Characterisation of a Multi-ligand Binding Chemoreceptor CcmL (Tlp3) of Campylobacter jejuni

Campylobacter jejuni is the leading cause of human gastroenteritis worldwide with over 500 million cases annually. Chemotaxis and motility have been identified as important virulence factors associated with C. jejuni colonisation. Group A transducer-like proteins (Tlps) are responsible for sensing the external environment for bacterial movement to or away from a chemical gradient or stimulus. In this study, we have demonstrated Cj1564 (Tlp3) to be a multi-ligand binding chemoreceptor and report direct evidence supporting the involvement of Cj1564 (Tlp3) in the chemotaxis signalling pathway via small molecule arrays, surface plasmon and nuclear magnetic resonance (SPR and NMR) as well as chemotaxis assays of wild type and isogenic mutant strains. A modified nutrient depleted chemotaxis assay was further used to determine positive or negative chemotaxis with specific ligands. Here we demonstrate the ability of Cj1564 to interact with the chemoattractants isoleucine, purine, malic acid and fumaric acid and chemorepellents lysine, glucosamine, succinic acid, arginine and thiamine. An isogenic mutant of cj1564 was shown to have altered phenotypic characteristics of C. jejuni, including loss of curvature in bacterial cell shape, reduced chemotactic motility and an increase in both autoagglutination and biofilm formation. We demonstrate Cj1564 to have a role in invasion as in in vitro assays the tlp3 isogenic mutant has a reduced ability to adhere and invade a cultured epithelial cell line; interestingly however, colonisation ability of avian caeca appears to be unaltered. Additionally, protein-protein interaction studies revealed signal transduction initiation through the scaffolding proteins CheV and CheW in the chemotaxis sensory pathway. This is the first report characterising Cj1564 as a multi-ligand receptor for C. jejuni, we therefore, propose to name this receptor CcmL, Campylobacter chemoreceptor for multiple ligands. In conclusion, this study identifies a novel multifunctional role for the C. jejuni CcmL chemoreceptor and illustrates its involvement in the chemotaxis pathway and subsequent survival of this organism in the host.


Introduction
Campylobacter jejuni is the one of the prevalent causes of acute human bacterial gastroenteritis worldwide [1][2][3]. C. jejuni is commonly found in the gastrointestinal tract of birds and poultry as commensal microbial flora [1], with infections in humans usually occurring from consumption of undercooked poultry, unpasteurised milk or untreated water. Symptoms include development of abdominal pains, fever and diarrhoea which can contain blood and leukocytes [4,5]. Additionally Campylobacter enteritis is associated with post infectious complications ranging from reactive arthritis or reactive myositis to the more severe Guillain-Barré syndrome [1,[6][7][8].
The development of disease depends on the ability of bacteria to adapt to the environment of the human gut [9]. To date, it is known that factors associated with virulence and pathogenicity of C. jejuni include iron acquisition, chemotaxis, adherence and lipooligosaccharides (LOS) [5,10,11]. C. jejuni is known to be highly motile in viscous environments, with motility and swimming velocity increasing with increasing viscosity [12].
Chemotaxis, the ability of bacterial cells to detect temporal changes in the chemical concentration of their surrounding environment, and flagella-mediated motility, have been reported to play an important role in the intestinal colonisation of avian and mammalian hosts, as well as the invasion of intestinal epithelial cells [10,[12][13][14][15][16][17]. Furthermore, Hugdahl et al. (1988) have identified chemoattractants that C. jejuni is preferentially motile towards, which include amino acids found in the gastrointestinal tract, organic acid intermediates of the TCA cycle and components of mucous such as mucin. [18]. The importance of C. jejuni chemotaxis and motility in colonisation and pathogenicity has previously been shown with non-motile mutants defective in chemotaxis, unable to colonise and cause disease in the gastrointestinal tract of mice [13,[19][20][21]. Furthermore, strains with mutations in chemotactic genes and flagella associated genes were not capable of colonising chicken caeca [14] and lost the ability to colonise rabbits [22] and ferrets [20]. Another important observation was that mutants defective in chemotactic motility also lost the ability to autoagglutinate, adhere and invade mammalian cells [12,15,16,23,24].
The fundamental components of the chemotaxis signalling pathway, which are conserved in all motile prokaryotes, consist of the chemoreceptors, a cytoplasmic histidine kinase, CheA, a coupling or scaffolding protein, CheW and/or CheV and a response regulator, CheY. In the well-characterised chemotaxis system of E. coli, CheA is regulated by the chemoreceptors through association with CheW and uses ATP to autophosphorylate a specific histidine residue. The phosphoryl group is subsequently transferred to the response regulator CheY [25]. Phospho-CheY interacts with the flagella motor to induce clockwise (CW) rotation resulting in a tumbling motion, where the cell momentarily stops and randomly reorientates [25][26][27][28]. Attractant-bound chemoreceptors inhibit CheA kinase activity, with a resulting decrease in the levels of phospho-CheY. In contrast, repellent-bound or empty chemoreceptors stimulate CheA kinase activity, thereby increasing phospho-CheY levels and tumbling events, orientating the cell in a new direction [25,26,28,29]. It is important to note that in C. jejuni, although the basic chemotaxis pathway backbone, consisting of Receptor-CheA-CheW(V)-CheY, is conserved, there are a number of differences to the paradigm E. coli model. C. jejuni encodes a two-domain CheA protein that includes a CheY-like response regulator domain in addition to the traditional histidine kinase domain, and a two domain CheV protein consisting of a CheWlike scaffolding domain and a CheY-like response regulator domain, as well as the paradigm CheW protein. C. jejuni also has a unique CheB protein that lacks a CheY-like response regulator domain found in all other bacterial chemotaxis pathways characterised to date [30].
Ten chemoreceptors have been identified in C. jejuni with homology to the methyl-accepting chemotaxis proteins (MCPs) in E. coli and have been designated Transducer-like proteins, Tlps. Group A Tlp receptors include CcaA (Tlp1), Tlps 2, 3, 4, 7 and 10 which have similarities to E. coli MCP structures and family A transducers of Halobacterium salinarium [30,31]. Group A Tlps are thought to sense extracellular ligands and consist of a periplasmic sensory domain which is variable between different receptors, two transmembrane domains, and the highly conserved C-terminal cytoplasmic signalling domain [30,32]. The sensory domain of each Tlp appears to be unique from that of non-Epsilon proteobacterial chemoreceptors; therefore sequence homology alone is not sufficient to determine specific ligand or ligands for each of the receptors. To date two of the group A Tlp receptors, Tlp1 and Tlp7, have been characterised [33]. CcaA was identified as the only receptor conserved in all sequenced strains of C. jejuni [34] and determined to be the receptor for aspartate [35] whereas Tlp7 binds to formic acid [36].
In this study we describe the characterisation of the periplasmic sensory domain of Cj1564 (Tlp3) chemoreceptor of C. jejuni strain NCTC 11168-O and characterise the ligand binding specificities of this chemoreceptor protein. We demonstrate the multifunctional role of Cj1564 in recognition of chemoattractants as well as chemorepellents, its association with the scaffolding proteins CheV and CheW and demonstrate the role of Cj1564 in cell to cell adhesion (autoagglutination) and biofilm formation, cell shape and host colonisation.

