- Email: [email protected]
Deer diet affects ribotype diversity of Escherichia coli for bacterial source tracking
Water Research 37 (2003) 3263–3268
Deer diet affects ribotype diversity of Escherichia coli for bacterial source tracking Peter G. Hartel*, Jacob D. Summer, William I. Segars Department of Crop and Soil Sciences, University of Georgia, 3111 Miller Plant Sciences Bld, Athens, GA 30602-7272, USA Received 6 January 2003; accepted 17 March 2003
Abstract Ribotyping is one of a number of genotypic methods for bacterial source tracking. This method requires a host origin database of one bacterial species be established in order to identify environmental isolates. Researchers establishing these databases have observed considerable ribotype diversity within a specific bacterial species. One source of this diversity may be diet. We determined the effect of diet on ribotype diversity for Escherichia coli in penned and wild deer (Odocoileus virginianus) in a 13-ha forested watershed. A total of 298 E. coli isolates was obtained, 100 from penned deer, 100 from wild deer, and 98 from the stream in the watershed to which all deer had access. The wild deer had significantly more ribotypes (35) than the penned deer (11 ribotypes, p ¼ 0:05). This result suggests that diet affected ribotype diversity, and that a host origin database for bacterial source tracking should contain bacterial isolates from wild rather than from captive animals. Also, 42 of 98 (42.9%) environmental isolates matched penned and wild deer ribotypes. If bacterial source tracking determines that fecal contamination is predominantly from wildlife, then it may be unnecessary to monitor these watersheds because control over wildlife is difficult. r 2003 Elsevier Science Ltd. All rights reserved. Keywords: Deer; Escherichia coli; Host origin; Microbial source tracking; Nonpoint source pollution; Water quality
1. Introduction In recent years, there has been considerable interest in developing phenotypic and genotypic methods for determining the host origin of fecal bacteria in contaminated waters, a technique commonly referred to as bacterial source tracking or microbial source tracking. All of these methods are based on the assumption that specific markers or strains of bacteria are associated with specific animal species (e.g., [1]). Most of the recent research on phenotypic (characteristics expressed by the bacterium) methods has concentrated on multiple antibiotic resistance (e.g., [2]), while most of the research on genotypic (DNA-based) methods has concentrated on ribotyping (e.g., [3]), *Corresponding author. Tel.: +1-706-542-0898; fax: +1706-542-0914. E-mail address: [email protected] (P.G. Hartel).
pulsed field gel electrophoresis (e.g., [4]), and various PCR methods (e.g., [5]). In this study, we used ribotyping because the section of DNA encoding ribosomal RNA (rRNA) is highly conserved. For this reason, ribotyping is considered one of the most reproducible of the molecular typing methods [6]. For ribotyping to work, a bacterial species must be selected. Among the best indicators of fecal contamination are the fecal coliforms, a group of nonpathogenic bacteria commonly found in the feces of warm-blooded animals [7]. Of fecal coliforms, all the bacterial source tracking studies have been done with Escherichia coli. Samadpour and Chechowitz [8] matched 421 of 589 E. coli ribotype patterns (71%) from Little Soos Creek (in Washington State) to cows, deer, dogs, ducks, horses, humans, swine, and chickens. Subsequent studies in US national parks and recreational areas also matched E. coli ribotypes to various animal hosts [9–11]. Ribotyping identified differences
0043-1354/03/$ - see front matter r 2003 Elsevier Science Ltd. All rights reserved. doi:10.1016/S0043-1354(03)00170-2
3264
P.G. Hartel et al. / Water Research 37 (2003) 3263–3268
between human and nonhuman sources of E. coli under conditions of a saltwater to freshwater gradient [3]. All ribotyping studies using E. coli for bacterial source tracking depend on a host origin database to identify environmental isolates. Researchers creating these host origin databases have observed considerable temporal [12] and geographic [13] diversity of their E. coli ribotypes. One reason for this diversity may be diet. Research on the general effect of diet on E. coli subspecies is limited. Diet played a prominent role in the adaptation of E. coli to native Australian bats, marsupials, and rodents [14]. Also, cattle fed grain had increased numbers of acid-resistant E. coli compared to cattle fed hay [15]. Several studies with enterohemorrhagic E. coli O157:H7 in cattle (e.g., [16]) suggest that numbers of this bacterium fluctuate with dietary stress. Research suggests, however, that deer are not a major reservoir of this E. coli strain [17]. Our objective was to determine the effect of diet on ribotype diversity of E. coli in deer. To accomplish this, we selected a 13.0-ha watershed containing a perennial first-order stream located in the Whitehall Experimental Forest, near the campus of the University of Georgia, Athens. The site is unusual in that it contains penned deer (Odocoileus virginianus) in a 2.2-ha portion of the watershed as part of the University of Georgia Whitehall Deer Research Facility. Bucks and does are kept separately, but are sometimes rotated among certain pens. Because of the number of penned deer, wild deer are attracted to the facility and their feces are abundant outside the pens. Because the penned deer are fed a known diet, this permitted us to assess the effect of diet on ribotype diversity of E. coli isolates obtained from the feces of penned and wild deer. Furthermore, because humans and domestic animals are excluded from the watershed, this site also represents a watershed where the sources of fecal coliforms are predominantly from wildlife. Some watersheds that exceed their limits for fecal coliforms (typically 200 fecal coliforms per 100 mL for recreational waters [7]) are also located in remote areas were the sources of fecal coliforms are predominantly from wildlife. If it could be firmly established that the sources of fecal coliforms are wildlife, then it may be unnecessary to monitor these watersheds because control over wildlife is difficult. Therefore, we also obtained E. coli isolates from the stream in the watershed during a runoff event and matched ribotypes of these isolates against ribotypes of E. coli isolates from penned and wild deer, as well as against E. coli ribotypes from cattle (Bos taurus), horses (Equus caballus), humans (Homo sapiens), and swine (Sus scrofa) contained in a host origin database from Athens, GA. In this manner, we determined the percentage contribution of human or nonhuman animals to the total E. coli (fecal coliform) load in the watershed.
