Diversity and Persistence of Salmonella enterica Strains in Rural Landscapes in the Southeastern United States

Salmonellosis cases in the in the United States show distinct geographical trends, with the southeast reporting among the highest rates of illness. In the state of Georgia, USA, non-outbreak associated salmonellosis is especially high in the southern low-lying coastal plain. Here we examined the distribution of Salmonella enterica in environmental waters and associated wildlife in two distinct watersheds, one in the Atlantic Coastal Plain (a high case rate rural area) physiographic province and one in the Piedmont (a lower case rate rural area). Salmonella were isolated from the two regions and compared for serovar and strain diversity, as well as distribution, between the two study areas, using both a retrospective and prospective design. Thirty-seven unique serovars and 204 unique strain types were identified by pulsed-field gel electrophoresis (PFGE). Salmonella serovars Braenderup, Give, Hartford, and Muenchen were dominant in both watersheds. Two serovars, specifically S. Muenchen and S. Rubislaw, were consistently isolated from both systems, including water and small mammals. Conversely, 24 serovars tended to be site-specific (64.8%, n = 37). Compared to the other Salmonella serovars isolated from these sites, S. Muenchen and S. Rubislaw exhibited significant genetic diversity. Among a subset of PFGE patterns, approximately half of the environmental strain types matched entries in the USA PulseNet database of human cases. Ninety percent of S. Muenchen strains from the Little River basin (the high case rate area) matched PFGE entries in PulseNet compared to 33.33% of S. Muenchen strains from the North Oconee River region (the lower case rate area). Underlying the diversity and turnover of Salmonella strains observed for these two watersheds is the persistence of specific Salmonella serovars and strain types that may be adapted to these watersheds and landscapes.


Introduction
Salmonella enterica are zoonotic bacteria associated with a wide range of animals, including humans, where they are a significant cause of enteric disease and often attributed to foodborne transmission [1]. The incidence of salmonellosis has decreased only slightly in the past 26 years (19 cases per 100,000 in 1987 versus 15.2 cases per 100,000 in 2013) [2,3]. While the implementation of HACCP (Hazard Analysis and Critical Control Points) for the food industry in the U.S. has reduced contamination of meats, milk, and eggs with foodborne pathogens [4], there has been increased recognition of Salmonella outbreaks associated with fresh produce [5][6][7][8][9][10][11][12][13][14][15]. The cultivation and processing of fruits and vegetables is intimately linked to the environment, where ample opportunities exist to introduce pathogens, through direct contamination from animals in crops and fields and the use of irrigation water that may be contaminated [5,6,8,9,16]. Although not widely linked to outbreaks, water itself is an important vehicle, especially in sporadic (non-outbreak associated) cases of gastrointestinal illnesses, including salmonellosis [17][18][19]. Sporadic cases are increasingly associated with a high burden of illnesses and are often not attributable to a particular source, such as food [20]. As a zoonotic agent, animals -livestock as well as wildlife-can be important contributors to the abundance and distribution of Salmonella in the environment and possible transmission to humans [21][22][23][24][25][26]. Salmonella enterica is intimately associated with the landscape, and its components (living and non-living). Efforts to examine the environmental ecology of this agent are needed to better understand how it may be controlled, especially in regards to non-foodborne and non-outbreak cases.
The notion that environmental parameters affect both the incidence and distribution of salmonellosis cases is illustrated in part by large regional differences in the rates of reported human cases in the U.S. Variations among states in reported cases are not always explained by differences in surveillance, demographics, patterns in food preparation, or food-distribution networks, suggesting that environmental and ecological factors could affect its relative distribution (e.g., biogeographical patterns [27]). Georgia remains among the states (all in the southeast) with the highest annual prevalence, at 24 cases per 100,000 [28], compared to the national average of roughly 15 cases per 100,000 [2] in the Foodborne Diseases Active Surveillance Network (FoodNet). The prevalence in the southern portion of the state, primarily in the Coastal Plain physiographic province, is markedly higher than that in the northern part of the state's Piedmont province. In 2011, there were 70.1 cases per 100,000 people in Georgia Public Health District 8-1 and 28.8 cases per 100,000 in Public Health District 10 in 2011, representing the south (Coastal Plain) and north (Piedmont), respectively (GA Dept. of Public Health; dph. georgia.gov). Regional patterns in the epidemiology of salmonellosis are reflected not only by significant differences in reported infection rates, but also in the distribution of specific Salmonella serovars between physiographic provinces and individual counties [28].