Identification of ligand binding specificity of Tlp3
To gain a greater understanding of the C. jejuni chemotactic pathway, characterisation of ligand binding potential for Cj1564 (Tlp3) was performed using recombinant Tlp3 periplasmic sensory domain peptide with all 20 amino acids and salts of organic acid arrays as well as 96-well binding assays ( Table 1). The array assay presents the compounds covalently bound to an epoxide group resulting in a high probability of a single molecular orientation. The plate assay relies on non-covalent charge and hydrophobic/ hydrophilic interactions for presentation of the molecules offering a different presentation to the array. A total of 12 interacting ligands were identified through the plate and array assays. The binding of the Tlp3 sensory domain to amino acids and salts was then confirmed by STD-NMR and SPR (Biacore, Table 1). SPR analysis found the highest affinity interactions of Tlp3 with lysine and glucosamine (K D ,10 mM; Table 1), additionally, biologically significant interactions (K D ,50 mM; Table 1, Figure S1) were observed for isoleucine, succinic acid, arginine, purine, malic acid and thiamine.
Mutation of tlp3 alters motility, autoagglutination and biofilm formation of C. jejuni Cj1564 (Tlp3) has previously been identified as a group A chemoreceptor for C. jejuni with homology in the signalling domain to other bacterial chemoreceptors such as that of E. coli and H. pylori [37]. The periplasmic domain of C. jejuni group A chemoreceptors are predicted to be involved in sensing and binding of extracellular ligands [37]. In order to identify biologically significant ligand binding specificity and function of Tlp3 in the chemotactic pathway, an insertionally inactivated isogenic tlp3 mutant and its complement were created. The mutant was constructed by deleting 52 central bp of the periplasmic domain of tlp3 and inserting a non-polar kanamycin resistance cassette which has the transcription terminator removed to minimise effects on genes downstream that may be transcribed in the same orientation as tlp3 (described in materials and methods: mutagenesis and complementation of tlp3). Microscopic analysis of Dtlp3 confirmed the presence of flagella and altered (loss of spiral shape) cellular morphology of all observed 11168-O Tlp3 mutant bacterial cells, with spiral morphology restored after complementation in approximately 50% of the complemented cells ( Figure 1A, B & C). Comparison of C. jejuni 11168-O and Dtlp3 motility shows that there was a 5-fold decrease in nondirected swimming motility of the mutant with motility partially restored to wild type levels in the complemented strain, Dtlp3c ( Figure 1D). Live imaging of fluorescently labelled C. jejuni isogenic strains allowed capture of bacterial cell motility and demonstrated a defect in motility illustrated by the inability of Dtlp3 cells to effectively 'swim' in a directional movement in absence of specific

Author Summary
Bacterial chemotaxis is an important part in initiation of colonisation and subsequent pathogenicity. In this study, we report direct evidence supporting the involvement of C. jejuni transducer-like protein Cj1564 (Tlp3) in the chemotaxis signalling pathway via small molecule arrays, surface plasmon and nuclear magnetic resonance (SPR and NMR) as well as chemotaxis assays of wild type and isogenic mutants. We further demonstrate its ability to interact with chemoattractants isoleucine, purine, malic acid and fumaric acid and chemorepellents lysine, glucosamine, succinic acid, arginine and thiamine. This is the first report identifying Cj1564 as a multi-ligand receptor for Campylobacter jejuni and its signal transduction initiation through the CheV and CheW proteins. Finally, our characterisation of C. jejuni Cj1564 provides additional basis for elucidating the roles of other group A chemoreceptors and their importance in the chemotaxis signalling pathway.
stimuli. Instead the cells were observed to 'twitch' and continuously 'tumble' without movement in any particular direction (Movie S1), thus demonstrating an altered ''random walk'' motility phenotype of the Dtlp3 mutant. The expression of the tlp3 gene in the isogenic mutant strain was reduced, as compared with the wild type and complemented strains, which is likely to be due to the interruption of the gene with the strongly promoted Km R cassette. Expression of tlp3 was analysed using quantitative real time PCR in 11168-O, Dtlp3 and Dtlp3c using primers that amplify the entire periplasmic domain gene region of tlp3. There was a 6.560.78 fold reduction observed in the expression of tlp3 in the Dtlp3 mutant strain compared to the wild type strain, conversely an 8.7560.98 fold increase was observed when screening for tlp3 expression in the complemented mutant compared to wild type (data not shown). This indicates that the inclusion of the resistance cassette into the tlp3 gene has negatively affected the expression of tlp3 from its own promoter. This indicated that the Kanamycin resistance cassette encodes the dominant promoter signal that inhibits production of RNA molecules containing the 59 end of the tlp3 gene region, upstream from the kanamycin resistance cassette. While the insertion of complementing tlp3 and chloramphenicol resistance cassette into the pseudogene (cj0046) results in over expression of tlp3 compared to the wild type strain. Further expression analysis was performed on the genes immediately upstream (cj1563c) and downstream (pflA) of the mutated tlp3 gene to confirm no polar effects were introduced as a result of mutagenesis. Expression of pgp1 and flaA were assessed to establish that the observed defect in motility is not due to changes in expression of genes involved in flagella development or function. No significant difference in expression (p.0.1) was observed for cj1563c, pflA, pgp1 and flaA (data not shown).
Biofilm formation and agglutination play an important role in the survival of bacterial cells. Consequently, we compared these characteristics between the wild type (11168-O), mutant (Dtlp3) and complemented (Dtlp3c) strains. Autoagglutination assays revealed that the Dtlp3 autoagglutinated while 11168-O and Dtlp3c had little to no autoagglutination (Figure 2A & B). It is also interesting to note that autoagglutination of the 11168-O Dtlp3 mutant was similar to that of C. jejuni 81-176 ( Figure S2), which carries a natural mutation in its tlp3 gene, revealed by the published genome sequence (Sanger, 2006). Furthermore, auto-agglutination was also shown to be independent of growth temperature at 25uC, 37uC and 42uC for all strains tested (data not shown). Additionally it appears that Dtlp3 autoagglutinates at a rate that is faster than normal gravitational pull. Biofilm formation, as assessed by crystal violet assay, indicated an approximate 1.5-fold increase in biofilm formation of Dtlp3 mutant compared to wild type, with the original levels in biofilm formation restored in Dtlp3c ( Figure 2C).

