Campylobacter is a leading cause of bacterial food-borne diarrhoeal disease worldwide, with symptoms ranging from mild to serious infections, which can result in permanent neurological damage; especially in elderly people (Silva et al., 2011). It is frequently found in poultry and chicken is an important source of Campylobacter infection. In the United Kingdom alone, Campylobacter is estimated to cause up to 700000 cases of infection and more than 100 deaths each year. Campylobacter infection costs the United Kingdom economy at least £900 million per year (DEFRA, 2012). In addition, Campylobacter jejuni is the most common species to cause a rare neuromuscular paralysis known as Guillain–Barré syndrome (Parkhill et al., 2000).

Campylobacter jejuni is pathogenic in humans and avian hosts although mechanistic understanding of differences is incomplete (Byrne et al., 2007; Jennings et al., 2011; Williams et al., 2013; Humphrey et al., 2014). Despite this and in limited isolates (such as M1, NCTC 11168, 13126, NCTC 12744) strain dependency is particularly well documented. Thus, individual C. jejuni genotypes have been shown to produce their own unique infection rates and in vivo behaviors in chickens when taken from the two major MLST clonal complexes (CC), CC-45 and CC-21 (Chaloner et al., 2014). This heterogeneity has also identified strains with an invasive phenotype that lead to extra-intestinal spread (Humphrey et al., 2015), and have been implicated in recent outbreaks (Harrison et al., 2013; Edwards et al., 2014). However, the differences between C. jejuni strains which cause invasive disease and those which remain localized in the gut are poorly understood. One recent explanation suggests that dysregulation of cytokine production leading to an over-exuberant pro-inflammatory response leads to gut damage and bacterial invasion (Humphrey et al., 2014). However, there is a relative paucity of data regarding the ability of individual C. jejuni, across the wide spectrum of different strains, to cause inflammation.

Genes important for C. jejuni virulence are associated with motility, adhesion, invasion and toxin production (Table 1). C. jejuni is a highly motile organism with bipolar flagella and motility is very important for colonization and infection in chickens and other animals (Guerry, 2007). Genes involved in motility include flaA, flaB, and flaC. The flaA gene is also important for invasion of epithelial cells, and is responsible for adherence and colonization by C. jejuni in the gastrointestinal tract (Guerry, 2007). In addition, flagella may help C. jejuni invasion mechanisms by serving as export apparatus in the secretion of non-flagellar proteins (Konkel et al., 2004), including the ability to deliver flaC and Campylobacter invasion antigen (cia) into the cell’s cytoplasm (Konkel et al., 2004). CiaC is required for bacterial invasion into host cells whereas ciaI has been reported to be required for intracellular survival of C. jejuni after invasion (Buelow et al., 2011; Eucker and Konkel, 2012). The ‘invasion associated protein’ is encoded by iamA and its exact role in this process is still unclear (Rivera-Amill et al., 2001). HtrA, a serine protease, may act as a chaperone protein, which affects folding of adhesins (Bæk et al., 2011). One of the main toxins produced by C. jejuni is cytolethal distending toxin (CDT), which causes direct DNA damage leading to the activation of DNA damage checkpoint pathways, resulting in cell death (Lee et al., 2003). CDT consists of three protein subunits (CdtA, CdtB, and CdtC), which are encoded by genes cdtA, cdtB, and cdtC. The expression of all three genes is required in order to produce an active form of CDT (Pickett et al., 1996).

Understanding the mechanisms behind Campylobacter interaction with the host has focussed attention on human intestinal epithelial cells (e.g., HT-29, T84, and CaCo-2) and has shown that bacterial internalization is very important in C. jejuni pathogenesis (Jin et al., 2003; MacCallum et al., 2006; Byrne et al., 2007; Larson et al., 2008; Friis et al., 2009; Li et al., 2011). C. jejuni invades intestinal epithelial cells in a microtubule-, microfilament- and caveolin-dependent manner with a distinct cell type specificity (Oelschlaeger et al., 1993; Byrne et al., 2007; Larson et al., 2008; Watson and Galán, 2008). Invasion of human intestinal epithelial cells by C. jejuni activates numerous downstream signaling pathways, including the MAP kinases, ERK and p38, leading to the production of the pro-inflammatory cytokine interleukin-8 (IL-8) (Hickey et al., 2000; Jin et al., 2003; Li et al., 2011) and the anti-inflammatory cytokine IL-10 in human systems (Li et al., 2011). Indeed, a relationship between IL-8 production and C. jejuni invasion has been previously proposed in human cells (Li et al., 2016). Whether similar responses are observed in avian epithelial cells is poorly understood and is complicated by the presence of two IL-8 orthologs, IL-8like1 (CXCLi1) and IL-8like2 (CXCLi2), which are both induced by Campylobacter (Larson et al., 2008).