2. Materials and methods 2.1. Deer age, genetic diversity, and diet The Whitehall Deer Research Facility contained 89 deer, ranging in age from >12 years old to fawns born in 2001 [18]. Many of the deer were born at the facility, but others were brought in from elsewhere (14 from South Carolina in 1995, 18 from North Carolina in 1997, and 16 from Stone Mountain, GA in 1997 [18]). In addition, one to three wild deer from Georgia are introduced into the facility each year. Therefore, the age and genetic basis of the penned deer are kept diverse and similar to that of wild deer. The penned deer were fed a standard diet (Big Buck Macho Feed [19]) comprising (%): ground corn (33.0), soybean hulls (23.4), alfalfa meal (20.0), soybean oil meal (13.9), wheat midds (5.8), dried molasses (2.0), animal and vegetable fat (1.0), limestone (0.2), and trace elements mix, vitamin mix, and salt (0.7). At the time of fecal sampling (mid-January), wild deer are likely eating any few remaining acorns from red or white oak (Quercus spp.) and leafy vegetation of evergreen vines, shrubs (e.g., greenbriar [Similax spp.], common privet [Ligustrum vulgare], yellow jessamine [Gelsemium sempervirens], and Japanese honeysuckle [Lonicera japonica] and other winter annual and perennial plants (e.g., henbit, [Lamium amplexicaule] and wild geranium [Geranium maculatum] [18]). 2.2. Selection and identification of E. coli isolates During a 2.92-cm rainfall event on January 19, 2001, 2.28 cm of rain (78%) fell between 1400 and 1600 h, and this was sufficient to cause visual signs of runoff throughout the watershed, including the penned area of the deer research facility. Duplicate water samples were aseptically collected at 1600 h from two locations in the watershed, one site 50 m upstream of the deer research facility and the other 100 m downstream. At the same time, freshly deposited fecal pellets from three different scat piles were obtained separately from bucks and does inside the pens as well as from wild deer outside the facility fences. Fecal pellets were aseptically collected with ethanol flame-sterilized forceps and placed in sterile Whirlpak bags. Water samples were processed within 6 h; fecal samples were processed within 24 h. In the case of the fecal pellets, three pellets from each scat pile were resuspended in 20 mL of 0.1% sterile peptone and were mixed outside of the bag by hand until the suspension was uniform. Each suspension was streaked on duplicate 5-cm Petri dishes containing mTEC agar (Difco Laboratories, Sparks, MD). In the case of the water samples, duplicate 10- and 100-mL water samples were filtered through separate sterile 0.45mm membranes and the membranes were transferred
P.G. Hartel et al. / Water Research 37 (2003) 3263–3268
aseptically to the mTEC agar plates. All plates were sealed inside triplicate Ziploc bags and were incubated submerged in a water bath at 44.570.2 C for 24 h according to Standard Methods for the Examination of Water and Wastewater [7]. A total of 300 randomly selected yellow isolates, consisting of 50 isolates each from the top and bottom of the watershed, 100 from penned deer, and 100 from wild deer were streaked onto tryptic soy agar (Difco). This total number was selected because it was consistent with the number of isolates obtained in other ribotyping studies (e.g., [3] 238 source isolates). Of the 100 penned deer isolates, 42 isolates were from a strictly buck pen, 38 isolates were from a strictly doe pen, and 20 isolates from a pen where bucks and does were rotated in and out. After incubation at 35 C for 24 h, the streaking was repeated twice on the same medium to ensure the purity of each isolate. Each isolate was inoculated into a 24-multiwell tissue culture plate containing separate 1-mL slants of Simmons citrate and urea agar (both Difco). Three bacterial species from the American Type Culture Collection (ATCC, Manassas, VA) were used as controls: E. coli ATCC #11775 (citrate negative, urea hydrolysis negative), Klebsiella pneumoniae ATCC #13883 (citrate positive, urea hydrolysis positive), and Enterobacter aerogenes ATCC #13048 (citrate positive, urea hydrolysis negative). Isolates that were citrate negative on Simmons citrate agar and urea hydrolysis negative on urea agar were subjected to an oxidase test. Isolates that were oxidase negative were considered E. coli. Of the 300 isolates, all but two (from the top of the watershed) were identified as E. coli. 2.3. DNA extraction, quantification, and ribotyping of E. coli isolates Each E. coli isolate was inoculated into Luria-Bertani broth contained in a test tube and incubated on a rotating shaker at 75 rpm at 35 C. After 18 h, a 2-mL sample of the culture was removed and the DNA extracted with a commercial kit (Qiagen DNeasy, Valencia, CA). The DNA was quantified with a fluorometer using standard DNA from E. coli strain B (Sigma Chemical Company, St. Louis, MO). Two 1-mg samples of DNA from each isolate and from the E. coli ATCC #11775 control were each separately digested overnight with the restriction enzymes EcoRI and PvuII according to the manufacturer’s directions (Roche Molecular Biochemicals, Indianapolis, IN). Loading dye (Sigma) was added to the digested DNA and the DNA was electrophoresed in a 1.0% agarose gel at 58 V for 3 h. Digoxigenin-labeled (DIGlabeled) Marker III (Roche) was the molecular weight marker and occupied every fifth lane of the gel. Control lanes contained no DNA (control) and DNA from E. coli ATCC #11775. DNA was transferred by Southern
3265
blotting to a nylon membrane with a vacuum blotting system (VacuGene, Amersham Pharmacia Biotech, Piscataway, NJ), and the DNA on the membrane was crosslinked with UV light. Following prehybridization at 42 C for 2 h, the membrane was hybridized at 42 C overnight to DIG-labeled cDNA from E. coli total rRNA (Sigma). Membranes were prepared for chemiluminescence by a series of washing steps before a chemiluminescent substrate for alkaline phosphatase (Roche) was added. Membranes were placed in an imager (FluorChem 8000, Alpha Innotech, San Leandro, CA) and images saved as a TIFF file. TIFF files were imported into GelCompar II (Applied Maths, Kortrijk, Belgium) for analysis. Typically, gels showed 9–11 bands for EcoRI and 11–13 bands for PvuII. DNA fragments o1375 base pairs were ignored because they were often indistinct. Lanes were normalized within the gel with the molecular weight marker and variations among the gels were assessed with the E. coli ATCC #11775 strain. Optimization (maximum percentage shift allowed between two different patterns for the patterns to still be considered a match) and tolerance (maximum percentage shift allowed between two bands on different patterns for the bands to still be considered a match) were each set at 1.00%. The normalized banding patterns for both enzymes were stacked with EcoRI on the top and PvuII on the bottom to create one combined ribotype pattern for each isolate; this was considered sufficient for good discrimination among ribotypes. Similarity indices were determined using Dice’s coincidence index and the distance among clusters calculated with the unweighted pair-group method using arithmetic averages (UPGMA). The banding pattern of the control E. coli ATCC #11775 strain varied from gel to gel and a similarity index of 95.0% was required for all the banding patterns to be considered the same ribotype. Based on variability of the inter-gel E. coli control, banding patterns of all other isolates had to have a similarity index of X95.0% to be considered the same ribotype. Five dendrograms were created: (a) penned bucks, (b) penned does, (c) all penned deer combined, (d) wild deer, and (e) all deer combined. When the data were combined, it was possible to determine the number of shared ribotypes. In addition, ribotype patterns from the environmental isolates were matched against ribotype patterns from penned and wild deer, and ribotype patterns of a host origin database (201 isolates total) from cattle, horses, humans, and swine living in the vicinity of Athens, GA. Because no assumptions can be made as to the distribution of the absolute difference in mean ribotype number between all penned deer combined (100 E. coli isolates) and wild deer only (100 E. coli isolates), the data were analyzed by an approximate randomization test [20]. Briefly, the ribotype patterns from each of the
3266
P.G. Hartel et al. / Water Research 37 (2003) 3263–3268
two groups were pooled and the patterns randomly selected without replacement to yield two data sets. Dendrograms were created from each data set and the number of ribotypes determined at X95% similarity. This process was repeated 20 times to construct an empirical distribution of the absolute difference in mean ribotype numbers. Repeating the process 20 times was the minimum necessary to yield a p value of 0.05. The null hypothesis was that numbers of ribotypes was the same between penned or wild deer. The approximate randomization test was repeated separately to determine the absolute difference in mean ribotype numbers between penned bucks only (42 E. coli isolates) and penned does only (38 E. coli isolates). The null hypothesis was that numbers of ribotypes was the same between bucks or does.
