Feral swine as sources of fecal contamination in recreational waters

Recreational waters are primary attractions at many national and state parks where feral swine populations are established, and thus are possible hotspots for visitor exposure to feral swine contaminants. Microbial source tracking (MST) was used to determine spatial and temporal patterns of fecal contamination in Congaree National Park (CONG) in South Carolina, U.S.A., which has an established population of feral swine and is a popular destination for water-based recreation. Water samples were collected between December 2017 and June 2019 from 18 surface water sites distributed throughout CONG. Host specific MST markers included human (HF183), swine (Pig2Bac), ruminant (Rum2Bac), cow (CowM3), chicken (CL), and a marker for shiga toxin producing Escherichia coli (STEC; stx2). Water samples were also screened for culturable Escherichia coli (E. coli) as part of a citizen science program. Neither the cow nor chicken MST markers were detected during the study. The human marker was predominantly detected at boundary sites or could be attributed to upstream sources. However, several detections within CONG without concurrent detections at upstream external sites suggested occasional internal contamination from humans. The swine marker was the most frequently detected of all MST markers, and was present at sites located both internal and external to the Park. Swine MST marker concentrations ≥ 43 gene copies/mL were associated with culturable E. coli concentrations greater than the U.S. Environmental Protection Agency beach action value for recreational waters. None of the MST markers showed a strong association with detection of the pathogenic marker (stx2). Limited information about the health risk from exposure to fecal contamination from non-human sources hampers interpretation of the human health implications.

www.nature.com/scientificreports/ animals; (2) constitute a relatively significant portion of the bacteria present in mammalian feces; (3) can be highly specific to particular species or taxonomic groups; and (4) are strict anaerobes. In warm waters their DNA may degrade beyond detection within 1 to 3 days 35,36 , which makes them indicative of recent contamination. Taken together, MST results can help identify mitigation targets to include in TMDLs.
In light of the challenges and opportunities at CONG, the objectives of this study were as follows: (1) measure the occurrence and concentrations of MST markers throughout CONG; (2) assess internal and external sources of fecal contamination; (3) assess the frequencies and distributions of detection of a pathogenic marker; and (4) investigate associations between MST marker concentrations and pathogenic marker presence.

Methods
Study site. Congaree National Park (CONG) in South Carolina is located in a broad floodplain between the Congaree and Wateree Rivers, and it includes over 11,000 acres (44.52 km 2 ) of old-growth forest and over 15,000 acres (60.70 km 2 ) of designated federal wilderness 37 . The Congaree River, which forms the southern park boundary ( Fig. 1), drains approximately 8,290 mi 2 (21,476 km 2 ) and has varied land use throughout the watershed. Multiple major municipal wastewater systems in the Congaree River basin are within 25 miles upstream from the Park boundary. The state capitol, Columbia, has a population of over 800,000 people in the metropolitan area and is located approximately 20 mi (32.19 km) upstream of the Park. The Wateree River forms a short stretch of the park's eastern boundary (Fig. 1) and drains over 5,600 mi 2 (14,500 km 2 ). Though not technically part of the park during normal in-bank streamflow conditions, the Congaree and Wateree Rivers (Fig. 1) are additional popular destinations for recreation 37 .
During normal flows on the Congaree and Wateree Rivers, the ecohydrology of the floodplain is dominated by several tributary systems including Cedar Creek, Tom's Creek, and Dry Branch. Cedar Creek (Fig. 1) is a particularly popular destination for canoeing, kayaking, and fishing. The lower portion of Cedar Creek that flows through the park (between Wise Lake (Site 5) and the Congaree River, Fig. 1, Table 1) is the only waterbody in South Carolina with the designation of Outstanding Natural Resource Water 38 , a state-level designation authorized under the Clean Water Act to recognize water bodies with high ecological importance and exceptional water quality. Land use in the Cedar Creek watershed upstream of the Park is predominantly agricultural (including silviculture, poultry, horses, and farming), with some light industry and rural housing.