Chemotactic behaviour and influence of Tlp3
In order to determine the biological relevance of Tlp3 in chemotactic motility, nutrient depleted chemotaxis assays were performed as previously described [35]. Chemotaxis towards a range of amino acids, glycans and other small molecules was investigated in 11168-O, Dtlp3 and the complemented mutant, Dtlp3c. A positive chemotaxis response was identified for 5 amino acids, suggesting these ligands are attractants ( Figure 3A; Table 1). A 4-log reduction in bacterial numbers was observed for the Dtlp3 (3.6 6 10 2 cfu/ml) compared to the wild type (6 6 10 6 cfu/ml) in migration towards isoleucine, additionally this was also observed for fumaric acid (9 610 2 and 4.14 610 6 cfu/ml respectively). For migration towards purine, a 1-log reduction for the Dtlp3 (1.08 6 10 5 cfu/ml) compared to the wild type (1.4 6 10 6 cfu/ml) was observed, with a 2-log reduction in migration towards malic acid (1.13 6 10 3 and 9 6 10 5 cfu/ml respectively) and a 3-log reduction in migration towards aspartate (3.7 6 10 3 and 7.8 6 10 6 cfu/ml respectively). The migration of wild type C. jejuni 11168-O towards aspartate was comparable to that previously published for nutrient depleted assay by Hartley-Tassell et al., 2010 [38]. Tlp3 was also found to mediate a repellent response to 5 amino acids ( Figure 3B; Table 1). Lysine mediated repellence was reduced for the Dtlp3 isogenic strain as a 2-log increase in viable bacterial numbers was detected around the lysine plug (2.4 6 10 5 cfu/ml) when compared to the wild type (2.2 6 10 3 cfu/ml). For migration of Dtlp3 towards glucosamine, a 3-log increase (7.8 6 10 6 cfu/ml) was detected, compared to that of the wild type (1.68 6 10 3 cfu/ml). In addition, for succinic acid, arginine and thiamine, a 1-log increase of bacterial numbers was observed in the Dtlp3 (2 6 10 7 , 6.8 6 10 7 and 9 6 10 6 cfu/ml, respectively) compared to the wild type strain (1.5 6 10 6 , 1.3 6 10 6 and 4 6 10 5 cfu/ml, respectively). Chemotaxis assay for alpha-ketogluta- rate was not definitive (data not shown), possibly due to the low affinity of this receptor for alpha-ketoglutarate.
Identification of chemoattractant and chemorepellent binding of Tlp3 by STD-NMR analysis 1 H STD-NMR analysis was used to investigate the epitope binding preferences of the chemoattractants (isoleucine, purine, malic acid) and chemorepellents (lysine, arginine, glucosamine) to the recombinant Cj1564 (Tlp3) periplasmic sensory domain peptide. Binding was observed with chemoattractants isoleucine and purine along with the chemorepellents lysine and arginine in accordance with the SPR analysis showing varying binding affinities. For the chemoattractant isoleucine (K D ,17 mM), the binding epitope appeared to be the methyl group, as strong methyl resonances were seen in the STD spectrum and only very weak signals for other side chain resonances ( Figure 4A). For purine (K D ,38 mM), the H6 and to a lesser extent the H2 protons (,5-fold) of the pyrimidine ring showed an STD effect compared to a negligible STD effect for the H8 proton of the imidazole ring. For the chemorepellent lysine (K D ,2.8 mM) an STD effect was clearly seen for all proton resonances along the side chain ( Figure 4B). A weaker STD effect (,4 fold) was seen for all proton resonances along the arginine (K D ,38 mM) side chain. No signals were detected for malic acid or glucosamine using STD-NMR. SPR analysis demonstrated that glucosamine and malic acid have slow disassociation rates explaining the absence of STD signals due to these amino acids as saturation transfer is a result of the spin diffusion process which requires fast exchange for magnetisation to be spread from the protein to the ligand in order to observe an STD effect.
Competition STD experiments were also performed to probe the binding preferences of Tlp3 and to unravel the nature of the binding interaction. Competition experiments testing the effect on binding to Tlp3 of attractant/attractant (isoleucine/purine), repellent/repellent (lysine/arginine) and attractant/repellent (isoleucine/arginine; isoleucine/lysine; purine/lysine; purine/arginine) were performed. For attractant/attractant (isoleucine K D ,17 mM/purine K D ,38 mM) there seemed to be a minor change in the STD effect seen for the ligands independently. For the repellent/repellent (lysine K D ,2.8 mM/arginine K D ,38 mM) however, there appeared to be a significant reduction in the STD effect of lysine in the presence of arginine whose weak STD effect appeared unchanged, even though the binding affinity of lysine is stronger (,14 fold) than that of arginine. These results demonstrate that Cj1564 (Tlp3) is able to bind to both chemoattractants and chemorepellents. No binding preference to Cj1564 (Tlp3) was observed in the presence of the chemoattractants isoleucine/ purine, compared with the ligands alone, however for the chemorepellent competition experiment even though lysine has a greater binding affinity (K D ,2.8 mM) compared with isoleucine (K D ,17 mM) a significant reduction in the STD effect of lysine was observed in the presence of arginine suggesting preferential binding of arginine over lysine. In the case of the attractant/repellent spectra: Isoleucine/ arginine -isoleucine (chemoattractant, K D ,17 mM) binds preferentially over arginine (chemorepellent, K D ,38 mM). For isoleucine/lysine -isoleucine (chemoattractant) binds preferentially over lysine (chemorepellent) with a significant reduction in the STD effect of lysine observed in the presence of isoleucine compared to the STD spectra of the ligands individually, even though the binding affinity of lysine (K D ,2.8 mM) is stronger than that of isoleucine (K D ,17 mM) ( Figure 4C & D). Purine/lysine -purine (chemoattractant) binds preferentially over lysine (chemorepellent) with a significant reduction in the STD effect of lysine in the presence of purine observed compared to the STD spectra of the ligands individually, even though the binding affinity of lysine (K D ,2.8 mM) is stronger than that of purine (K D ,38 mM). Purine/arginine -purine (chemoattractant) binds preferentially over arginine (chemorepellent), a negligible STD effect of arginine in the presence of purine was observed with these ligands determined to have comparable binding affinities (K D ,38 mM). These results demonstrate that there is preferential binding of chemoattractants over the chemorepellents regardless of the individual binding affinity of the ligands to Cj1564 (Tlp3).

