Determining the source of fecal contamination in recreational waters.

Introduction

Microbiological contamination as a result of human activity impairs water quality and is one of the greatest threats to natural and poorly managed artificially constructed aquatic areas. Escherichia coli, an indicator of fecal coliform bacteria, is still one of the most important and widely used measures of water quality because it is an indicator of fecal pollution (U.S. Environmental Protection Agency [U.S. EPA], 2000). Fecal pollution of surface waters can pose a serious public health threat if the fecal material contains human pathogens; therefore, it is necessary to improve methods of identifying sources of potential and actual microbial pathogens.

According to the Centers for Disease Control and Prevention, recreational water-associated outbreaks of both gastrointestinal and nongastrointestinal illness are at the highest two-year level ever reported (Yoder et al., 2004). In Iowa, for example, 57 percent of state beaches had at least one high Escherichia coli result during weekly monitoring conducted at 37 beaches in 2003. No sources have been pinpointed as the specific cause in any of these areas; thus, it is not known whether human pathogens are entering surface water (O'Brien, 2004). It is also not clear whether the recent increases in "fecal contamination" are due to 1) increased monitoring for coliforms, fecal coliforms, and E. coli; 2) increases in the number of these organisms entering surface waters from nonhuman sources; or 3) a combination of these two factors. For nearly 75 years, tests for coliform bacteria, fecal coliform bacteria, E. coli, or combinations of these bacteria have been used to evaluate various food products, inanimate objects, and water bodies for contamination from fecal material. It has now become clear that there may also be nonhuman hosts for some strains of enteropathogenic E. coli (such as those that cause hemolytic uremic syndrome) and related species detected by the traditional tests for coliforms, fecal coliforms, or both. Thus, it has become more important than ever to determine the source and strain of enteric organisms reaching surface waters.

Several methods have been used to accomplish this task. Ribotyping and repetitive-element anchored polymerase chain reaction (PCR) (genotypic testing) have been used to distinguish fecal E. coli of human origin from E. coli isolates of nonhuman origin (Carson, Shear, Ellersieck, & Asfaw, 2001; Parveen, Portier, Robinson, Edminston, & Tamplin, 1999). It is hoped that a genotypic reference catalog will eventually aid in the identification of bacterial sources to surface waters nationwide. A recent review, however, has found that the current state of progress in the application of genetic typing (bioinformatics) for tracking the source of bacteria is still problematic. Harmsen and Karch (2004) report that a major technical limitation of ribotyping lies in multiple interpretations of the results. For example, there are many RNA sequence databases worldwide but the sequences for individual strains vary among these databases. Ideally, for recreational waters, a single reference data bank containing all strains of E. coli would be made available for diagnostic purposes and would be queried for source tracking. Such a data bank does not yet exist, and comparisons of strains that yield 100 percent similarities are rare. Thus, there is still the problem of interpreting strain relatedness on the basis of sequence similarities. The problem "How close is close enough?" also needs to be resolved (Harmsen & Karch, 2004). In addition, Johnson and co-authors (2004) and McLellan (2004) indicate both temporal and geographic variations of E. coli strains within and between animal species, as well as variations in E. coli strain with diet and even multiple strains within single animal species. Also genotypic testing is often slow and quite expensive for environmental testing. Although phenotypic approaches suffer from limitations similar to those of genotypic approaches (Stoeckel et al., 2004), the authors and others have had success using phenotypic tests to differentiate E. coli isolates from multiple sources (human, waterfowl, livestock, wildlife) (Hagedorn et. al., 1999; O'Brien, 2004; Uzoigwe, 2004) in discreet watersheds. The phenotypic tests are rapid and comparatively inexpensive, and when done in limited geographic areas and time frames, they can differentiate among highly suspected source phenotypes.

