Salmonella serotype diversity and seasonality in urban and rural streams

Authors


Correspondence

Janis L. Thomas, Ontario Ministry of the Environment, 125 Resources Road, Toronto, ON, Canada. E-mail: janis.thomas@ontario.ca

Abstract

Aims

To investigate the prevalence, seasonality and genetic diversity of Salmonella enterica serotypes, particularly those of human and veterinary health significance, in urban and rural streams.

Methods and Results

Using a swab collection technique and multiple culture media for isolation, Salmonella were detected in 78·4% of water samples (November 2003 to July 2005) taken from urban and rural/agricultural streams in the Grand River watershed (Ontario, Canada). Among 235 isolates, there were 38 serotypes, with the predominant serotypes and phagetypes (PT) being Salmonella Typhimurium PT 104 and Salmonella Heidelberg PT 19. These are also the most common Salmonella serotypes found in humans and farm animals locally and across Canada, a trend not commonly reported. The urban stream had more frequent Salmonella occurrence, greater serotype diversity and greater genetic variability (based on pulsed field gel electrophoresis) of specific strains compared with the rural/agricultural streams. Distinct seasonality in serotypes of health significance was observed only in the rural/agricultural streams, which is likely a reflection of seasonal source inputs in these watersheds. Despite the lower occurrence of these strains in stream water in the colder months, laboratory studies did not support reduced survival of Salm. Typhimurium and Salm. Heidelberg at lower temperatures, although survival differences were observed with other serotypes.

Conclusions

A diverse range of Salmonella serotypes and PT were obtained from both urban and rural/agricultural streams, with the predominant strains being those most frequently associated with human and veterinary disease in Canada.

Significance and Impact of the Study

The ubiquitous nature of Salmonella in water and the predominance of serotypes/PT of human or veterinary health significance suggest that the aquatic environment is a reservoir for this bacterium and could be involved in the transport and dissemination of this pathogen between hosts.

Introduction

Salmonellosis is a major public health burden in many countries, including developed countries such as Canada and the USA (Voetsch et al. 2004; Thomas et al. 2006). It is estimated that only a small portion of infections are clinically recognized, and as a result, the rates of salmonellosis are greatly underestimated (Thomas et al. 2006). The majority of salmonellosis cases are considered to be sporadic or endemic, meaning they are independent of outbreaks and travel-related illness (PHAC 2007a). The sources of endemic illness are considered of domestic origin and may include local food or water (PHAC 2007a,b).

Foodborne transmission is considered the predominant exposure route for Salmonella (Mead et al. 1999; CDC 2009a); however, interest of the waterborne aspect of this pathogen remains, as several drinking water outbreaks have been attributed to this bacterium (CDC 2011). Most known drinking water outbreaks have occurred in small unchlorinated groundwater supplies; however, the extent of water-related health impacts remains elusive and many questions remain as to the role that water plays in sporadic or endemic cases of salmonellosis. Recently, Denno et al. (2009) conducted a matched case–controlled study on sporadic salmonellosis in children and reported that nonfoodborne exposure might be as important as foodborne exposure. In their study, exposure to specific water sources, including drinking untreated water from private wells and recreation in surface waters, were risk factors for sporadic salmonellosis in children. In recognition of the gaps that exist in the study of waterborne Salmonella, the USEPA included this bacterium on the Contaminant Candidate List 3 (CCL 3; USEPA 2009b).

The occurrence levels of nontyphoidal Salmonella in river water range greatly, with frequencies reported between 3% and 79·2% (Johnson et al. 2003; Gannon et al. 2004; PHAC 2007b; Simental and Martinez-Urtaza 2008; Haley et al. 2009; Gorski et al. 2011; Edge et al. 2012). However, most studies report frequencies to be ≤20%, including most Canadian studies (Johnson et al. 2003; Gannon et al. 2004; PHAC 2007a,b; Wilkes et al. 2009; Edge et al. 2012). Variable survival rates for Salmonella in the aquatic environment have been reported, depending on the conditions of the study conducted (Johnson et al. 1997), although it is generally assumed that Salmonella can survive for considerable periods of time in water, or at least as long or longer than faecal indicator bacteria (Wright 1989; Catalao Dionisio et al. 2000; USEPA 2009a).

The reported occurrence of waterborne Salmonella, as well as its survival capability in water, implies that the aquatic environment may play a role in the transmission of Salmonella between host animals. However, few studies have reported similarities between the predominant Salmonella serotypes/phagetypes (PT) obtained from the aquatic environment and those of clinical relevance in humans (e.g., Polo et al. 1999; Simental and Martinez-Urtaza 2008). This lack of association leaves many unanswered questions regarding epidemiological connections between water and human health.

To fully understand the transmission of Salmonella, in particular serotypes of human and veterinary health significance, the environmental occurrence of strains needs to be further defined. Knowledge of the waterborne occurrence of Salmonella serotypes and PT, both geographically and temporally, as well as the genetic variability of strains, is a first step and principle factor in understanding the risk to the population that uses these waters for recreation and as a source of drinking water and is critical information needed prior to the implementation of prevention and control strategies for the future.

The aim of this study was to determine the spatial and temporal distributions of Salmonella serotypes in three tributaries, two dominated by agricultural/rural activities and one urban, as well as to examine the genetic relatedness of Salmonella serotypes of human health importance that are circulating within these tributaries. To further probe the mechanism underlying seasonal differences in the occurrence of certain serotypes, cold temperature survival experiments were also performed.

Materials and methods

Study area

The Grand River watershed, which drains to Lake Erie, is the largest watershed in south-western Ontario (Canada) at approximately 6800 km2 (Fig. 1a). A source water protection area was created to protect the five surface water intakes and over 200 municipal wells that supply water to 86% of the population living in the watershed, which is just under 900 000 (GRCA 2010). In addition to its use for drinking water, the Grand River and its tributaries are used for numerous recreational activities including fishing, canoeing and swimming, as well as for agricultural activities including livestock watering and irrigation.

