Recovery of thermophilic Campylobacter by three sampling methods from river sites in Northeast Georgia, USA, and their antimicrobial resistance genes.

Abstract

Sixteen sites in the watershed of the South Fork of the Broad River (SFBR) in Northeastern Georgia, USA, were sampled in two seasons to detect Campylobacter. Sites were classified as mostly influenced by forest, pasture, wastewater pollution control plants (WPC) or mixed-use. Sampling was repeated in the late spring and late fall for two years for a total of 126 samples. Free-catch water and sediment grab samples were taken at each site; Moore’s swabs were placed for up to three days at most sites. A total of 56 isolates of thermophilic Campylobacter were recovered. Thirteen samplings were positive by two or three methods and 26 samplings were positive by only one method; once by Moore’s swab only and 25 times by free-catch water only. Campylobacter was detected at 58% of cattle pasture sites, 30% of forested sites and 81% of WPC sites. Twenty-one of the isolates carried antimicrobial resistance genes, mostly bla-OXA-61. Free-catch water samples were more efficient than Moore’s swabs or sediment samples for recovery of Campylobacter, which was more likely to be detected in streams near cattle pastures and human communities than in forested land.

INTRODUCTION:

Campylobacter jejuni and C. coli are enteropathogens that cause about 1.3 million human infections a year in the United States (Scallan et al. 2011). The major source of infection is agreed to be foodborne with poultry products carrying the majority of the blame (Wagenaar et al. 2013). Flowing water could serve as a carrier for migration of Campylobacter, but this hypothesis needs to be further tested. Exposure to or drinking environmental water is associated with an increased odds ratio for human infection (Kapperud et al. 2003; Carrique-Mas et al. 2005; MacDonald et al. 2015). However, methods for studying the role of fresh-water aquatic environments in transmitting Campylobacter to humans by direct exposure or infect the livestock that are in the human food chain have not been fully evaluated.

Source attribution for bacterial infections is commonly performed by linking genetic types found in human infections with the types found in potential sources. For instance, analyses of multi-locus sequence typed (MLST) C. jejuni have re-affirmed the association of poultry with human infection (Mughini Gras et al. 2012). The association of types of Campylobacter from environmental water with human disease has been weak (Mughini Gras et al. 2012). However, natural water environments are complex and the sampling method may be biased in ways that affect linkage analyses.

The research reported here was intended to determine if isolation of Campylobacter spp. differed by sampling method of a watershed in Northeastern Georgia, USA. Three sampling methods were tested: filtration of free-catch water, sediment, and Moore’s swab (Moore 1948; Barrett et al. 1980). Each method surveyed different microhabitats with the potential for recovering different Campylobacter populations. Population genetic analyses of Campylobacter recovered by these methods were used to determine sequence type (ST) and presence of antimicrobial resistance gene markers to determine if different microhabitats with different populations of Campylobacter were present in the surveyed watershed. The alternative hypothesis is that there is free migration between the tested microhabitats yielding an undifferentiated admixture in the aquatic ecosystem.

RESULTS AND DISCUSSION:

A total of 59 Campylobacter isolates were recovered (Figure 1). Free-catch water yielded 40 positive samples out of 126 samples (31%); sediment yielded 9 positive samples from 126 samples (7%); and Moore’s swab yielded 10 positive samples from 109 samples (9%) (water significantly greater than sediment or Moore’s swab, chi square P < 0.01; sediment not significantly different from Moore’s swab, chi square > 0.05). Four isolates were C. lari and 54 were C. jejuni. Seventeen sequence types (ST) were found (C. lari were not typed), ten were found only once and six were previously unknown types. ST21 (detected 4 times) is the most common type in the PubMLST database; ST61 (detected 11 times) is the seventh most common in the database and is often associated with cattle (some of the pasture sites had cattle present). Free-catch water was the sample most frequently Campylobacter positive. In a prior study it was shown that as few as 10 cells of Campylobacter per liter of water could be detected by the procedure with free-catch water (Meinersmann et al. 2008). Fifty-six percent of the samplings adjacent to WPC were positive by at least one recovery method, 33% of pasture samplings were positive and 22% of forest samplings were culture positive. Campylobacter was detected at least once from all sites except for two forest sites.

Figure 1:

A. Isolates: The dates of isolation are noted on the left, followed by the three methods of collection (free-catch water, sediment grab sample, or Moore’s swab). Sites are listed across the top characterizing major land use at each site (SFBR-60 is mixed use), and are roughly in order of upstream on left to downstream on right with vertical lines separating samples on separate tributaries (see Figure 2). The numbers are the Campylobacter jejuni sequence types (ST) assigned to the isolate in the sample. Types found only once have a white background. Types found multiple times have a unique color background for each occurrence. Black hash background indicates samples not collected. Numbers in parentheses identify isolates bearing antimicrobial resistance genes (1 = blaOXA-61, 2 = tet(O), 3 = aph). B. Clonal complexes (CC) of each of the occurring STs (na = no applicable CC).

blaOXA-61 was found in 21 isolates. One of these isolates also carried a gene for tetracycline and another carried genes for tetracycline and aminoglycoside resistance as well as the blaOXA-61 (Figure 1).