Although nearly every state beach in Iowa tested in 2003 indicated some form of fecal pollution, the Iowa Department of Natural Resources (IDNR) has, on the basis of historical testing, identified eight state-owned public beaches that are particularly vulnerable to fecal pollution, including Wyth Lake in George Wyth State Park in East Central Iowa. Wyth Lake Beach has been monitored for indicator bacteria since the summer of 1999. Historically, bacteria numbers at Wyth Lake are highest mid-July through August. The numbers are generally the lowest from early summer through the beginning of July. In January of 2003, George Wyth Park staff met with representatives from the Iowa Geological Survey, IDNR Wildlife and Environmental Protection divisions, and conservation officers to try to solve the problem. Four permanent monitoring wells were placed between the restroom facilities and the beach, leak tests were performed on the sewer pipes, historical aerial photos were studied for information on past land use, and many samples from throughout the watershed were tested for indicator bacteria. Before the study, samples from the monitoring wells had not shown high numbers of E. coli, suggesting that humans may not be the primary source of bacteria in the water at the swimming beach. Much survey work had been done, but no efforts had concentrated on using specific bacterial-source-tracking methodologies to trace and determine the source of fecal indicator bacteria in the swimming area. Wyth Lake is an artificial basin with no surface water connections to other lakes or streams. There is no direct agricultural drainage from livestock or row crops. Thus, the potential sources of fecal contamination were quite limited.

Proper identification of host sources of bacteria entering surface water, such as humans and wild animals, was necessary to solve the problem and provided an opportunity to test the universality of one inexpensive type of phenotypic source tracking: nutrient utilization profiles.

Materials and Methods

Water Collection and Filtration

To approximate the sampling protocol carried out by IDNR, water samples were collected weekly from May 14, 2004, to July 13, 2004. Samples were taken from nine lake water sites within the swimming area in Wyth Lake. Points included water samples from ankle, knee, and chest depths from the left, middle, and right locations of the beach. Water samples were also collected at a fishing jetty opposite the beach and from four permanent monitoring wells placed between restroom facilities and the beach.

All water samples were collected in sterile plastic bottles, and within one hour of collection, the water samples were processed for E. coli counts with a membrane filtration method. Ten milliliters of lake water was placed in 90 mL of sterile water and filtered via vacuum filtration through sterile 0.45-[micro]m membrane filters. Each filter was then placed on membrane-thermotolerant Escherichia coli (mTec) agar, and this preparation was incubated at 37[degrees]C for 2 hours, then at 44.5[degrees]C for 22 hours (American Public Health Association [APHA], 1998).

Fecal Collection

Fecal samples were collected from potential sources of E. coli, which, for this watershed, were only humans and geese. Fecal samples from geese were collected from fecal material present on the beach. A sterile loop was used to collect the fecal material, which was then transferred directly to MacConkey's agar (Fisher, Chicago, Illinois) and incubated at 37[degrees]C for 24 hours.

Human fecal samples were collected from volunteers attending a summer workshop and from university students by use of a sterile cotton swab after defecation. The swab was then streaked directly onto MacConkey's agar and incubated at 37[degrees]C for 24 hours.

E. coli Counts and Isolation

After 24-hour incubation of the filtered water samples, the filter paper was placed on a pad soaked with urea. Bacterial colonies that are urease positive will turn purple in the presence of urea. Bacterial colonies will stay yellow if they are urease negative. Urease-negative colonies were suspected to be E. coli, so these were counted (APHA, 1998), and representative colonies were transferred to MacConkey's agar and incubated at 37[degrees]C for 24 additional hours.

The following procedure was used to process fecal samples and water samples that would undergo further phenotypic testing. On MacConkey's agar, characteristic E. coli colonies are bright pink and doughnut shaped. The authors completed a further confirmation test of suspect colonies by transferring each characteristic E. coli colony to EC-MUG liquid medium (Fisher, Chicago, Illinois). After incubation for 24 hours at 37[degrees]C, positive cultures produced [beta]-glucoronidase, which caused the medium to fluoresce under a black light. All isolates testing positive on EC with MUG medium were classified as E. coli and stored on trypticase soy agar (TSA) slants for later phenotypic characterization.

