. M. E. Wertheimvandillen, and J. Vandernoordaa. 1990. Rapid and simple method for purification of nucleic acids. J. Clin. Microbiol. 28:495–503. CCLEAN. 2007. Central Coast Long-term Environmental Assessment Network program overview. CCLEAN, Santa Cruz, CA. Clean Water Act. 1948. Federal Water Pollution Control Act. 62 Stat. 1155– 1328, Public Law 845. U.S. Congress, Washington, DC. Colwell, R. R. 1996. Global climate change and infectious disease: the cholera paradigm. Science 274:2025–2031. Cooper, I. R., H. D. Taylor, and G. W. Hanlon. 2007. Virulence traits associated with verocytotoxigenic Escherichia coli O157 recovered from freshwater biofilms. J. Appl. Microbiol. 102:1293–1299. Corso, P. S., M. H. Kramer, K. A. Blair, D. G. Addiss, J. P. Davis, and A. C. Haddix. 2003. Cost of illness in the 1993 waterborne Cryptosporidium outbreak, Milwaukee, Wisconsin. Emerg. Infect. Dis. 9:426–431. Curriero, F. C., J. A. Patz, J. B. Rose, and S. Lele. 2001. The association between extreme precipitation and waterborne disease outbreaks in the United States, 1948–1994. Am. J. Public Health 91:1194–1199. Dick, L. K., A. E. Bernhard, T. J. Brodeur, J. W. Santo Domingo, J. M. Simpson, S. P. Walters, and K. G. Field. 2005. Host distributions of uncultivated fecal Bacteroidales bacteria reveal genetic markers for fecal source identification. Appl. Environ. Microbiol. 71:3184–3191. el-Sherbeeny, M. R., C. Bopp, J. G. Wells, and G. K. Morris. 1985. Comparison of gauze swabs and membrane filters for isolation of Campylobacter spp. from surface water. Appl. Environ. Microbiol. 50:611–614. Faruque, S. M., M. J. Albert, and J. J. Mekalanos. 1998. Epidemiology, genetics, and ecology of toxigenic Vibrio cholerae. Microbiol. Mol. Biol. Rev. 62:1301–1314. Fayer, R. 2004. Cryptosporidium: a water-borne zoonotic parasite. Vet. Parasitol. 126:37–56. Field, K. G., A. E. Bernhard, and T. J. Brodeur. 2003. Molecular approaches to microbiological monitoring: fecal source detection. Environ. Monit. Assess. 81:313–326. Field, K. G., and M. Samadpour. 2007. Fecal source tracking, the indicator paradigm, and managing water quality. Water Res. 41:3517–3538. Fujioka, R. S. 2001. Monitoring coastal marine waters for spore-forming bacteria of faecal and soil origin to determine point from non-point source pollution. Water Sci. Technol. 44:181–188. Gannon, V. P., S. D’Souza, T. Graham, R. K. King, K. Rahn, and S. Read. 1997. Use of the flagellar H7 gene as a target in multiplex PCR assays and improved specificity in identification of enterohemorrhagic Escherichia coli strains. J. Clin. Microbiol. 35:656–662. Gourmelon, M., M. P. Caprais, R. Segura, C. Le Mennec, S. Lozach, J. Y. Piriou, and A. Rince. 2007. Evaluation of two library-independent microbial source tracking methods to identify sources of fecal contamination in French estuaries. Appl. Environ. Microbiol. 73:4857–4866. Gregory, J. B., R. W. Litaker, and R. T. Noble. 2006. Rapid one-step quantitative reverse transcriptase PCR assay with competitive internal positive control for detection of enteroviruses in environmental samples. Appl. Environ. Microbiol. 72:3960–3967. Harwood, V. J., A. D. Levine, T. M. Scott, V. Chivukula, J. Lukasik, S. R. Farrah, and J. B. Rose. 2005. Validity of the indicator organism paradigm for pathogen reduction in reclaimed water and public health protection. Appl. Environ. Microbiol. 71:3163–3170. Hrudey, S. E., P. Payment, P. M. Huck, R. W. Gillham, and E. J. Hrudey. 2003. A fatal waterborne disease epidemic in Walkerton, Ontario: comparison with other waterborne outbreaks in the developed world. Water Sci. Technol. 47:7–14. Hu, Y., Q. Zhang, and J. C. Meitzler. 1999. Rapid and sensitive detection of Escherichia coli O157:H7 in bovine faeces by a multiplex PCR. J. Appl. Microbiol. 87:867–876. Isobe, K. O., M. Tarao, M. P. Zakaria, N. H. Chiem, L. Y. Minh, and H. Takada. 2002. Quantitative application of fecal sterols using gas chromatography-mass spectrometry to investigate fecal pollution in tropical waters: western Malaysia and Mekong Delta, Vietnam. Environ. Sci. Technol. 36: 4497–4507. Jofre, J., E. Olle, F. Ribas, A. Vidal, and F. Lucena. 1995. Potential usefulness of bacteriophages that infect Bacteroides fragilis as model organisms for monitoring virus removal in drinking water treatment plants. Appl. Environ. Microbiol. 61:3227–3231. Juni, E., and A. Janik. 1969. Transformation of Acinetobacter calco-aceticus (Bacterium anitratum). J. Bacteriol. 98:281–288. Karim, M. R., F. D. Manshadi, M. M. Karpiscak, and C. P. Gerba. 2004. The persistence and removal of enteric pathogens in constructed wetlands. Water Res. 38:1831–1837. Keymer, D. P., M. C. Miller, G. K. Schoolnik, and A. B. Boehm. 2007. Genomic and phenotypic diversity of coastal Vibrio cholerae strains is linked to environmental factors. Appl. Environ. Microbiol. 73:3705–3714. Kildare, B. J., C. M. Leutenegger, B. S. McSwain, D. G. Bambic, V. B. Rajal, and S. Wuertz. 2007. 16S rRNA-based assays for quantitative detection of

