Applied Microbiology and Biotechnology

, Volume 76, Issue 4, pp 927–934

Microbial source tracking in a small southern California urban watershed indicates wild animals and growth as the source of fecal bacteria

  • Sunny C. Jiang
  • Weiping Chu
  • Betty H. Olson
  • Jian-Wen He
  • Samuel Choi
  • Jenny Zhang
  • Joanne Y. Le
  • Phillip B. Gedalanga
Environmental Biotechnology

DOI: 10.1007/s00253-007-1047-0

Cite this article as:
Jiang, S.C., Chu, W., Olson, B.H. et al. Appl Microbiol Biotechnol (2007) 76: 927. doi:10.1007/s00253-007-1047-0

Abstract

Three independent microbial source tracking (MST) methods were applied to a small urban subwatershed in Orange County, California. Fifty-seven water samples collected over summer 2002 were analyzed for human adenovirus and enterovirus. Enterococci and E. coli were isolated for antibiotic resistance analysis (ARA) and for PCR identification of human- and animal-specific toxin genes, respectively. All water samples were PCR negative for human enteroviruses and E. coli human-specific toxin gene. E. coli toxin markers revealed the presence of toxin genes specific to bird, rabbit, and cow. Enterococci ARA results supported this conclusion and indicated that fecal bacteria from bird and wild animal feces as well as soil were the predominant source found in the watershed. An E. coli, isolated from the watershed and inoculated back into the heat-sterilized storm drain water, increased 4 log units within 6 days. Collectively, these results suggest that bird and wild animal feces, soil amendments, and/or fecal coliform growth in the storm drain are the major contributors to the fecal bacterial pollution in downstream areas. However, human adenoviruses were detected on two occasions. Fecal bacterial concentrations were not elevated on these two occasions, suggesting that the elevated levels of fecal indicator bacteria in this small watershed could be unrelated to the source of human adenovirus.

Keywords

Microbial source tracking (MST) Urban watershed Human virus E. coli biomarkers Adenovirus Enterovirus Antibiotic resistance analysis (ARA) 

Introduction

Microbial source tracking, designed to identify the source of fecal contamination in the environment, has received much attention in recent years. Several comprehensive reviews have compared and summarized the pros and cons of most of the current technologies (Sinton et al. 1998; Scott et al. 2002; Simpson et al. 2002; Field et al. 2003a). These methodologies are generally classified into three different categories: library/database-dependent genotypic and phenotypic typing methods and library-independent biomarker assays. At least three multilaboratory, multi-investigator, large-scale method comparison studies have been conducted using blind samples for evaluation of the accuracy of each method (Griffith et al. 2003; Stoeckel et al. 2004; Blanch et al. 2006). So far, there has not been a method that is applicable to all environmental settings due to the complexity of environmental conditions. These diverse source tracking technologies are described as tools within a toolbox. It is recommended that multiple tools should be selected and applied to the contamination site to improve the accuracy and resolution of the pollution source identification (Stewart et al. 2003).

Antibiotic resistance analysis (ARA) is among the simplest and the least expensive microbial source tracking methods developed for watershed studies. This library-dependent phenotypic typing method is based on bacterial resistance to antibiotics. ARA has been effectively used to identify the source of microbial contamination in several earlier studies (Hagedorn et al. 1999; Wiggins et al. 1999; Harwood et al. 2000). This method relies on statistical discrimination between resistant types from a large library of organisms. Recently, the method has been met with criticisms. For example, databases developed may be only representative of the geographic location in which it was developed (Ebdon and Taylor 2006). The degree of confidence of source classification decreases with increases in environmental complexity and temporal scales (Moore et al. 2005; Anderson et al. 2006). Other researchers also argued that the link between bacterial antibiotic resistance properties and their prior exposure history is questionable. However, Gaun et al. (2002) showed 95% agreement between amplified fragment length polymorphisms and ARA profiles of wild animal fecal sources.

Viruses are host specific, such that human viruses only infect and are shed from human hosts. Research on detection of human viruses in the past has solely focused on identifying human pathogens in the water and the potential health risk. In more recent microbial source tracking comparison studies, human viruses were identified as a conservative marker for human fecal pollution in water samples (Griffith et al. 2003; Noble et al. 2003). However, virus detection by PCR technology does not indicate the infectivity. Human viruses, i.e., adenovirus, decay extremely slowly in the environment (Enriquez et al. 1995; Gerba et al. 2002) and, therefore, provide no information on the age of the contamination (Choi and Jiang 2005).