Sample collection and processing. One-liter water samples were collected and analyzed from 18 surface-water sites distributed throughout CONG (Fig. 1, Table 1). The sampling strategy was designed to determine if contamination was coming from inside or outside the Park by capturing primary inflows (Congaree River, Cedar Creek, Dry Branch, Myers Creek, Tom's Creek, Wateree River) at respective upstream and downstream Park boundaries as well as to characterize a cross section of the principal hydrologic features (tributaries, wetland, oxbow lakes; Fig. 1) found within CONG. Eleven sample collection events were conducted between December 2017 and June 2019 to investigate temporal and spatial variability in sources of fecal contamination. Logistical constraints existed that included funding limitations, remote site access, and flooding thus resulting in different frequencies of sampling across study sites. The number of sites sampled during each collection event www.nature.com/scientificreports/ varied from 4 to 16. Five sites (Sites 3, 4, 6, 9, and 10; Fig. 1) were targeted for repeated sampling (sampled during each sample collection event as conditions allowed) to address temporal variability in sources; the remaining 13 sites were sampled intermittently (one to four times; Table 1) to inform spatial variability in sources. Grab samples for molecular analysis were stored in coolers on ice until they were transferred to a refrigerator at the laboratory until processing, which occurred within 48 h of sample collection. Samples were filtered on a vacuum manifold through a 47-mm polycarbonate filter with 0.4-µm pore size (Millipore, Bedford, MA, USA). Samples were filtered in 100-mL aliquots or to the point of the filter clogging (minimum volume 20-mL 39 ). Six filter replicates were collected per sample that included two filters that were enriched for E. coli with modified mTEC agar following the incubation protocol in EPA Method 1603 40 to enhance detection of shiga-toxin producing E. coli STEC 41,42 and four unenriched filters. Two unenriched filters and one enriched filter were processed for molecular analysis. The remaining filter replicates were retained as backups for use if needed. All filter replicates were stored at − 80 °C until DNA extraction. Sterile de-ionized water (sterile DI; autoclaved 15 PSI, 121 °C for 15 min per 1-L of sterile DI 43 ) controls were filtered in conjunction with each sample event to determine presence of potential contamination during sample collection and/or laboratory processing and were treated identically to filtered samples in downstream processing.
Citizen science Escherichia coli analysis. Beginning in October 2018, select duplicate water samples were collected for E. coli analysis at the CONG Old-Growth Bottomland Forest Research and Education Center as part of a Citizen Science methods development program by CONG personnel as described in McCarthy 24 . Samples were processed within 6 h of collection and kept on wet ice during the time between collection and processing. Each sample was screened at three dilutions to ensure accurate and quantifiable results. Samples were tested for E. coli concentrations and enumerated using the most probable number (MPN) method with IDEXX Colilert-18 (IDEXX Laboratories, Inc. Westbrook, ME) for enumerating E. coli. A laboratory negative control (sterile DI) was analyzed alongside samples for each sample event to test for contamination in reagents or contamination potentially introduced during sample processing. Resulting data were used in a post-hoc analysis to investigate relationships between E. coli MPN and MST marker concentrations.

DNA extraction and quantitative PCR. DNA was extracted using the Qiagen DNeasy PowerLyzer
PowerSoil DNA Isolation Kit (Qiagen; Germantown, MD) 44 following manufacturer's instructions. We did not explicitly test the efficiency of the DNA extractions; however, this isolation kit has been shown to efficiently extract Bacteroides and E. coli DNA from stool samples 45 . Samples were tested for PCR inhibition following Bradshaw, et al. 46 using the Sketa22 assay 47 . Samples were screened with probe-based quantitative PCR (qPCR) MST markers for humans (HF183 48 , Table 2), ruminants (Rum2Bac 49 , Table 2), chickens (CL 50 , Table 2), and swine (Pig2Bac 49 , Table 2). Samples that tested positive for ruminants were also screened for cows (CowM3 51 , Table 2). PCR efficiency ranged from 82 to 102% 39 . Enriched samples were screened for stx2 52 (Table 2). Pig-2Bac has been found to be highly sensitive, but the MST marker has been found to occur in dog feces, human feces, and septage 32 . To test for possible cross amplification of Pig2Bac with human contamination, we calculated Spearman's ρ for the correlation between concentrations of HF183 and Pig2Bac. A significant correlation between Pig2Bac and HF183 could indicate that Pig2Bac concentrations included detection of human contamination.
