Genomic and phenotypic characteristics of Swedish C. jejuni water isolates

Campylobacter jejuni is the most common cause of bacterial gastroenteritis. Major reservoirs are warm-blooded animals, poultry in particular, but Campylobacter can also be transmitted via water. In this paper, we have taken a closer look at the biology and potential virulence of C. jejuni water isolates. Seven C. jejuni isolates from incoming surface water at water plants in Sweden were characterized with whole genome sequencing and phenotypical testing. Multi locus sequence typing analysis revealed that these isolates belonged to groups known to include both common (ST48CC) and uncommon (ST1275CC, ST683, ST793 and ST8853) human pathogens. Further genomic characterization revealed that these isolates had potential for arsenic resistance (due to presence of arsB gene in all isolates), an anaerobic dimethyl sulfoxide oxidoreductase (in three isolates) and lacked the MarR-type transcriptional regulator gene rrpB (in all but one isolate) earlier shown to be involved in better survival under oxidative and aerobic stress. As putative virulence factors were concerned, there were differences between the water isolates in the presence of genes coding for cytolethal distending toxin (cdtABC), Type VI secretion system and sialylated LOS, as well as in biofilm formation. However, all isolates were motile and could adhere to and invade the human HT-29 colon cancer cell line in vitro and induce IL-8 secretion suggesting potential to infect humans. This is, to the best of our knowledge, the first study where C. jejuni water isolates have been characterized using whole genome sequencing and phenotypical assays. We found differences and shared traits among the isolates but also potential to infect humans.


Introduction
Campylobacter is the most common cause of bacterial gastroenteritis in most parts of the world and C. jejuni is the predominant pathogen [1]. Warm-blooded animals, such as poultry, pigs and ruminants are major reservoirs for Campylobacter and the bacteria are thought to be PLOS  mainly transmitted via handling and eating raw or undercooked meat [2,3]. Campylobacter can also be transmitted via environmental pathways, such as water, and waterborne outbreaks of Campylobacter are not uncommon [4][5][6][7]. In the Nordic countries, groundwater for drinking water is usually not treated and reports from Finland describe Campylobacter outbreaks where heavy rain has led to contamination of groundwater wells [6,7]. It has also been shown that cattle drinking untreated water from lakes or private water supplies are more likely to test positive for Campylobacter [8,9]. Studies have shown C. jejuni water survival time at a low temperature to vary between two weeks and four months [10][11][12][13] and strains isolated from different sources have been noted to show different survival potentials [13][14][15]. These inter-strain differences have been suggested to be caused by variations in genetic content [12], however, whole genome sequencing has rarely been used to study waterborne C. jejuni. The population structure of C. jejuni is highly diverse and described using Multi Locus Sequence Typing (MLST). To date there are almost 9000 STs, divided into several clonal complexes (CCs; http://pubmlst.org/campylobacter/, last accessed 22 May 2017). The CCs of C. jejuni can be considered to consist of generalists when able to colonize various hosts or specialists if host-or niche-specific [16].
C. jejuni are fastidious bacteria needing a microaerobic environment for survival and growth, which optimally occurs at 42˚C. However, C. jejuni have been isolated from environmental sources where neither the atmosphere nor the temperature have been optimal [11][12][13]. For survival in the environment, bacteria can use strategies such as biofilm formation and motility [17,18]. By forming biofilm, bacteria can provide themselves with a protective environment under harsh conditions and also allow for dispersal of the bacteria. Motility has been shown to be important for biofilm formation and the ability to form biofilm has been noted to vary between different lineages of C. jejuni [19][20][21].
To study the potential of Campylobacter to cause human infections, in vitro infection models with human epithelial cell lines are frequently used to determine adherence, invasion and downstream cellular responses. Several studies have demonstrated differences between C. jejuni strains in the ability to adhere to and invade cell lines in vitro and it has been suggested that C. jejuni strains causing more severe symptoms also adhere to and invade cells in vitro at a higher grade than strains causing milder symptoms [22]. However, other studies have shown that both C. jejuni strains causing mild and those causing severe human infections adhere to and invade cell lines in vitro to a similar extent [23]. C. jejuni have also been shown to be capable of inducing downstream cell responses such as interleukin 8 (IL-8) secretion [24,25], an early signal for acute inflammatory response to bacterial infection [24]. The IL-8 response caused by C. jejuni has been suggested to be correlated to the capability of the strains to adhere and invade [24].
