Comparative genotypic and pathogenic examination of Campylobacter concisus isolates from diarrheic and non-diarrheic humans

Background Campylobacter concisus is an emerging enteric pathogen, yet it is commonly isolated from feces and the oral cavities of healthy individuals. This genetically complex species is comprised of several distinct genomospecies which may vary in pathogenic potential. Results We compared pathogenic and genotypic properties of C. concisus fecal isolates from diarrheic and healthy humans residing in the same geographic region. Analysis of amplified fragment length polymorphism (AFLP) profiles delineated two main clusters. Isolates assigned to AFLP cluster 1 belonged to genomospecies A (based on genomospecies-specific differences in the 23S rRNA gene) and were predominantly isolated from healthy individuals. This cluster also contained a reference oral strain. Isolates assigned to this cluster induced greater expression of epithelial IL-8 mRNA and more frequently contained genes coding for the zonnula occludins toxin and the S-layer RTX. Furthermore, isolates from healthy individuals induced greater apoptotic DNA fragmentation and increased metabolic activity than those from diarrheic individuals, and isolates assigned to genomospecies A (of which the majority were from healthy individuals) exhibited higher haemolytic activity compared to genomospecies B isolates. In contrast, AFLP cluster 2 was predominated by isolates belonging to genomospecies B and those from diarrheic individuals. Isolates from this cluster displayed greater mean epithelial invasion and translocation than cluster 1 isolates. Conclusion Two main genetically distinct clusters (i.e., genomospecies) were identified among C. concisus fecal isolates from healthy and diarrheic individuals. Strains within these clusters differed with respect to clinical presentation and pathogenic properties, supporting the hypothesis that pathogenic potential varies between genomospecies. ALFP cluster 2 isolates were predominantly from diarrheic patients, and exhibited higher levels of epithelial invasion and translocation, consistent with known roles for these factors in diarrhoeal disease. Conversely, isolates from healthy humans and AFLP cluster 1 or genomospecies A (which were predominantly isolated from healthy humans) exhibited increased haemolytic ability, apoptotic DNA fragmentation, IL-8 induction, and/or carriage of toxin genes. Given that this cluster contains an oral reference strain, it is possible that some of the AFLP cluster 1 isolates are periodontal pathogens and may cause disease, albeit via a different mechanism than those from AFLP cluster 2.


Background
Campylobacter species are one of the most common causes of human enteritis in North America (Centers for Disease Control and Prevention, U.S. Department of Agriculture, and Food and Drug Administration Collaborating Sites Foodborne Disease Active Survey Network [FoodNet]; Public Health Agency of Canada website, http://dsol-smed.phac-aspc.gc.ca/dsol-smed/ ndis/diseases/camp_e.html). While Campylobacter jejuni and Campylobacter coli are the most commonly isolated species, studies have also implicated 'cryptic' species within the genus, such as Campylobacter concisus, as causal agents of acute enteritis [1][2][3][4]. Compared to C. jejuni, C. concisus is fastidious to isolate as it is often sensitive to selective antimicrobial agents commonly-used in conventional isolation media, and generally requires a hydrogen-enriched atmosphere and a prolonged incubation period for growth [5]. As such, it is rarely cultured by standard isolation methods employed by many diagnostic facilities. Although knowledge of its clinical importance is limited, C. concisus has been cited as an emerging human pathogen [5,6].
Campylobacter concisus was originally isolated from periodontal lesions [7]. However, its pathogenic role in oral cavity infections remains uncertain, since it can also be isolated from healthy gingiva [8]. Additionally, C. concisus has been isolated from the feces of diarrheic patients [1][2][3][4], often in the absence of known pathogens. However, the bacterium is also frequently isolated from feces of asymptomatic patients, which has lead to the conclusion that it may be part of the normal intestinal microbiota [9,10]. Some evidence indicates that C. concisus may be an opportunistic pathogen. For example, Engberg et al. [9] observed that C. concisus was predominantly isolated from pediatric, elderly, and immunocompromised patients, in contrast to C. jejuni and C. coli which are typically isolated from diarrheic patients of all ages. Consequently because of its association with diarrheic, healthy, and immunocompromised patients, the specific role of C. concisus as a primary intestinal pathogen has yet to be firmly established.