3. Results When the number of ribotypes was determined from separate dendrograms of penned bucks only and penned does only, both groups had six ribotypes. Under the conditions of the approximate randomization test, the null hypothesis was not rejected. Therefore, sex did not affect ribotype number. When the ribotype patterns of E. coli isolates from penned bucks and penned does were combined to determine the amount of sharing, 12 different ribotypes were observed (Fig. 1, Table 1a). Two of the 12 ribotypes (16.7%) were shared and these represented 24 of 80 isolates (30.0%). Bucks had the majority of the ribotypes (6 of 12) and almost a majority of the isolates (36 of 80 isolates, 45.0%). When the number of ribotypes was determined from separate dendrograms of penned deer only and wild deer only, the penned and wild deer had 11 and 35 ribotypes, respectively. Under the conditions of the approximate randomization test, the null hypothesis was rejected, and the difference between the number of ribotypes for two groups (n ¼ 24) was statistically significant (p ¼ 0:05). When the ribotype patterns of E. coli isolates from penned and wild deer were combined to determine the amount of sharing, 44 different ribotypes were observed (dendrogram not shown, Table 1b). Only six ribotypes (13.6%) were shared, but these six ribotypes represented almost a majority of the isolates (93 of 200 isolates, 46.5%). Of the remaining unshared ribotypes, 27 (61.4%) were found in wild deer and 11 (25.0%) were Fig. 1. Dendrogram of the ribotype patterns of 80 E. coli isolates from bucks and does from the University of Georgia Whitehall Deer Research Facility near Athens, GA. The similarity index is given on the top scale. Based on the variability of the inter-gel control, E. coli ATCC #11775, the cutoff to distinguish differences among ribotypes was 95% (dashed vertical line).
P.G. Hartel et al. / Water Research 37 (2003) 3263–3268 Table 1 Number and percent of shared and unshared Escherichia coli ribotypes and isolates between (a) penned bucks and does, and (b) penned and wild deer Ribotypes Number
Isolates %
(a) Penned bucks versus penned does Buck 6 50.0 Shared 2 16.7 Does 4 33.3 Total
12
100.0
(b) Penned deer versus wild deer Wild deer 27 61.4 Shared between 6 13.6 wild and penned deer Penned deer 11 25.0 Total
44
100.0
Number
%
36 24 20
45.0 30.0 25.0
80a
100.0
74 93
37.0 46.5
33
16.5
200
100.0
a
20 isolates were not included because they came from a pen where bucks and does were rotated in and out.
found in penned deer. Therefore, although 100 E. coli isolates were each obtained from penned and wild deer, penned deer had less than half the E. coli ribotypes of wild deer. When the 98 environmental isolates were matched against the 200 deer isolates, 19 of 48 (39.6%) and 23 of 50 (46.0%) isolates were identified as deer isolates from the top and bottom of the watershed, respectively. Of the 42 total matched isolates, 27 (64.3%) were from wild deer, 12 (28.6%) were from penned deer, and three (7.1%) were shared between wild and penned deer. When the 98 environmental isolates were matched against the host origin database from Athens, Ga., three isolates matched a horse ribotype and one matched a human ribotype, for total of four matches (4.1%).
3267
other captive animals fed a standard diet, then the database may not represent the ribotype diversity in the same wild animals. It is unlikely that geographic variability is a confounding factor here because the maximum home range for white-tailed bucks and does in suburban Georgia is only 644 and 334 ha, respectively [21]. There was no significant difference in the number of ribotypes between penned bucks and does, suggesting that sex differences were minimal. This result was expected because of the close proximity of the two populations in the pens, and suggests that host origin databases may not have to include isolates from both male and female animals. Nevertheless, the penned animals had significantly fewer ribotypes than the wild animals and it may be necessary to test other penned deer populations to confirm this. Of the 98 environmental E. coli isolates, 42 (42.9%) matched deer ribotypes. This matching was expected in the watershed because the large number of penned deer (n ¼ 89), the number of wild deer attracted to these penned deer, and the runoff conditions all guaranteed a large number of fecal bacteria from this source. Wild deer must have been attracted to the penned deer because almost as many environmental isolates matched deer isolates from above the deer pens (19 of 48) as below (23 of 50) the deer pens. When the environmental isolates were also tested against a host origin database, three isolates matched horses and one-matched humans. This matching was also expected because some ribotypes are shared among different animal species. The total percentage sharing here was 4.1%, consistent with the maximal 9.5% of BOX PCR DNA fingerprint sharing among cows, chickens, and swine [5]. Finally, the majority of the isolates (52 of 98, 53.1%) were unidentified. This high percentage of unknowns was also expected because E. coli isolates were not obtained from other wild animal species in the watershed. Ribotyping could identify deer as a source of fecal contamination, and given a sufficiently large host origin database, it may be possible to identify most of the remaining unidentified environmental isolates.