The application of E. coli host-specific toxin gene as a biomarker for source identification is based on diarrheic E. coli, which is a pathogenic subset of E. coli known to be associated with the gut of warm-blooded animals. The diarrheic E. coli causes diarrhea and related illnesses in a variety of animals and humans. Factors, making this group worthy for microbial source tracking, are the relationships between specific toxins and host species. A variety of fimbriae, used to attach to intestinal cells of explicit hosts, provide certain diarrheic E. coli carrying particular toxin host specificity. Previous research has identified potential biomarkers based on enterotoxins for fecal source tracking (Field et al. 2003b). To date, these include the pig, bird, cow, and human biomarkers (Olson and Oshiro 1997; Khatib et al. 2002, 2003; Chern et al. 2004). Toxin genes specific to dog and rabbit have also been identified although they have not yet been used for microbial source tracking effort (Adams et al. 1997; Johnson et al. 2000).

Laguna Niguel, a small urban watershed in southern California, had experienced chronic fecal coliform and enterococci contamination, with concentrations on average of 2–4 orders of magnitude greater than State of California-established type 2 recreational standards. San Diego Regional Water Quality Control Board issued a Cleanup and Abatement Order for the area. However, the cleanup effort was hindered due to the unknown pollution sources. Here, we applied three MST tools, namely, ARA, human viruses, and E. coli toxin biomarkers to identify these sources. The results of this study demonstrate that multiple source tracking approaches converge and improve the confidence of pollution source identification.

Materials and methods

Study site

The Orange County, California, subwatershed under investigation covers approximately 0.8 square miles of hilly terrain with significant open space and ∼1,500 single- and multifamily homes. There are no commercial, industrial, agricultural, or equestrian uses in the subwatershed. The climate is Mediterranean, with about 12″ of rainfall per year, mostly between October and April. During the summer dry season, the source of surface water in the subwatershed is primarily runoff from landscape irrigation, flowing via street gutters and slope v-ditches to catch basins and to the underground storm drain system. Groundwater comprises ∼30% of the total discharge from the storm drain.

Sample collection

Water samples were collected approximately twice weekly on randomly selected days of the week between May and August 2002 from two locations in the subwatershed. One site was at the curbside on the street near one of the catch basins (Basin 13) in the watershed. The second site was at the end of the storm drain outlet collecting water from all underground storm drain system in the watershed. On each sampling day, the curbside was sampled once (morning) while the storm drain outlet was twice (morning and afternoon) to capture the signal of ground water seepage and the differences in the volume of runoff. Approximately 15 l of water sample was taken using sterilized carboys and was transported to UCI laboratories for immediate processing. A total of 57 water samples were collected and analyzed during the study period.

Feces from wild birds, rabbits, domestic dogs, cats, and unidentified wild animals were collected from the neighborhood drainage area on the same days as the water samples. Bird feces were swiped using sterilized cotton Q-tips and were resuspended in sterile saline buffer. Other fecal materials were collected using sterile plastic scoops and were placed in sterile sampling tubes. In addition, 11 soil amendment samples were collected from lawn and landscape areas where organic soil amendment was applied in the drainage area. Sewage samples (15 in total) were taken through the underground sewer manholes near Catch Basin 13 in the neighborhood during the study period. Only freshly excreted fecal samples were used for enterococci isolation.

Enterococci isolation and enumeration

Water and sewage samples were serially diluted in sterile phosphate-buffered saline (PBS), and enterococci were enumerated using the membrane filtration technique following the EPA protocol 1601. Fecal and soil samples were weighed, homogenized with sterile saline buffer, and settled to remove large debris, and supernatants were removed and serially diluted for enterococci enumeration and isolation. The final concentration was presented as colony-forming unit (CFU) per gram of feces or soil (wet weight) or CFU per milliliter of sewage/water. Twenty to 50 isolates were picked from each fecal sample for ARA analysis (described below).

Antibiotic resistance analysis (ARA) and discriminant analysis

The ARA is essentially the same as described in the paper of Choi et al. (2003). The selection and concentration of antibiotics followed previous reports (Harwood et al. 2000). The numbers of isolates used for the analysis from each fecal category are: 971 from rabbits (37 fecal samples), 872 from dogs (36 fecal samples), 922 from birds (23 fecal samples), 48 from cats (1 fecal sample), 224 from unknown animal feces (8 fecal samples), 132 from soil (11 samples), 685 from sewage (15 samples), and 1,676 from water samples (57 samples).