Campylobacter concisus is a heterogeneous species complex comprised of several phenotypically indistinguishable but genetically distinct taxa ("genomospecies"). Numerous methods can be used to genetically separate the genomospecies, including PCR analysis of the 23S rRNA gene [11] and cluster analysis of amplified fragment length polymorphism (AFLP) or random amplified polymorphic DNA (RAPD) profiles [1,2]. Based on these typing methods, at least two main C. concisus genomospecies have been identified [1,2,4,11].
Differences in pathogenicity amongst distinct genomospecies of some bacterial taxa [12,13] support the notion that certain C. concisus genomospecies may be more likely than others to cause intestinal disease. While an early study by Van Etterijck et al. [10] concluded that C. concisus was not pathogenic given similar isolation rates from diarrheic and healthy children, genetic diversity of the isolates with respect to clinical presentation was not considered. A more recent study showed that isolates from healthy individuals were genetically distinct from those of diarrheal origin; however, differences in epithelial cytotoxicity between the two groups were not evident [2]. Additionally, cluster analysis of diarrheic isolate AFLP profiles delineated two main C. concisus genomospecies (designated genomospecies 1 and 2), which where characterized by type strains of oral and diarrheal origin, respectively [1]. Genomospecies 2 isolates were more frequently isolated from the stool of patients presenting with diarrhea in which no other pathogens were found, and bloody diarrhea was associated only with genomospecies 2 isolates. While these studies suggest that distinct C. concisus genomospecies may vary in their pathogenic ability, this has yet to be empirically examined.
Our understanding of Campylobacter pathogenesis is based primarily on C. jejuni. Its small, spiral shape coupled with flagella-mediated motility, allow C. jejuni to penetrate intestinal mucus [14] where it can then adhere to and invade intestinal epithelial cells. This bacterium can also translocate across the intestinal epithelium via a paracellular mechanism involving disruption of epithelial tight junctions [15,16] or via a lipid raft-mediated transcellular mechanism [17]. C. jejuni also causes cellular cytotoxicity through the production of various toxins; cytolethal distending toxin (CDT) is a well-characterized toxin produced by most strains. The cytolethal distending toxin blocks cell proliferation in the G2/M phase resulting in cellular distension leading to the induction of apoptotic cell dealth [18]. This bacterium also induces intestinal epithelial secretion of interleukin-8 (IL-8), a pro-inflammatory chemoattractant that recruits neutrophils to the site of infection [19]. Cytolethal distending toxin-like activity has been reported for a majority of clinical C. concisus isolates [2], however the role that this toxin plays in pathogenesis is unknown. A membranebound haemolytic phospholipase is also produced by most clinical C. concisus isolates [20]. In addition, C. concisus genes coding for zonnula occludins toxin (zot) and a surface-layer protein belonging to the RTX (repeats in the structural toxins) family (S-layer RTX) have been recently identified [21]. Zonnula occludins toxin was first recognized as a toxin of Vibrio cholera, and disrupts the integrity of the intestinal epithelial barrier by targeting tight junctions [22]. S-layer RTX is a pore-forming toxin that is also found in Campylobacter rectus [23], and toxins within this family are recognized as important virulence factors [24].
The present study examines the hypothesis that the two main C. concisus genomospecies exhibit differences in pathogenicity. To address this hypothesis, we compared genotypic and pathogenic properties of C. concisus fecal isolates from diarrheic and asymptomatic ("healthy") humans. Specifically, genotypes of isolates were compared by AFLP analysis and a genomospeciesspecific 23S rRNA gene PCR assay. Numerous pathogenic properties were also assessed including: (i) intestinal epithelial adherence, invasion, and translocation; (ii) ability to disrupt epithelial permeability, cause apoptotic DNA fragmentation, affect metabolic activity, and induce IL-8; hemolytic and cytotoxic activities; and (iii) carriage of toxin genes encoding CDT, ZOT, and Slayer RTX proteins.