Phenotypic Characterization of E. coli Isolates

A sterile solution of 0.4 percent saline was prepared and transferred to test tubes so that each tube contained 22 mL of the solution. To each tube of saline, 3 drops of thioglycolate were added. A loopful of E. coli taken from TSA was mixed with the saline, and the concentration of E. coli was adjusted to a transmittance ([lambda]530 nm) between 59 percent and 61 percent. Biolog GN2 96 Well Microplates (Hayward, California) were inoculated with the prepared inocula and incubated for 24 hours at 37[degrees]C. A plate reader set at [lambda]590 nm was then used to measure the reduction of tetrazolium dye in wells selected by discriminate analysis to determine the nutrient utilization patterns (phenotype) of each isolate. Approximately 60 of 95 phenotypic traits represented in the wells of Biolog GN2 Microplates are the same for all isolates, which is not surprising since the isolates are all strains of the same species (E. coli). The remaining wells, however, could be used to obtain "nutrient utilization patterns" (NUPs) characteristic of isolates from the known sources used in the study (Table 1). The NUPs determined for known sources were then compared with isolates from surface water and monitoring wells at Wyth Lake to determine the source of contamination.

Results

Using Nutrient Utilization Patterns (NUPs) that were developed specifically for the study, the authors determined that at the fishing jetty, 0 percent of the E. coli isolates were of human origin, 83 percent of the isolates were from geese, and 17 percent of the isolates were from an unidentified source (Figure 1). At the swimming beach, 31 percent of the isolates were of human origin, and 35 percent of the isolates were from geese, but 34 percent of the isolates did not fit either goose or human NUP (Figure 2). In the monitoring wells, 54 percent of the E. coli that was isolated originated from humans, 23 percent came from geese, and 23 percent was from an unidentified contributor (Figure 3).

Between May 21, 2004, and May 23, 2004, a rainfall event totaling 5.28 inches occurred, causing flooding throughout the watershed area of Wyth Lake. The week before the flooding, E. coli numbers ranged from 0 to 40 (average 15) colony forming units (CFUs) per 100 mL of water at nine locations along the swimming beach. During the flooding, E. coli numbers rose significantly, averaging 370 CFUs per 100 mL of water (range: 90-680 colonies). The week following the flooding, E. coli found in the lake water decreased to an average of 130 CFUs (range: 70-400 CFUs). Two weeks after the flood, the number of E. coli declined even further, to an average of 38 (range: 0-110 colonies per 100 mL of lake water) (Figure 4). The percentage of E. coli isolates from the surface water that had NUPs of human origin rose during the flooding from 0 percent to 57.1 percent and then quickly dropped the week after flooding to 14.3 percent (Figure 5).

Discussion

At Wyth Lake, IDNR monitors E. coli counts in lake water every week during the summer. Because of high levels of E. coli in the beach area, Wyth Lake had to post a swimming advisory for 17 percent of the summer in 2003, and in 2004, the advisory was posted for 62 percent of the study period (May 14-July 13). High counts of fecal coliform bacteria, including E. coli, have caused several periodic closure advisories not only at Wyth Lake, but also at a number of other recreational beaches in Iowa. These high counts, however, give no indication of the sources of fecal contamination or whether human pathogens are present.

Because of success identifying sources of fecal pollution to small watersheds as recently as 2002, investigators suggested the possibility of constructing a universal phenotypic or genotypic (or both) database, which could be referenced for identification of sources for all surface water (Scott, Rose, Jenkins, Farrah, & Lukasik, 2002). Several detailed studies conducted after the work of Scott and co-authors, however, found genotypic diversity in E. coli isolates from animals (Johnson et al., 2004), humans (Escobar-Paramo et al., 2004), and surface waters (McLellan, 2004). In addition. Stoeckel and co-authors (2004) evaluated seven genotypic and phenotypic source-tracking protocols with known source isolates. All of the protocols used methods to compare known source isolates with library databases in blind tests, and the study concluded that none of the protocols are field ready (Stoeckel et al., 2004). Such results show the use of universal reference databases is still problematic.

In a previous study in the authors' lab, multiple-antibiotic resistance (MAR) and nutrient utilization patterns (NUPs) were used as phenotypic methods of bacterial-source tracking in the Silver Lake, Iowa, watershed. Both phenotypic techniques were used on the same isolates from potential sources, including geese and humans, and the two phenotypic tests had similar results in identifying the source of isolates found in Silver Lake water (Uzoigwe, 2004).