Figure 1.

Grand River watershed located in south-western Ontario (a), and sample locations within the three tributaries included in this study: Conestogo River, Canagagigue Creek and Laurel Creek (b). (▲) Sewage treatment and (image_n/jam12079-gra-0001.png) urban/built up area.

Three tributaries that flow into the Grand River were chosen for study: Canagagigue Creek (CAN), Conestogo River (CON) and Laurel Creek (LC), each with two to three sampling locations per tributary (Fig. 1b).

These rivers are adjacent tributaries that flow into the Grand River. CAN and CON are both located in rural areas characterized by intensive farming (i.e. crops and livestock) and are highly (60%) tile drained (Dorner et al. 2007). The LC watershed includes largely urban and suburban areas, but with no point sources of faecal pollution (i.e. no sewage treatment discharge); however, it is impacted by waterfowl and stormwater runoff throughout the year (Table 1).

Table 1. General land use and stream information for each tributary
TributaryTotal area (m2)Percentage of land coveraAverage stream discharge (m3 s−1)b
UrbanAgri-cultureForest/rangeWetland
  1. a

    Per cent land cover comes from a GIS layer created by the Ontario Ministry of Natural Resources and includes Ontario Land Cover data derived from LANDSAT Thematic Mapper data, values represent land-use upstream of the last sampling location on each stream.

  2. b

    Dorner et al. (2007).

Laurel Creek7·6 × 10738441610·1
Canagagigue Creek1·1 × 10848712<11
Conestogo River8·2 × 108<1821713

Samples were collected between November 2003 and June 2005 from the tributaries, with the aim of collecting several samples per season throughout the year, including the winter months. Samples were usually taken from each tributary two times per month.

Salmonella collection and isolation

A swab collection technique was used for sample collection, as described in Standard Methods (APHA et al. 2005). These swabs were suspended below the water surface for three to 5 days. Following collection, the swabs were transported on ice in sterile bags to the laboratory for analysis within 24 h. Each swab was pre-enriched overnight in buffered peptone water at 37°C in 1-l sterile glass bottles. Following overnight incubation, each bottle was shaken and 1 ml was placed into 9 ml of each enrichment broth: Rapapport Vassiliadis broth and tetrathionate broth (42°C for 24 h). Portions of these enrichment broths were streaked onto Brilliant Green Sulfa Agar (37°C for 24 h) and a 200-μl portion was placed onto modified semi-solid Rapapport Vassiliadis agar (42°C for 24 h). Presumptive isolates were streaked onto MacConkey agar (37°C for 24 h) and subjected to further biochemical testing on triple sugar iron, lysine iron agar and urea broth. Following biochemical confirmation, isolates were confirmed through agglutination using Salmonella O antiserum Poly A-I & Vi (Difco, MD, USA).

Several wildlife faecal samples were collected in the urban watershed (LC). On 11 February 2004, faecal samples (n = 6) from mallard ducks (Anas platyrhynchos) were collected. These samples were freshly obtained from ice or snow at sample point LC-3, as wildlife tend to congregate at this location as a result of easy access to the stream. Faecal samples (n = 10) of unknown origin were also collected on 7 May 2007, on the banks of LC at sample point LC-3. In all instances, faecal material was collected with a sterile cotton-tipped swab and placed in a sterile bag. Faecal samples were placed directly into pre-enrichment tubes and isolation continued as described above.

Isolates were sent to the Office International des Epizooties (OIÉ) Reference Laboratory for Salmonellosis at the Laboratory for Foodborne Zoonoses, Public Health Agency of Canada (Guelph, Ontario) for serotyping and selected Salmonella serotypes (Salmonella Typhimurium, Salmonella Heidelberg, Salmonella Enteriditis and monophasic strains) were phagetyped. Isolates that reacted with phages, but did not conform to any recognized PT, were designated as atypical. Strains that did not react with any of the phages were designated as untypable.

Pulsed field gel electrophoresis

Salmonella Typhimurium, Salm. Heidelberg and several monophasic isolates were subjected to pulsed field gel electrophoresis (PFGE) following the standardized PulseNet method with slight modifications (Ribot et al. 2006; CDC 2009b). PFGE was performed using XbaI (New England Biolabs, Whitby, Ontario, Canada). Electrophoresis was carried out for 19 h with an initial switch time of 2·2 s and a final switch time of 63·8 s at 6 V using the CHEF-DR III electrophoresis system (Bio-Rad, Hercules, CA, USA).

Resulting banding patterns were analysed using BioNumerics software (Applied Maths, Sint-Martens-Latem, Belgium). Each image was normalized to a lambda ladder (CHEF DNA size standard blocks; Bio-Rad), and relatedness of each pattern was calculated using Dice coefficients with a 1·5% tolerance and optimization. Dendograms were generated using an unweighted pair group method using arithmetic averages. Banding patterns were considered identical if the similarity was ≥99% and considered closely related if similarities were ≥80%, which would show few differences in banding patterns.

Stream conditions and water quality observations

As swabs were placed in water for several days, it was not possible to use stream discharge as a direct variable for analysis; therefore, samples were designated as being collected during base-flow or high-flow (event-flow) conditions following examination of the hydrograph for each stream during the time a swab was suspended in the water column. To sort the samples during base-flow or event-flow conditions, data were examined from flow gauge stations in each of the three tributaries. Data were extracted from the Water Survey of Canada's (Environment Canada) collection of discharge or level data from HYDAT stations (CDRom, Hydat Version 2005 – 2.04, June 2007) and imported into the Web-based Hydrograph Analysis Tool (WHAT; Purdue University, West Lafayette, IN, USA) to determine the base-flow separation. Results were imported into Microsoft Excel sheet where they were visually examined after plotting hydrographs (water discharge or depth against time). Average daily discharge data were used for sites in CAN and CON. Both of these tributaries demonstrate a sluggish response to events that could be easily observed in the hydrograph. Determining whether an event occurred in the urban tributary, LC, was more difficult as there were rapid responses to smaller precipitation events. Therefore, data for precipitation and field observations were used in conjunction with the level data to determine whether event flow was occurring in the stream. A sample was designated as being taken during an event if the flow or level data were above base flow or level during all or part of the time that the swab was in the water. Swabs that were placed in the water during the falling limb of the hydrograph were considered a base-flow sample.