Genotypes
Sequence analysis to confirm the identities of the clinical isolates indicated >99% 16S rRNA gene sequence similarity (near full-length) between the type strain C. concisus LMG7788 and all of the clinical isolates (GenBank accession numbers are listed in Table 1). Based on the genomospecies-specific PCR assay of the 23S rRNA gene [11], six and 12 of the 22 clinical C. concisus isolates were assigned to genomospecies A and B, respectively (Table 1). Three isolates generated PCR products for both genomospecies A and B primer sets (designated "A/B"), and one isolate did not amplify with either primer set (designated "X"). The type strain, LMG7788, was assigned to genomospecies A, consistent with previous observations [2]. Campylobacter concisus-specific PCR of the cpn60 gene was strongly positive for 21 isolates including the type strain and weakly positive for two isolates. Weak PCR products were likely due to mismatching of the PCR primers with their target gene (due to DNA sequence divergence), resulting in inefficient PCR amplification.
AFLP analysis indicated considerable genetic variability existed among the C. concisus isolates ( Figure 1). Reproducibility between duplicate independent analyses of each isolate was 93.1 ± 3.6% (mean ± SD; Additional file 1). The isolates clustered into two phylotypes distinguished from each other at the 34% similarity level. All isolates assigned to AFLP cluster 1 belonged to genomospecies A and included the type strain plus five isolates that were obtained from healthy (n = 4) and diarrheic (n = 1) humans. Of the seventeen isolates assigned to AFLP cluster 2, 94% (16/17) were isolated from diarrheic stools, and 71% belonged to genomospecies B (n = 12) while 17% belonged to genomospecies A/B (n = 3), 6% belonged to genomospecies A (n = 1), and one isolate was unassigned.
Epithelial translocation was not different for any of the C. concisus isolates relative to C. jejuni 81-176 ( Table 2). The mean translocation of C. concisus genomospecies B isolates was greater than isolates belonging to genomospecies A (4.46 ± 0.20 log 10 CFU/ml versus 3.99 ± 0.09 log 10 CFU/ml, respectively; P = 0.048), and isolates assigned to AFLP cluster 2 relative to cluster 1 (4.58 ± 0.16 log 10 CFU/ml versus 3.99 ± 0.11 log 10 CFU/ml, respectively; P = 0.03). Mean translocation between isolates from diarrheic and healthy humans did not differ (4.48 ± 0.15 log 10 CFU/ml versus 4.17 ± 0.25 log 10 CFU/ml, respecively; P = 0.26). Translocation and invasion were weakly correlated (R 2 = 0.27; P = 0.059). There were no differences in the initial or final TER between Campylobacter-treated monolayers and the sterile broth control, and the final TER of all monolayers remained greater than 1000 Ω × cm 2 (Additional file 2) indicating that translocation occurred across an intact epithelial barrier. Also, FITC-dextran permeability did not differ for monolayers treated with C. concisus relative to the sterile broth control (Additional file 2).

Toxin genes
The CDT B gene was not detected in any of the C. concisus isolates, but was present in C. jejuni 81-176 (Additional file 3). The zot gene was detected in 80% of C. concisus isolates from healthy humans (i.e., four of five isolates), 22% of isolates from diarrheic humans (i.e., two of nine isolates), and the type strain. The S-layer RTX gene was present in C. concisus CHRB3287 and CHRB2004, although amplification was weak for the latter isolate. The zot and S-layer RTX genes were not detected in C. jejuni 81-176.