4. Discussion The data suggest that diet significantly affected the number of E. coli ribotypes in deer. It is unlikely that deer age or genetic diversity affected the number of E. coli ribotypes because the penned deer were managed to mimic wild deer populations. The idea that diet affects the number of E. coli ribotypes is consistent with reports that diet changes numbers of enterohemorrhagic E. coli O157:H7 in cattle (e.g., [16]) and specific E. coli subspecies in certain Australian mammals [14]. In our study, the penned deer had fewer than half of the ribotypes of wild deer. If researchers construct a host origin database of E. coli isolates obtained from zoo or
5. Conclusions Diet significantly affected the ribotype diversity of E. coli in deer. Penned deer, which were fed a standard diet, had fewer than half the E. coli ribotypes of wild deer. This result suggests that isolates for bacterial source tracking databases should come from wild rather than from captive animals. During a runoff event, deer contributed 42.9% of the fecal contamination (as E. coli) to the stream in the watershed. As host origin databases for bacterial source tracking become more comprehensive both with regards to number of isolates
3268
P.G. Hartel et al. / Water Research 37 (2003) 3263–3268
as well as the animal species tested, it may be possible to identify an even greater percentage of isolates, even to the point of identifying specific animal operations. If it can be shown in certain instances that the vast majority of isolates come from wildlife, then it may be unnecessary to monitor the watersheds because control over wildlife is difficult.
[8]
[9]
Acknowledgements [10]
We thank Bruce Beck, Demetrius Cox, Dan Hall, Robin Kuntz, David Osborn, Todd Rasmussen, Ken Roberts, and Karen Rodgers for their advice and assistance. This research was partially supported by grants from the USDA–CSREES Special Grants Program and the USDA–CSREES through the Texas Agricultural Extension Service.
[11]
References
[13]
[1] Amor K, Heinrichs DE, Frirdich E, Ziebell K, Johnson RP, Whitfield C. Distribution of core oligosaccharide types in lipopolysaccharides from E. coli. Infect Immun 2000;68(3):1116–24. [2] Wiggins BA, Andrews RW, Conway RA, Corr CL, Dobratz EJ, Dougherty DP, Eppard JR, Knupp SR, Limjoco MC, Mettenburg JM, Rinehardt JM, Sonsino J, Torrijos RL, Zimmerman ME. Use of antibiotic resistance analysis to identify nonpoint sources of fecal pollution. Appl Environ Microbiol 1999;65(11):3483–6. [3] Parveen S, Portier KM, Robinson K, Edmiston L, Tamplin ML. Discriminant analysis of ribotype profiles of E. coli for differentiating human and nonhuman sources of fecal pollution. Appl Environ Microbiol 1999;65(7): 3142–7. [4] Parveen S, Hodge NC, Stall RE, Farrah SR, Tamplin ML. Phenotypic and genotypic characterization of human and nonhuman E. coli. Water Res 2001;35(2):379–86. [5] Dombek PE, Johnson L-AK, Zimmerley ST, Sadowsky MJ. Use of repetitive DNA sequences and the PCR to differentiate E. coli isolates from human and animal sources. Appl Environ Microbiol 2000;66(6):2572–7. [6] Farber JM. An introduction to the hows and whys of molecular typing. J Food Prot 1996;59(10):1091–101. [7] Clesceri LS, Greenberg AE, Eaton AD. Standard methods for the examination of water and wastewater, 20th ed.
[12]
[14]
[15]
[16]
[17]
[18]
[19]
[20] [21]
Washington, DC: American Public Health Association, American Water Works Association, and Water Environment Federation, 1998. Samadpour M, Chechowitz N. Little Soos Creek microbial source tracking. Report to Surface Water Management Division, King County Department of Public Works, Seattle, WA, 1995. Berghoff K. Beach sediment bacterial contamination and microbial source tracking study. 1997 Year End Summary Report, Glen Canyon National Recreational Area, National Park Service, UT, 1998. Farag A, Goldstein JN, Woodward DF, Samadpour M. Water quality in three creeks in the backcountry of Grand Teton National Park, USA. J Freshwater Ecol 2001;16(1): 135–43. Tippets N. Backcountry water quality testing in Grand Teton National Park–1998 Summer Season. Report from the Environmental and Contaminants Research Center, USGS, 1999. Jenkins MB, Hartel PG, Olexa TJ, Stuedemann JA. Putative temporal variability of E. coli ribotypes from yearling steers. J Environ Qual 2003;32(1):305–9. Hartel PG, Summer JD, Hill JL, Collins JV, Entry JA, Segars WI. Geographic variability of E. coli ribotypes from animals in Idaho and Georgia. J Environ Qual 2002;31(4):1273–8. Pupo GM, Lan R, Reeves PR, Baverstock RB. Population genetics of E. coli in a natural population of native Australian rats. Environ Microbiol 2000;2(6):594–610. Diez–Gonzalez F, Callaway TR, Kizoulis MG, Russell JB. Grain feeding and the dissemination of acid-resistant E. coli from cattle. Science 1998;281(5383):1666–8. Cray Jr WC, Casey TA, Bosworth BT, Rasmussen MA. Effect of dietary stress on fecal shedding of E. coli O157: H7 in calves. Appl Environ Microbiol 1998;64(5):1975–9. Fischer JR, Zhao T, Doyle MP, Goldberg MR, Brown CA, Sewell CT, Kavanaugh DM, Bauman CD. Experimental and field studies of E. coli O157: H7 in white-tailed deer. Appl Environ Microbiol 2001;67(3):1218–24. Osborn D. Personal communication. School of Forest Resources, 3–324 Forest Resources, University of Georgia, Athens, GA 30602-2152, 2001. Roberts KR. Personal communication. Poultry Science Research College, South Milledge Ave., University of Georgia, Athens, GA 30602-2772, 2001. Noreen EW. Computer intensive methods for testing hypotheses: an introduction. New York: Wiley; 1989. Rogers CL. Utilization of cedar glades by white-tailed deer at Chickamauga Battlefield Park. MS thesis, University of Georgia, Athens, 1996.