DNA from two filter replicates per sample were screened for each marker with two qPCR replicates for a total of four qPCR replicates per sample per marker. Reactions were carried out in 96-well qPCR plates in 20-µL reaction volumes. Final concentrations of reagents in the assays were 1-µM forward and reverse primers, 80-nM 6-carboxy-fluorescein FAM-labeled TaqMan probe, 0.02-mg/mL BSA (Life Technologies), 1 × DNA TaqMan Fast Universal PCR Master Mix (Life Technologies), and 4-µL of template (i.e., genomic DNA, nanopore water as a no template control, or MST marker standard concentrations of 10 to 10 6 copies). Synthetic gene fragments with the marker insert sequences were used to create the standard curves for all MST markers and as a qPCR positive template control for stx2. Microbial source tracking markers with low specificity can reduce confidence that a detection is indicative of contamination from the targeted host species. To reduce the likelihood of spurious detection of an MST marker from a non-target host species, samples were considered absent for a marker when amplification occurred in zero or one of the four qPCR replicates. Samples absent for an MST marker were assigned an estimated concentration of 0 gene copies. The stx2 assay was analyzed as presence/absence where amplification in one or both qPCR replicates was considered evidence of presence. With the exception of the standard curve qPCR replicates for MST markers and positive template control reactions for the stx2 qPCR assay, no procedural positive controls were included during sample processing. Therefore, the influence of variable enrichment efficiencies for E. coli or DNA extraction efficiencies among samples on MST marker concentration or stx2 presence/absence results cannot be ruled out. Across all MST markers, R 2 values for standard curves ranged from 0.987 to 0.999, PCR amplification efficiencies ranged from 82 to 102% for all markers, and slopes ranged from -3.85 to -3.27. The limit of quantification (LOQ) for each MST marker was determined as the lowest standard curve copy number with a coefficient of variation (CV = 100*σ/µ; where σ is the standard deviation, and µ is the mean number of estimated gene copies) less than 35%. Back calculated gene copy estimates were based on the equation X 0 = E AMP (b-C T ) where X 0 is the initial number of target copies in the qPCR; E AMP is the exponential amplification value, which is 1 + the amplification efficiency (e.g., if the amplification efficiency is 0.94, E AMP = 1.94); b is the y-intercept of the standard curve; and C T is the cycle number when the amplification curve crosses the threshold line that distinguishes fluorescence intensity of a reaction from background levels.

Statistical analysis.
A statistical technique similar to McKee, et al. 33 was used to account for error in qPCR estimates below LOQ. If a qPCR estimated gene copy number was greater than 0 and below LOQ, a random www.nature.com/scientificreports/ number was selected from a normal distribution centered at the qPCR estimated copy number. Randomly selected negative values were set to 0 and randomly selected values above LOQ were set to the LOQ. To determine the standard distribution of the error estimates for each marker, a trendline was created with the standard deviations of the back-calculated gene copy number estimates for the standard curve gene copy numbers at and below LOQ (Table A.1). An additional data point at the origin (0, 0) served as a value for the no template controls. The equation for that trendline was used to calculate the standard deviation for each randomly selected number, which was a function of the qPCR estimated gene copy number. Randomly selected numbers were treated the same as qPCR measured concentrations above the LOQ for statistical analysis. All analyses were conducted in JMP v 14.2.0 (SAS Institute Inc.). Recursive partitioning was conducted to determine the MST marker and concentration that best predicted stx2 detection and E. coli density splits. Recursive partitioning is a nonparametric data-mining analysis that splits the response variable into two categories based on an explanatory variable cutting value. The cutting value is determined with a partition algorithm that finds the explanatory variable split that maximizes the difference in the response frequencies between the two nodes of the split. Spearman's ρ were used to assess correlative relationships between MST marker concentrations and E. coli MPN.
To account for multiple comparisons of the same data, we used p < 0.01 to indicate statistical significance for all Spearman's rank correlations.