In this study, we characterized C. jejuni water isolates using a genotypic and phenotypic approach in order to determine features important for water survival and to reveal the potential of the isolates to infect human cells. For these purposes, whole genome sequencing, phenotypic analyses for motility and biofilm formation and an in vitro infection model were used.

Bacterial isolates
Seven Campylobacter isolates collected by the National Food Agency in 2000 from raw (incoming) surface water at water plants in Sweden were characterized in this study ( Table 1). The isolates were identified as C. jejuni using MALDI-TOF Biotyping (Microflex, Bruker, Billerica, Massachusetts, US). In addition, the earlier characterized clinical strains C. coli 76339 [26] and C. jejuni 76577 [19] and the C. jejuni reference strains NCTC 11168 and 81-176 were used as experimental controls.

Genomics
The isolates were cultured for 24-48h on blood agar (Columbia agar plates supplemented with 5% horse blood; Oxoid, Basingstoke, UK) in a microaerobic atmosphere (Campygen, Oxoid) at 42˚C. The DNA was extracted from bacterial cultures using MagNa Pure Compact Nucleic Acid isolation kit I (Roche, Penzberg, Germany) according to manufacturer´s protocol version 12. The isolates were whole genome sequenced with an Illumina HiSeq 2500 platform with a 2 x 300 paired end run. For assembly of the reads into contigs, Geneious (version 8.1.5.) [27] with the Mira plugin (version 1.0.1.) was used and merged contigs were assembled with Geneious. The assembled sequences were annotated by RAST [28] and the translated coding sequences (tCDS) were extracted. To determine orthologues clusters, a reciprocal blastp query was performed using an E-value of 10 −5 and the OrthAgogue [29] and MCL-edge tools [30].
Plasmids were assembled from the raw sequence reads with Geneious de novo assembler version 8.1.9 [27].
PCR. The lipooligosaccharide (LOS) locus class was determined with PCR as previously described [25] using primers described by Parker et al [31]. For the isolates that remained unassigned to a LOS locus class after PCR typing a manual search for previously sequenced LOS locus classes [32] was perfomed.
Phenotyping. All phenotypic experiments described below were performed in duplicate and repeated at least three times for each strain. Biofilm. The isolates were tested for biofilm formation as previously described by Revez et al. [19] with minor modifications. A bacterial suspension with a concentration of 4 x 10 6 CFU/ml was prepared by harvesting bacteria from blood agar and resuspending in PBS. Of the bacterial suspension, 10 μl was added to Brucella broth (Becton, Dickinson and Company, Franklin Lakes, New Jersey, US) in glass tubes and incubated at 37˚C in a microaerobic environment for 48 h. After carefully removing the broth, the tubes were stained with 1% crystal violet solution. The isolates were identified as positive if a stained band was seen at the air-liquid interface. The C. jejuni strain 76577, previously identified as positive for biofilm formation [19], was used as a positive control and broth without addition of bacteria was used as a negative control.
Motility. The motility of the isolates was tested according to Szymanski et al. [34], with minor modifications. The isolates were cultured in Brucella broth at 42˚C in a microaerobic atmosphere for 17-18 h. The bacteria were centrifuged at 8000 x g for 5 min, the supernatant removed and the pellets resuspended in PBS to a concentration of approximately 10 8 CFU/ml. Of the bacterial suspension, 5 μl was stabbed into a Brucella soft agar plate (0.4%) and swarming zones were measured after incubation at 42˚C in a microaerobic atmosphere for 48 h. The clinical C. coli clade 3 strain 76339 was included for comparison.
Adhesion/invasion assay. The HT-29 human colon cancer cell line (ECACC 91072201) was maintained in RPMI 1640 media (Gibco by life technologies, Carlsbad, California, US) supplemented with 2 mM glutamine (Swedish National Veterinary Institute, Uppsala, Sweden), 10% Fetal bovine serum (FBS, Gibco by Life Technologies), 100 U/ml penicillin and 100 μg/ml streptomycin (Swedish National Veterinary Institute). Overnight bacterial cultures were centrifuged, diluted in cell culture media and added to low passage HT-29 cells grown in RPMI 1640 supplemented with 1% FBS at a MOI of 100. At indicated time points, media was collected for IL-8 ELISA. For adhesion/invasion assays, cells were washed four times in PBS to remove non-adhered bacteria and lysed in 20 mM Tris, pH 7.5, 150 mM NaCl and 0.15% Triton X-100. The lysate was diluted 10 and 100 times for qPCR analysis of the 16S rRNA gene together with 10 000 times-diluted starting cultures to determine the adhesion/invasion percentage.