5814

35.

36.

37. 38.

39. 40.

41.

42. 43.

44.

45.

46.

47.

48.

49.

50.

51.

SCHRIEWER ET AL.

universal, human-, cow-, and dog-specific fecal Bacteroidales: a Bayesian approach. Water Res. 41:3701–3715. Koch, W. H., W. L. Payne, B. A. Wentz, and T. A. Cebula. 1993. Rapid polymerase chain reaction method for detection of Vibrio cholerae in foods. Appl. Environ. Microbiol. 59:556–560. Kreader, C. A. 1995. Design and evaluation of Bacteroides DNA probes for the specific detection of human fecal pollution. Appl. Environ. Microbiol. 61:1171–1179. Kreader, C. A. 1998. Persistence of PCR-detectable Bacteroides distasonis from human feces in river water. Appl. Environ. Microbiol. 64:4103–4105. Kueh, C. S. W., T. Y. Tam, T. Lee, S. L. Wong, O. L. Lloyd, I. T. S. Yu, T. W. Wong, J. S. Tam, and D. C. J. Bassett. 1995. Epidemiological study of swimming-associated illnesses relating to bathing-beach water quality. Water Sci. Technol. 31:1–4. Luther, K., and R. Fujioka. 2004. Usefulness of monitoring tropical streams for male-specific RNA coliphages. J. Water Health 2:171–181. Miller, W. A., M. A. Miller, I. A. Gardner, E. R. Atwill, M. Harris, J. Ames, D. Jessup, A. Melli, D. Paradies, K. Worcester, P. Olin, N. Barnes, and P. A. Conrad. 2005. New genotypes and factors associated with Cryptosporidium detection in mussels (Mytilus spp.) along the California coast. Int. J. Parasitol. 35:1103–1113. O’Reilly, C. E., A. B. Bowen, N. E. Perez, J. P. Sarisky, C. A. Shepherd, M. D. Miller, B. C. Hubbard, M. Herring, S. D. Buchanan, C. C. Fitzgerald, V. Hill, M. J. Arrowood, L. X. Xiao, R. M. Hoekstra, E. D. Mintz, and M. F. Lynch. 2007. A waterborne outbreak of gastroenteritis with multiple etiologies among resort island visitors and residents: Ohio, 2004. Clin. Infect. Dis. 44:506–512. Pruss, A. 1998. Review of epidemiological studies on health effects from exposure to recreational water. Int. J. Epidemiol. 27:1–9. Rajal, V. B., B. S. McSwain, D. E. Thompson, C. M. Leutenegger, B. J. Kildare, and S. Wuertz. 2007. Validation of hollow fiber ultrafiltration and real-time PCR using bacteriophage PP7 as surrogate for the quantification of viruses from water samples. Water Res. 41:1411–1422. Rajal, V. B., B. S. McSwain, D. E. Thompson, C. M. Leutenegger, and S. Wuertz. 2007. Molecular quantitative analysis of human viruses in California stormwater. Water Res. 41:4287–4298. Reischer, G. H., J. M. Haider, R. Sommer, H. Stadler, K. M. Keiblinger, R. Hornek, W. Zerobin, R. L. Mach, and A. H. Farnleitner. 2008. Quantitative microbial faecal source tracking with sampling guided by hydrological catchment dynamics. Environ. Microbiol. 10:2598–2608. Rhodes, M. W., and J. Kator. 1999. Sorbitol-fermenting Bifidobacteria as indicators of diffuse human faecal pollution in estuarine watersheds. J. Appl. Microbiol. 87:528–535. Ristori, C. A., S. T. Iaria, D. S. Gelli, and I. N. G. Rivera. 2007. Pathogenic bacteria associated with oysters (Crassostrea brasiliana) and estuarine water along the south coast of Brazil. Int. J. Environ. Health Res. 17:259–269. Robertson, L. J., A. T. Campbell, and H. V. Smith. 1992. Survival of Cryptosporidium parvum oocysts under various environmental pressures. Appl. Environ. Microbiol. 58:3494–3500. Samadpour, N., J. Stewart, K. Steingart, C. Addy, J. Louderback, M. McGinn, J. Ellington, and T. Newman. 2002. Laboratory investigation of an E. coli O157:H7 outbreak associated with swimming in Battle Ground Lake, Vancouver, Washington. J. Environ. Health 64:16–20. Santo Domingo, J. W., D. G. Bambic, T. A. Edge, and S. Wuertz. 2007. Quo vadis source tracking? Towards a strategic framework for environmental monitoring of fecal pollution. Water Res. 41:3539–3552. Savichtcheva, O., and S. Okabe. 2006. Alternative indicators of fecal pollu-


52.