Salmonella is ubiquitous in fresh and marine environmental surface waters [2,29,30], but contamination may come from many different routes such as effluent from wastewater treatment plants, contaminated runoff from urban or agricultural areas, overburdened septic systems, or local and migratory fauna [30][31][32][33][34]. Contamination of environmental waters with Salmonella may be of a greater public health concern than previously thought due to the ability of it to persist and, in some cases, grow outside of a host organism [29]. This characteristic increases the probability of survival between hosts [35]. The environment, including surface waters, can be considered as a part of the lifecycle of Salmonella, and therefore influences the biogeographical patterns of these pathogens.
Previous studies in the Atlantic Coastal Plain (south Georgia and north Florida) indicate that Salmonella are commonly detected in the environment, i.e., streams and ponds [36][37][38][39].
Frequency of detection ranges from 29% to 96% of samples among these studies, with concentrations reaching 5,400 MPN L -1 in the southern reaches of the Upper Suwanee Watershed, which spans southern Georgia and north Florida [36,39]. At the uppermost reaches of the Upper Suwanee Watershed in south Georgia, the Little River is typical of the heavily vegetated, slow-moving stream systems in this region. We have previously shown that 79% of sites along the Little River were positive for Salmonella [37] with levels ranging from 2.5 MPN L -1 to 36.3 MPN L -1 . Salmonella densities are positively impacted by precipitation and temperature (e.g., summer season) when serovars associated with human cases are also more likely to be detected [38]. Here we expanded on this work and examined the distribution and diversity of Salmonella serovars and strains across geographic space, time, and between water and wildlife reservoirs, which may affect exposure routes and transmission to humans.

Description of study sites
The sample areas were located in two distinct physiographic regions of Georgia, USA-the low lying Coastal Plain and the higher elevation Piedmont physiographic province (Fig 1). The regions were also distinguished by prevalence of salmonellosis. Otherwise the selected areas were similar in watershed size, land use, population, and median incomes.
The 334 km 2 Little River watershed is near Tifton, Georgia in the South Atlantic Coastal Plain, and forms the headwaters for the larger Upper Suwannee Watershed. The Little River watershed is typified by broad floodplains with very poorly defined stream channels and gently sloping uplands. Approximately 45% of the watershed is woodland, 37% crops, 4% pastures, 7% idle, and 7% roads, urban, and water (as described in [37]). Swamp hardwoods occur along the stream edges and are often accompanied by thick undergrowth forming the riparian vegetation boundary along stream networks. Three sampling stations were selected representing first to fourth stream orders with varying levels of flow throughout the year. First order streams are headwaters and are small and narrow whereas fourth order streams are fed by multiple tributaries and are larger and broader. Stations were located in Tift County (upstream of the City of Tifton) in GA Public Health District 8-1. Tift County is rural but is considered to be urbanizing. The 2010 population was 40,118 (59.8 people km -2 ) with 6.9% of the population under the age of 5 [40]. The per capita income was $18,928 [40]. This district has the highest case rates for salmonellosis in the state (70.1 cases per 100,000 in 2011).
The Oconee River Basin consists of two headwater tributaries, the North Oconee River and the Middle Oconee River, which originate at the northern end of the basin in the Piedmont Upland physiographic province, at an elevation of about 305 m above mean sea level. These headwater streams are generally well entrenched, flow through narrow floodplains, and have steep gradients ranging from 0.15 to 1.4 m km -1 . These rivers flow for approximately 100 km to a point just south of Athens, Georgia, where they join to form the Oconee River. Land use in the upper portion of the basin is primarily rural, with poultry farming, dairy farms and grazing for beef production as primary uses. Three sampling stations were selected along the North Oconee River and its tributaries in Jackson County, upstream of the city of Athens, and included first to fourth ordered streams. Jackson County recorded a population of 60,485 in 2010 (68.8 people km -2 ) with 6.8% of the population under the age of 5 [40]. The per capita income was $22,830 [40]. Jackson County is located in GA Public Health District 10, which is a lower case rate area for salmonellosis in Georgia (28.9 cases per 100,000 in 2011).