Involvement of tlp3 in host bacterial interactions
In vitro invasion in the epithelial cell line Caco-2 is significantly reduced in Dtlp3. The intestinal epithelial cell line Caco-2 was used in the gentamicin protection assay to assess the role of Tlp3 in attachment and subsequent invasion in vitro using an isogenic Dtlp3 and Dtlp3c strain. The isogenic Dtlp3 mutant displayed a 6-fold reduced level of adherence and 13-fold reduction in invasion of the human intestinal Caco-2 cell surface, compared to the wild type strain or complemented mutant strain Dtlp3c ( Figure 5A and 5B). This finding further suggests that a mutation in tlp3 affects the ability of the bacteria to actively adhere and invade cultured epithelial cells.
Dtlp3 does not affect chick colonisation. The effect of tlp3 deletion from the C. jejuni 11168-O chemoreceptor repertoire in vivo was explored using a C. jejuni chick colonisation model [39]. One-day old chicks were infected orally with a dose of 10 8 C. jejuni wild type 11168-O, Dtlp3 or Dtlp3c, and caecal contents were assessed for C. jejuni after 5 days post infection, once complete colonisation had been reached [39]. The Dtlp3 isogenic mutant strain exhibited no significant difference in colonisation ability compared to wild type ( Figure S3). This interesting result, together with the increased Dtlp3 autoagglutination data, led us to hypothesise that the Dtlp3 mutant may aggregate on the surface of the caeca. To confirm this, we used scanning electron microscopy to visualise Dtlp3 isogenic mutant strain cells in vivo on the chicken caeca. Figure S4A shows C. jejuni 11168-O wild type cells loosely scattered on the surface of the caecal tissue while Figure S4B shows a tightly compacted aggregate of Dtlp3 isogenic mutant cells. As there was no difference between colonisation ability of the wild type bacteria and Dtlp3, we can speculate that in vivo in an avian host, the Dtlp3 may be able to detect a chemotactic signal to follow in the lumen or caeca, allowing subsequent aggregation of the bacteria on the caecal surface coincidently resulting in similar colonisation levels.
The signalling domain of Tlp3 interacts with other proteins and protein domains of the C. jejuni chemotaxis signalling pathway To further investigate the role of Tlp3 in the chemotaxis signal transduction pathway in C. jejuni, the predicted cytoplasmic signalling domain of Tlp3 was analysed for protein-protein interactions with C. jejuni NCTC11168-O chemotaxis proteins using the yeast two-hybrid and three-hybrid systems as previously described [35]. The yeast two-hybrid system was used because of a lack of similar genetic manipulation systems available for C. jejuni. We have previously demonstrated biological validity of protein interactions identified by the yeast two-hybrid system via pulldown assays for the components of the C. jejuni aspartate receptor, CcaA [35] signal transduction. The yeast system allows detection of interacting proteins in vivo by utilising the two separable domains of the GAL-4 transcription factor, the DNA-binding domain (DNA-BD) and the transcription activation domain (AD). This system relies on the reconstitution of the GAL-4 transcription factor when two proteins of interest, the 'bait' protein fused to the DNA-BD and the 'target' protein fused to the AD, interact, thus allowing activation of reporter gene expression [40], Residues 517-662 of Tlp3, encompassing the region homologous to the highly conserved MCP signalling domain, were selected for analysis. The signalling domain is an independent structural motif, which interacts with the dimeric CheA and the scaffolding protein CheW to form MCP-CheW-CheA ternary signalling complexes that regulate CheA histidine kinase activity [41][42][43]. It is necessary to note that this region of Tlp3 is identical to residues 513-659 of Tlp2 (Cj0144) and 520-665 of Tlp4 (Cj0262) [37] and thus is referred to as Tlp sig 2.3.4 in this study ( Figure S5). The Tlp sig 2.3.4 was found to interact with a number of the C. jejuni NCTC11168-O chemotaxis proteins and protein domains, as shown in Table 2. A medium strength interaction was detected As it has previously been shown that CcaA (Tlp1) of C. jejuni preferentially binds to CheV over CheW [35], we utilised the yeast three-hybrid system to determine if the signalling domain of Tlp3 (which is identical to the signalling domains of Tlp2 and Tlp4) displayed similar binding preferences for the scaffolding proteins. In the three-hybrid system, a third protein can be conditionally expressed and its role in the interaction between the AD and DNA-BD fusion proteins can be examined [44]. We specifically looked at the effect of presence of both CheW and CheV proteins on their interactions with Tlp sig 2.3.4 . It was found that the coexpression of native CheW from pBrTlp234IWII (Table 3) had no effect on the medium strength interaction observed between Tlp sig 2.3.4 and CheV (BD-Tlp sig 2.3.4 and AD-CheV). Analysis of interactions between the reciprocal combinations of fusion proteins using pBrVIWII ( and AD-CheW). This was confirmed in the analysis of the