Air temperature and precipitation data at the Region of Waterloo International Airport were extracted from Environment Canada's web site (www.climate.weatheroffice.gc.ca). Due to the distance between the precipitation collection point and the tributaries under study (c. 8–25 km), the precipitation data collected may not accurately represent specific events in each the tributaries. Therefore, these were averaged by month.

Differential survival studies

Several serotypes obtained from river samples were examined for survival differences at 4°C. Each isolate used to inoculate microcosms was taken from frozen (−80⁰C) cultures and grown in LB broth at 37°C for 16 h. One mL of the Salmonella culture was centrifuged (maximum speed, 10 min; Eppendorf Centrifuge 5415 D, Mississauga, ON, Canada), and the cells were resuspended and washed twice with sterile saline (0·85%). Washed cells were then added to the microcosms containing 100 ml of sterile 0·85% saline (pH 8·04) in a 500-ml glass flask.

Two separate trials were conducted: the first trial examined the survival of environmentally derived Salm. Typhimurium (isolate no. 3A2) and Salmonella Montevideo (no. 3C4) for up to 94 days, and the second trial included Salm. Heidelberg (no. 19D1) and Salmonella Agona (no. 21C1) for up to 121 days. Each serotype was placed into separate microcosms in a cold room (4°C) and continuously shaken at 30 rev min−1 on an orbital shaker for the duration of the trial. Each trial was conducted in replicate.

Direct counts on LB agar (trial one) or on LB containing ampicillin (64 μg ml−1, trial two) were taken to determine the concentration of culturable cells in each flask at time zero. The starting cell concentrations for each microcosm were approximately 106–107 ml−1. Aliquots were taken several additional times within the first 24 h of the study. Subsequently, samples were taken more frequently over the first 2 weeks and less frequently up to the end of the study (weekly to monthly). At each sampling time, aliquots were taken out, diluted in 0·85% saline and plated in duplicate.

Data analysis

Chi-square tests were used to determine whether differences existed in the occurrence of Salmonella among streams and sample locations, seasonally and between flow conditions. Fisher's exact test was performed if the expected value was <5 or total value was <50. The level of significance was set at a P < 0·05. Spearman rank correlations of monthly Salmonella occurrence data vs monthly precipitation, air temperature and water temperature were also conducted. Differences in survival among various serotypes were determined by examining differences between slopes of log concentrations (cells ml−1) vs time using ancova.

Results

Occurrence and spatial variability of Salmonella serotypes and PT

A total of 116 swabs were analysed for the presence of Salmonella between November 2003 and June 2005, and 91 (78·4%) of these samples were found to be positive for Salmonella isolates. The frequency of Salmonella varied among tributaries (< 0·001) with LC and CAN demonstrating the highest occurrence at 89% (56/63) and 82% (23/28), respectively. Samples taken from CON resulted in the lowest recovery of Salmonella at 48% (12/25).

A total of 235 waterborne Salmonella isolates were obtained, with 38 different serotypes (Table 2). The five most frequently observed serotypes were Salm. Typhimurium, Salm. Heidelberg, Salmonella Thompson, Salmonella Infantis and Salmonella Kentucky, together comprising 43% of the total isolates observed (101/235). Half (17/34) of the Salm. Typhimurium isolates were designated as variant five strains (Var. 5-), which are O:5 negative serological variants of Salm. Typhimurium.

Table 2. Occurrence of Salmonella serotypes overall and within each tributary
SerotypePer cent occurrence (number of isolates)
OverallLaurel CreekCanagagigue CreekConestogo River
  1. Bold values represent the predominant serotype observed in each tributary.

  2. a

    Includes Salmonella Typhimurium Var. 5- and nonvariant strains. Var. 5- represents 5 of 16 of isolates in Laurel Creek, 10 of 14 in Canagagigue Creek and 2 of 4 in Conestogo River; Var. 5- refers to the O:5-negative variant of Salm. Typhimurium (formerly Var. Copenhagen; CDC 2009a).

Typhimuriuma14·5 (34)12·6 (16) 18·7 (14) 12·1 (4)
Heidelberg8·1 (19)5·5 (7)4·0 (3) 27·3 (9)
Thompson7·7 (18) 14·2 (18)   
Infantis6·4 (15)4·7 (6)12·0 (9) 
Kentucky6·4 (15)4·7 (6)6·7 (5)12·1 (4)
Agona5·1 (12)7·1 (9)4·0 (3) 
Oranienburg4·7 (11)4·7 (6)5·3 (4)3·0 (1)
Kiambu3·8 (9)7·1 (9)  
Senftenberg3·8 (9)2·4 (3) 18·2 (6)
Montevideo3·4 (8)3·1 (4) 12·1 (4)
Tennessee3·4 (8)1·6 (2)8·0 (6) 
Mbandaka3·0 (7)3·9 (5) 6·1 (2)
Berta2·6 (6)4·7 (6)  
Putten2·6 (6) 8·0 (6) 
Uganda2·6 (6) 8·0 (6) 
4, 5, 12:b:–2·1 (5)3·9 (5)  
Newport2·1 (5)1·6 (2)4·0 (3) 
Anatum1·7 (4)0·8 (1)4·0 (3) 
Hadar1·7 (4)3·1 (4)  
Saintpaul1·7 (4)3·1 (4)  
Derby1·3 (3)2·4 (3)  
4,5, 12:i:–1·3 (3) 1·3 (1)6·1 (2)
28:y:–1·3 (3) 4·0 (3) 
Indiana1·3 (3) 4·0 (3) 
Orion1·3 (3)2·4 (3)  
Ago0·9 (2)1·6 (2)  
Enteritidis0·9 (2)0·8 (1)1·3 (1) 
Give0·4 (1)  3·0 (1)
Hartford0·4 (1)0·8 (1)  
19:–:–0·4 (1)0·8 (1)  
4, 12:–:–0·4 (1) 1·3 (1) 
23:d:-0·4 (1) 1·3 (1) 
Rough-O:fgt:–0·4 (1)0·8 (1)  
Rough-O:d:l,w0·4 (1) 1·3 (1) 
Litchfield0·4 (1)0·8 (1)  
Muenchen0·4 (1)0·8 (1)  
Pomona0·4 (1) 1·3 (1) 
Worthington0·4 (1) 1·3 (1) 