Discussion
The observed high level of genetic diversity amongst the isolates of C. concisus is in agreement with previous studies [1,2,10], and highlights the complex nature of this species. Cluster analysis of AFLP profiles indicated that the isolates examined in the current study comprised at least two distinct clusters. Similarly, Aabenhus et al. [1] denoted four AFLP clusters among 62 C. concisus isolates of which the majority (n = 56) were assigned to one of two main clusters. Results of PCR assays targeting the 23S rRNA and cpn60 genes largely corresponded with the AFLP grouping, and lend support to the suggested genetic relationship between the isolates. As C. concisus is a common inhabitant of the oral cavity, it is to be expected that it may be isolated from both healthy and diarrheic individuals. Examining isolates from healthy individuals, it was observed that the majority of isolates belonged to genomospecies A and their AFLP profiles clustered together (AFLP cluster 1) along with the type stain of oral origin. This AFLP cluster also included one genomospecies A isolate (CHRB 1609) from a diarrheic individual. Further studies are needed to determine whether this group of isolates represents inhabitants of the oral cavity that have survived gastrointestinal transit or whether they are intestinalassociated.
The majority (94%) of isolates from diarrheic individuals were assigned to ALFP cluster 2. Among these isolates, 71% were assigned to genomospecies B, while only 11% of diarrheic isolates belonged to genomospecies A. Engberg et al. [2] reported a similar predominance of genomospecies B isolates among diarrheic fecal isolates, of which 33% and 67% were assigned to genomospecies A and B, respectively. Likewise, Aabenhus et al. [1] reported that 34% and 53% of fecal isolates from diarrheic patients were assigned to genomospecies A and B, respectively. Our observation that isolates from genomospecies B were exclusively obtained from diarrheic individuals suggests a potential role for these isolates in intestinal disease. A comparative molecular examination of strains belonging to genomospecies A and B may shed light on their respective pathogenic potential.
Examination of the pathogenic properties amongst C. concisus isolates determined that epithelial invasion and translocation were higher for isolates assigned to AFLP cluster 2 (of which 94% were from diarrheic individuals). Additionally, epithelial translocation was higher for isolates belonging to genomospecies B (of which all isolates were from diarrheic individuals). This is of potential clinical relevance as invasiveness and translocation ability are the only factors definitively correlated with enteritis in C. jejuni-infected patients [26] and are likely associated with inflammatory responses and occasional bacteraemia observed with C. concisus infections [27]. To our knowledge, this is the first study to report differences in pathogenicity between the two main C. concisus genomospecies, further supporting the likelihood that isolates belonging to AFLP cluster 2/genomospecies B incite enteritis in humans. All of the clinical C. concisus isolates examined in the current study caused hemolysis of sheep erythrocytes, consistent with previous observations of hemolytic phospholipase activity in all C. concisus genomospecies A and B isolates from diarrheic children [20]. As such, hemolytic activity appears to be a general characteristic of this species. Hemolysins are involved in pathogenesis and host colonization in other taxa [28], thus it was an unexpected observation that C. concisus genomospecies A isolates exhibited greater mean hemolysis than isolates belonging to genomospecies B. We also observed that hemolytic activity by C. concisus was inversely correlated with epithelial adherence and invasion. Staphylococcus aureus exhibits a similar inverse correlation that is attributed to interference of its α-hemolysin with epithelial β1-integrins that mediate host-cell interactions [29]. Moreover, lower amounts of α-hemolysin are produced by invasive S. aureus isolates from endocarditis patients compared to less-invasive isolates from open wounds [30]. Whether C. concisus hemolysin also interferes with epithelial receptors that promote adherence and invasion is unknown, and additional studies are warranted.
Another unexpected finding was that isolates from healthy individuals induced greater mean epithelial DNA fragmentation and metabolic activity compared to those from diarrheic individuals, and these variables were positively correlated. DNA fragmentation is used as an indicator of cell death. The two primary modes of cell death, namely apoptosis and necrosis, can be distinguished on the basis of physiologic features. DNA fragmentation can be present in both processes; however, during apoptosis, cell membranes typically remain intact, whereas during necrosis, cellular integrity is rapidly disrupted leading to the release of cytoplasmic contents (including lactate dehydrogenase) into the surrounding environment ("cytotoxicity"). Based on this definition, four isolates from healthy individuals (CHRB2004, CHRB3235, CHRB3287, and CHRB3290) and two isolates from diarrheic humans (CHRB2370 and CHRB3152) induced cell death consistent with apoptosis. While apoptosis can be mediated via CDT [18], it is unlikely that CDT was involved in the apoptosis given that we did not detect the CDTB gene in any of the strains included in our study. Furthermore, we did not find any genes with similar sequence to the CDTB gene using a BLAST search of the published C. concisus genome (NCBI accession number NC_009802), indicating that other factors (i.e. opposed to the CDT) may be responsible. The role that Campylobacter-induced epithelial cell death plays in pathogenesis is currently poorly understood; hence, the clinical significance of these findings for C. concisus remains to be determined.