Because of the results of the previous studies, the study reported here also attempted to use the same NUPs that identified the sources of fecal pollution in Silver Lake to identify the sources of fecal pollution in Wyth Lake in Waterloo, Iowa. The authors soon found, however, that nutrient utilization patterns obtained with Biolog GN2 microtiter plates that best discriminated among isolates from humans and geese with respect to Wyth Lake (the study reported here) were not the same as those that best discriminated between humans and geese in the Silver Lake study. Thus, the results of the study reported here also show that, even for watersheds less than 100 miles apart, phenotypic libraries may not be universally applicable. That does not mean, however, that phenotypic methods cannot be used to identify sources of fecal contamination in surface waters.

Through selection of 12 definitive nutrient sources for humans and 14 definitive nutrient sources for geese, the authors were able to construct an NUP library specific to the Wyth Lake watershed and use that library to determine the host origin of E. coli isolated from Wyth Lake surface water.

Since E. coli isolates from human and geese at Wyth Lake were not correctly classified by the NUPs established for isolates from the Silver Lake watershed, the authors concluded that it is necessary to obtain E. coli isolates from known suspect sources at each site in order to design an NUP that will correctly classify the E. coli isolates from unknown sources. It is also important to know of potential sources before beginning the characterization so that isolates from suspect sources may be obtained for the purpose of forming nutrient utilization (or other phenotype) patterns. This need has also been shown for genotypic methods used for source tracking (Johnson et al., 2004; McLellan, 2004). For example, the Silver Lake watershed has major potential impacts from humans, geese, pigs, and cows, whereas the major potential sources at Wyth Lake are limited primarily to humans and geese. Even so, as many as 35 percent of the unknown isolates from lake water still could not be classified as originating from either humans or geese.

With the NUPs used in the study reported here, the average rate of correct classification for human-derived phenotypes was 75 percent (known sources were compared with NUP patterns assigned to all isolates from humans). Human-source samples were collected from short-term visitors from several states. Different diets and physical environments may lead to phenotypically different strains of E. coli, and in fact, just as the group of people from which the E. coli was collected was very diverse, so was the E. coli in their intestinal tracts (Escobar-Paramo et al., 2004). The NUP isolated from a diverse population of humans was, however, distinct from the NUP isolated from goose feces collected at Wyth Lake.

The average rate of correct classification for geese-derived phenotypes collected throughout the summer was 70.8 percent (known sources were compared with NUP patterns assigned to all isolates from geese). When the NUP established for geese fecal material collected at the beginning of the summer was applied to fecal samples collected at the end of the summer, the rate of correct classification dropped to a point that was no longer acceptable. This result showed that during the summer, E. coli present in the intestinal tract of geese had phenotypically changed. Similarly, O'Brien (2004) found changing antibiotic resistance patterns for isolates from pigs as their feeding habits changed. These outcomes are evidence that a phenotypic database needs to be temporally, as well as spatially, quite limited.

Flooding occurred in the park that surrounds Wyth Lake the week of May 21, 2004, and the park, swimming beach, and restroom facilities were closed during that time. Although a dike between the lake and the flooding river (the Cedar River) was not breached, the water level of Wyth Lake rose significantly. The rise in water level was probably due not only to the excessive amounts of rain, but also to a rising water table from base flow of the Cedar River. There is no apparent surface water connection between the river and the lake, and yet during the flooding event, the percentage of E. coli isolated from lake water which tracked to human sources increased from 0 percent to 57.1 percent. Since the park was closed at this time, there were no humans swimming in the lake or using the restroom facilities. This evidence supports the supposition that there is a link between Wyth Lake and a source of human fecal contamination from a sewer connection that occasionally overflows directly into the lake. This finding will help the Iowa Department of Natural Resources implement a solution to the chronic bacterial problem at the Wyth Lake swimming beach.