Discussion
In this study we have shown the role of Tlp3 in C. jejuni 11168-O chemotaxis as a multiple ligand-binding protein, capable of detecting numerous chemoattractants and chemorepellents using a range of confirmatory methodologies such as small molecule arrays, SPR and STD-NMR. The substrates identified as having a strong binding affinity include lysine and glucosamine with biologically significant interactions observed for isoleucine, aspartate, succinic acid, arginine, purine, malic acid, and thiamine. Therefore we propose to name this receptor CcmL; Campylobacter chemoreceptor for multiple ligands.
Similar observations have previously been made for H. pylori isolates, where positive chemotaxis was identified for the sugars and amino acids phenylalanine, aspartic acid, glutamic acid, isoleucine and a negative chemotaxis for leucine and tyrosine [45]. Isoleucine has previously been demonstrated to be a chemoattractant for other organisms including B. subtilis [46], H. salinarum [47], P. aeruginosa [48] and H. pylori [45]. In B. subtilis, the McpC chemoreceptor binds isoleucine weakly, yet with sufficient affinity to suggest direct binding. It is interesting to note that in our study, glucosamine has been identified as a repellent, while in other bacteria such as E. coli [49] and B. burgdorferi it has been determined to be an attractant, and has been classified as a non-essential nutrient [48,50]. Lysine has been demonstrated as an attractant for P. aeruginosa [48] which is opposite to our findings which show lysine as a chemorepellent for C. jejuni. Furthermore, in B. subtilis, the McpC chemoreceptor did not show binding to lysine, however, it was suggested that McpC binds lysine by an indirect method most likely involving ancillary proteins, further suggesting McpC may be a universal chemoreceptor able to respond to numerous amino acids [51]. It had been shown in E. coli that even though chemoreceptors are sensitive to a particular ligand, they can also detect a large number of structurally related amino acids and their analogues [52].
C. jejuni is highly adapted to the environment of the avian gut and as a result, uses efficient chemotactic motility to colonise the mucous-filled crypts of the lower gastrointestinal tract [53]. It is reasonable to hypothesise that Tlp3 may be involved in interacting with ligands to sense the external environment in order to navigate inside the host and may be involved directly or indirectly in C. jejuni metabolic and catabolic pathways. It is interesting to note that aspartate, glutamate, proline and serine are the most abundant amino acids found in chicken excreta [54], and serine catabolism has been reported to be essential for colonisation of the avian gut by C. jejuni [55]. However, C. jejuni lacks the key glycolytic enzyme 6-phosphofructokinase as well as alternative pathways for sugar catabolism [56] so it utilises amino acids and Krebs cycle intermediates for energy production and encodes all of the enzymes required for a complete oxidative TCA cycle [37,57,58]. C. jejuni needs to be able to sense and move towards amino acids and small organic acids such as aspartate, asparagine and serine in order to catabolise these compounds to use as the sole source of reduced carbon and energy due to its inability to utilise glucose [59]. Additionally, a recent study has indicated that energy taxis may also be one of the driving forces behind movement to optimal conditions for energy generation and subsequent colonisation [60,61].
We have shown that a mutation in the C. jejuni ccmL chemoreceptor gene lead to alteration of phenotypic characteristics of the bacteria, such as cellular morphology, autoagglutination behaviour and biofilm formation, highlighting its role in the C. jejuni life cycle. Furthermore, the signalling domain of this chemoreceptor also interacts and binds with both CheV and CheW scaffolding proteins.
Additionally, we used STD-NMR analysis to confirm the binding of amino acid and salts of organic acid arrays, plate binding assays and SPR analysis. STD-NMR indicated that the Tlp3 receptor binds chemoattractants preferentially over chemorepellents irrespective of binding affinity, however, is still able to recognise both attractants and repellents in isolation. This may indicate that upon binding an attractant, the repellent binding site of CcmL through allostery prevents the binding of a repellent ligand to the protein. Furthermore, it appears that the binding site may be able to accommodate more than one chemoattractant at a time as competition tests between two chemoattractants results in positive binding for both attractants present; indicating that one chemoattractant is not preferred over another regardless of the measured affinity. These results indicate a much more complex   interaction between chemotaxis receptor proteins and ligands than previously thought, with a complex milieu of attractants acting in concert rather than the receptor binding preferentially to a specific ligand based on the hierarchy of affinities. Furthermore, the Tlp3 predicted structure indicates the presence of a single Cache_1 (PFO2743) (calcium channels and chemotaxis receptors) domain, representing a single ligand binding pocket that accommodates multiple ligands with varying affinity. This is in agreement with our STD-NMR data that shows interchange of the ligands binding to purified periplasmic domain of CcmL in competitive assays. CcmL (Cj1564) shares complete homology in the cytoplasmic domain with Tlp2 (Cj0144) and Tlp4 (Cj0262), but has less than 50% homology with the cytoplasmic domains of Tlp1 (Cj1506), Tlp7 (Cj0951/52) and Tlp10 (Cj0019). Tlp3 has the greatest homology to Tlp2 with 72% identity across the entire protein and 38% identify/59% similarity across the periplasmic domain. The greatest stretch of similarity between Tlp3 and Tlp2 in the periplasmic binding region is within the cache domain that is present in both proteins. Homology between the periplasmic domain of Tlp3 and the periplasmic domains of Tlp1 and 4 is below 33%. No homology was detected between Tlp3 periplasmic domain and the periplasmic domains of Tlp7 and 10.
To further characterise interactions of Tlp3 with components of the C. jejuni chemotaxis signalling pathway, a well-established yeast-two-hybrid and three-hybrid protein-protein interaction system was used. The two-and three-hybrid system data suggests that both the scaffolding proteins CheW and CheV are capable of binding to the Tlps2, 3 and 4 signalling domains with no obvious preference for either protein observed. CheW and CheV may form mixed multi-protein complexes with these receptors, as the expression of CheV was found to strengthen the interaction between the signalling domain of Tlp sig 2.3.4 and CheW. This appears to vary from the observations for the CcaA (Tlp1) signalling domain, which has previously been shown to preferentially bind CheV [35]. While the signalling domains of Tlps 2, 3 and 4 and of Tlp1 are very similar, some amino acid differences do exist which may account for the differences in the binding preferences of these Tlps for the CheW and CheV scaffolding proteins. These amino acid differences are within the region homologous to residues 350-471 of the E. coli serine chemoreceptor, Tsr, a fragment which has been shown to be capable of mediating CW-biased signals and is therefore predicted to contain residues involved in the binding of CheW and CheA in this species [62] (Figure S5). This data lead us to speculate that in C. jejuni, the group A Tlp2, 3 and 4 signal via alternative or mixed scaffolding proteins and that different binding affinities of the chemoreceptors with CheV and CheW may control the composition of receptor clusters.
The mutation in the tlp3 (cj1564) gene resulted in an altered cellular morphology of the bacterial cell and inability of Dtlp3 cells to 'run'. Increased formation of biofilm and the high rate of autoagglutination may indicate a possible increased response to stress by the tlp3 mutant. A previous study by Vegge et al. (2009) reported that mutation of the tlp3 gene along with other genes does not affect the motility of C. jejuni in a rich medium [63] however; our results indicate mutation of tlp3 significantly reduces motility when compared to the wild type and complemented strains. This is in agreement with a study carried out by   [64] where mutation of tlp3 reduced motility and complementation restored the wild type phenotype [23]. Furthermore, reduced motility is also likely due to an inability of the bacteria to engage in efficient chemotaxis signalling. Further inspection of the isogenic Dtlp3 mutant cells by microscopy indicated that the bacteria retained the ability to twitch and tumble in a stationary position suggesting a bias toward clockwise rotation of the flagella, subsequently leading to a higher frequency of tumbles and inhibited smooth swimming. It appears however, that the bacterial cell was still capable of directional movement when a stimulus was supplied in vivo. Additionally, scanning electron microscopy and quantitative PCR analysis confirmed the presence of flagella indicating that the lack of motility was not due to a defect in the flagella development; however, the helical shape of the bacteria was altered. The helical shape of C. jejuni has long been associated with pathogenesis but the genetic components involved in modulating C. jejuni morphology have only recently been identified, where the protein termed Pgp1 (peptidoglycan peptidase 1) in C. jejuni 81-176 was identified to be involved in maintenance of the helical shape [65]. To date no correlation has been found between chemoreceptors and cellular morphology.
In the in vitro model of infection, it appears that Tlp3 may be a critical factor for invasion as Dtlp3 displayed a markedly reduced level of adherence and invasion compared to the wild type strain. However, in vivo assessment of the Dtlp3 mutant and wild type strain showed no differences in colonisation ability of the avian caeca based on cell counts, even though ability to adhere and invade caco-2 cells in vitro was significantly reduced. This finding agrees with the previously published data on avian colonisation for the Tlp3 mutant of C. jejuni 81-176 [66]. However, it is important to consider that subsequent publication of the genome sequence of 81-176 (Sanger, 2006) revealed a natural mutation of the Tlp3 homologue in 81-176 which exists as 2 separate reading frames: CJJ81176_1548 and CJJ81176_1549 encoding the majority of the periplasmic and cytoplasmic domains of Tlp3, respectively. There is a nucleotide deletion in the sequence of CJJ81176_1548 at position 1467624 (A) in the genome subsequently altering the amino acid sequence, consequently changing the reading frame of the transmembrane and cytoplasmic domains (CJJ81176_1549) resulting in incorrect translation and production of a nonfunctional Tlp3 protein in 81-176. There are an additional number of mutations in both the remainder of the periplasmic and cytoplasmic domains with several intervening stop codons. The absence of a functional Tlp3 in 81-176 has previously been reported [67].
It is interesting to note that C. jejuni 81-176 has the same complement of Tlp genes as 11168 and 81116, with the exception of Tlp3, which has a naturally occurring mutation in 81-176. When phenotypic characteristics pertaining to biofilm formation and autoagglutination of 81-176, and 11168-O are compared, a wild type 81-176 shows higher levels of aggregation, biofilm formation (data not shown) and autoagglutination, similar to that observed for Tlp3 mutant of 11168 ( Figure S2).
Characterising the function of C. jejuni chemosensory proteins, as described in this study, will contribute to understanding chemotaxis signalling pathways which are involved in colonisation and further identify chemoreceptor ligand specificity of individual group A Tlp receptors and their involvement in the chemotaxis pathway and its importance in the survival of this organism. The findings in this study also provide insight into the complexity of chemotaxis receptor protein-ligand interactions with implications not just for C. jejuni chemotaxis but for all bacterial chemotaxis.