53.

54.

55. 56.

57.

58. 59.

60. 61.

62.

63.

64.

65.

66.

tion: relations with pathogens and conventional indicators, current methodologies for direct pathogen monitoring and future application perspectives. Water Res. 40:2463–2476. Savichtcheva, O., N. Okayama, and S. Okabe. 2007. Relationships between Bacteroides 16S rRNA genetic markers and presence of bacterial enteric pathogens and conventional fecal indicators. Water Res. 41:3615–3628. Seurinck, S., T. Defoirdt, W. Verstraete, and S. D. Siciliano. 2005. Detection and quantification of the human-specific HF183 Bacteroides 16S rRNA genetic marker with real-time PCR for assessment of human faecal pollution in freshwater. Environ. Microbiol. 7:249–259. Shanks, O. C., J. W. Santo Domingo, R. Lamendella, C. A. Kelty, and J. E. Graham. 2006. Competitive metagenomic DNA hybridization identifies host-specific microbial genetic markers in cow fecal samples. Appl. Environ. Microbiol. 72:4054–4060. Slifko, T. R., H. V. Smith, and J. B. Rose. 2000. Emerging parasite zoonoses associated with water and food. Int. J. Parasitol. 30:1379–1393. Stewart, J. R., R. J. Gast, R. S. Fujioka, H. M. Solo-Gabriele, J. S. Meschke, L. A. Amaral-Zettler, E. del Castillo, M. F. Polz, T. K. Collier, M. S. Strom, C. D. Sinigalliano, P. D. R. Moeller, and A. F. Holland. 2008. The coastal environment and human health: microbial indicators, pathogens, sentinels and reservoirs. Environ. Health 7:1–14. Stricker, A. R., I. Wilhartitz, A. H. Farnleitner, and R. L. Mach. 2008. Development of a Scorpion probe-based real-time PCR for the sensitive quantification of Bacteroides sp. ribosomal DNA from human and cattle origin and evaluation in spring water matrices. Microbiol. Res. 163:140–147. Teunis, P. F., C. L. Chappell, and P. C. Okhuysen. 2002. Cryptosporidium dose response studies: variation between isolates. Risk Anal. 22:175–183. Till, D., G. McBride, A. Ball, K. Taylor, and E. Pyle. 2008. Large-scale freshwater microbiological study: rationale, results and risks. J. Water Health 6:443–460. U.S. EPA. 2005. Method 1623: Cryptosporidium and Giardia in water by filtration/IMS/FA. U.S. EPA, Washington, DC. U.S. EPA. 2006. Method 1682: Salmonella in sewage sludge (biosolids) by modified semisolid Rappaport-Vassiliadis (MSRV) medium. EPA-821-R06-14. U.S. EPA, Washington, DC. Vaneechoutte, M., D. M. Young, L. N. Ornston, T. De Baere, A. Nemec, T. Van Der Reijden, E. Carr, I. Tjernberg, and L. Dijkshoorn. 2006. Naturally transformable Acinetobacter sp. strain ADP1 belongs to the newly described species Acinetobacter baylyi. Appl. Environ. Microbiol. 72:932–936. Vezzulli, L., E. Pezzati, M. Moreno, M. Fabiano, L. Pane, and C. Pruzzo. 2009. Benthic ecology of Vibrio spp. and pathogenic Vibrio species in a coastal Mediterranean environment (La Spezia Gulf, Italy). Microb. Ecol. 58:808–818. Vogel, J. R., D. M. Stoeckel, R. Lamendella, R. B. Zelt, J. W. S. Domingo, S. R. Walker, and D. B. Oerther. 2007. Identifying fecal sources in a selected catchment reach using multiple source-tracking tools. J. Environ. Qual. 36: 718–729. Walters, S. P., V. P. J. Gannon, and K. G. Field. 2007. Detection of Bacteroidales fecal indicators and the zoonotic pathogens E. coli O157:H7, Salmonella, and Campylobacter in river water. Environ. Sci. Technol. 41:1856– 1862. Wiedenmann, A., P. Kruger, K. Dietz, J. M. Lopez-Pila, R. Szewzyk, and K. Botzenhart. 2006. A randomized controlled trial assessing infectious disease risks from bathing in fresh recreational waters in relation to the concentration of Escherichia coli, intestinal enterococci, Clostridium perfringens, and somatic coliphages. Environ. Health Perspect. 114:228–236.