[FIGURE 4 OMITTED]

[FIGURE 5 OMITTED]

Conclusion

The results of the study reported here show that phenotypic bacterial categorization of E. coli isolates from contaminated waters can improve investigators' ability to determine the source of fecal contamination. The authors found that the same phenotypic database could not, however, be used to predict sources in different watersheds. Thus, surveillance and monitoring are still very important components in determining the source of contamination in specific surface water bodies so that proper corrective actions may be implemented.

TABLE 1 Substrates (Wells) Used for Phenotypic Profiles and Growth of
Geese and Human E. coli Isolates on Each Substrate*

                                        Reaction for    Reaction for
Well     Phenotypic Identification      Human           Goose
number   Source in Well                 Identification  Identification

A7       N-acetyl-D-galactosamine       +               NU
A8       N-acetyl-D-glucosamine         +               NU
A10      L-arabinose                    +               NU
B3       L-fucose                       NU              +
B11      D-mannitol                     NU              +
B12      D-mannose                      NU              +
C6       D-sorbitol                     +               +
C8       D-trehalose                    +               NU
D6       D-galacturonic acid            +               NU
D9       D-glucuronic acid              +               NU
F1       Bromosuccinic acid             -               +
F3       Glucuronamide                  NU              +
F6       L-alanine                      NU              +
F8       L-asparagine                   +               NU
F11      Glycyl-L-aspartic acid         NU              +
H2       Inosine                        NU              +
H3       Uridine                        NU              +
H4       Thymidine                      NU              +
H5       Phenylethylamine               -               NU
H6       Putrescine                     -               NU
H7       2-aminoethanol                 -               NU
H9       Glycerol                       NU              +
H11      [alpha]-D-glucose-l-phosphate  NU              +
H12      D-glucose-6-phosphate          NU              +

Well     Phenotypic Identification                       Goose
number   Source in Well                  Human Isolates  Isolates

A7       N-acetyl-D-galactosamine         95.80%         NU
A8       N-acetyl-D-glucosamine           91.66%         NU
A10      L-arabinose                     100%            NU
B3       L-fucose                        NU              100%
B11      D-mannitol                      NU               93.54%
B12      D-mannose                       NU               87.19%
C6       D-sorbitol                       95.80%          93.50%
C8       D-trehalose                      95.80%         NU
D6       D-galacturonic acid              91.66%         NU
D9       D-glucuronic acid                95.80%         NU
F1       Bromosuccinic acid               79.16%          58.06%
F3       Glucuronamide                   NU               87.10%
F6       L-alanine                       NU               98.54%
F8       L-asparagine                     91.66%         NU
F11      Glycyl-L-aspartic acid          NU              100%
H2       Inosine                         NU               93.54%
H3       Uridine                         NU               80.65%
H4       Thymidine                       NU               87.10%
H5       Phenylethylamine                100%            NU
H6       Putrescine                       95.80%         NU
H7       2-aminoethanol                  100%            NU
H9       Glycerol                        NU               96.77%
H11      [alpha]-D-glucose-l-phosphate   NU               90.32%
H12      D-glucose-6-phosphate           NU               96.77%

*NU indicates that the wells were not used.

FIGURE 1 Sources of E. coli at Jetty

Human         83%
Goose         17%
Unidentified   0%

The relative percentages of E. coli isolates from geese and humans found
at a fishing jetty at Wyth Lake.

Note: Table made from pie chart.

FIGURE 2 Sources of E. coli at Swimming Beach

Human         31%
Goose         35%
Unidentified  34%

The relative percentages of E. coli isolates from geese and humans found
at the swimming beach at Wyth Lake.

Note: Table made from pie chart.

FIGURE 3 Sources of E. coli in Monitoring Wells

Human         54%
Goose         23%
Unidentified  23%

The relative percentages of E. coli isolates from geese and humans found
in monitoring wells between the bathhouse and the swimming beach at Wyth
Lake.

Note: Table made from pie chart.

Acknowledgements: This work was supported, in part, by funding from the Roy J. Carver Charitable Trust, the Iowa Space Grant Consortium, and the National Atmospheric Space Administration. The authors thank the Iowa Department of Natural Resources for its help and cooperation.