Data for ARAs were analyzed using SAS software (version 8; SAS Institute) essentially as described by Wiggins et al. (1999). Correct classification rate (CCR) is defined as the number of isolates (from a given source) that are placed in the correct source category by the discriminant analysis. It was expressed as percentage of chance for correct assignment. The misclassification rate (MCR) is calculated based on the miss-assignment of the test subjects to the wrong source category. The average frequency for a source to be miss-assigned to other sources was used as the measure of the chances of miss-assignment (percentage). In this study, we introduce the concept of MCR cutoff, which is defined as the average MCR to each source category plus two standard deviations. An unknown source is rejected from the assigned source category if the classification rate for the category is no greater than the MCR cutoff. This is to increase the confident level of unknown source assignment.

E. coli isolation, enumeration, and DNA extraction

E. coli was isolated and enumerated from water samples using membrane filtration method as previously described (Khatib et al. 2002). Three dilutions and five replicates of each dilution were used to obtain the most probable number (MPN) as previously described (Khatib et al. 2002). The filters were placed on both mTEC (Difco, Detroit, MI, USA) and mTEC supplemented with Congo Red (Difco) to harvest E. coli carrying different toxin genes. After enumeration of colonies on the filter, cells were scraped from the filter, were resuspended in 1× PBS buffer on the filter, were pelleted by centrifugation, and were extracted for total DNA extraction using freeze–thaw and phenol–chloroform methods as described (Tsai and Olson 1991). The DNA extracts were quantified spectrophotometrically and were immediately stored at −80°C until PCR assay. If necessary, total DNA was further purified using Eppendorf miniprep spin column (Eppendorf, Westbury, NY, USA) to eliminate PCR interferences.

Nested PCR for E. coli toxin genes

Nested-PCR primers for the cow (LTIIa gene) and human (STh gene) biomarkers were adopted from published research (Khatib et al. 2002, 2003; Chern et al. 2004). Biomarkers for rabbit- (ralG gene), bird- (tsh gene), and dog- (papG III gene) specific toxin genes were based on published reports (Provence and Curtiss 1994; Adams et al. 1997; Johnson et al. 2000; Johnson and Stell 2001) and were screened for host specificity as described in the work of Khatib et al. (2002, 2003). All nested-PCR primers were designed based on these gene sequences retrieved from NCBI GenBank using Primer Express II (Applied Biosystems). The PCR amplicons were confirmed either by restriction enzyme analysis or Southern hybridization with internal probes. The information on primer sequences, amplification protocols, probe sequences, restriction enzymes, and hybridization conditions is presented in the electronic supplementary material.

Quantification of positive results for a specific toxin gene was by MPN of E. coli determined from multiple filters and dilutions (Khatib et al. 2002, 2003). The relative contribution of each specific toxin gene to the overall fecal source is described by the percentage of MPN of E. coli with the specific toxin gene among the E. coli containing all toxin genes.

Molecular detection of human viruses

Ten liters of water from each site was concentrated, and nucleic acids were extracted and subjected to PCR detection of human adeno- and enteroviruses as previously described (Jiang et al. 2001; Jiang and Chu 2004). Primers, probes, and PCR protocol are presented in the electronic supplementary material.

Growth potential of E. coli in the subwatershed

An assay to assess the ability of fecal coliform bacteria to grow on nutrients contained in the dry weather runoff water was based on the method of Rajkowski and Rice (1999), designated as fecal coliform growth potential assay. Briefly, water samples were collected on three different dates from both Catch Basin and Storm Drain, which were heat-treated (pasteurized) to kill bacteria. After testing to ensure the complete kill, E. coli isolated from the local sample was inoculated back into each water sample to give a final concentration of approximately 10–100 CFU/100 ml. Samples were incubated in the dark at room temperature (21°–25°C) and were assayed for bacterial numbers on trypticase soy agar (TSA, Difco) at 35°C at 2- to 3-day intervals until no further increases in CFU were observed (6 to 14 days).

Quality assurance and quality control procedures

QA/QC of laboratory reagents and instruments were performed before sample processing and analyses. Details of the QA/QC procedures are presented in the electronic supplementary material.