Deer diet affects ribotype diversity of Escherichia coli for bacterial source tracking Peter G. Hartel*, Jacob D. Summer, William I. Segars Department of Crop and Soil Sciences, University of Georgia, 3111 Miller Plant Sciences Bld, Athens, GA 30602-7272, USA Received 6 January 2003; accepted 17 March 2003
Abstract Ribotyping is one of a number of genotypic methods for bacterial source tracking. This method requires a host origin database of one bacterial species be established in order to identify environmental isolates. Researchers establishing these databases have observed considerable ribotype diversity within a specific bacterial species. One source of this diversity may be diet. We determined the effect of diet on ribotype diversity for Escherichia coli in penned and wild deer (Odocoileus virginianus) in a 13-ha forested watershed. A total of 298 E. coli isolates was obtained, 100 from penned deer, 100 from wild deer, and 98 from the stream in the watershed to which all deer had access. The wild deer had significantly more ribotypes (35) than the penned deer (11 ribotypes, p ¼ 0:05). This result suggests that diet affected ribotype diversity, and that a host origin database for bacterial source tracking should contain bacterial isolates from wild rather than from captive animals. Also, 42 of 98 (42.9%) environmental isolates matched penned and wild deer ribotypes. If bacterial source tracking determines that fecal contamination is predominantly from wildlife, then it may be unnecessary to monitor these watersheds because control over wildlife is difficult. r 2003 Elsevier Science Ltd. All rights reserved. Keywords: Deer; Escherichia coli; Host origin; Microbial source tracking; Nonpoint source pollution; Water quality
1. Introduction In recent years, there has been considerable interest in developing phenotypic and genotypic methods for determining the host origin of fecal bacteria in contaminated waters, a technique commonly referred to as bacterial source tracking or microbial source tracking. All of these methods are based on the assumption that specific markers or strains of bacteria are associated with specific animal species (e.g., [1]). Most of the recent research on phenotypic (characteristics expressed by the bacterium) methods has concentrated on multiple antibiotic resistance (e.g., [2]), while most of the research on genotypic (DNA-based) methods has concentrated on ribotyping (e.g., [3]), *Corresponding author. Tel.: +1-706-542-0898; fax: +1706-542-0914. E-mail address: [email protected] (P.G. Hartel).
pulsed field gel electrophoresis (e.g., [4]), and various PCR methods (e.g., [5]). In this study, we used ribotyping because the section of DNA encoding ribosomal RNA (rRNA) is highly conserved. For this reason, ribotyping is considered one of the most reproducible of the molecular typing methods [6]. For ribotyping to work, a bacterial species must be selected. Among the best indicators of fecal contamination are the fecal coliforms, a group of nonpathogenic bacteria commonly found in the feces of warm-blooded animals [7]. Of fecal coliforms, all the bacterial source tracking studies have been done with Escherichia coli. Samadpour and Chechowitz [8] matched 421 of 589 E. coli ribotype patterns (71%) from Little Soos Creek (in Washington State) to cows, deer, dogs, ducks, horses, humans, swine, and chickens. Subsequent studies in US national parks and recreational areas also matched E. coli ribotypes to various animal hosts [9–11]. Ribotyping identified differences
0043-1354/03/$ - see front matter r 2003 Elsevier Science Ltd. All rights reserved. doi:10.1016/S0043-1354(03)00170-2
3264
P.G. Hartel et al. / Water Research 37 (2003) 3263–3268
between human and nonhuman sources of E. coli under conditions of a saltwater to freshwater gradient [3]. All ribotyping studies using E. coli for bacterial source tracking depend on a host origin database to identify environmental isolates. Researchers creating these host origin databases have observed considerable temporal [12] and geographic [13] diversity of their E. coli ribotypes. One reason for this diversity may be diet. Research on the general effect of diet on E. coli subspecies is limited. Diet played a prominent role in the adaptation of E. coli to native Australian bats, marsupials, and rodents [14]. Also, cattle fed grain had increased numbers of acid-resistant E. coli compared to cattle fed hay [15]. Several studies with enterohemorrhagic E. coli O157:H7 in cattle (e.g., [16]) suggest that numbers of this bacterium fluctuate with dietary stress. Research suggests, however, that deer are not a major reservoir of this E. coli strain [17]. Our objective was to determine the effect of diet on ribotype diversity of E. coli in deer. To accomplish this, we selected a 13.0-ha watershed containing a perennial first-order stream located in the Whitehall Experimental Forest, near the campus of the University of Georgia, Athens. The site is unusual in that it contains penned deer (Odocoileus virginianus) in a 2.2-ha portion of the watershed as part of the University of Georgia Whitehall Deer Research Facility. Bucks and does are kept separately, but are sometimes rotated among certain pens. Because of the number of penned deer, wild deer are attracted to the facility and their feces are abundant outside the pens. Because the penned deer are fed a known diet, this permitted us to assess the effect of diet on ribotype diversity of E. coli isolates obtained from the feces of penned and wild deer. Furthermore, because humans and domestic animals are excluded from the watershed, this site also represents a watershed where the sources of fecal coliforms are predominantly from wildlife. Some watersheds that exceed their limits for fecal coliforms (typically 200 fecal coliforms per 100 mL for recreational waters [7]) are also located in remote areas were the sources of fecal coliforms are predominantly from wildlife. If it could be firmly established that the sources of fecal coliforms are wildlife, then it may be unnecessary to monitor these watersheds because control over wildlife is difficult. Therefore, we also obtained E. coli isolates from the stream in the watershed during a runoff event and matched ribotypes of these isolates against ribotypes of E. coli isolates from penned and wild deer, as well as against E. coli ribotypes from cattle (Bos taurus), horses (Equus caballus), humans (Homo sapiens), and swine (Sus scrofa) contained in a host origin database from Athens, GA. In this manner, we determined the percentage contribution of human or nonhuman animals to the total E. coli (fecal coliform) load in the watershed.