The number of serotypes obtained from each stream was also different (P < 0·001). The greatest diversity was in LC, where 27 different serotypes were observed. Of these serotypes, 14 were never isolated from CAN or CON. Salmonella Thompson was the most frequently isolated serotype in LC. In CAN, 20 different serotypes were obtained with Salm. Typhimurium being the most frequently isolated. In CON, only nine different serotypes were observed, with Salm. Heidelberg being the most common (Table 2).

Faecal samples from mallard ducks collected along the banks of LC were all positive and all isolates were found to be a monophasic serotype: 4, 5, 12:b:–. This serotype was only isolated from water at LC, where it was observed at two separate sample locations and dates in the same year. Sampling of faecal material from animals of unknown species in the LC sub-watershed failed to isolate Salmonella.

Eight different PT of Salm. Typhimurium (Fig. 2a) were observed, PT 104 being the predominant and representing 53% of the Salm. Typhimurium isolates. Over half of these PT 104 isolates (11/18) were designated as Var. 5 – strains. Six different PT of Salm. Heidelberg were observed with the most common being PT 19 (Fig. 2b). Both of these predominant PT, Salm. Typhimurium PT 104 and Salm. Heidelberg PT 19, were isolated from all three tributaries.

Figure 2.

Phagetypes of Salmonella Typhimurium (a) and Salmonella Heidelberg (b) observed in three tributaries of the Grand River: Laurel Creek (■), Conestogo River (image_n/jam12079-gra-0002.png) and Canagagigue Creek (image_n/jam12079-gra-0003.png).

Temporal and seasonal variability of Salmonella serotypes

A total of 26 high-flow events were captured during this study. The remaining 34 sampling dates were classified as nonevent or base flow. Of the 116 swabs collected, 43 were classified as event samples and 73 were classified as base-flow samples. Although there was a greater proportion of swabs positive for Salmonella during event flow vs base flow (Table 3), the difference was not significant (P = 0·41).

Table 3. Occurrence of Salmonella during event-flow and base-flow conditions throughout the study for each tributary
TributaryEvent flowBase flow
No. positiveNo. negativePercent positiveNo. positiveNo. negativePercent positive
Laurel Creek24485·732391·4
Conestogo River5183·371236·8
Canagagigue Creek7277·816384·2
Overall36783·7551875·3

Spearman rank correlations between Salmonella occurrence, and average monthly precipitation, air temperature and water temperature were rs = 0·56, rs = 0·39 and rs = 0·53, respectively. None of these Spearman rank correlations are significant at the 5% level (critical rs = 0·648). The lowest occurrence of Salmonella did coincide with the month with the lowest observed precipitation, air temperature and water temperature, which was February when only 25% of samples were positive for Salmonella.

Overall, significant differences were not observed among seasons (P > 0·8) with samples positive for Salmonella in the fall, winter, spring and summer at 78·4% (29/37), 71·4% (10/14), 72·7% (16/22) and 80% (28/35), respectively. However, significant differences (P = 0·002) were observed in the serotype diversity between seasons (data not shown). Summer and fall had the highest diversity of serotypes (22 and 23), whereas the winter and spring months had the lowest diversity (11 and 10).

The occurrence of Salm. Typhimurium and Salm. Heidelberg appeared to vary by season (Fig. 3). Salmonella Typhimurium was found in all seasons but had a higher incidence in summer (28% or 21/75) and a lower incidence in the spring (3% or 1/34). Salmonella Typhimurium was the most frequently isolated serotype in both the summer and fall.

Figure 3.

Occurrence of Salmonella Typhimurium (a) and Salmonella Heidelberg (b) by season, overall (□) and in the three tributaries: Laurel Creek (■), Conestogo River (image_n/jam12079-gra-0002.png) and Canagagigue Creek (image_n/jam12079-gra-0003.png).

Overall, Salm. Heidelberg was the second most common serotype isolated in the spring and summer months where it represented 24% (8/34) and 9% (7/75) of the isolates obtained, respectively. In the fall months, this serotype was the fifth most common serotype observed at 5% (4/82). It was not observed in the winter months.

At the tributary level, both Salm. Typhimurium and Salm. Heidelberg demonstrated different seasonal trends (Fig. 3a,b). Salmonella Typhimurium was observed in all seasons in LC, where it represented the third and fifth most commonly isolated serotype in winter and spring at 13% (2/16) and 7% (1/14), respectively, and it was the most commonly observed serotype in summer and fall at 14% (5/35) and 16% (7/43), respectively. In contrast to LC, this serotype was not observed in all seasons in CAN and CON. In CAN, Salm. Typhimurium was observed only in the summer, where it was the predominant serotype representing 47% (14/30) of the isolates. Similarly, in CON, this serotype was not observed in the winter and spring, but emerged in the summer and fall where it represented the third and second most common serotype, respectively.

During the summer months, the majority of the Salm. Typhimurium isolates obtained were Var. 5- (100% were Var. 5- in CAN, 100% in CON and 60% in LC). No isolates in the spring and fall months were characterized as Var. 5- and only two isolates were characterized as variant strains in winter, both of which were found in LC.