Ethics statement
Animal experiments were carried out in strict accordance with the Griffith University Animal Ethics Committee guidelines and assigned approval number BDD/02/11. All procedures involving animals were reviewed and approved by National Health and Medical Research Council Australian code of practice for the care and use of animals for scientific purposes 7th edition 2004.

Expression and purification of the sensory domain of tlp3
The DNA fragment encoding the periplasmic sensory domain (amino acids 43-290) of Tlp3 was amplified using primers incorporating start and stop codons (Table S2) and ligated into pGEM-TEasy to form pGU0815 (Table S1). The insert in pGU0815 was restricted at primer-specific NdeI and XhoI sites and subcloned into pET-19b to form pGU0816 (Table S1). For recombinant protein expression, competent E. coli BL21(DE3) cells were transformed with pGU0816. An overnight culture of BL21(DE3)/pGU0816 was used to inoculate LB containing ampicillin (100 mg/ml) that was incubated at 37uC with aeration. Protein expression was induced using 1 mM IPTG when OD 600 nm reached 0.4-0.6. Expression of the Tlp3 peri -His fusion protein was verified by SDS-PAGE and Western Blot analysis using Anti-His mouse IgG (Cell Signalling) as shown in Figure S6. The cell pellet was resuspended in PBS containing 6 M urea, lysozyme (0.2 mg/ml) and protease inhibitor cocktail mix (50 ml/ ml) and incubated at room temperature for 1 h using a rotational mixer. The cells were sonicated and an additional freeze/thaw step performed to aid in cell lysis. The insoluble cell debris was removed by centrifugation at 38 000 rpm for 90 min. The clarified supernatant was added to 1.5 ml of TALON Metal Affinity Resin (Clontech) and rotated overnight at 4uC using a rotational mixer. The slurry mix was then packed by gravity into a 10 ml Bio-Rad chromatography column. The column was washed twice with PBS, then washed with PBS containing 20 mM imidazole, washed three times with PBS and the bound His-tagged protein was eluted with PBS containing 150 mM imidazole. Residual imidazole was removed from the sample using Econo-Pac 10DG Desalting Columns (Bio-Rad) according to manufacturer's specification. Purity was confirmed by analysis of samples by SDS-PAGE and Western Blot using anti-His antibodies (Bio-Rad). To ensure that the recombinant protein was properly folded, its properties were compared to denatured protein. Denatured protein recombinant Tlp3 was not soluble in PBS or TBS buffers, unlike correctly folded protein. Moreover, the binding affinities of recombinant protein following purification without urea were also tested. No change in binding affinities could be observed (data not shown).