2. Materials and methods 2.1. Deer age, genetic diversity, and diet The Whitehall Deer Research Facility contained 89 deer, ranging in age from >12 years old to fawns born in 2001 [18]. Many of the deer were born at the facility, but others were brought in from elsewhere (14 from South Carolina in 1995, 18 from North Carolina in 1997, and 16 from Stone Mountain, GA in 1997 [18]). In addition, one to three wild deer from Georgia are introduced into the facility each year. Therefore, the age and genetic basis of the penned deer are kept diverse and similar to that of wild deer. The penned deer were fed a standard diet (Big Buck Macho Feed [19]) comprising (%): ground corn (33.0), soybean hulls (23.4), alfalfa meal (20.0), soybean oil meal (13.9), wheat midds (5.8), dried molasses (2.0), animal and vegetable fat (1.0), limestone (0.2), and trace elements mix, vitamin mix, and salt (0.7). At the time of fecal sampling (mid-January), wild deer are likely eating any few remaining acorns from red or white oak (Quercus spp.) and leafy vegetation of evergreen vines, shrubs (e.g., greenbriar [Similax spp.], common privet [Ligustrum vulgare], yellow jessamine [Gelsemium sempervirens], and Japanese honeysuckle [Lonicera japonica] and other winter annual and perennial plants (e.g., henbit, [Lamium amplexicaule] and wild geranium [Geranium maculatum] [18]). 2.2. Selection and identification of E. coli isolates During a 2.92-cm rainfall event on January 19, 2001, 2.28 cm of rain (78%) fell between 1400 and 1600 h, and this was sufficient to cause visual signs of runoff throughout the watershed, including the penned area of the deer research facility. Duplicate water samples were aseptically collected at 1600 h from two locations in the watershed, one site 50 m upstream of the deer research facility and the other 100 m downstream. At the same time, freshly deposited fecal pellets from three different scat piles were obtained separately from bucks and does inside the pens as well as from wild deer outside the facility fences. Fecal pellets were aseptically collected with ethanol flame-sterilized forceps and placed in sterile Whirlpak bags. Water samples were processed within 6 h; fecal samples were processed within 24 h. In the case of the fecal pellets, three pellets from each scat pile were resuspended in 20 mL of 0.1% sterile peptone and were mixed outside of the bag by hand until the suspension was uniform. Each suspension was streaked on duplicate 5-cm Petri dishes containing mTEC agar (Difco Laboratories, Sparks, MD). In the case of the water samples, duplicate 10- and 100-mL water samples were filtered through separate sterile 0.45mm membranes and the membranes were transferred
P.G. Hartel et al. / Water Research 37 (2003) 3263–3268
aseptically to the mTEC agar plates. All plates were sealed inside triplicate Ziploc bags and were incubated submerged in a water bath at 44.570.2 C for 24 h according to Standard Methods for the Examination of Water and Wastewater [7]. A total of 300 randomly selected yellow isolates, consisting of 50 isolates each from the top and bottom of the watershed, 100 from penned deer, and 100 from wild deer were streaked onto tryptic soy agar (Difco). This total number was selected because it was consistent with the number of isolates obtained in other ribotyping studies (e.g., [3] 238 source isolates). Of the 100 penned deer isolates, 42 isolates were from a strictly buck pen, 38 isolates were from a strictly doe pen, and 20 isolates from a pen where bucks and does were rotated in and out. After incubation at 35 C for 24 h, the streaking was repeated twice on the same medium to ensure the purity of each isolate. Each isolate was inoculated into a 24-multiwell tissue culture plate containing separate 1-mL slants of Simmons citrate and urea agar (both Difco). Three bacterial species from the American Type Culture Collection (ATCC, Manassas, VA) were used as controls: E. coli ATCC #11775 (citrate negative, urea hydrolysis negative), Klebsiella pneumoniae ATCC #13883 (citrate positive, urea hydrolysis positive), and Enterobacter aerogenes ATCC #13048 (citrate positive, urea hydrolysis negative). Isolates that were citrate negative on Simmons citrate agar and urea hydrolysis negative on urea agar were subjected to an oxidase test. Isolates that were oxidase negative were considered E. coli. Of the 300 isolates, all but two (from the top of the watershed) were identified as E. coli. 2.3. DNA extraction, quantification, and ribotyping of E. coli isolates Each E. coli isolate was inoculated into Luria-Bertani broth contained in a test tube and incubated on a rotating shaker at 75 rpm at 35 C. After 18 h, a 2-mL sample of the culture was removed and the DNA extracted with a commercial kit (Qiagen DNeasy, Valencia, CA). The DNA was quantified with a fluorometer using standard DNA from E. coli strain B (Sigma Chemical Company, St. Louis, MO). Two 1-mg samples of DNA from each isolate and from the E. coli ATCC #11775 control were each separately digested overnight with the restriction enzymes EcoRI and PvuII according to the manufacturer’s directions (Roche Molecular Biochemicals, Indianapolis, IN). Loading dye (Sigma) was added to the digested DNA and the DNA was electrophoresed in a 1.0% agarose gel at 58 V for 3 h. Digoxigenin-labeled (DIGlabeled) Marker III (Roche) was the molecular weight marker and occupied every fifth lane of the gel. Control lanes contained no DNA (control) and DNA from E. coli ATCC #11775. DNA was transferred by Southern
3265
blotting to a nylon membrane with a vacuum blotting system (VacuGene, Amersham Pharmacia Biotech, Piscataway, NJ), and the DNA on the membrane was crosslinked with UV light. Following prehybridization at 42 C for 2 h, the membrane was hybridized at 42 C overnight to DIG-labeled cDNA from E. coli total rRNA (Sigma). Membranes were prepared for chemiluminescence by a series of washing steps before a chemiluminescent substrate for alkaline phosphatase (Roche) was added. Membranes were placed in an imager (FluorChem 8000, Alpha Innotech, San Leandro, CA) and images saved as a TIFF file. TIFF files were imported into GelCompar II (Applied Maths, Kortrijk, Belgium) for analysis. Typically, gels showed 9–11 bands for EcoRI and 11–13 bands for PvuII. DNA fragments o1375 base pairs were ignored because they were often indistinct. Lanes were normalized within the gel with the molecular weight marker and variations among the gels were assessed with the E. coli ATCC #11775 strain. Optimization (maximum percentage shift allowed between two different patterns for the patterns to still be considered a match) and tolerance (maximum percentage shift allowed between two bands on different patterns for the bands to still be considered a match) were each set at 1.00%. The normalized banding patterns for both enzymes were stacked with EcoRI on the top and PvuII on the bottom to create one combined ribotype pattern for each isolate; this was considered sufficient for good discrimination among ribotypes. Similarity indices were determined using Dice’s coincidence index and the distance among clusters calculated with the unweighted pair-group method using arithmetic averages (UPGMA). The banding pattern of the control E. coli ATCC #11775 strain varied from gel to gel and a similarity index of 95.0% was required for all the banding patterns to be considered the same ribotype. Based on variability of the inter-gel E. coli control, banding patterns of all other isolates had to have a similarity index of X95.0% to be considered the same ribotype. Five dendrograms were created: (a) penned bucks, (b) penned does, (c) all penned deer combined, (d) wild deer, and (e) all deer combined. When the data were combined, it was possible to determine the number of shared ribotypes. In addition, ribotype patterns from the environmental isolates were matched against ribotype patterns from penned and wild deer, and ribotype patterns of a host origin database (201 isolates total) from cattle, horses, humans, and swine living in the vicinity of Athens, GA. Because no assumptions can be made as to the distribution of the absolute difference in mean ribotype number between all penned deer combined (100 E. coli isolates) and wild deer only (100 E. coli isolates), the data were analyzed by an approximate randomization test [20]. Briefly, the ribotype patterns from each of the
3266
P.G. Hartel et al. / Water Research 37 (2003) 3263–3268
two groups were pooled and the patterns randomly selected without replacement to yield two data sets. Dendrograms were created from each data set and the number of ribotypes determined at X95% similarity. This process was repeated 20 times to construct an empirical distribution of the absolute difference in mean ribotype numbers. Repeating the process 20 times was the minimum necessary to yield a p value of 0.05. The null hypothesis was that numbers of ribotypes was the same between penned or wild deer. The approximate randomization test was repeated separately to determine the absolute difference in mean ribotype numbers between penned bucks only (42 E. coli isolates) and penned does only (38 E. coli isolates). The null hypothesis was that numbers of ribotypes was the same between bucks or does.
