Helicobacter pylori chemoreceptor TlpC mediates chemotaxis to lactate

It is recently appreciated that many bacterial chemoreceptors have ligand-binding domains (LBD) of the dCACHE family, a structure with two PAS-like subdomains, one membrane-proximal and the other membrane-distal. Previous studies had implicated only the membrane-distal subdomain in ligand recognition. Here, we report the 2.2 Å resolution crystal structure of dCACHE LBD of the Helicobacter pylori chemoreceptor TlpC. H. pylori tlpC mutants are outcompeted by wild type during stomach colonisation, but no ligands had been mapped to this receptor. The TlpC dCACHE LBD has two PAS-like subdomains, as predicted. The membrane-distal one possesses a long groove instead of a small, well-defined pocket. The membrane-proximal subdomain, in contrast, had a well-delineated pocket with a small molecule that we identified as lactate. We confirmed that amino acid residues making contact with the ligand in the crystal structure—N213, I218 and Y285 and Y249—were required for lactate binding. We determined that lactate is an H. pylori chemoattractant that is sensed via TlpC with a K D = 155 µM. Lactate is utilised by H. pylori, and our work suggests that this pathogen seeks out lactate using chemotaxis. Furthermore, our work suggests that dCACHE domain proteins can utilise both subdomains for ligand recognition.

Chemoreceptors have been classified according to the size of their LBD into cluster I (~150 amino acids) or cluster II (~250 amino acids) 15 . Much of what is known about bacterial chemoreceptors comes from studies on cluster I chemoreceptors with a 4HB LBD. However, more recent studies have shown that extra-cytoplasmic LBDs of chemoreceptors from different bacteria vary largely in their amino acid sequence and three-dimensional structure 16 and, to date, additional structural families have been identified, including single CACHE (sCACHE) [17][18][19] , helical biomodular (HBM) 20,21 and double CACHE (dCACHE) domains 19,22-25 . In H. pylori, four chemoreceptors have been identified based on full genome sequence analysis: TlpA, TlpB, TlpC, and TlpD. TlpD is a soluble, cytoplasmic chemoreceptor that is involved in energy taxis 26 and the repulsion response to reactive oxygen species 27 and acid 28 . TlpA, TlpB, and TlpC are integral membrane proteins 29,30 . TlpA has been linked to recognition of bicarbonate and arginine as attractants 31 and acid as a repellent 28 , whilst TlpB has been reported to detect acidic pH 17,30 and the quorum-sensing molecule autoinducer-2 (AI-2) 32 as repellents and direct the chemoattraction response to urea 33 . No signals have been associated with TlpC.
Amongst the four H. pylori chemoreceptors, only the periplasmic LBD of TlpB has been structurally characterised 17 . It is a homodimer of sCACHE modules 17,19 -a feature that contrasts to the helical-bundle (4HB) modules of the extensively characterised aspartate receptor Tar from Salmonella typhimurium 37 , the serine receptor Tsr from E. coli 38 , and the McpS chemoreceptor from Pseudomonas putida 20 . It is now recognised that the CACHE domain, either in its single sCACHE or double dCACHE form, is the most abundant extracellular sensing domain in prokaryotes, and is commonly found in two-component histidine kinases and chemoreceptors 16,19,22,23,[39][40][41] .
TlpC is the least characterised chemoreceptor in H. pylori, and its natural ligand was unknown. H. pylori tlpC mutants are outcompeted by wild type during stomach colonisation, and TlpC modulates the chemotactic response to acid 29,36 . A BLAST search with the sequence of the sensing domain of TlpC against the structures deposited in the Protein Data Bank (PDB) identified no structural homologues of this domain. However, a pairwise comparison of profile Hidden Markov Models using the HHpred server 42 predicted homology at the level of secondary structure to the sensing domains of family 1 histidine kinases (PDB entries 3lia, 3lib, 3lic, 3lid, 3lif) 43 and chemoreceptors Tlp1 and Tlp3 from Campylobacter jejuni (PDB entries 4wy9 and 4xmr) 23,24 . These sensing modules belong to the recently redefined dCACHE_1 structural family 19 .
dCACHE domains consist of two structurally similar subdomains that each adopt a PAS-domain-like fold and are arranged in tandem, with one membrane-proximal and the other membrane-distal. dCACHE domain proteins can recognise their signal molecules directly or indirectly. Directly recognised ligands include amino acids 22,23,[44][45][46] , pyrimidines 47 and purines 48 . In all previously characterised dCACHE domains, direct sensing involves binding of the signal molecule to the membrane-distal, rather than membrane-proximal, subdomain, and no role for the membrane-proximal subdomain has been determined 22,23,43,44,49 .
In this paper, we report the crystal structure of LBD of H. pylori TlpC in complex with a small-molecule ligand that co-purified with the protein. The ligand was bound to the membrane-proximal subdomain. Based on the analysis of the electron density maps and the chemical nature of the ligand-binding pocket, we hypothesised and confirmed the ligand to be lactate. The location of the binding site has been validated by mutagenesis. We furthermore verified that lactate acts as an attractant for H. pylori, and that TlpC mediates the chemoattractant response. To the best of our knowledge, this is the first example of the dCACHE domain that directly recognises its ligand via the membrane-proximal module.