Metabolic activity can be measured using the MTT assay in which metabolically active epithelial cells reduce a yellow tetrazolium salt (MTT) to purple formazan crystals that can be spectrophotometrically quantified. All of the isolates that we examined, except one isolate that caused epithelial sloughing (CHRB6), induced higher MTT values (> 130%) than the control, indicating that epithelial metabolic activity is increased by C. concisus. Some clinical strains of C. jejuni have also been reported to cause similar increases in epithelial MTT values [31]. Given the short incubation period for the MTT assay, we conclude that the increased values most likely reflect an increase in metabolic activity due to cellular stress rather than an increase in epithelial cell numbers due to proliferation. The observed correlation between metabolic activity and DNA fragmentation may be a consequence of the increased energy demands required to sustain the apoptotic process (i.e., apoptotic DNA fragmentation is an ATP-dependent process [32]). The chemokine, IL-8 is a major mediator of inflammation. In the current study, all C. concisus isolates induced transcription of IL-8 in epithelial monolayers (> 2-fold) as has been previously reported for C. jejuni [19] and C. concisus [33]. Campylobacter jejuni induces epithelial IL-8 secretion by at least two independent mechanisms, one of which requires invasion and the other that is CDT-dependent [19,34]. We observed that induction of IL-8 transcription by C. concisus was not correlated with invasion. Man et al. also recently showed that three C. concisus strains stimulated production of IL-8 in intestinal epithelial irrespective of their invasive ability [33]. Thus in contrast to C. jejuni, it appears that factors other than invasion or CDT (which appears to be lacking in this species) are responsible for the up-regulation of IL-8 incited by C. concisus. The observation that expression of IL-8 mRNA was greater in epithelial cells treated with isolates from AFLP cluster 1 compared to isolates from cluster 2 was unexpected and suggests that these isolates may have pathogenic potential.
We identified genes encoding S-layer RTX and the zonnula occludins toxin in some of the isolates, confirming initial reports of these toxin genes in C. concisus [21]. Surprisingly, the zot gene was more prevalent in isolates from healthy (80%) compared to diarrheic (22%) humans. The clinical significance of this observation remains to be determined. Although Zot has been shown to disrupt epithelial tight junctions, we did not observe any changes in permeability or TER of epithelial monolayers throughout the 3 h incubation period for any of the isolates. This is contrary to the observation of Man et al., that C. concisus caused increased epithelial permeability, decreased TER, and loss of membraneassociated zonnula occludens and occludin in epithelial monolayers [33]. Possible reasons for this discrepancy include variation in methodology between the two studies (i.e., Man

Conclusion
In summary, two main genomospecies were identified among fecal isolates of C. concisus from healthy and diarrheic individuals. The genomospecies differed with respect to clinical presentation and pathogenic properties, which is consistent with the hypothesis that certain genomospecies have different pathogenic potential. AFLP cluster 2 was predominated by isolates belonging to genomospecies B and those from diarrheic individuals. Isolates from this cluster displayed higher mean epithelial invasion and translocation than cluster 1 isolates, consistent with a potential role in inflammatory diarrhea and occasional bacteraemia. In contrast, isolates assigned to AFLP cluster 1 belonged to genomospecies A and were predominantly (but not strictly) isolated from healthy individuals. Isolates assigned to this cluster induced greater expression of epithelial IL-8 mRNA and more frequently contained genes coding for the zonnula occludins toxin and the S-layer RTX. Furthermore, isolates from healthy individuals induced greater apoptotic DNA fragmentation and increased metabolic activity than did isolates from diarrheic individuals, and isolates assigned to genomospecies A (of which the majority were from healthy individuals) exhibited higher haemolytic activity compared to genomospecies B isolates. This suggests that isolates from this cluster may also cause disease, albeit via different mechanisms than isolates from AFLP cluster 2. AFLP cluster 1 contains a reference strain isolated from the oral cavity, thus it is possible that this cluster contains isolates that are primarily periodontal pathogens.