3. Results When the number of ribotypes was determined from separate dendrograms of penned bucks only and penned does only, both groups had six ribotypes. Under the conditions of the approximate randomization test, the null hypothesis was not rejected. Therefore, sex did not affect ribotype number. When the ribotype patterns of E. coli isolates from penned bucks and penned does were combined to determine the amount of sharing, 12 different ribotypes were observed (Fig. 1, Table 1a). Two of the 12 ribotypes (16.7%) were shared and these represented 24 of 80 isolates (30.0%). Bucks had the majority of the ribotypes (6 of 12) and almost a majority of the isolates (36 of 80 isolates, 45.0%). When the number of ribotypes was determined from separate dendrograms of penned deer only and wild deer only, the penned and wild deer had 11 and 35 ribotypes, respectively. Under the conditions of the approximate randomization test, the null hypothesis was rejected, and the difference between the number of ribotypes for two groups (n ¼ 24) was statistically significant (p ¼ 0:05). When the ribotype patterns of E. coli isolates from penned and wild deer were combined to determine the amount of sharing, 44 different ribotypes were observed (dendrogram not shown, Table 1b). Only six ribotypes (13.6%) were shared, but these six ribotypes represented almost a majority of the isolates (93 of 200 isolates, 46.5%). Of the remaining unshared ribotypes, 27 (61.4%) were found in wild deer and 11 (25.0%) were Fig. 1. Dendrogram of the ribotype patterns of 80 E. coli isolates from bucks and does from the University of Georgia Whitehall Deer Research Facility near Athens, GA. The similarity index is given on the top scale. Based on the variability of the inter-gel control, E. coli ATCC #11775, the cutoff to distinguish differences among ribotypes was 95% (dashed vertical line).
P.G. Hartel et al. / Water Research 37 (2003) 3263–3268 Table 1 Number and percent of shared and unshared Escherichia coli ribotypes and isolates between (a) penned bucks and does, and (b) penned and wild deer Ribotypes Number
Isolates %
(a) Penned bucks versus penned does Buck 6 50.0 Shared 2 16.7 Does 4 33.3 Total
12
100.0
(b) Penned deer versus wild deer Wild deer 27 61.4 Shared between 6 13.6 wild and penned deer Penned deer 11 25.0 Total
44
100.0
Number
%
36 24 20
45.0 30.0 25.0
80a
100.0
74 93
37.0 46.5
33
16.5
200
100.0
a
20 isolates were not included because they came from a pen where bucks and does were rotated in and out.
found in penned deer. Therefore, although 100 E. coli isolates were each obtained from penned and wild deer, penned deer had less than half the E. coli ribotypes of wild deer. When the 98 environmental isolates were matched against the 200 deer isolates, 19 of 48 (39.6%) and 23 of 50 (46.0%) isolates were identified as deer isolates from the top and bottom of the watershed, respectively. Of the 42 total matched isolates, 27 (64.3%) were from wild deer, 12 (28.6%) were from penned deer, and three (7.1%) were shared between wild and penned deer. When the 98 environmental isolates were matched against the host origin database from Athens, Ga., three isolates matched a horse ribotype and one matched a human ribotype, for total of four matches (4.1%).
3267
other captive animals fed a standard diet, then the database may not represent the ribotype diversity in the same wild animals. It is unlikely that geographic variability is a confounding factor here because the maximum home range for white-tailed bucks and does in suburban Georgia is only 644 and 334 ha, respectively [21]. There was no significant difference in the number of ribotypes between penned bucks and does, suggesting that sex differences were minimal. This result was expected because of the close proximity of the two populations in the pens, and suggests that host origin databases may not have to include isolates from both male and female animals. Nevertheless, the penned animals had significantly fewer ribotypes than the wild animals and it may be necessary to test other penned deer populations to confirm this. Of the 98 environmental E. coli isolates, 42 (42.9%) matched deer ribotypes. This matching was expected in the watershed because the large number of penned deer (n ¼ 89), the number of wild deer attracted to these penned deer, and the runoff conditions all guaranteed a large number of fecal bacteria from this source. Wild deer must have been attracted to the penned deer because almost as many environmental isolates matched deer isolates from above the deer pens (19 of 48) as below (23 of 50) the deer pens. When the environmental isolates were also tested against a host origin database, three isolates matched horses and one-matched humans. This matching was also expected because some ribotypes are shared among different animal species. The total percentage sharing here was 4.1%, consistent with the maximal 9.5% of BOX PCR DNA fingerprint sharing among cows, chickens, and swine [5]. Finally, the majority of the isolates (52 of 98, 53.1%) were unidentified. This high percentage of unknowns was also expected because E. coli isolates were not obtained from other wild animal species in the watershed. Ribotyping could identify deer as a source of fecal contamination, and given a sufficiently large host origin database, it may be possible to identify most of the remaining unidentified environmental isolates.