IL-8 ELISA. The IL-8 levels in the media were measured using the IL-8 ELISA Kit (Thermo Fisher Scientific, Waltham, Massachusetts, US) according to the manufacturer's instructions. Media was diluted four to ten times prior to the assay. A standard of known concentration (included in the kit) was used to assess variations between infections. Results are expressed as fold increase over uninfected (mock) cells.

Genomics
Seven C. jejuni isolates, collected from incoming water at water surface plants in Sweden, were whole genome sequenced ( Table 1). Analysis of MLST identified ST for six of the seven isolates; ST48 (three isolates), ST637 (one isolate), ST683 (one isolate) and ST793 (one isolate), respectively. One isolate was submitted to PubMLST and assigned to the new ST8853. The ST48 isolates belonged to the ST48CC, the ST637 to ST1275CC, and the remaining three isolates were unassigned to any CC at the time of the analysis (Table 1).
To identify traits involved in virulence and survival of C. jejuni water isolates, the genomes were clustered into orthologue groups and annotated in RAST. Annotations of groups not shared by all C. jejuni water isolates, but shared by isolates from more than one ST, were manually analyzed (S1 Table). The genomes were also manually searched for genes known to be involved in specific traits of interest, such as virulence and motility.
A previously described [35] arsenic resistance gene cluster consisting of four genes (arsP, arsR, arsC and acr3) was found in the isolate VA33 (ST683) and in the three ST48 isolates. However, the gene coding for the membrane transporter arsP [36] was fragmented in all the four isolates and the gene coding for the efflux pump protein Acr3 was disrupted by a premature stop codon in the ST48 isolates ( Table 2). The arsP gene was also identified in VA12 (ST1275CC, Table 2). The arsB gene, which is also coding for an efflux pump for arsenic [37], but not included in the gene cluster mentioned above, was detected among all of the water isolates studied ( Table 2).
Intact cdtABC genes coding for cytolethal distending toxin (CDT) were found in the C. jejuni ST48 water isolates and in the isolate VA33 (ST683, Table 2), which was also verified with PCR (data not shown). The putative virulence genes ciaB, pldA, cadF and ceuE were identified in the genomes of all isolates, with only minor sequence variations in the predicted amino acid sequences (data not shown).
Most of the major flagellar motility genes, such as fla, flg, fli, motA and motB, were present in all C. jejuni water isolates. For VA12 (ST1275CC), VA33 (ST683), VA48 (ST793) and VA49 (ST8853), deletions and insertions were detected in the nucleotide sequences of the flaA, flaB, flaG and flgL genes when compared to the corresponding gene sequences in the C. jejuni reference strain NCTC 11168. However, these particular deletions and insertions did not result in any frameshifts and thus, all ORFs seemed to be intact. The flgD, fliD and fliK genes were detected in all C. jejuni water isolates, but variations in the sequences resulting in disrupted ORFs were detected in some of the isolates (Table 3). Furthermore, the pseE gene involved in flagellar modification was identified in all of the water isolates, but an intact ORF was only identified in five of the isolates (Table 3). Two genes implicated in biofilm formation, csrA and peb4, were present in all water isolates and showed almost no differences in the gene and translated amino acid sequences. In four of the isolates, VA1 (ST48), VA25 (ST48), VA52 (ST48) and VA33), the luxS gene involved in quorum sensing and shown to be important for biofilm formation [39] was detected.
A plasmid of 20801 bp was identified in three of the C. jejuni water isolates (ST48CC) and annotated against the plasmid sequence of C. jejuni F38011 (S2 Table, NCBI Acc. No. CP00 6851), which gave the best hit (99.4%) when the sequence was blasted against all available entries in the NCBI database. Among the annotations an oxidoreductase (7-alpha-hydroxysteroid dehydrogenase) and a betalactamase were identified as well as genes possibly involved in virulence, such as prevent-host-death protein and ABC transporter permease.
LOS locus classes were assigned to five of the seven C. jejuni water isolates using PCR (data not shown). The three ST48 isolates were assigned to LOS locus class B2, VA12 (ST1275CC) to LOS locus class C and VA33 (ST683) to LOS locus class F, J or S whereas the two isolates VA48 (ST793) and VA49 (ST8853) were untypeable ( Table 2). The isolates VA48 and VA49 remained unassigned even after a manual search for previously sequenced LOS locus classes.