Spatial and temporal variability in sources throughout the Park and at Cedar Creek. HF183
and Pig2Bac were detected broadly across the study sites (Figs. 2, 3 and 4), with detections at sites inside, outside, and on the boundary of CONG (Figs. 3, 4). The highest concentrations of HF183 detected were on the Congaree River boundary waters (Site 1, Congaree R @ gage: 26 copies/mL; and Site 15, Congaree R @ SC601: 12 copies/mL, Fig. 2 Fig. 5). The stx2 marker was detected at boundary water sites, as well as sites within and upstream of CONG (Fig. 6). Of the internal sites, stx2 was detected most frequently in samples from Cedar Cr @ Kingsnake (Site 10, five of 11 samples, Fig. 6). Temporally, Pig2Bac was the most consistent marker to be detected, because it was present in one or more samples for 10 of 11 sample events (Fig. 4). HF183 was detected at one or more sites for 6 of 11 sample events www.nature.com/scientificreports/     www.nature.com/scientificreports/ dictor of stx2 detection across all samples was Pig2Bac concentrations ≥ 43 copies/mL (R 2 = 0.135). The Pig2Bac split correctly identified 5 of 15 samples with stx2 detections. Samples with Pig2Bac concentrations ≥ 43 copies/ mL were associated with an average 51% greater probability of stx2 detection compared to only a 14% probability of stx2 detection in samples with Pig2Bac concentrations < 43 copies/mL. All samples with stx2 detections that were positive for HF183 were also positive for Pig2Bac, whereas seven samples with stx2 detections were positive for Pig2Bac but not HF183. Three of the four samples with stx2 detections that were not positive for either HF183 or Pig2Bac were from Myers Creek. stx2 was only detected in two samples for which Rum2Bac was also detected. Posthoc analysis with culturable E. coli data from citizen science collected samples (Table A.2) also indicated that Pig2Bac concentrations ≥ 43 copies/mL were the best recursive partitioning split for E. coli MPN (R 2 = 0.331, root mean square error = 137.6, AICc = 324.3). When Pig2Bac concentrations were ≥ 43 copies/mL, E. coli concentrations had a mean of 357 MPN (± 195 standard deviation). In contrast, mean E. coli concentration for samples with Pig2Bac concentrations < 43 copies/mL was 115 MPN (± 130), indicating an average increase in E. coli MPN of 242 in samples containing ≥ 43 copies/mL of Pig2Bac. No significant correlations were detected between MST markers and E. coli MPN (Table 3).

Discussion
Results of this study suggest that humans are an intermittent source of bacterial contamination and swine are primary contributors of fecal contamination in CONG. Infrequent detection of the ruminant marker and no detections of the cow or poultry marker suggest that wildlife, cattle, and poultry operations were not major contributors of contamination in the Park at the time of sample collection.
Internal versus external sources of contamination. Human MST marker data are consistent with upstream inputs at park boundaries as well as some backcountry locations, but show no evidence that the park's primary visitor center septic system is a major source of fecal contamination in recreational waters within CONG. Feral swine markers, by contrast, are consistent with inputs from feral swine internal to CONG. Possible sources of contamination from outside the Park include failing septic systems and leaking wastewater conveyances. Sources of human contamination from within the park could include on-site septic systems as well as local backcountry contributions from hikers and campers. HF183 was detected at all four Cedar Creek sites, however, concentrations were never at levels above 10 copies/mL and all were below the LOQ. The highest concentrations of the human marker were observed from Congaree River samples, consistent with external municipal or residential wastewater conveyances as a potential source. Human contamination was not detected at the Muck Swamp site, located adjacent to the Park's Visitor Center septic system, suggesting that the on-site septic system was likely not a major net source of bacterial contamination during this study. It is possible that lack of septic contamination here could be related to its filtering through organic rich soils in Muck Swamp, as these types of tannin-rich settings have been noted to have natural anti-bacterial properties 54 . Another previously suggested possible source of human waste contamination within the Park is direct local inputs of contamination associated with visitation in the backcountry. Contaminants indicative of anthropogenic contamination have been found in water samples from non-stream sites at CONG, including samples from Muck Swamp, Wise Lake, and Weston Lake, under non-flood conditions (i.e., in the absence of a surface-water connection to extra-Park contaminant sources on the Congaree and Wateree Rivers or tributaries) 55 . In contrast, under flood conditions, interior Park locations can be hydrologically connected to the Congaree and Wateree Rivers and the associated external human wastewater sources. Two U.S. Geological Survey real-time gaging stations on the Congaree River (Congaree R @ Gage, site 1; Fig. 1) and Cedar Creek (Cedar Cr @ Gage, site 6; Fig. 1) are used to inform the public when the Park is under flood conditions based on gage height (15 feet or greater at Congaree R @ Gage, 8 feet or greater for Cedar Cr @ Gage; www. nps. gov/ cong/ plany ourvi sit/ condi tions. htm, accessed Feb. 28, 2020). Data from April 1, 2019 (waterdata.usgs.gov/, accessed Feb. 28, 2020) indicate CONG was not under flood conditions when the human MST marker was detected at Stump Gut, Horseshoe Lake, and Cedar Creek @ Gage suggesting contamination may have come from internal inputs. Other detections of human contamination at the Cedar Cr. @ Gage site were concurrent with upstream detections at the Cedar Creek upstream Park boundary sites at Cedar Cr. @ Myers and Myers Cr., consistent with downstream transport from external sources during nonflood conditions. All other detections of human contamination were at boundary water sites, or sites proximal to the boundary and downstream from potential external sources.