4. Discussion The data suggest that diet significantly affected the number of E. coli ribotypes in deer. It is unlikely that deer age or genetic diversity affected the number of E. coli ribotypes because the penned deer were managed to mimic wild deer populations. The idea that diet affects the number of E. coli ribotypes is consistent with reports that diet changes numbers of enterohemorrhagic E. coli O157:H7 in cattle (e.g., [16]) and specific E. coli subspecies in certain Australian mammals [14]. In our study, the penned deer had fewer than half of the ribotypes of wild deer. If researchers construct a host origin database of E. coli isolates obtained from zoo or
5. Conclusions Diet significantly affected the ribotype diversity of E. coli in deer. Penned deer, which were fed a standard diet, had fewer than half the E. coli ribotypes of wild deer. This result suggests that isolates for bacterial source tracking databases should come from wild rather than from captive animals. During a runoff event, deer contributed 42.9% of the fecal contamination (as E. coli) to the stream in the watershed. As host origin databases for bacterial source tracking become more comprehensive both with regards to number of isolates
3268
P.G. Hartel et al. / Water Research 37 (2003) 3263–3268
as well as the animal species tested, it may be possible to identify an even greater percentage of isolates, even to the point of identifying specific animal operations. If it can be shown in certain instances that the vast majority of isolates come from wildlife, then it may be unnecessary to monitor the watersheds because control over wildlife is difficult.
[8]
[9]
Acknowledgements [10]
We thank Bruce Beck, Demetrius Cox, Dan Hall, Robin Kuntz, David Osborn, Todd Rasmussen, Ken Roberts, and Karen Rodgers for their advice and assistance. This research was partially supported by grants from the USDA–CSREES Special Grants Program and the USDA–CSREES through the Texas Agricultural Extension Service.
[11]
References
[13]
[1] Amor K, Heinrichs DE, Frirdich E, Ziebell K, Johnson RP, Whitfield C. Distribution of core oligosaccharide types in lipopolysaccharides from E. coli. Infect Immun 2000;68(3):1116–24. [2] Wiggins BA, Andrews RW, Conway RA, Corr CL, Dobratz EJ, Dougherty DP, Eppard JR, Knupp SR, Limjoco MC, Mettenburg JM, Rinehardt JM, Sonsino J, Torrijos RL, Zimmerman ME. Use of antibiotic resistance analysis to identify nonpoint sources of fecal pollution. Appl Environ Microbiol 1999;65(11):3483–6. [3] Parveen S, Portier KM, Robinson K, Edmiston L, Tamplin ML. Discriminant analysis of ribotype profiles of E. coli for differentiating human and nonhuman sources of fecal pollution. Appl Environ Microbiol 1999;65(7): 3142–7. [4] Parveen S, Hodge NC, Stall RE, Farrah SR, Tamplin ML. Phenotypic and genotypic characterization of human and nonhuman E. coli. Water Res 2001;35(2):379–86. [5] Dombek PE, Johnson L-AK, Zimmerley ST, Sadowsky MJ. Use of repetitive DNA sequences and the PCR to differentiate E. coli isolates from human and animal sources. Appl Environ Microbiol 2000;66(6):2572–7. [6] Farber JM. An introduction to the hows and whys of molecular typing. J Food Prot 1996;59(10):1091–101. [7] Clesceri LS, Greenberg AE, Eaton AD. Standard methods for the examination of water and wastewater, 20th ed.
[12]
[14]
[15]
[16]
[17]
[18]
[19]
[20] [21]
Washington, DC: American Public Health Association, American Water Works Association, and Water Environment Federation, 1998. Samadpour M, Chechowitz N. Little Soos Creek microbial source tracking. Report to Surface Water Management Division, King County Department of Public Works, Seattle, WA, 1995. Berghoff K. Beach sediment bacterial contamination and microbial source tracking study. 1997 Year End Summary Report, Glen Canyon National Recreational Area, National Park Service, UT, 1998. Farag A, Goldstein JN, Woodward DF, Samadpour M. Water quality in three creeks in the backcountry of Grand Teton National Park, USA. J Freshwater Ecol 2001;16(1): 135–43. Tippets N. Backcountry water quality testing in Grand Teton National Park–1998 Summer Season. Report from the Environmental and Contaminants Research Center, USGS, 1999. Jenkins MB, Hartel PG, Olexa TJ, Stuedemann JA. Putative temporal variability of E. coli ribotypes from yearling steers. J Environ Qual 2003;32(1):305–9. Hartel PG, Summer JD, Hill JL, Collins JV, Entry JA, Segars WI. Geographic variability of E. coli ribotypes from animals in Idaho and Georgia. J Environ Qual 2002;31(4):1273–8. Pupo GM, Lan R, Reeves PR, Baverstock RB. Population genetics of E. coli in a natural population of native Australian rats. Environ Microbiol 2000;2(6):594–610. Diez–Gonzalez F, Callaway TR, Kizoulis MG, Russell JB. Grain feeding and the dissemination of acid-resistant E. coli from cattle. Science 1998;281(5383):1666–8. Cray Jr WC, Casey TA, Bosworth BT, Rasmussen MA. Effect of dietary stress on fecal shedding of E. coli O157: H7 in calves. Appl Environ Microbiol 1998;64(5):1975–9. Fischer JR, Zhao T, Doyle MP, Goldberg MR, Brown CA, Sewell CT, Kavanaugh DM, Bauman CD. Experimental and field studies of E. coli O157: H7 in white-tailed deer. Appl Environ Microbiol 2001;67(3):1218–24. Osborn D. Personal communication. School of Forest Resources, 3–324 Forest Resources, University of Georgia, Athens, GA 30602-2152, 2001. Roberts KR. Personal communication. Poultry Science Research College, South Milledge Ave., University of Georgia, Athens, GA 30602-2772, 2001. Noreen EW. Computer intensive methods for testing hypotheses: an introduction. New York: Wiley; 1989. Rogers CL. Utilization of cedar glades by white-tailed deer at Chickamauga Battlefield Park. MS thesis, University of Georgia, Athens, 1996.