Sample collection
Water samples were collected monthly from December 2010 to November 2011 at each of the six stations (Fig 1), which were accessible on foot with public access. Sterile 1-L polypropylene bottles were filled by hand at the deepest part of the stream. For a limited number of collection events (n = 3 in each watershed), samples were obtained from well water, after flushing the spigot for 5 min, at farms near the sampling sites (in coordination with owners). All samples were maintained on ice, transported to the laboratory, and processed within 6 h of collection. Wildlife near each of the six sites was also sampled. Songbirds, raccoons, and opossums were the focus of the active surveillance. Mist nets were used as previously described [41] to capture birds. Birds were held in individual disposable paper bags until they defecated. After 1 h, even if the bird had not defecated, it was released to avoid capture-related mortality. Opossums and raccoons were captured with baited live box traps (Havahart, Woodstream Corp, Lititz, PA). Briefly, traps baited with sardines were deployed at sunset and checked at dawn. If an animal was captured, the trap was turned so that it would stand vertically and the animal was gently forced to the bottom and hand injected with an anesthetic. Once the animal had reached a light plane of anesthesia, it was removed from the trap, and approximately 1 g of feces was removed directly from the rectum by digital extraction. Fresh fecal material was immediately immersed in 10 ml of dulcitol selenite broth and maintained at room temperature until submitted to the Athens Diagnostic Laboratory for isolation. Samples were submitted to the laboratory within 48 h. The University of Georgia Animal Care and Use and Procedures Committee approved protocols involving capture and handling of animals associated with this project (AUP # A2010 08-159-Y3-A0).

Salmonella isolation
Each water sample (100 ml) was filtered in duplicate onto 0.45 μm 47-mm mixed cellulose ester membranes, inserted into a sterile 50-ml centrifuge tube containing 20 ml of 1% sterile peptone broth and incubated overnight at 37°C. A 100-μl aliquot of turbid broth culture was used to inoculate a15-ml tube 10 ml Rappaport-Vassiliadis (RV) broth, which was then incubated at 37°C for~24 h. One loopful of overnight growth from RV broth was spread onto XLD agar plates and incubated for 24 h. Presumptive Salmonella colonies (H 2 S positive) were picked and transferred to LB agar stabs. Cultures were streaked for isolation three times before final Salmonella confirmation and serovar determination (see below). Each sample was scored as positive or negative for Salmonella presence following confirmation steps.
All H 2 S positive colonies were further characterized biochemically to identify Salmonella. Real time PCR was used to screen Selenite broths that were culture negative for Salmonella [44]. PCR template was prepared from a pool of three Selenite broth cultures. DNA was isolated from 1 ml of the pooled enrichment using Ultra Clean Fecal DNA Kit (MO Bio Inc., Carlsbad, CA). Positive pools were then tested individually by PCR and subcultured simultaneously onto Salmonella selective media. A delayed secondary enrichment was done for culture negative, PCR positive Selenite enrichment broths [43]. Any suspect, biochemically atypical Salmonella were confirmed by PCR [44]. Salmonella isolates were forwarded to the National Veterinary Service Laboratory (Ames, IA) for serotyping.

Molecular typing of Salmonella isolates by pulsed-field gel electrophoresis
Pulsed-field gel electrophoresis (PFGE) was used to determine the genetic relatedness [45][46][47] among Salmonella isolates obtained during this study, archived isolates from environmental and animal sources, and human isolates represented in the PulseNet USA national database, which included reports from states in the southeast US (GA, FL, SC, and AL). A master database of Salmonella PFGE patterns was generated in BioNumerics (Applied Maths; Austin, TX). Comparisons were made between PFGE patterns using Dice coefficient [48] and unweighted pair group method of arithmetic averages (UPGMA) clustering. Clusters were identified based on a 75% similarity cut-off.