Feral swine are well documented throughout the southeastern U.S. and particularly in the study region of the Coastal Plain in South Carolina 5 . The lack of hog farms within the Congaree River watershed indicates feral swine as the main source of the swine MST marker in CONG. Further, while a comparison of swine marker loads (concentration per unit time) would be necessary for verification, higher swine MST marker concentrations at Cedar Creek sites within the Park compared to sites at the upstream boundary of the Park suggests swine contamination is predominantly, but not exclusively, occurring within CONG.
The home ranges of feral swine follow similar patterns to other mammals and tend to be larger for males than females. Feral swine home ranges can vary annually and seasonally and are generally smaller when food sources are more abundant 56,57 . Previous studies on feral swine home ranges in South Carolina have indicated that in coastal marshes, the average home range sizes were 2.26 km 2 for males and 1.81 km 2 for females 58 and 3.89 km 2 on average for males and females in the Piedmont 59 . In-house data from Park files and partner research suggest that local swine can move up to four miles per day and swim the Congaree River multiple times per day, indicating that detections of the swine MST marker external to CONG could be from feral swine the home ranges of which include the Park. www.nature.com/scientificreports/ There are a couple limitations to the results from this study. One limitation is that the methods did not test for markers from the American alligator (Alligator mississipianensis). Research suggests that crocodilian feces are exceptionally rich in FIB 60 , although limited research has been conducted on pathogenic E. coli in reptiles 61 . Research at the Park suggests that the local alligator populations are low density and transient, but they could be disproportionately represented in some samples. Most alligators, however, are concentrated along the Congaree and Wateree Rivers at the southeast end of the park and well away from the Cedar Creek locations central to this analysis. Additionally, analysis for specific markers for turkey and other avian sources were not conducted, and these animals can represent a significant portion of the non-swine and non-deer megafauna in the Park. A second limitation to the study is that we cannot rule out non-specific amplification of the swine marker as a factor in the frequency of detection and estimated concentrations of the swine marker over the course of the study. We did not detect a correlation between the human and swine marker concentrations, which we interpreted as evidence that non-specific amplification from the swine marker was not occurring when human fecal contamination was present. However, the lack of correlation may have been caused by differences in stability of the signal between the human and swine markers due to factors such as differences in decay rates among markers from different sources 62 . Differences in environmental persistence among MST markers and markers from different sources may also explain why we did not detect a significant association between any of the MST markers and the pathogenic E. coli marker.
Human health risk from exposure to water with fecal contamination. We did not find evidence of a strong association between sources of fecal contamination and presence of STEC. Possible methodological explanations for the lack of a strong association between stx2 presence and any of the MST markers include variable efficiencies in bacterial recovery from environmental samples, different DNA extraction efficiencies among samples, and suboptimal incubation conditions for STEC. Unlike most other fecal coliforms, which are characterized by their ability to grow at 44.5-45.5 °C, studies have suggested that E. coli O157:H7, the most common serotype of STEC to cause infection 63,64 , does not grow well at temperatures above 41 °C 64,65 . However, Duris, et al. 42 detected STEC in more than 50% of river water samples from Michigan and Indiana using a similar incubation protocol as used in the study herein. Results from our study and Duris, et al. 42 indicate that incubating enriched samples at 44.5 °C does not inherently prevent detection of STEC, although presence may be underestimated.