Results

Enterococci bacteria among fecal source samples

Enterococcus spp. was detected in all source categories sampled during this study although concentrations among individual samples varied significantly (the standard deviation was greater than the mean value for several source categories). Rabbit feces contained the highest concentration of enterococci, followed by bird and dog feces (Fig. 1). Rabbit and bird feces were the most frequently observed animal feces throughout the subwatershed, suggesting that bird and wild animal feces were important environmental sources of enterococci.
Fig. 1

Geometric mean concentration of Enterococcus spp. in source samples collected during the study period. Source samples include feces from dog, rabbit, wild bird, unknown wild animals, and soil amendment samples from the neighborhood under investigation. Sewage samples were collected from the sewer system through underground manholes on the street near Catch Basin 13. Water samples collected from the end of the storm drain outlet (drain), both in the morning and the afternoon, and from the street curbside near catch Basin 13 (basin) were also used for enterococci enumeration and isolations. Standard deviation was not plotted on the graph due to the large variable of concentration within each source category

Antibiotic resistant analysis

Enteococci isolates (3,851 in total) from eight individual source categories (described above) were used to establish ARA database/library. The discriminant analysis results showed that the CCR ranged from 28% for sewage samples to 96% for isolates from cat feces (Fig. 2a). Enterococci from amended soil yielded a CCR of 64% match, followed by birds (60%) and rabbits (51%). However, enterococci from dog feces and local sewage yielded CCR less than 30% (Fig. 2a). As our goal was to determine the contribution of wild vs domestic as well as sewage sources to fecal pollution, several sources were combined. After grouping isolates from rabbit, bird, and unknown wild animals as “wild”, the overall CCR improved to 78%, and the category “domestic” including feces of dogs and cats yielded a CCR of 48% (Fig. 2b). The average MCR with standard deviation is presented in Fig. 2c. The average MCR is the highest for “wild” source, indicating that there was an average of 30% chance of isolates from other sources to be miss-assigned as “wild”.
Fig. 2

Correct and misclassification rates for antibiotic resistance analysis. a Individual source category. b Combined categories. c Average misclassification rates for combined categories. The individual categories include feces of bird, rabbit, unknown wild animals, domestic cat, dog, soil amendment samples, and sewage collected from sewer manholes in the watershed. The wild category combines feces from bird, rabbit, and unknown wild animals, while the domestic category includes cat and dog feces. The rate was an expression of percentage of chance that a sample was correctly assigned to its source category. The average misclassification rate was the average percentage of chance that a sample was miss-assigned to categories other than itself

Fig. 3

Growth potential of E. coli in runoff water collected from street curbside near a catch basin and the end of the storm drain outlet. The E. coli used for inoculation was originally isolated from one of the storm drain water samples

Application of the ARA library to unknown isolates from water samples indicated matches for “wild” and “soil”. Sewage was excluded from all waters because the classification rates to sewage fell below MCR cutoff (20%). “Domestic” was excluded from water samples collected from the morning samples for the same reason (MCR cutoff is 23%), but was included as a match for samples collected from the storm drain in the afternoon (25%).

Human- and animal-specific E. coli toxin gene markers

All samples were negative for human toxin gene marker while positive for four other toxin genes tested (Fig. 3). Bird-specific toxin genes were the most frequently detected among all water samples, accounting for nearly 90% of all toxin genes detected in some samples. The cow-waste-specific toxin gene was also prevalent in the earlier samples collected over the study. There were no dairy or cattle grazing areas in the subwatershed, but mulching practices suggest a source for this toxin gene. Rabbit fecal marker was also found in all water samples tested, ranging from <1 to 49% of the total toxin genes. Dog-specific marker contributed a small but consistent fraction of toxin genes to the overall positive genes (two water samples collected in May contain <0.07% as shown in Fig. 3).

Growth of fecal indicator bacteria in the subwatershed

Figure 4 shows that E. coli started to grow immediately upon inoculation into the pasteurized water collected from the subwatershed. The E. coli numbers increased to 4–5 log units within 6 to 7 days of incubation in the dark, but between days 7 and 12, stationary growth phase was attained, suggesting that nutrients were depleted from the water. The generation time calculated using the exponential phase of the growth curve was 5.4 h.
Fig. 4

Distribution of host-specific toxin genes among E. coli isolated from subwatershed. The genes were detected by PCR using specific primers for each toxin gene

Detection of human enterovirus and adenovirus

Human enteroviruses are negative by RT-PCR and internal probe hybridization for all samples tested. However, two of the samples, both collected from the storm drain outlet in the afternoon, were PCR positive for adenoviruses. Bacterial indicators (enterococci) in the water samples were not elevated on either of these dates.

Discussion

Using three independent microbial source tracking methods, the results of this study indicate that human sewage was not a major contributor of fecal bacterial impairment in this small urban watershed. E. coli human toxin gene and enterovirus tests by PCR were negative. E. coli toxin genes indicated that birds were a major source of fecal pollution in the watershed. In water samples, “wild” and “soil” were the two meaningful categorical matches for the ARA. Excellent independent agreement in this grouping between the ARA and the toxin gene biomarkers was achieved.