No studies have investigated whether similar mechanisms exist across collections of C. jejuni strains that define the species and have been isolated from relevant environmental, veterinary or clinical sources. In addition, previous work alluded to above in human cells, and the very limited work in chicken epithelial cells, do not represent the diversity across the Campylobacter species (<8 strains). In this work, we investigated the response of 100 C. jejuni strains that have been characterized previously at the genome level (Sheppard et al., 2013) by investigating their inflammatory (cytokine), adhesion, invasion, toxicity and signalling responses in 8E11 (avian) intestinal epithelial cells and compared these responses with human intestinal epithelial cells (HT-29). The host responses measured here suggest that avian and human epithelial cells share common mechanisms to combat C. jejuni but there is exceptional phenotypic diversity in the bacterial population.

The data in this paper add to the growing body of evidence that supports the idea that C. jejuni is a pathogen in chickens (Neill et al., 1984; Byrne et al., 2007; Jennings et al., 2011; Williams et al., 2013; Humphrey et al., 2014). Thus, C. jejuni induce inflammatory and toxicity responses and can also invade human and importantly avian epithelial cell lines. While there was little difference between human and avian cell responses, there was a wide range across all bacterial strains studied. We deliberately chose 100 strains of C. jejuni from a variety of sources (including chicken, human, cattle, and wild birds) and across sequence types to give a good representation of strains studied previously at the genomic level although we did not detect differences between these groups (Sheppard et al., 2011, 2013). Coincidentally, the reference strains M1 and NCTC11168 produced responses equivalent to the average for the whole study population. The question remains as to the source of the variation across the whole population. Interestingly, the presence of key virulence factors associated with C. jejuni pathogenesis was relatively high (>80%, Table 1) and thus differential gene expression may be responsible for the dramatic differences in epithelial cell responses. We did not measure individual virulence factor expression in this study but previous work on the transcriptome of a variety of Campylobacter strains showed that they are subject to complex regulation (Dugar et al., 2013).

Epithelial-derived IL-8 production is vital for early neutrophil infiltration into the gut in humans (Bennett et al., 2010) and chickens (Humphrey et al., 2014). This study confirms the importance of human IL-8 and identifies both CXCLi1 and CXCLi2 as important early chemokines induced in avian epithelial cells following C. jejuni infection. This is consistent with studies in chicken LMH epithelial cells (Larson et al., 2008). Interestingly, CXCLi2 (like human IL-8) was significantly higher than CXCLi1. To our knowledge this is the first time that a study using a large collection of C. jejuni strains has found differences in these two chicken IL-8 homologs. On a smaller scale (n = 6 strains), Larson et al. (2008) found the opposite effect. It is interesting to speculate that CXCLi2 (and not CXCLi1) may be the functional equivalent to IL-8 in humans. This is supported by structural data on amino acid similarity where CXCLi1 and CXCLi2 are 48 and 67%, respectively, identical to human IL-8 (Sick et al., 2000; Gupta et al., 2008).

The three cellular responses defined by ‘IL-8/CXCLi1/CXCLi2 production,’ ‘cellular toxicity,’ and ‘Campylobacter invasion’ are not always inextricably linked as demonstrated by the correlation curves (Figure 5). However, clear links were demonstrated between increased cytokine output (IL-8, CXCLi1 and 2) and ‘toxicity’ suggesting similar mechanisms of induction. Indeed, Hickey et al. (1999, 2000) suggested two mechanisms of Campylobacter-induced IL-8 production in human cells involving; (i) adherence and or invasion; and (ii) cdt expression. While this known link between IL-8 and invasion was demonstrated in INT407 human epithelial cells, we did not find a significant correlation between them in HT-29 cells. This was also the case for CXCLi2 production and invasion in avian cells. In contrast, a strong positive correlation was made in avian cells between CXCLi1 and invasion, again supporting an important role for these cytokines in early Campylobacter responses in chickens.

Numerous human cell lines have been used to study Campylobacter pathogenesis, including T84 (Zheng et al., 2008), INT407 (Borrmann et al., 2007), HT-29 (Bahrami et al., 2011), and CaCo-2 (Man et al., 2010) intestinal epithelial cells. There is very little data in avian intestinal cell systems. These cell lines are particularly useful for high throughput studies with many strains such as the current study. When considering the three cellular endpoints of IL-8 expression, cellular toxicity, and Campylobacter invasion none of these cells mimic equivalent in vivo responses exactly. Campylobacter invasion into CaCo-2 cells show very good correlation to in vivo invasive potential in chickens (Hanel et al., 2004) and humans (Everest et al., 1992) but produce limited cytokine responses (MacCallum et al., 2006) whereas HT-29 and T84 produce robust cytokine responses but no good evidence of relevance to in vivo invasion responses (MacCallum et al., 2006). In our hands, we could detect cellular invasion in both HT-29 cells and the 8E11 avian cells at a level of 1–3% of the initial inoculum with sufficient robustness to differentiate strains and for consistency over three replicate experiments. Levels of invasion are particularly dependent on time and the starting inoculum, but our results are in keeping with levels of invasion of up to 4% in CaCo-2 cells shown previously (Hanel et al., 2004). This suggests great potential in the avian cell line used in this study for investigating the diversity of C. jejuni responses in vitro.