Serovar distribution in Salmonella isolated from the Little River and Oconee River Basins
In total, 1,029 isolates from the two watersheds (water and animals) and archived environmental and animal samples were processed for serovar and PFGE pattern. These included 355 isolates from water and animals collected in the Little River (Upper Suwannee) and Oconee River watersheds (2005-2011) and an additional 674 archived isolates from animal sources (various species) obtained from the Salmonella Reference Collection (SARA 1-72) [49] and past studies [21,37,48,[50][51][52][53]. Thirty-seven unique serovars were identified from salmonellae isolated from water and animals in the Little River and Oconee River watersheds. These included 15 serovars that were ranked among the top 20 for human cases in the US and in Georgia [31]. Eighteen serovars were recovered only from water (represented by 53 isolates) and included six of the top 20 ranked serovars in human cases (U.S. and Georgia). Only two serovars, Braenderup and Paratyphi B var L (+) tartate +, were found in both watersheds and none of the PFGE types were shared between the two regions ( Table 1). Only five serovars were recovered solely from animals (cattle, hogs, opossum, raccoons or song birds) and none included serovars commonly associated with human cases. Two serovars were found in animals from both watersheds (Dublin and Muenster) ( Table 1). Most isolates (82%; 291/355) were associated with serovars found in both water and animals, including 4 isolates from shallow wells in the Little River watershed. In all, 14 serovars were recovered from both sources and nine of these were among the top 20 for human cases (US and Georgia) ( Table 1). Unlike water-only and animalonly serovars, most of the serovars from both sources were also found in both watersheds (11 were shared).
Regardless of source, Salmonella enterica Rubislaw, Give, Hartford, Braenderup, and Muenchen were the serovars most frequently isolated from either region, accounting for 53% and 62% of total isolates from the Little River and Oconee River watersheds, respectively. Salmonella Muenchen and Rubislaw were the most frequently encountered serovars in both river basins. With the exception of S. Braenderup, most other Salmonella serovars encountered in both watersheds were transient, being isolated only once for a given year.

Strain distribution in Salmonella isolated from the Little River and Oconee River Basins
Salmonella PFGE patterns for environmental and archived isolates were compared to each other to evaluate trends in strain persistence and relatedness from the two sample sites. In addition to patterns in serovar distribution (Table 1), there was also considerable strain diversity in Salmonella isolated from the two study sites. Among 37 Salmonella serovars isolated from either region, there were 204 unique PFGE patterns for the water and animal isolates analyzed. Frequently PFGE patterns clustered together by serovar (75% similarity) (Fig 2); however, there were several Salmonella serovars where PFGE patterns with <75% similarity generated two or more clusters (S. Bareilly-2 clusters; S. Gaminara-6 clusters; S. Give-4 clusters; S. Meleagridis-2 clusters). Two Salmonella serovars, S. Muenchen and S. Rubislaw, exhibited the greatest diversity in PFGE patterns, necessitating separate cluster analyses (Figs 3 and 4). Salmonella Muenchen PFGE patterns fell into one of 16 clusters; clusters I & II represented 41% and 16% of PFGE profiles, respectively. Salmonella Rubislaw demonstrated even greater diversity in PFGE profiles, with patterns falling into one of twenty-three clusters. No single S. Rubislaw PFGE cluster accounted for more than 15% of the total patterns. While there was a high level of diversity in PFGE profiles for both serovars, more patterns matched human clinical cases entered in PulseNet for S. Muenchen (44%; n = 25) than for S. Rubislaw (18%; n = 22).  While there were several Salmonella serovars common to both watersheds, only 5 of 204 PFGE types observed were shared among isolates from the two regions (Table 1). Most Salmonella PFGE types (82%) were rare, only appearing once. There were 11 unique PFGE types that were identified in isolates from both water and animals within the same watershed sampling area ( Table 2). For eight of these, water and animal isolates were collected within the same season or year and included Salmonella serovars S. Anatum, S. Bareilly, S. Give, S. Hartford, S. Mbandaka, S. Montevideo, and S. Newport. Most PFGE types were specific to water or to animals.