Recursive partitioning of E. coli indicated that swine MST marker levels above 43 copies/mL were associated with E. coli levels above the EPA beach action value (BAV) for recreational freshwaters of 235 colony forming units/100 mL (CFU/100 mL; conversion between CFU to MPN is 1:1). This suggests that when Pig2Bac levels are above a certain threshold, FIB levels are likely to exceed BAV levels. The E. coli BAV for recreational freshwaters is based on investigations of the frequency of gastro-intestinal (GI) illness in swimmers exposed to water downstream of a sewage treatment facility and thus primarily exposed to human contamination 14 . Research suggests that, in general, exposure to human fecal contamination may be a greater human-health risk than exposure to fecal contamination from non-human sources [66][67][68] . This is likely due in part to the host specificity of many viruses 69 that are the etiologic agents most frequently responsible for human illness associated with exposure to recreational waters 70 . Therefore, the health risk from exposure to recreational freshwater contaminated by non-human sources may not be accurately represented by the BAV. Additionally, the BAV was developed to indicate the likelihood of illness from recreational exposure to contaminated waters. Feral swine zoonoses caused by exposure to contaminated freshwater may present increased variability in the severity of illness that is not represented by the BAV.
Studies suggest that exposure to fecal contamination from domestic pigs presents a lower health risk than exposure to human fecal contamination 66,71 . Health risks from exposure to recreational waters with FIB densities of 35 CFU/100 mL enterococci and 126 CFU/100 mL E. coli from domestic pig fecal contamination were estimated to be substantially lower than the risk from exposure to human sources, with the median risk from domestic pigs two orders of magnitude lower than the human-based benchmarks 66 . Based on these findings, a new water-quality benchmark for domestic pigs should not be ruled out 66 .
If health risks from exposure to domestic pig and feral swine fecal contamination are similar, the frequency of detection of the feral swine MST marker in CONG relative to the human MST marker may indicate the health risk is lower than would be predicted if humans were the predominant source of contamination. However, illness rates from exposure to feral swine fecal contamination may differ from illness rates due to exposure to domestic pig contamination. Feral swine and domestic pigs have been found to serve as reservoirs to different enteropathogenic bacterial strains 72 , suggesting that health risks from exposure to domestic pig fecal contamination may not accurately represent the risk from exposure to feral swine. Feral swine are known to be hosts for numerous zoonotic pathogens, including Mycobacterium avium complex 73 , which can cause respiratory illness and be transmitted through recreational waters 74 . The application of quantitative microbial risk assessments 75 may be useful for predicting the likelihood of illness from zoonotic pathogens in recreational waters with feral swine fecal contamination. Further research is needed to determine the human health risks from exposure to feral swine contamination and to identify other locations where human exposure to feral swine fecal contamination is likely to occur from contact with contaminated recreational waters.
Regulatory implications of wildlife contamination in impaired recreational waters. Study data suggesting significant wildlife inputs (relative to humans) have potential regulatory implications. Current (2020) EPA Recreational Water Quality Criteria (RWQC) 14 acknowledge the potential for differences in human health risks from exposure to recreational waters contaminated by human versus non-human sources but do not provide water-quality standard recommendations that take sources of fecal contamination into account. Instead, www.nature.com/scientificreports/ the 2012 RWQC describes methods that can be used to develop site-specific standards for states that desire to address the variability in human health risks associated with different sources of fecal contamination. In Florida, MST was used to assess sources of fecal contamination in a wildlife conservation-managed watershed that was included on the 2010 Florida Department of Environmental Protection 303(d) list of impaired waters for FIB 76 .
Results from the MST analysis suggested that birds, not humans, were the main source of contamination 76 . A natural source exclusion status was obtained for the water body, which was subsequently removed from the Florida 303(d) list. At least 13 other states have natural resource-or "natural conditions"-exclusions in their administrative code, although the conditions for these exceptions and how they are applied can vary by state 77 . Intermittent detection of human contamination in South Carolina 303(d) listed impaired recreational waters in CONG may preclude these sites from a natural source exclusion. Further research would be necessary to determine if a natural resource exclusion would be appropriate or possible for the impaired streams in CONG.

Data availability
Quantitative PCR data for this study are available in the associated USGS data release 39 in the USGS ScienceBase-Catalog, which is the authoritative repository of these data.