Similar to what was observed with Salm. Typhimurium, Salm. Heidelberg showed a marked seasonal trend in CAN and CON compared with LC. This serotype showed the greatest occurrence in the spring (Fig. 3b) in CAN and CON and was not observed in the fall or the winter months. Similar to the observation seen with Salm. Typhimurium, Salm. Heidelberg showed more consistent levels of occurrence in LC; however, no Salm. Heidelberg was observed in this tributary in the winter months.

Genetic relatedness of waterborne Salmonella Heidelberg isolates

Six unique PFGE banding patterns were observed in Salm. Heidelberg isolates following treatment with XbaI, all of which showed a ≥86·9% similarity (Fig. 4). Two main clusters, showing a >90% similarity between isolates, were observed in the analysis, with PT 26 predominating in cluster A and PT 19 predominating in cluster B.

Figure 4.

Dendogram based on pulsed field gel electrophoresis patterns following digestion with XbaI for various phagetypes of Salmonella Heidelberg obtained from stream water. Clusters are denoted as a and b.

Several isolates of the same PT that were obtained from different locations or sampling dates within the same tributary, as well as between different tributaries, demonstrated indistinguishable restriction patterns following digestion with XbaI. This was observed in isolates in cluster A and B. In addition, several isolates representing different PT also demonstrated indistinguishable restriction patterns (cluster B).

As observed in the predominant cluster (cluster A, Fig. 4), indistinguishable XbaI patterns were observed between several PT 19 isolates and several other PT, including PT 46, PT 18, PT 29a, an atypical PT and one untypable isolate. Isolates demonstrating this identical banding pattern were obtained from all three tributaries and at various sampling locations.

Genetic relatedness of waterborne Salmonella Typhimurium isolates

A total of 16 unique banding patterns were observed in the 33 isolates of Salm. Typhimurium and 3 monophasic 4, 5 12:i:– isolates, which are considered variants of Salm. Typhimurium. PFGE with XbaI grouped the majority of these isolates into two major clusters (clusters A–B; Fig. 5).

Figure 5.

Dendogram based on pulsed field gel electrophoresis patterns following digestion with XbaI for Salmonella Typhimurium and monophasic isolates obtained from stream water. Clusters are denoted as a and b.

Cluster A, where isolates demonstrated a ≥76·8% similarity, was composed of seven distinct PT (including one atypical PT) of Salm. Typhimurium and two PT of serotype 4, 5, 12:i:–. Most isolates found to be of the same PT demonstrated indistinguishable restriction patterns (e.g., PT 41, PT 22 and PT 108). Many of these isolates that shared identical restriction patterns and PT were found at different sample locations and/or on different dates. Cluster A also contained several variant strains (Var. 5-) of Salm. Typhimurium, many of which showed identical XbaI patterns to nonvariant isolates designated with the same PT (e.g. PT 41).

Cluster B, which demonstrated a >80% similarity between PFGE profiles, was dominated by Salm. Typhimurium PT 104 (and PT 104a) strains. In total, 18 of the 19 PT 104 isolates obtained from water were found in this cluster and showed a high degree of similarity between isolates (>86·5%), with the majority of isolates showing indistinguishable restriction patterns following XbaI digestion. Many variant (Var. 5-) and nonvariant isolates of PT 104 showed identical restriction patterns even when obtained from different tributaries, sample locations and times of the year. Overall, more distinguishable XbaI patterns were observed in PT 104 isolates obtained from LC than from the other tributaries with eight isolates showing five distinguishable patterns (Fig. 5). One PT 104 isolate (19A2a from LC) showed the most unique profile and fell outside of the two main clusters. This isolate demonstrated the lowest similarity (<60%) to other waterborne Salm. Typhimurium or monophasic serotypes.

All monophasic serotype 4, 5, 12:b:– isolates demonstrated >85% similarity following XbaI digestion (data not shown). The majority of serotype 4, 5, 12:b:– isolates (five of six total) from fresh duck faeces demonstrated indistinguishable restriction patterns. Identical (100%) or highly similar patterns (>86% similarity) between isolates from faeces and those obtained from LC waters were also observed.

Differential Salmonella survival at 4°C

Comparisons of survival of various environmentally derived Salmonella serotypes at 4°C in 0·85% saline are shown in Fig. 6. In both trails, low rates of mortality were observed in the initial stages. Culturability declined more rapidly after a month. In all instances, culturable cells were detected at the end of the trial (94–121 days).

Figure 6.

Survival in saline water at 4°C for Salmonella Agona (○), Salmonella Heidelberg (Δ), Salmonella Typhimurium (□) and Salmonella Montevideo (×). Standard deviations for replicate trials are shown.

The slope of culturable cells vs time showed no significant differences between replicate trials for each isolate (P = 0·50–0·98); however, differences were observed among serotypes, with Salm. Agona demonstrating differences when compared to the three other serotypes (P < 0·001). Survival differences were not observed between Salm. Heidelberg, Salm. Typhimurium and Salm. Montevideo (P = 0·35–0·97).

Discussion

The frequency of Salmonella in water is highly variable in time and space, as with faecally derived pathogens in surface water. To help reduce this variability and enhance recovery, we used a swab collection technique and several media combinations. Using these techniques, the ubiquitous nature of Salmonella in the aquatic environment was revealed, as isolates were observed throughout the year and in all tributaries, and a higher occurrence of Salmonella was observed compared with many other studies. Salmonella was obtained in 78·4% of the water samples, compared with 3–21% reported from studies in other Canadian watersheds (Johnson et al. 2003; Gannon et al. 2004; PHAC 2007a,b; Wilkes et al. 2009; Edge et al. 2012). While most studies world-wide describe similar occurrence levels to the other Canadian studies, a recent study by Haley et al. (2009) reported Salmonella at comparable levels (79·2%) to this study in highly agricultural rivers in Georgia, USA.