The pair-wise Fst values for different populations are shown in Tables 1 and 2. Fst values that have a significant difference from zero indicate different populations that may be due to isolation or selection. The populations stratified by collection method (Table 1) were indistinguishable. Stratification of the populations by type of site (Table 2) had small but significant differentiation of forest and pasture sites but no significant differentiation of WPC sites from either forest or pasture sites.

Table 1:

Pair-wise Fst values (P value [significance]) of populations by method of recovery.

	Water	Sediment	Swab
Water	0
Sediment	0.0242 (0.22523)	0
Swab	0.00792 (0.32432)	−0.07191 (0.90090)	0

Table 2:

Pair-wise Fst values (P value [significance]) of populations by type of site collected.

	Forest	Pasture	WPC
Forest	0
Pasture	0.14569 (0.0)	0
WCPC	0.01071 (0.32432)	0.03021 (0.19820)	0

Campylobacter has been reported to grow well in biofilms (Indikova et al. 2015). Therefore we expected sediment and/or Moore’s swabs, which may have surface biofilms, to be enriched for Campylobacter and consequently have a higher recovery of the organism. However, this was not the case and recovery was highest in the free-catch water samples. This can be explained as either differences in the sensitivity of the sampling methods or differences in the micro-habitats that select for populations of different recoverability. This contrasts with Fernandez et al. (2003) who reported greater recovery with Moore’s swab than by filtration. However, we used a different filtration method that was demonstrated to be very sensitive (as low as 10 cells per liter) and applicable to turbid water (Meinersmann et al. 2008).

The Fst test for population differentiation indicated that the different methods did not select for differing populations. Thus free-catch water was probably more sensitive and this could be simply a factor of sample volume. Approximately one liter of water was sampled by filtration while for the sediment and Moore’s swab about 10 ml of water-equivalent was sampled. Any biofilm enrichment in the latter two samples was less than enough to give equivalent recovery from water. Results of the Fst analysis imply that Campylobacter strains are free to migrate between water and the habitats created by Moore’s swab or sediment.

Fst analysis did show differentiation of the pasture site population from the forest site population. It is noteworthy that a substantial number of isolates from pasture sites were ST-61. This is a type that is considered to be a specialist that is found at a higher prevalence in ruminants (Sheppard et al. 2014). Some of the pastures along the sampling sites were inhabited by cattle. Bradshaw et al. (2016) found that there was a positive correlation of isolation of Campylobacter with Escherichia coli and microbial source tracking markers for cattle (Rum-2-Bac and CowM3) in the water and sediment at pasture sites.

Most types found in this study are not represented in the PubMLST database in great enough numbers to accurately conclude if they are specialists or generalist. However, ST267, ST50 and ST21 are clearly generalist having been found in a wide variety of hosts (Gripp et al. 2011). Some types represented in the database with fewer than 25 isolates had high correlation with environmental water (ST3889 [83%], ST1224 [100%], ST2524 [100%]). ST475 had never been found in environmental water before despite being represented by 154 isolates in the database.

All sample sites were in lotic environments; therefore, it was expected that the location of Campylobacter sub-types could change over time. However, mixing along the river was insufficient to homogenize the populations. This could be due to strains dying out or becoming too dilute to detect as they move downstream. Campylobacter are considered to be fragile and short-lived in the environment (Indikova et al. 2015). The current data may indicate that there are static and long-term sources of introduction of specific types into the tributaries of the South Fork of the Broad River. Campylobacter could be introduced directly from nearby animals (e.g. cattle, wild birds) or sheltered habitats may exist that constantly leach Campylobacter into the river. Alternatively, it is possible that there are types of Campylobacter that have an enhanced ability (adaptation) to survive in water. It may be expected that types that shared this phenotype would be closely related but the diversity of the recovered isolates was large. In the PubMLST/Campylobacter database there were 404 types that were only found in water and 354 (87%) of these were only found once, supporting the conclusion of high diversity of isolates in environmental waters.

The number of positive WPC samples may have been insufficient to show a statistical differentiation by Fst.

Unlike Khan et al. (2014) in Canada, we did not find any C. coli isolates. Swine are not as common along the South Fork of the Broad River in Georgia as along some of the rivers that Khan surveyed. Overall, our study reinforces the conclusion of Khan et al. (2014) that Campylobacter found in the rivers reflect the neighboring animal agricultural activities. However, we also found isolates that have been associated with various sorts of wild birds (goose ST692, ST699; other wild bird ST692, ST699, ST267, ST50, ST21).