While in vitro pathogenicity assessments are informative, they do not necessarily correspond with the ability of an isolate to cause disease in vivo. Clearly, further studies, particularly in vivo, are needed to confirm that these genetically distinct groups of C. concisus indeed differ in their ability to cause intestinal disease. In this regard, comparative genomic and pathogenicity examinations using animal models have been initiated.

Bacterial isolates and growth conditions
A total of 23 C. concisus isolates recovered from different individuals were used in this study (Table 1). These included five isolates recovered from the stools of healthy volunteers (i.e., individuals that exhibited no recent symptoms of enteritis), 17 isolates obtained from stools of diarrheic humans, and the type strain of C. concisus isolated from the oral cavity of a healthy human (LMG7788; = CCUG 13144; = ATCC 33237). Isolates were collected from people residing in the Chinook Health Region of Southwestern Alberta, Canada. These isolates were originally collected as part of a larger study [35]. Scientific and ethics approval for stool collection was obtained from the Regional Ethics Committee of the former CHR and from the University of Lethbridge Human Subject Research Committee. Campylobacter jejuni 81-167 [36] was used as a positive pathogen control for all pathogenicity assays. In addition, the non-pathogenic Escherichia coli HB101 was used as a negative pathogen control for measuring epithelial IL-8 expression in response to the presence of bacteria. Isolates were stored at -80°C in Columbia broth (Difco, Detroit, MI) containing 40% glycerol. With the excepiton of E. coli which was grown in an aerobic enviornment, inocula of C. concisus for cell culture assays were prepared by growing isolates for 14-16 h in Columbia broth (37°C, 100 rpm) in a microaerobic atmosphere (consisting of 5% O 2 , 10% CO 2 , 30% H 2 and balance nitrogen).

16S rRNA gene sequence
Genomic DNA was extracted using a DNAeasy Tissue kit (Qiagen Inc., Mississauga, ON) according to the manufacture's instructions. The 16S rRNA gene was PCR amplified using the primers UNI27F and UNI1492R [37] (Table 5) and the resultant product was used as template for sequencing. A BigDye Terminator kit (Applied Biosystems, Foster City, CA) along with universal primers (Table 5) were used for sequencing the near full-length 16s rRNA gene according to the manufacturer's instructions. Sequence reactions were separated with an ABI 3130 automated DNA sequencer (Applied Biosystems). Sequences were analyzed using Sequencher software (Gene Codes, Ann Arbor, MI) and compared directly with the NCBI non-redundant nucleotide database using BLASTN.

Amplified fragment length polymorphism analysis
Campylobacter concisus isolates were genotyped using the AFLP protocol described by Kokotovic and On [38]. Briefly, genomic DNA (125 ng) was digested with Cps6I (10 U) in Y+/Tango Buffer (MBI) for 1 h at 37°C. BglII (10 U) was then added, and digestion was continued for one additional hour. Restriction site-specific adaptors ( Table 5) were then ligated to the digested fragments for 2 h at room temperature. PCR amplification of the ligation mixture (diluted 10-fold) was carried out using primers BGL2F-0 and CSP6I-A (Table 5) for 35 cycles with an annealing temperature of 54°C. The final products were separated with an ABI 3130 automated DNA sequencer (Applied Biosystems). To analyze AFLP profiles, fragments ranging from 75 to 500 bp and the 500LIZ Genescan molecular mass standard were imported and compared using the BioNumerics 4.01 software (Applied Maths, Kortrijk, Belgium). Relationship of AFLP profiles ("curves") were inferred by use of the Pearson-product-moment correlation coefficient (applying 2% optimization) and clustered by the unweighted pair group with mathematical average (UPGMA) method. To ensure reproducibility, AFLP analysis was conducted twice for isolate, and one representative of each AFLP profile was used for cluster analysis.