The swab collection technique probably increased the likelihood of detection and also helped to increase the overall number of isolates obtained to allow for a more comprehensive evaluation of the serotypes present. Identification of the different Salmonella strains at the serotype level is critical to understand whether clinically important isolates are present. This study observed a close relationship between the predominant serotypes in water and those observed in animals and humans in the region of study (Region of Waterloo) as well as in Canada (PHAC 2007a,b). Overall, Salm. Typhimurium and Salm. Heidelberg were the most common serotypes observed in surface water samples.

During the time of this study, the C-EnterNet program (National Integrated Enteric Pathogen Surveillance Program) operating in the Region of Waterloo found the majority of Salmonella infections in humans to be associated with three serotypes: Salm. Typhimurium, Salm. Heidelberg and Salm. Enteriditis (PHAC 2007a,b). A comparison of travel-related vs endemic human cases of Salmonella indicated that all of the Salm. Typhimurium and Salm. Heidelberg were of domestic origin in the region, whereas over half of the Salm. Enteriditis cases were travel-related (PHAC 2007b). Endemic cases of Salm. Typhimurium and Salm. Heidelberg might indicate that local environmental sources, including water, may play a role in exposure to the population. Salmonella Enteriditis does not appear to be endemic in the farm animals in the region or in the country (Guerin et al. 2005; PHAC 2007b), which possibly explains the low occurrence of Salm. Enteriditis observed in the tributaries in this study, <1%.

Both Salm. Typhimurium and Salm. Heidelberg are predominant serotypes observed in farm animals, with Salm. Typhimurium being the most common serotype observed in swine and cattle in the region (PHAC 2007a,b). Salmonella Heidelberg predominates among poultry isolates (Government of Canada 2005); however, it is also found among the top ten isolates in cattle and swine in Canada (Government of Canada 2005). The predominance of these serotypes in farm animals is likely why these two serotypes were the most frequently isolated in the rural/agricultural tributaries (CAN and CON).

Few studies have found an association between predominant serotypes in water and those observed in humans and animals (e.g. Polo et al. 1999; Simental and Martinez-Urtaza 2008). The reason for this disparity might be related to the limited number of Salmonella isolates generally obtained from water in most studies compared with the larger, and therefore more representative, number obtained from other sources such as faecal samples.

Waterborne isolates of Salm. Typhimurium and Salm. Heidelberg demonstrated a diverse range of PT, with PT 104 and PT 19 predominating, respectively. During the time of this survey, these PT represented the most common Salm. Typhimurium and Salm. Heidelberg PT obtained from human clinical samples within Canada (Government of Canada 2005, 2006, 2007; PHAC 2007c). Similarly, these PT are also the most commonly reported among nonhuman isolates in Canada, where Salm. Typhimurium PT 104 is frequently obtained from cattle and swine and Salm. Heidelberg PT 19 is commonly found in chicken (Government of Canada 2005, 2007). Both of these PT have also been reported in a variety of other animals in Canada, although at a lower frequency (PHAC 2007c). Few studies have reported the predominance of these PT in water (Martinez-Urtaza et al. 2004a; PHAC 2007a,b).

While Salm. Typhimurium and Salm. Heidelberg were the predominant serotypes observed, a diverse group of serotypes were obtained from the aquatic environment, suggesting that many hosts are contributing to the loading of water with these pathogens. Of the 38 serotypes obtained, several serotypes rarely reported in humans and farm animals were observed, although at a lower frequency (e.g. Salmonella Pomona, Salmonella Kiambu and Salmonella Uganda). The occurrence of these strains may be attributed to their occurrence in wildlife. In addition, several ‘rough’ strains and incomplete serotypes were observed in water. A similar finding was reported by Baudart et al. (2000) in river waters in France. These authors suggested that the occurrence of these strains of Salmonella in the natural environment may be related to resulting genetic modifications caused by exposure to environmental stresses. Future consideration should be given to the phenomenon of naturalization, as has been suggested in recent years with Escherichia coli (Byappanahalli et al. 2003; Whitman and Nevers 2003; Kon et al. 2007).

Pathogen monitoring studies are commonly conducted in streams impacted by agriculture because of the prevalence of pathogens such as Salmonella in farm animals. For similar reasons, streams impacted by sewage treatment effluent are frequently monitored. Urban streams, on the other hand, are generally overlooked as significant sources of pathogens, particularly if there are no point sources of contamination. Interestingly, compared with the other tributaries, the urban stream demonstrated the highest level of Salmonella occurrence, the greatest diversity of serotypes and a greater genetic variability within select PT following PFGE. Of the 28 serotypes observed, 14 were only found in the urban stream. The great diversity of serotypes observed likely reflects the large number and variety of host species, in particular wildlife, that act as reservoirs of Salmonella in this tributary.

Salmonella has a broad host range and is shed by a variety of wild animals, including birds (Hall and Saito 2008; Dolejska et al. 2009; Gorski et al. 2011), reptiles (Chambers and Hulse 2006), deer (Branham et al. 2005; Gorski et al. 2011) and small mammals (Gorski et al. 2011), as well as domestic pets (Carter and Quinn 2000). A similar finding was recently reported by Patchanee et al. (2010), who observed a higher diversity of serotypes in a North Carolina watershed characterized as residential/industrial, compared with others designated as agricultural (crop or swine production) or forested. These authors also speculated that the greater diversity of serotypes observed is related to the larger variety of hosts in urban watersheds.

The significance of avian carriers of Salmonella has been recognized with increasing frequency, although much remains unknown regarding the link between these sources and public health risk (Hall and Saito 2008). Although birds can carry strains that are less commonly associated with human health, many birds are carriers of Salm. Typhimurium (Dolejska et al. 2009; Gorski et al. 2011). The urban creek in this study is impacted by many types of birds, including nonmigratory geese and ducks that roost throughout the year. In addition, these waters are very turbid and allow limited light penetration which might enhance the survival of these bacteria once they have entered the water.