Identification of ligand binding potential of Tlp3
Amino acid and small molecule arrays and plate based binding assays were performed as previously described. [35]. The ligands investigated by Surface Plasmon Resonance (SPR) included alphaketoglutarate, glucosamine, lysine, purine (basic ring), malic acid and L-isoleucine (Sigma). SPR experiments were performed using a Biacore T100 biosensor system (GE-healthcare) at 25uC in 16 PBS pH 7.2 at a flow rate of 30 ml/min. Purified His-Tlp3 was diluted to 0.15 mM in 16 PBS pH 7.2 and loaded on flow cell 2 (FC2) of a Ni 2+ NTA sensor chip with 5 min contact time. FC1 had no protein loaded and was used as reference. Amino acids were prepared in 1 6PBS pH 7.2 and serially diluted from 0.0125 to 0.2 mM. The amino acids were loaded to the sensor chip using single-cycle kinetics, i.e. after the injection of five amino acid dilutions the chip was regenerated with EDTA. Subsequently, the chip was re-loaded with Ni 2+ and His-Tlp3 before the injection of the next amino acid to be tested. A 10 min dissociation time was allowed after the addition of each analyte. SPR signals were analyzed using the Biacore Evaluation software to determine K D . to give a total volume of 250 mL in a 3 mm NMR tube for NMR analysis. Control samples were prepared in an identical manner without Tlp3 added. All NMR experiments were performed on a Bruker Avance 600 MHz spectrometer, equipped with a 5-mm TXI probe with triple axis gradients at 283 K without sample spinning. 1 H NMR spectra were acquired with 32 scans, a 2 s relaxation delay over a spectral width of 6000 Hz. Solvent suppression of the residual HDO peak was achieved by continuous low-power presaturation pulse during the relaxation delay. In the STD-NMR experiments of the chemoattractants (isoleucine, purine and malic acid) and chemorepellents (lysine, arginine and glucosamine) in complex with Tlp3, the protein was saturated at 20.5 ppm in the aliphatic region of the spectrum and off-resonance at 33 ppm with a cascade of 40 selective Gaussianshaped pulses of 50 ms duration (50 dB), which correlates to a strength of 190 Hz. A 100 ms delay between each soft pulse was applied, resulting in a total saturation time of 2 s and 2 K scans.

STD NMR spectroscopy
Data were obtained with an interspersed acquisition of pseudotwo-dimensional on-resonance and off-resonance spectra in order to minimize the effects of temperature and magnet instability. Onand off-resonance spectra were processed separately, and the final STD-NMR spectrum was obtained by subtracting the individual on-and off-resonance spectra, resulting in less subtraction artefacts. Relative STD effects were calculated according to the equation A STD = (I 0 2I sat )/I 0 = I STD /I 0 by comparing the intensity of the signals in the STD-NMR spectrum (I STD ) with signal intensities of a reference spectrum (I 0 ). The STD signal with the highest intensity was set to 100% and other STD signals were calculated accordingly. A spin lock field of 10 ms was applied to remove unwanted background protein signals. Increased spin lock fields resulted in artefacts and reduced ligand signal intensities. Control STD-NMR experiments were performed using an identical experimental setup and the same ligand concentration but in the absence of the protein.

Mutagenesis and complementation of tlp3
The tlp3 gene (cj1564) in C. jejuni 11168-O was inactivated by inverse PCR mutagenesis [69] to produce the strain 11168-ODtlp3::Km (Table S1). The tlp3 periplasmic domain (amino acids 43-290) was amplified from the 11168-O genome and cloned into pGEM-TEasy to produce pGU0815 (Table S1). Inverse PCR with primers designed to incorporate a BglII restriction site and delete 52 bp of the tlp3 periplasmic domain was followed by insertion of a non-polar kanamycin resistance cassette with a consensus campylobacter promoter from pMW10 [70] in the same orientation as tlp3 to generate plasmid pGU0817 (Table S1). A kanamycin resistance cassette which lacks the transcription terminator was used in order to minimise effects on genes downstream of tlp3. Each construct was verified by DNA sequencing and subsequently electro-transformed into the motile variant of 11168 original clinical isolate, donated by D.G Newell, VIR, London [71]. Replacement of the mutant allele was verified by PCR and DNA sequencing. The complement of this mutant was generated by inserting the complete tlp3 gene including promoter into the pseudogene cj0046 using the plasmid pC46 (Table S1). The tlp3 complement plasmid was electro-transformed into C. jejuni 11168-O with insertion into the pseudogene cj0046 locus to produce strain 11168-ODtlp3::KmVcj0046::Cm (Table S1). Complementation was confirmed by PCR and sequence analysis.