Phenotyping
The ability of the water C. jejuni isolates to form biofilm was tested in three independent experiments and the isolates belonging to ST48 (ST48CC), ST637 (ST1275CC) and ST683 (CC ua) were able to form biofilm (Fig 1).
The motility assay showed that all isolates were motile, however, VA12 (ST1275CC) displayed the lowest motility (Fig 1). When the motility results were compared with the genome sizes, it was shown that isolates with larger genomes were less motile and a higher motility was detected among the isolates with smaller genomes (R 2 = 0.61; Fig 2).
To assess the potential virulence of the water C. jejuni isolates and see whether they would be able to infect human cells, an in vitro infection model with HT-29 colon cancer cells was used. All isolates were able to infect HT-29 cells (Fig 3A), and adhered bacteria could be detected already at 30 minutes post-infection (data not shown). At 1 h, mean levels ranged between 0.02 and 0.12% for the C. jejuni water isolates (Fig 3A). The level of adherence/invasion for the isolates belonging to ST48 varied among the isolates assigned to this ST. At 8 h, the levels had reached 1% for some strains (data not shown), possibly due to bacterial growth. The levels of adhered/invaded bacteria of the water isolates were consistently lower than those of the C. jejuni NCTC 11168 reference strain at all time points.
To further assess the potential ability of the water isolates to cause inflammation, IL-8 levels in the cell media were measured. IL-8 started to be detected at 2 hours post-infection and  increased during the course of the infection (data not shown). In general, all water isolates induced lower levels of IL-8 than NCTC 11168 and no significant differences could be seen between the water isolates ( Fig 3B). In conclusion, no clear correlation could be seen between the levels of adhered/invaded bacteria and induced IL-8 levels.

Discussion
In this study, whole genome sequencing, phenotypical tests and in vitro infection results were used to characterize C. jejuni isolates collected from raw surface water at water plants in Sweden. The aim was to detect unique traits among these isolates to gain a better understanding of environmental survival of C. jejuni and the potential of the water isolates to cause human infection. Our results showed that the seven C. jejuni water isolates were genetically diverse and belonged to ST48CC (three isolates), ST1275CC (one isolate) or were unassigned to any CC (three isolates). ST48CC is one of the most common CCs among human C. jejuni isolates but also contains isolates from various other sources, including the environment [40,41]. In contrast, ST1275CC isolates have been reported to be commonly found in environmental waters and wild birds [40,42,43]. Furthermore, a search in the Campylobacter PubMLST database showed that ST1275CC isolates had been detected from wild birds and environmental waters, but only rarely from human samples. Moreover, ST683 and ST793, unassigned to any CC, had previously been found in samples from environmental waters and wild birds only, according to PubMLST database. Thus, according to the MLST, our water isolates included both common and uncommon human pathogens. Whole genome sequencing revealed the gene arsB, coding for an efflux transporter for inorganic arsenic, in all our C. jejuni water isolates. In four of our C. jejuni water isolates (ST683 and ST48) an arsenic resistance gene cluster associated with high level resistance to inorganic arsenic [35] and with a fragmented arsP gene was identified. In addition, the arsP gene coding for an efflux transporter specific for organic arsenic [36] was found in VA12 (ST1275CC). Toxic arsenic exists in the natural environment and in order to survive in its presence bacteria have developed resistance mechanisms [35]. Organoarsenic compounds have also previously been used as growth promoters in poultry production and as antimicrobials to control coccidiosis caused by parasites [44]. However, in this study, we did not test the resistance of our isolates to arsenic compounds and therefore the role of the genes described remains to be studied.
For Campylobacter to survive in water, it would be beneficial to handle oxidative and aerobic stress. The MarR-type transcriptional regulator gene rrpB has previously been shown to be involved in C. jejuni oxidative and aerobic stress response [45]. Wildtype strains lacking rrpB have been shown to survive better under oxidative and aerobic stress than strains with rrpB [46]. The rrpB has also been shown to be more uncommon in C. jejuni strains associated with water and wildlife than other strains associated with bovines [46]. In our study, rrpB was only present in isolate VA12 (ST1275, Table 2), which suggests that this particular isolate might be more sensitive to oxidative and aerobic stress but otherwise water isolates would be more adapted to handle oxidative and aerobic stress. However, further testing is needed to confirm this finding.