Salmonella Thompson was the most common serotype obtained from the urban stream. This serotype was not observed in the other two tributaries, which implies that the source of this serotype is absent in rural/agricultural tributaries. This serotype is less common in humans and domestic animals compared with Salm. Typhimurium and Salm. Heidelberg; however, it is often reported among the top ten serotypes in the Canadian population (Health Canada 2003; Government of Canada 2006). The prevalence of Salm. Thompson in wildlife is unknown; however, this serotype is more commonly reported in poultry (Guerin et al. 2005; Government of Canada 2006), and therefore, it might be associated with birds or waterfowl in the urban tributary.

Several other serotypes obtained in this tributary suggest avian sources, including Salm. Agona, Salm. Montevideo, Salmonella Senftenberg and Salmonella Litchfield (Nesse et al. 2005; Hall and Saito 2008). In addition, monophasic serotype 4, 5, 12:b:-, which was only observed in the urban stream, was isolated from duck faecal samples collected adjacent to it. Following PFGE all serotype 4, 5, 12:b:– isolates taken from faecal samples of ducks showed indistinguishable XbaI patterns, which were also identical to many waterborne isolates obtained within the same tributary, suggesting a direct link between wildlife and water quality in these urban waters.

Birds, such as ring-billed gulls, mallard ducks and Canada geese, adapt well to urban environments. These sources could explain why Salmonella was commonly observed year round in the urban stream, especially in the cold winter months, compared with agricultural/rural streams. Regardless of the sources of Salmonella obtained from the urban creek, all of the serotypes obtained can pose a health risk to the human population.

In the rural/agricultural watersheds, Salm. Typhimurium was the predominant serotype in one stream (CAN) and Salm. Heidelberg in the other (CON), although both serotypes were observed in all three tributaries. The predominance of these two serotypes in these tributaries is likely a result of the intensive agricultural activities and the high levels of manure production in these watersheds (Dorner et al. 2004).

Over half of the Salm. Typhimurium isolates in this study were defined as variants (Var. 5-) that lack the O:5 antigen. Although the epidemiology of Salm. Typhimurium Var. 5- is poorly understood, studies have shown that Var. 5- is common in farm animals, particularly in cattle and swine (Frech et al. 2003; Zhao et al. 2005; Government of Canada 2006). Its higher occurrence in agricultural animals might explain why these variant strains were more predominantly obtained from the agricultural tributaries (n = 12), over the urban tributary (n = 5) in this study. Following PFGE, Var. 5- strains of Salm. Typhimurium did not cluster together, but were scattered throughout the dendogram among other nonvariant strains. Var. 5- and nonvariant strains commonly showed identical XbaI patterns and tended to cluster around the PT classification. This is consistent with the findings of Zhao et al. (2005) who reported identical PFGE patterns among Var. 5- and nonvariant isolates from a diverse group of farm animals in the USA. In our waterborne isolates, it was common to see Var. 5- and nonvariant isolates with identical XbaI patterns from different tributaries. This may indicate that limited genetic variability exists between Var. 5- and nonvariant strains, regardless of the original source of the strain.

Limited genetic heterogeneity was observed among waterborne isolates of Salm. Heidelberg following digestion with XbaI (≥86·9% similarity observed between all isolates). Many isolates demonstrated indistinguishable restriction patterns within and between PT, which demonstrates a close genetic relatedness among PT. In addition, several isolates of the same PT, which varied spatially and/or temporally, demonstrated indistinguishable XbaI patterns, which might suggest that clones are circulating within and between tributaries. The use of an additional enzyme for PFGE could provide additional information on the clonality of Salm. Heidelberg in these waters.

Waterborne Salm. Typhimurium isolates were more genetically diverse compared with Salm. Heidelberg, with overall similarities in XbaI PFGE patterns >60% (Fig. 5). Compared to Salm. Heidelberg, Salm. Typhimurium isolates revealed a higher degree of genetic heterogeneity between PT. Similar to Salm. Heidelberg, many isolates of the same PT shared indistinguishable PFGE patterns. High levels of genetic relatedness and clonality within specific Salm. Typhimurium PT appears to be consistent with other studies that examined isolates from a range of samples, including human clinical samples (Baggesen et al. 2000; Guerri et al. 2004), faecal samples from farm animals (Baggesen et al. 2000) and shellfish (Martinez-Urtaza et al. 2004a).

Based on the PFGE result, PT 104 was genetically distinct from other PT of Salm. Typhimurium. Integrated prophage elements and the Salmonella genomic island 1 (SGI-1), which encodes drug resistance genes, are thought to contribute to the genetic diversity in this PT (Guerri et al. 2004; Cooke et al. 2008).

Similar to this study, Martinez-Urtaza et al. (2004a) demonstrated distinct clustering around PT following PFGE of Salm. Typhimurium from shellfish taken from marine waters off the coast of Spain, with many isolates of the same PT exhibiting identical XbaI profiles. They suggested that a common source might be responsible for the contamination as isolates with indistinguishable PFGE patterns were obtained at the same location over many different sampling dates. In the current study, strains with identical patterns were observed within the same tributary on different sampling dates, which supports the possibility that a common source exists. However, these identical patterns were also observed between strains obtained from different tributaries, which may suggest that there is a common source that is highly mobile in nature (e.g. waterfowl). It is also important to consider that the method of analysis in this study (i.e. PFGE) does not have sufficient strain discrimination power to measure smaller genetic differences in the population, in which case concrete assumptions regarding common sources within and among tributaries may not be possible.