Phenotypic characterisation: Motility, agglutination and biofilm formation
Motility assays were performed as previously described [23] with the following modifications: C. jejuni strains were grown microaerobically at 42uC for 24 h in 10 ml Brucella broth (Oxoid) on a shaker (50 rpm/min). Cells were collected by centrifugation at 3000 g for 5 min and washed once with Brucella broth. Equal numbers of bacterial cells (5 ml) of 1 610 9 cfu/ml of each strain of C. jejuni were stabbed on top of 0.35% Mueller Hinton Agar plates (MHA Oxoid) and incubated microaerobically at 42uC for 48 h. Autoagglutination (AAG) was performed as previously described [72]. Briefly the cells were harvested with 2 ml of phosphatebuffered saline (PBS pH 7.2) and OD 600 nm adjusted to 1.0. The bacterial suspension was poured into sterile glass tubes (13 6 100 mm) and incubated at 25uC, 37uC or 42uC for 24 h. 1.0 ml of the upper aqueous phase was carefully aspirated and the OD 600 nm measured. Additionally viable bacteria were enumerated by plate counts. The lower 1 ml of solution containing the majority of the autoagglutinated cells were analysed by scanning electron microscopy (SEM). Biofilm assays were performed according to Reeser et al [73] and Tram et al [74].

In vitro adherence and invasion in epithelial cell line
Assays were performed as previously described [20] with modifications. Approximately 100 ml of bacterial suspension containing 2 6 10 8 -4 6 10 8 bacteria per ml was inoculated into 24-well plate containing a confluent monolayer of Caco-2 cells. A 5 min centrifugation step at 500 g was included to facilitate the movement of C. jejuni onto the surface of the cell monolayer. The cells were incubated for 45 min at 37uC in a 5% CO 2 humidified atmosphere to allow passive adherence and internalization. For the adherence assay, the total number of bacteria associated with the cell layer was enumerated by viable count. For the invasion assay, infected cells were washed three times with PBS and 1 ml of MEM (Dibco) containing 400 mg/ml gentamicin was added to the cell monolayer for 3 h. The monolayer was washed three times with PBS and the cells lysed with 200 ml of 0.2% Triton X-100; intracellular bacterial counts were enumerated by viable count. Each assay was performed in triplicate.

SEM of C. jejuni
Cells were grown microaerobically for 24 h at 42uC, collected and washed three times with Brucella broth by centrifugation at 1000 g for 4 min. The final wash step was carried out using PBS and the OD 600 nm adjusted to 0.025. Bacteria were fixed on plastic cover slips with 2% glutaraldehyde and 5% formaldehyde solution for 10 min. Slides were washed three times with H 2 O and dehydrated in gradient steps of water/ethanol (15,30,50,75,90 and 100%). Last steps of dehydration were performed in 50% HMDS/ethanol solution (Hexamethyldisilazane), followed by a final step of 100% HMDS. Slides were left to air dry and subsequently coated with gold (6 nm) prior to analysis on Jeol 5000 Scanning Electron Microscope.
RNA extraction, cDNA synthesis and RT-qPCR RNA was extracted using RNeasy kit according to the manufacturer's protocol (Qiagen). cDNA synthesis and RT-qPCR was performed using 11168-O, Dtlp3 and Dtlp3c as previously described [33].

Modified chemotaxis nutrient depleted agar assay
Nutrient depleted chemotaxis assay was performed as previously described [35] with the following modifications. One or two plugs (6 mm diameter) equidistantly apart (60 mm) were removed from Petri dishes containing 0.5% agar in H 2 O. Each well was filled with 0.5% agar containing 2 mM of selected amino acid. The plates were overlayed with 0.1% agar bacteriological (Oxoid) without nutritional supplements and left for 2 h to allow for diffusion of amino acids to create a chemical gradient. Cultures of C. jejuni were adjusted to OD 600 nm 1.8 (10 9 cfu/ml). 100 ml drop of bacterial suspension was inoculated using a micropipette on top of the sloppy agar (0.1% agar bacteriological in H 2 O) in the centre of the Petri dish with the plates incubated at 37uC for 4 h to allow chemotactic migration of the bacteria. To determine the number of viable bacteria associated with each amino acid plug, a 5 mm area around and including each plug was removed and placed into a microcentrifuge tube containing 900 ml of Brucella broth. These were incubated microaerobically for 1 h at 37uC to allow the bacteria to dissociate from the plug into the media. Viable counts were performed with enumeration by serial dilution followed by plate counts. C. jejuni 81116 flaA 2 /flaB 2 isogenic mutant was used as a non-motile, non-chemotactic control; additionally agar plugs containing no amino acid were used as a negative control.

Yeast two-hybrid and three-hybrid analyses
Yeast two-hybrid and three-hybrid analyses of protein interactions were performed as described previously [35]. Figure S1 Single cycle SPR curves showing concentration dependent interactions between Tlp3 and small molecules. A, Interaction between Tlp3 and isoleucine; B, Interaction between Tlp3 and lysine; C, Interaction between Tlp3 and glucosamine; D, Interaction between Tlp3 and succinic acid; E, Interaction between Tlp3 and arginine; F, Interaction between Tlp3 and purine; G, Interaction between Tlp3 and malic acid; H, Interaction between Tlp3 and thiamine; I, Interaction between Tlp3 and aspartate; J, Interaction between Tlp3 and a-ketoglutarate; K, No interaction observed between Tlp3 and alanine; L, No interaction observed between Tlp3 and asparagine. All interactions present were tested with a two-fold serial dilution between 12.5 and 200 mM. Underlined residues represent the region homologous to the shortest fragment of the E. coli serine chemoreceptor, Tsr (residues 350-471), required for CW-signalling and stimulation of CheA histidine kinase activity and are, therefore, involved in binding CheW [41]. Alignment was performed using CLUSTALW 2. Movie S1 Assessment of motility in fluorescently labelled C. jejuni 11168-O, Dtlp3 and Dtlp3c. Live imaging of culture grown C. jejuni was performed in a heated culture chamber set to 37uC using a Nikon Eclipse Ti microscope. Bacteria were grown and prepared as described in materials and methods. C. jejuni 11168-O, Dtlp3 and Dtlp3c cells were fluorescently labelled with FITC (Fluorescein-5-EX, Succinimidyl Ester, Molecular Probes) in Brucella broth and visualised on the BSA blocked glass-bottom dish at 606. Movements were captured for 15-20 s, with video footage showing normal run and tumble bias for C. jejuni 11168-O (Movie S1A), inhibition of smooth swimming with enhanced rate of tumbles by the Dtlp3 (Movie S1B) and the restored run and tumble movement by the complemented mutant Dtlp3c (Movie S1C). (ZIP)