PCR for 23S rRNA, cpn60, CDT B, S-layer RTX, and zot genes Primers for PCR are listed in Table 5. PCR amplification of the 23S rRNA gene was conducted according to the method of Bastyns et al. [11], except that the two reverse primers (CON1 and CON2) were used independently rather than as a mixture. Isolates amplifying with either MUC1/CON1 or MUC1/CON2 primers were assigned to genomospecies A or B, respectively. Campylobacter concisus-specific nested-PCR amplification of the chaperonin gene (cpn60) was conducted using the primers Ccon-cpn_66f and Ccon_cpn_423r for 25 cycles with an annealing temperature of 53°C [35]. The resultant PCR product was used as a template for a second round of PCR with the nested primers Ccon_cpn_72f and Ccon_cpn_342r for 30 cycles with an annealing temperature of 53°C. PCR for the CDT B gene was conducted using the primers P5 and P6 (Table 5) for 30 cycles with an annealing temperature of 65°C. PCR for the S-layer RTX gene was conducted using the primers FCCC13826_1838 and RFCCC13826_1838 (Table 5) for 30 cycles with an annealing temperature of 58°C. PCR for the zot gene was conducted using the primers FCCC13826_2075 and RFCCC13826_2075 for 30 cycles with an annealing temperature of 56°C.

Intestinal epithelial cell culture and inoculation
T84 human colonic epithelial cells (passages 7 to 20; ATCC, Manassas, VA) were grown in DMEM/Ham F-12 plus 10% fetal bovine serum, 200 mM L-glutamine, 100 U/ml penicillin, 100 μg/ml streptomycin, 80 μg/ml tylosin (all from Sigma, Oakville, ON), and incubated at 37°C and 5% CO 2 . For cell culture assays, confluent T84 monolayers were washed twice and media was replaced with antibiotic-free DMEM/Ham F12. Monolayers were inoculated with sterile Columbia broth (= control) or Campylobacter to achieve a multiplicity of infection of 100 CFU per epithelial cell, and incubated for 3 h at 37°C . Due to the intensive nature of the assays for assessment of pathogenic potential (i.e., adherence, invasion, translocation, hemolytic ability, and cytotoxicity), representative isolates of C. concisus from diarrheic and healthy humans were examined for pathogenicity (n = 5 from AFLP cluster 1, n = 9 from AFLP cluster 2).

Adherence and invasion
T84 enterocyte monolayers were grown in 24-well plates and inoculated as described above. Following incubation, monolayers were washed three times with PBS. To assess adherence, monolayers were lysed with 0.1% Triton X-100 in PBS for 10 min at room temperature on an orbital shaker. Following lysis, bacteria were enumerated by plating ten-fold serial dilutions onto Karmali agar (Oxoid, Nepean, ON). Invasion was determined using a gentamicin protection assay. After incubation, monolayers were washed three times with PBS. Monolayers were then incubated for 2 h with fresh tissue culture medium containing gentamicin (500 μg/ml) to kill extracellular bacteria as previously described [39]. Following incubation, monolayers were washed, lysed and bacteria were enumerated as for the adherence assay.
A preliminary experiment was conducted to ensure that a bactericidal concentration of gentamicin was used for the invasion assay.
Translocation and epithelial permeability T84 cells were seeded onto Transwell filters at 4 × 10 5 cells/filter (5 μm pore size, 1.13 cm 2 ; Costar, Corning Inc. Corning, NY) and cultured as described above. Transepithelial electrical resistance (TER) was monitored with an electrovoltohmeter (World Precision Instruments, Sarasota, FL), and monolayers were used at confluence (TER >1000 Ω × cm 2 ). Monolayers were inoculated as described above. Following incubation, the basolateral medium was serially diluted, spread onto Karmali agar and incubated microaerobically at 37°C. Permeability was assayed as described previously [25]. Briefly, monolayers were washed with Ringer's solution, and a 3 kDa FITC-dextran probe (500 μl, 100 mM in Ringer's solution; Molecular Probes, Eugene, OR) was added to the apical compartment and 1 ml of Ringer's solution was also added to the basal compartment. Following incubation for 3 h at 37°C, samples were collected from the basal compartment and absorbance at 485 nm was measured.