Several studies have observed positive relationships between precipitation and occurrence of pathogens in the aquatic environment (e.g. Baudart et al. 2000; Simental and Martinez-Urtaza 2008; Setti et al. 2009; Wilkes et al. 2009); however, no significant relationship was observed in our study. While many authors have acknowledged that the relationship is complex and not always strong (e.g. Martinez-Urtaza et al. 2004b; Wilkes et al. 2009; Gorski et al. 2011), it is likely that the method of sample collection used in our study was a restricting factor for this comparison. Regardless of the explanation, the similarity between the occurrence levels and serotype diversity among various flow conditions speaks to the ubiquitous nature of Salmonella in all three of these tributaries. The occurrence of Salmonella in these waters under various flow conditions also indicates that protection of source waters through the implementation of management practices should not be restricted to anthropogenic activities that only generate run-off following precipitation events.

Similar frequencies of occurrences were observed over all seasons, including winter. This is in contrast to other studies that have reported low levels or no occurrence of waterborne Salmonella in the winter or colder months (e.g. Arvanitidou et al. 2005; Wilkes et al. 2009). While no seasonal trends in occurrence of Salmonella were observed in our study, there was a marked temporal diversity of serotypes with season. The greatest diversity was seen in the summer and fall months when there were twice as many serotypes observed compared with the winter and spring months. These findings are consistent with results reported by Haley et al. (2009) who observed a greater diversity of serotypes in surface waters in the summer months and lowest in the winter months in Georgia, USA.

A distinct seasonality was also observed in serotypes of human and veterinary health significance – Salm. Typhimurium and Salm. Heidelberg – both showed higher frequencies in the summer and spring months, respectively. A similar seasonal trend was also noted by Martinez-Urtaza et al. (2004b) in marine waters in Spain, where the incidence of Salm. Typhimurium in water was significantly higher in the summer. Although data are not available on the seasonality of these serotypes in the human population in the area, these trends correspond to the peaks in overall Salmonella infections in the region (PHAC 2007a). Serotyping data at the national level, however, reveal that both Salm. Heidelberg and Salm. Typhimurium can show strong seasonal trends in infection in the Canadian population (Ford et al. 2003; Health Canada 2003), with maximum frequency of infection generally between June and September. Although it is difficult to ascertain a relationship between environmental exposure and resulting illness, these data indicate that water may play a role in the overall distribution of these serotypes among various hosts at specific times of the year.

The seasonal maxima for both Salm. Typhimurium and Salm. Heidelberg was evident only in the rural/agricultural streams. These peaks were not observed in the urban stream, which showed a consistent occurrence of these serotypes throughout the year. The difference might reflect changes in farm practices in the cooler months or the seasonal shedding of certain serotypes in domestic farm animals, whereas the consistent occurrence in the urban stream, particularly in the winter months, may reflect the continuous low level shedding of these serotypes by wildlife and/or outputs from other nonpoint sources. Farm animals are less apt to be a source of waterborne Salmonella in the colder months as pasturing is limited and manure is generally stored. Also run-off in these months is lower and the movement of Salmonella to the watercourse is minimal. Although manure spreading is less in the winter months, it still occurs in the CAN and CON watersheds.

It is possible that survival differs among serotypes, particularly those of human health significance. As found by Baudart et al. (2000) and Haley et al. (2009), serotypes of human health significance were either absent or found at a lower prevalence in water at various times of the year. In our study, Salm. Typhimurium only represented 8% of the isolates obtained in winter and dropped to 3% in the spring. Salmonella Heidelberg was absent from winter samples; however, it was observed in the spring as the second most common serotype. However, differences were not observed among the survival rates of environmental isolates of Salm. Typhimurium, Salm. Heidelberg and Salm. Montevideo at 4°C. Comparable survival between serotypes of human health significance and Salm. Montevideo, which was the most common serotype obtained in the winter months, indicates that low water temperature alone is not responsible for the lower frequency of Salm. Typhimurium and Salm. Heidelberg at colder times of the year. This supports the notion that less shedding of these specific serotypes in farm animals or the reduced transport to the watercourse is most likely the explanation for the lower observed frequency in the colder months.

Survival of Salm. Agona was significantly greater at 4°C compared with the other serotypes tested, in contrast to the results for strains of human health significance. Although these survival studies were conducted under controlled conditions and do not reflect the actual survival rates in the aquatic environment where many other environmental factors can influence survival, including predation, sunlight intensity and water chemistry (Winfield and Groisman 2003; USEPA 2009c), this finding suggests that survival differences may exist between certain Salmonella serotypes in the aquatic environment at lower temperatures. Further experiments with other isolates of Salm. Agona, as well as other serotypes obtained from water, would help to determine the extent of these differences.

The detection of waterborne pathogens, including Salmonella, remains a challenging task. However, to represent the predominant serotypes circulating within the aquatic system, it is critical to maximize the number of isolates obtained. Obtaining more isolates in this study demonstrated that serotypes that predominate in humans and farm animals, Salm. Typhimurium and Salm. Heidelberg, were also the most common in water, a similarity rarely reported. Although this does not indicate a direct correlation between exposure and illness, these data indicate that water may play an important role in the transmission of these serotypes among hosts and may be a first step in understanding whether an environmental exposure source exists for Salmonella.

The ubiquitous nature of Salmonella in water and the presence of serotypes of human and veterinary health significance, as well as the long-term survival of Salmonella, suggest that environmental exposure through consumption or contact with contaminated water is plausible. As our understanding of the occurrence of waterborne Salmonella increases, as well as the implementation of risk based assessment advances, our knowledge of the overall contribution of waterborne pathogens to enteric disease will expand. Surveillance and monitoring of Salmonella in environmental sources, such as water, is a step towards improving the understanding of the epidemiology of salmonellosis, as well as developing control strategies for the future.

Acknowledgements

Sincere thanks to the staff at the Public Health Agency of Canada (OIÉ Reference Laboratory for Salmonellosis, Guelph Ontario; C.A. Muckle, K. Rahn, L. Cole, B. Wilkie, K. Mistry, A. Perets) for serotyping and phagetyping isolates and to the Ontario Ministry of the Environment (S. Weir, A. Abbey) for the use of the PFGE system.

Conflict of interest

No conflict of interest is declared by the authors.

Ancillary