Despite turnover of Salmonella strains, there were several Salmonella strains isolated at least twice from either watershed, representing 19 serovars and 34 PFGE types. Most of the PFGE types were encountered sporadically, often isolated once in a given year. Just over half (18/34) were encountered multiple years; 14 that were originally isolated in 2005 were still present in  The incidence of salmonellosis in Georgia is skewed within the state, with the highest incidence occurring in the southern region. To address whether the distribution of specific serovars or strains in the environment might be associated with these trends in human cases, select Salmonella PFGE types identified between Little River and Oconee River isolates were compared to CDC PulseNet database for matching PFGE patterns among human isolates. Due to the high number of PFGE patterns identified, only a subset were compared, representing 1) the most common Salmonella serovars in human cases, 2) a Salmonella strain that was persistent in either watershed, or 3) present in both water and wildlife from that locale (n = 113). Approximately, half of the Salmonella isolated from water and wildlife sources had matching PFGE patterns with PulseNet database of human isolates (Table 4). Human cases from the state of Georgia were reported for 7 and 4 of the Salmonella PFGE types from Little River and Oconee River watersheds, respectively (represented by serovars: S. Anatum, S. Braenderup, S. Hartford, S. Montevideo, S. Muenchen, S. Newport, S. Paratyphi B, and S. SaintPaul). There was no significant difference in preponderance of Salmonella strains that had matching PFGE patterns with human isolates in PulseNet between the two river basins (Table 4); however, there was a difference in the distribution of S. Muenchen strains associated with human cases. Ninety percent of S. Muenchen strains (n = 10) from the Little River had matching PFGE profiles with PulseNet entries, while only one third of S. Muenchen PFGE patterns for Oconee River isolates (n = 24) had matches with PulseNet database.

Discussion
The overall Salmonella serovar composition noted in this study and in prior work was similar between these two rural watersheds [36,37,53]. Salmonella serovars generally associated with food animals were rare in both watersheds (e.g., S. Newport, S. Enteritidis, S. Typimurium) [54][55][56], whereas elsewhere in the United States and Canada, such Salmonella serovars have been frequently isolated from watersheds [57,58]. The serovar composition found in the present study also differed from the findings in agricultural ponds within the coastal plain of Georgia, where serovars associated with food production were common and displayed a relatively low diversity of serovar type [39]. In the natural and flowing river systems that were the focus of this study, serovar diversity was high and only two serovars common in food were observed, and were found only rarely (S. Newport and S. Saintpaul) [59,60]. Here, S. Muenchen and S. Rubislaw were the most commonly detected isolates in this study, with 47 and 70 isolates, respectively. These serovars have also been found in other natural waters [53,61], including other areas of the Atlantic Coastal Plain [36,53,61]. These serovars are also often associated with human cases in southern Georgia, where they ranked 4 th and 11 th in reported cases between 2000 and 2006 (Georgia Dept. of Health; District 8-1). Interestingly, common poultry Salmonella serovars such as S. Enteritidis and S. Heidelberg [62] were very rarely isolated even though north Georgia has the greatest concentration of poultry production in the state [63] and application of poultry manures to fertilize pasture land is a common practice in many areas of the state [38,64]. Fourteen of the 37 serovars identified in this study, representing 82% of the isolates, were recovered both from surface waters and wildlife captured nearby (one serovar, Rubislaw, was also identified from well water). Eleven PFGE types were identical between water and animals. In contrast, only five serovars were recovered from animals alone. While 18 serovars were found solely in water, these represented only 15% of the isolates. Salmonella isolation was especially common in opossums and raccoons caught in proximity to the water collection sites in both watersheds. It is unknown how much these animal species and others [65,66] contribute to Salmonella loading in either river basin, but the results indicate that a significant population of Salmonella strains may be moving between wildlife hosts and the environment, including water. Interestingly, many of the Salmonella serovars identified in the Little River (southern Georgia) have been associated with outbreaks epidemiologically linked to fresh produce [15]. There is a significant delineation in agricultural output within the state of Georgia, where the southern region is noted more for produce production. As water is especially important in cultivation and production of produce, the watersheds in southern Georgia are potentially important conduits for introducing Salmonella contamination to food products. Irrigation ponds in the region have been shown to harbor Salmonella, but at a lower diversity than noted in the watershed studies presented here [39]. Differences between a pond environment and a flowing river system may reflect differences in loading and contamination and may also reflect ecological differences in the system.