REFERENCES

American Public Health Association. (1998). Standard methods for the examination of water and wastewater (20th ed.). Washington, DC: Author.

Carson, A.C., Shear, B.L., Ellersieck, M.R., & Asfaw, A. (2001). Identification of fecal Escherichia coli from humans and animals by ribotyping. Applied and Environmental Microbiology, 67, 1503-1507.

Escobar-Paramo, P., Grenet, K., Menac'h, A., Rode, L., Salgado, E., Amorin, C., Gouriou, S., Picard, B., Rahimy, M.C., Andremont, A., Denamur, E., & Ruimy, R. (2004). Large-scale population structure of human commensal Escherichia coli isolates. Applied and Environmental Microbiology, 70(9), 5698-5700.

Hagedorn, C., Robinson, S. L., Filtz, J.R., Grubbs, S.M., Angier, T.A., & Reneau, R.B. (1999). Determining sources of fecal pollution in a rural Virginia watershed with antibiotic resistance patterns in fecal streptococci. Applied and Environmental Microbiology, 65(12), 5522-5531.

Harmsen, D., & Karch, H. (2004). 16s rDNA for diagnosing pathogens: a living tree. American Society for Microbiology News, 70(1), 19-24.

Johnson, L.K., Brown M.B., Carruthers E.A., Ferguson J.A., Dombek P.E., & Sadawsky, M.J. (2004). Sample size, library composition, and genotypic diversity among natural populations of Escherichia coli from different animals influence accuracy of determining sources of fecal pollution. Applied and Environmental Microbiology, 70(8), 4478-4485.

McLellan, S.L. (2004). Genetic diversity of Escherichia coli isolated from urban rivers and beach water. Applied and Environmental Microbiology, 70(8), 4658-4665.

O'Brien, E.H. (2004). Multiple antibiotic resistance for Escherichia coli source tracking in Iowa. Unpublished master's thesis, University of Northern Iowa, Cedar Falls.

Parveen, S., Portier, K.M., Robinson, K., Edminston, L., & Tamplin, M.L. (1999). Discriminant analysis of ribotype profiles of Escherichia coli for differentiating human and nonhuman sources of fecal pollution. Applied and Environmental Microbiology, 65, 3142-3147.

Scott, T.M., Rose, J.B., Jenkins, T.M., Farrah, S.R., & Lukasik, J. (2002). Microbial source tracking: Current methodology and future directions. Applied and Environmental Microbiology, 68(12), 5796-5803.

Stoeckel, D.M., Mathes, M.V., Hyer, K.E., Hagedorn, C., Kator, H., Lukasik, J., O'Brien, T.L., Fenger, T.W., Samadpour, M., Strickler, M., & Wiggins, B.A. (2004). Comparison of seven protocols to identify fecal contamination sources using Escherichia coli. Environmental Science Technology, 38, 6109-6117.

U.S. Environmental Protection Agency. (2000). Improved enumeration methods for the recreational water quality indicators: Enterococci and Escherichia coli (EPA/821/R-97004). Washington DC: Author, Office of Science and Technology.

Uzoigwe, J.C. (2004). Using nutrient utilization patterns to determine the source of Escherichia coli found in surface water. Unpublished master's thesis, University of Northern Iowa, Cedar Falls.

Yoder, J.S., Blackburn, B.G., Craun, G.F., Hill, V., Levy, D.A., Chen, N., Lee, S.H., Calderon, R.L., & Beach, M.J. (2004). Surveillance for waterborne-disease outbreaks associated with recreational water--United States, 2001-2002. Centers for Disease Control and Prevention Surveillance Summaries, 53(SS08), 1-22.

K.J. Meyer

C.M. Appletoft

A.K. Schwemm

J.C. Uzoigwe

E.J. Brown, Ph.D.

Corresponding Author: Edward Brown, Professor of Biology and Environmental Science, University of Northern Iowa, 127 MSH-0421, Department of Biology, Cedar Falls, IA 50614-0421. E-mail: [e-mail].