This study confirmed the importance of canonical pro- (ERK) and anti-inflammatory (PI 3-Kinase -Akt) pathways for the C. jejuni-induced production of IL-8 and IL-10 in human epithelial cells (Watson and Galan, 2005; Li et al., 2011). Furthermore, ERK-dependent Campylobacter invasion is also supported by previous studies (Jin et al., 2003; Hu et al., 2006; Samuelson and Konkel, 2013; Samuelson et al., 2013). The importance of these pathways in avian cells is a novel result of this study and confirms that the underlying mechanisms are similar between human and chicken cells. In addition, the significantly higher IL-10 responses in the avian cells suggest that the avian gut may produce IL-10 as a method to tolerate large doses of Campylobacter and is supported by previous in vivo studies showing that some breeds of chicken produce more ‘regulated’ responses (Humphrey et al., 2014). While the avian cells used here are derived from leghorn chickens it would be interesting to speculate on the breeds used previously (Humphrey et al., 2014). There is evidence using Bayesian structural modeling of in vivo responses in chickens that IL-10 profiles are indeed different between breeds and this needs exploring further (Williams et al., 2013; Reid et al., 2016).

Study of Campylobacter uptake by endocytic pathways confirmed the requirement of lipid rafts (Lin et al., 2011) in human cells and extended this role to avian epithelial cells. We also found that microfilaments were required for invasion of both HT-29 cells and the avian ones. Interestingly, at least two mechanisms exist as INT407 (Konkel and Cieplak, 1992) and CaCo-2 (Russell and Blake, 1994) cells show microfilaments/microtubules-dependent and -independent mechanisms, respectively, suggesting that the role of cytoskeleton may be strongly cell dependent. Further work into mechanisms of uptake, in the present study, confirmed roles for dynamin and clathrin which has not been documented to date. Interestingly Cdt uptake into cells does involve clathrin coated pits (Thelestam and Frisan, 2004) and may be the mechanisms observed here. We could not confirm a consistent role for caveolins (using filipin and genistein) in the uptake process during this study. Indeed, this is consistent with a recent publication which suggests that C. jejuni invasion is independent of caveolins (Konkel et al., 2013).

One interesting consequence of inhibiting endocytic pathways was the concurrent reduction in cell toxicity confirming that Campylobacter is responsible for the toxicity and the endocytic inhibitors have negligible effects on cell viability. Another consequence of inhibiting endocytosis was the ‘extra’ effect of inhibiting cytokine production. Previously, De Zoete et al. (2010) established that live Campylobacter are very weak stimulators of both human and chicken TLR-2, -4, and -5. In striking contrast, lysed Campylobacter induce strong NF-kappaB activation through human TLR1/2/6 and TLR4 and chicken TLR2t2/16 and TLR4 but not via TLR5 of either species (De Zoete et al., 2010). Our results support the concept that ‘invasion’ or ‘internalization’ of some kind is necessary for cytokine production. Indeed, Hickey et al. (2000) also suggest an ‘invasion’ dependent pathway for IL-8 induction.

These results are novel because of the number and diversity of relevant C. jejuni strains used on an avian cell line to determine pathogenic mechanisms. However, we realize that there are certain limitations that provide an opportunity to improve our study in future work. Firstly, avian 8E11 cells are derived from small and large intestines of white leghorn chickens and are positive for enterocyte markers, villin, E-cadherin, and cytokeratin. Previous studies have confirmed the importance of broiler breed to the final inflammatory response and suggest that cells from commercial fast and slower growing breeds may be more relevant to study avian gut responses (Kaiser et al., 2016). Indeed, recent technologies point to the precision modeling of chicken intestinal slices (Punyadarsaniya et al., 2015). Secondly, with respect to assay methodology, we recognize that 6 h is probably not the optimal time to measure adhesion and earlier timepoints should be assessed in the future. In addition the gentamicin protection assay is subject to numerous limitations and artifacts as reviewed in detail previously (Friis et al., 2005). Finally, here we focused our attention on using C. jejuni strains that define the species, from a variety of sources and sequence types, as published previously (Sheppard et al., 2013). Recent Campylobacter outbreaks have highlighted the importance of invasion from the gut (e.g., to the liver) suggesting the importance of focussing on groups of invasive and non-invasive Campylobacter in future studies.

These novel data suggest that avian systems are likely to use similar host defense pathways to humans in response to Campylobacter spp. However, the sheer diversity and range of responses suggests that ‘a one strain fits all approach’ to in vivo experimental infection would not give meaningful data for the study of Campylobacter pathogenesis.

DJ was responsible for data generation and analysis across all figures of the manuscript. Design of signaling aspects of the work was developed and overseen by VK. Design of invasion assays was lead by LW. TH and TW were responsible for re-drafting the work and revising it critically for important intellectual content. All authors approved the final submitted draft and had opportunity for editing the document.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. The reviewer PN and handling Editor declared their shared affiliation.