The overall level of Salmonella and their diversity in natural watersheds supports the idea that landscapes may be an important feature in sporadic transmission to humans. Similar disparities in the geographic distribution of human illnesses associated with other zoonotic bacteria such as Escherichia coli O157:H7 and Campylobacter sp. have been observed elsewhere, where higher illness rates are found in regions with high livestock densities, likely due to increased animal contact and environmental exposure [18,19,67,68]. Evidence from recent work suggests that direct environmental exposures may be important in non-outbreak scenarios. For example, the prevalence of private wells for drinking water and use of septic systems are noted risk factors for non-outbreak associated salmonellosis, especially among children [17,20]. Similar risk factors were also noted for specific serovars common in the southeast U.S. [20] While both of the rural areas investigated in this study rely on septic systems for waste disposal, there is a higher rate of private (untreated) wells as the source of drinking water in the southern part of the state (28.7% of the total population and 95% of the rural population [69].
Salmonella enterica is a commonly detected pathogen in the waters of Georgia, and here showed a very high level of diversity both at the serovar and PFGE-type strain level, especially in the Little River watershed in the Atlantic Coastal Plain. Similar observations have been made regarding the genetic diversity of Salmonella isolated from other river systems [58], including the Suwanee watershed of South Georgia/North Florida [36,58]. This temporal turnover of Salmonella strain types in the two river basins may follow point-source contamination and possible impact of land use on prevalence and diversity within these environs. While there was significant genetic diversity in each of these two river basins, we did identify matching PFGE patterns between Salmonella recently isolated from the Oconee River and Little River and those from earlier studies of the two watersheds [37,53]. This suggests certain Salmonella strains persist in this environment due to either continued contribution by an animal population or long-term persistence within this aquatic environment, suggesting a specific niche within the river or watershed. For example, evidence indicates that sediment may support long-term survival and persistence for Escherichia coli O157:H7 [70], Salmonella [70,71], and Campylobacter [72]. Storm events can churn up this sediment and reintroduce these dormant, persistent Salmonella strains into the water column. Salmonella levels within the water column significantly increase in the river following storm events and there is a positive correlation in Salmonella prevalence, concentration, and rainfall [37].
Over all serovars tested, there were no differences between the Oconee River and the Little River basin in the percentage of environmental Salmonella strains that matched human clinical strains (~50% each, determined by identical PFGE patterns with human isolates reported to the PulseNet database), despite the fact that prevalence of salmonellosis is higher in the south (including the Little River area). When focusing on specific Salmonella serovars, we identified distinct differences between the watersheds in the proportions of strains matching those of human isolates in PulseNet. For example, 90% of the S. Muenchen isolates from the Little River watershed matched entries in PulseNet versus only 33% from the Oconee River watershed. The greater preponderance of matches between environmental and human isolates for the Little River and the Oconee River is especially significant due to the high level of genetic diversity inherent in these populations and may suggest differences in climate, landscape, and human activities between the two watersheds. In addition to epidemiological studies that may help to determine where and how humans may be exposed through the environment (including water), in the future whole genome sequencing of these environmental, animal, and human isolates, with matching PFGE patterns, will allow us to discern genetics that underlie the pathogenic potential of environmental Salmonella and genetic markers that identify point source for contamination.

Conclusion
Studies examining risk factors for salmonellosis can no longer focus only on impacts associated with food production. Salmonella is a broad zoonotic agent that is likely part of the ecology of the landscape, with high rates of exchange probable between humans, water, and wildlife. There are several inherent differences between North and South Georgia in its geography, geology, land use, and ecology that may be driving the rates of salmonellosis within the state. The recent publication of the Salmonella Atlas for 32 major Salmonella serovars by the CDC further supports geographical differences in the incidence of disease in the United States [73]. Understanding ecological interactions between pathogens, the environment, and humans is essential for reducing the burden of human illnesses due to Salmonella.