The successful application of ARA is dependent on the availability of sufficient database and watershed scales. The small subwatershed investigated here fits these criteria. Although past studies included bird feces as a source for ARA (Choi et al. 2003; Moore et al. 2005), rabbit feces are not commonly tested. This study showed that rabbit feces contain one of the highest concentrations of Enterococcus spp. per unit weight. The high CCR among cat feces found in this study is due to a small sample size.

The prevalence of E. coli marker specific to cow waste in the watershed may be explained by organic mulch applied locally to landscape. Cow manure is one of the waste sources used to amend mulch. Although this material is composted, E. coli within the manure may not always be completely killed. This hypothesis is supported by an independent study conducted by the city, which demonstrated high concentrations of fecal coliform bacteria in organic compost collected around the neighborhood (Board 2003). The enterococci ARA results also supported this hypothesis, indicating a linkage between isolates from soil amendments and runoff waters. Perhaps organic compost and irrigation practices provide a favorable environment for survival and replication of some strains of E. coli. We only now begin to understand the ecology of E. coli in the environment (Topp et al. 2003; Berry and Miller 2005; Ishii et al. 2006). LeJeune et al. (2001) have shown that E. coli 0157:H7 survives in excess of 200 days in cow droppings on pasturelands. The elevated concentration of cow biomarkers in the initial part of this study could indicate one or a combination of factors: soil or amendments being washed into the storm drain with associated bacteria, differential growth rates in the environment influencing toxin concentrations or mulch amendment using a herd waste with a high toxin concentration (Khatib et al. 2002, 2003). Thus, the high portion of the cow-specific E. coli toxin genes observed should not be simply interpreted as the high number of E. coli of cow origin. Further investigation is necessary to understand the contribution of organic fertilizers to bacterial sources either via amendment or through growth of fecal bacteria. Amendments and dog feces are areas where inputs can be managed by city or county ordinances, while other inputs from birds or rabbits are nearly impossible to control.

The water temperature in the storm drain in this southern California neighborhood is approximately 23°–25°C in the summer, which is slightly greater than the temperature used in our growth experiments. Storm drains, as in this study, are underground and shaded from direct sunlight exposure preventing photoinactivation of fecal bacteria. The replication rate of E. coli in our laboratory simulations should be taken as a maximum growth potential for these environments as other competing bacteria and protozoa predators were excluded in the lab setting. Other work indicated that inoculated E. coli die-off is complete in 2 weeks in creek water receiving dry flow runoff due to predation (Khatib et al. 2002). The doubling rate found here suggests a constant level of fecal bacteria input from “pockets” of storm drain system. It is also interesting to observe the lack of a lag phase in the growth experiment suggesting that the organisms in the system have adapted to the nutrients available. The growth of the bacteria in the storm drain may be favored by the low flow rate. Further computation of dry weather runoff resident time within the watershed coupled with doubling rate may shed light on the proportion of fecal indicator bacteria from regrowth.

The occurrence of human adenoviruses detected by PCR, albeit with no information on infectivity, is intriguing as most waterborne outbreaks occur sporadically and are often from an unidentified waste source (Craun and Calderon 2006) as did our findings for adenovirus and the cow biomarker. The PCR results, although not quantitative, yielded very weak amplicons after nested PCR suggesting either a low level of adenovirus or PCR inhibitors in the DNA extracts of the samples. Although these findings may appear aberrant, both deserve more understanding. Regardless of the specific meaning of these findings, we concluded that sewage is not the contributor to the fecal indicator bacteria impairment in this watershed because of transient adenovirus detection and based on the findings from combining the three source tracking methodologies.

Acknowledgments

This investigation was supported by a fund from the County of Orange, California. The field sampling effort was supported by Public Facility and Resource Department of County of Orange. Special thanks go to Chris Crompton, Bruce Moore, and Mary Brill at County of Orange for their valuable input on research design and field logistics support; to Nancy Palmer at City of Laguna Niguel for providing background data on the research site and valuable input on the study design; and to Erica Dunbar at UCI for the assistance with growth experiment.

Supplementary material

253_2007_1047_MOESM1_ESM.doc (76 kb)
ESM 1(DOC 75.5 kb)

Copyright information

© Springer-Verlag 2007

Authors and Affiliations

  • Sunny C. Jiang
    • 1
  • Weiping Chu
    • 1
  • Betty H. Olson
    • 1
  • Jian-Wen He
    • 1
  • Samuel Choi
    • 1
  • Jenny Zhang
    • 1
  • Joanne Y. Le
    • 1
  • Phillip B. Gedalanga
    • 1
  1. 1.Civil and Environmental EngineeringUniversity of CaliforniaIrvineUSA

Personalised recommendations