Hemolysis
Hemolysis of sheep erythrocytes was measured as previously described [20]. In brief, C. concisus cells cultured in Columbia broth as described above were centrifuged (8000 × g, 3 min) and cell pellets were washed with sterile PBS, suspended in PBS to 1 × 10 9 CFU/ml, and then serially diluted 2-fold in PBS. Equal volumes (100 μl) of cell suspension and sheep erythrocytes (2% vol/vol in PBS) were mixed in a U-bottom 96-well plate. The plate was then incubated at 37°C under microaerobic conditions for 18 h. A comparative negative control (without bacteria) was also incubated under similar conditions. A positive control for total hemolysis (100%) was performed by replacing the same volume of bacterial cell suspension with distilled water. After incubation, the tubes were centrifuged at 1000 × g for 5 min, and the OD 490 of the supernatants for the 1/3 dilution were measured. Data were reported as the percent total hemolysis of sheep erythrocytes (compared to the positive control).
DNA fragmentation, cytotoxicity, and metabolic activity T84 monolayers were grown in 24-well plates and inoculated as described above. Control monolayers were also treated with camptothecin (4 μM), hydrogen peroxide (H 2 O 2 , 0.5 mM), or sterile broth. Following incubation, DNA fragmentation was measured using a Cellular DNA Fragmentation ELISA kit (Roche Applied Science, Laval, QC) according to the manufacturer's protocol. Lactate dehydrogenase released into the surrounding tissue culture was measured using a Cytotoxicity Detection kit (Roche) according to the manufacturer's protocol. Metabolic activity (i.e. MTT assay) was measured using a Cell Proliferation Kit I (Roche) according to the manufacturer's protocol, except that gentamicin (500 μg/ml) was incorporated into the MTT solution.
Interleukin-8 real-time quantitative PCR T84 monolayers were grown in six-well plates and inoculated with C. concisus and C. jejuni as described above. In addition, monolayers were inoculated at an MOI of 100 with E. coli HB101. Following incubation, the culture medium was removed and replaced with RNAlater (3 ml/ well; Qiagen), and cells were stored at 4°C until processed for RNA extraction (< 1 week). Total RNA was isolated using the RNeasy mini kit (Qiagen), according to the manufacturer's protocol. RNA was reverse transcribed using a QuantiTect reverse transcription kit (Qiagen) according to the manufacturer's protocol. PCR was conducted using an Mx3005P Stratagene thermocycler (Stratagene, Cedar Creek, TX). All PCR reactions were carried out in 20 μl volumes and contained 1X QuantiTect SYBR Green PCR Master Mix (Qiagen), forward and reversed primers (0.5 μM each; Table 5) and 2 μL of RT reaction. Amplification conditions were 1 cycle at 95°C for 15 min followed by 40 cycles of 94°C for 15 s, 60°C for 30 s, and 72°C for 30 s. Melting curve analysis was conducted over a range of 55 to 95°C to assess specificity of amplification. Interleukin-8 expression was normalized to the housekeeper gene, C1orf33, and fold changes in expression relative to the sterile-broth control was calculated using the 2 -ΔΔCT method.
the C. concisus-specific cpn60 primers. We also thank Kathaleen House for isolating and conducting the initial characterization of C. concisus isolates. We wish to thank the anonymous reviewers of this manuscript for their insightful and constructive comments. This work was supported by a Peer Review Grant from Agriculture and Agri-Food Canada (Growing Forward initiative).
Authors' contributions LDK participated in the design of the study, performed experiments, conducted data analysis, and drafted the manuscript. GDI participated in the design of the study and edited the manuscript. All authors approved the final manuscript.