The pasture water samples were significantly more likely to carry resistance than were forest water samples (P = 0.049) but there were not enough isolates in any other category to establish significant differences. blaOXA-61 was the most commonly found antimicrobial resistance gene detected and was especially prevalent in isolates from pasture sites. Resistance to ampicillin can be conferred in Campylobacter by blaOXA-61 (Devi et al. 2019). The blaOXA-61 was always found on the chromosome. The ST-61 isolate from pasture 2 on 5-June-2013 carried a plasmid that bore both tet(O) and aph. The plasmid was less than 0.05% different from C. jejuni subsp. jejuni 00–2544 plasmid (Genbank CP006710) isolated in Canada in the year 2000 (Clark et al. 2014). The ST-50 isolate from forest 1 on 21-May-2013 had sequences that matched about 80% of the same plasmid among three contigs, but lacked the region bearing aph. Ampicillin is not indicated to treat Campylobacter infections but the antimicrobial is commonly used in beef cattle for a number of infections.

In conclusion, the sampling methods reported here did not distinguish different populations of Campylobacter from river water. The use of free-catch water was the most sensitive method and no evidence of concentration of Campylobacter was noted by use of sediment or Moore’s swabs. Segregation of populations was evident based on the neighboring land use suggesting that recoverable Campylobacter do not migrate far via river water and that more than one source of introduction into the river water exists.

MATERIALS AND METHODS:

The surveyed watershed has been described in detail by Bradshaw et al. (2016) including a land-use map and the physiochemical parameters that existed at the time of sampling. Briefly, the South Fork of the Broad River watershed contains about 583 km² with mixed land use, approximately 55% forest and 42% agriculture. Cattle-on-pasture use the largest area with approximately 17 head per km² and poultry in modern grow-out houses have the greatest density at about 10,678 broilers per km² (Anon. 2012). Wastewater pollution control plants (WPC) were just upstream of two sample sites. Each sampling site was classified by the predominant land use in the immediate riparian zone: forest, pasture, or downstream of a WPC (see Figure 1). Figure 2 shows the direction of water flow between the sample sites to demonstrate how water may be shared by the sites.

Figure 2:

Upstream/downstream relationships of sites. Arrows indicate direction of water flow.

Collection dates are indicated in Figure 1. Sample dates were changed when necessary to avoid any rain fall event of 0.64 cm or more. Three collection methods were used: 1. One liter of free-flowing water was captured in a sterile bottle, the water was filtered and the filter put into 25 ml Bolton’s enrichment broth (BEB) as described by Meinersmann et al. (2008); 2. Approximately 100 g of sediment from the top of the creek bed was gathered into a sterile tube and 10 g was added to 25 ml BEB; 3. Moore’s swabs (non-medicated cotton tampons) were attached to weights by a nylon monofilament line approximately 10 cm long and placed midstream three to five days before collection. The recovered swab was placed into a sterile Whirl-pak bag (Nasco, Fort Atkinson, WI, USA), manually agitated to release attached cells and express water, and 1 ml was aseptically placed into 10 ml BEB. All the BEB tubes had the covering airspace replaced with Campy-gas (2.5% O₂, 7% H₂, 10% CO₂ and the balance N₂) and then were incubated at 42° C for 24 hour. Aliquots were streaked onto Campy-Cefex plates (Stern et al. 1992) that were incubated at 42°C in the atmosphere described above for up to 48 h and one colony was harvested.

Isolates presumptively identified as Campylobacter were subcultured on Brucella agar and the growth was collected for DNA purification using a kit following the manufacturer’s instructions (UltraClean® microbial DNA isolation Kit, Mo Bio Laboratories Inc., Carlsbad, CA, USA). Sequencing libraries were prepared using the Nextera XT sample preparation kit (Illumina, San Diego, CA, USA). The genomic DNA sequence of each isolate was determined using the Illumina MiSeq platform with a 2 × 250 paired end run according to manufactures instructions (Illumina, San Diego, CA). Raw sequence read files for each isolate were mapped to a reference sequence of the atpA gene of C. jejuni using Geneious Mapper (Biomatters Ltd., Aukland, NZ) and the consensus of the mapping was used to identify the species by comparing to a local database of Campylobacteraceae atpA gene sequences (Miller et al. 2014). The 16S rRNA sequences were also extracted to confirm putative species identification. Raw sequence reads were further mapped to multi locus sequence type (MLST) loci described by Dingle et al. (2001) and the strict consensus for each mapping was trimmed to match the reference and then submitted to the website pubmlst.org/campylobacter/ (Jolley and Maiden 2010) to obtain the allelic identifiers and the sequence type (ST) of the isolate.

Raw sequence reads were submitted to ResFinder (https://cge.cbs.dtu.dk/services/ResFinder/) to determine antimicrobial resistance genes carried by the isolates (Zankari et al. 2012).

The program Arlequine ver. 3.5.2 (Excoffier et al. 2005) was used to perform Fst analysis of population differentiation by site type or collection.

SIGNIFICANCE AND IMPACT OF THE STUDY:

The role of environmental water in transmitting Campylobacter was investigated and methods for recovery of the organism were compared. The sequence types of recovered Campylobacter correlated with adjacent land use without regard to the method used to isolate the organisms. Sequence types and antimicrobial resistance genes associated with cattle were most prevalent near pastures. Even though types were recurrent at a given site, types appeared to be lost or replaced as the water flowed downstream.