DMSO exists naturally in freshwater and marine environments, in rainwater and in the atmosphere [47]. C. jejuni have been shown to be capable of utilizing DMSO as an electron acceptor under low-oxygen conditions [48]. Here, an anaerobic DMSO-reductase coded by dmsABC was found in three of our C. jejuni water isolates (ST1275CC, ST683 and ST793). This DMSO-reductase has been shown to be common in C. jejuni isolates from humans and chicken [49] but uncommon in bovine isolates [49].
The majority of our isolates were able to form biofilm under the conditions tested (Fig 1). By forming biofilm, bacteria can create a protective environment for survival under harsh conditions [18]. As motility has been shown to be important for the formation of biofilm [19][20][21], the results from the biofilm assay were compared to the motility results. However, we could not see any correlations between the results. Although variations were seen in the motility genes between the C. jejuni water isolates, none of them could fully explain the differences in motility and biofilm formation. PseE mutations have earlier been shown to result in a nonmotile phenotype [50], but here we could not see any motility loss for the isolates that seemed to lack an intact pseE gene. However, an intact pseE was only identified in the biofilm positive isolates suggesting that pseE-dependent flagellar modification is important for biofilm formation. In addition to motility, quorum sensing has been shown to affect the biofilm formation and studies have shown a decreased biofilm formation and motility in mutants lacking luxS [39,51]. Here, luxS was detected in four out of five water isolates that formed biofilm, but was lacking in the biofilm negative isolates. The swarming zones of the isolates were also compared with the genome sizes and the results showed that the isolate with the smallest genome had the highest motility and the isolate with the largest genome had the smallest swarming zone (Fig  2). These results may suggest that the C. jejuni that are less motile either contain more genetic material to compensate for the disadvantage or that the larger (heavier) genome directly and negatively impact the ability of the bacteria to move. Together, these findings suggest that the motility and biofilm phenotypes are multigenic and attributed to variations in a number of different genes.
The potential virulence of the C. jejuni water isolates was assessed in an in vitro infection model. The ability of the isolates to adhere to/invade and induce an immune response in a human colon cancer cell line was analyzed. All of the water isolates were able to both adhere/ invade the cultured cells and to induce an IL-8 response, which suggests a potential to infect and cause inflammation in humans. However, as compared to the C. jejuni NCTC 11168 reference strain, the levels of adherence/invasion and IL-8 response were considerably lower. CDT is a toxin that has been shown to induce an IL-8 response [52] and in four of the C. jejuni water isolates, belonging to ST48 and ST683, the intact gene locus cdtABC was identified. The level of adherence/invasion for the isolates with intact cdtABC (ST48) varied, however, all the ST48 isolates induced the same level of IL-8 response (Fig 3). In one study the in vitro IL-8 response was not affected by CDT when the C. jejuni were able to adhere/invade [52], which might explain why isolates lacking CDT were able to induce similar levels of IL-8 as ST48 isolates. Together, these results suggest that the IL-8 response can be induced by different bacterial virulence mechanisms. As all isolates had intact genes for the putative virulence genes ciaB, pldA, cadF and ceuE, a more thorough analysis of expression levels of these genes might give an indication of the specific virulence strategy for each isolate.
A gene cluster coding for T6SS, a structure through which bacteria can deliver effector proteins into adjacent prokaryotic and eukaryotic cells [53], was identified in three of the C. jejuni water isolates ( Table 2). The T6SS may be advantageous for the bacteria in several ways, e.g. to outcompete or to interact with other bacteria in the gut or in the environment [53]. However, as the T6SS gene cluster was not present in all of the water isolates, this structure does not seem to be critical for the survival in water.
The ST48 water isolates were assigned to LOS locus class B2 in accordance with a previous report [54]. C. jejuni with potential to sialylate LOS (LOS locus A, B and C) have been proposed to cause more severe symptoms [55] and also to adhere and invade cultured intestinal epithelial cell lines at a higher level than isolates without this characteristic [56,57]. In this study, no differences in the in vitro infections were seen between the isolates with or without the potential to sialylate their LOS. This is in line with our previous study [25] where C. jejuni with potential to sialylate LOS did not induce a higher IL-8 response than those without the ability to sialylate their LOS.
In conclusion, this is, to the best of our knowledge, the first study where C. jejuni isolated from water have been characterized using both whole genome sequencing and phenotypical assays. Although our collection of water isolates was limited, we found both characteristics common for all water isolates but also interesting genotypical and phenotypical differences between the isolates that might influence their potential to survive in the environment and transmit to human hosts.
Supporting information S1