Results
Overall structure of TlpC LBD. The three-dimensional structure of recombinant H. pylori TlpC LBD (residues 34-297 plus six additional residues (GIDPFT) at the N-terminus, introduced as an artifact of the cloning procedure), was determined by X-ray crystallography using a single-wavelength anomalous dispersion (SAD) technique to a resolution of 2.2 Å. The TlpC LBD crystals (hereafter referred to as form A) belonged to the space group C2, with three molecules in the asymmetric unit related to each other by a three-fold pseudo-symmetry. The coordinates of these molecules were refined independently, and in the final model, they showed very similar backbone conformations that could be superimposed in a pairwise fashion with an overall root mean square deviation (r.m.s.d.) for the C α atoms of 0.5-0.7 Å. Disordered regions 170-175, 271-274 and 295-297 were not seen in the electron density maps and could not be modelled.
In common with family 1 histidine kinases, the TlpC LBD has a dCACHE fold, and is composed of a membrane-proximal and membrane-distal PAS-like modules folding against the N-terminal and C-terminal halves of a long stalk helix, respectively (Fig. 1b). The TlpC LBD structure comprises six α-helices and 11 β-strands (Fig. 1c). The membrane-distal subdomain (residues 63-186) contains a six-stranded antiparallel β-sheet with the strand order 2′ 2 1 5 4 3, which is flanked on one side by an antiparallel two-helix bundle formed by helix α2 and the C-terminal half of helix α1, and on the other side by helix α3. The membrane-proximal subdomain (residues 34-62, 189-292) contains a five-stranded antiparallel β-sheet with the strand order 7 6 10 9 8. This β-sheet is flanked by an antiparallel two-helix bundle formed by helix α4 and the N-terminal half of helix α1 on one side, and by helices α5 and α5′ on the other side. Finally, an additional helix α6 forms an extension of strand β10 at the C-terminal end of TlpC LBD. The membrane-distal and membrane-proximal subdomains are intimately associated with each other, with a total buried surface area of 1169 Å 2 , which is equivalent to 16% of the total buried surface area (BSA) of an individual subdomain.
Analysis of the packing of monomers in the crystal lattice revealed head-to-tail arrangement of molecules, where the membrane-proximal subdomain of one subunit packs against the membrane-distal subdomain of the other, which is not likely to represent a physiologically relevant assembly. To determine the oligomeric state of TlpC LBD in solution, size-exclusion chromatography coupled to multi-angle light scattering (SEC-MALS) experiments were performed. TlpC LBD eluted as a single monodispersed peak in all conditions tested. The derived molecular weight of 28.8 kDa was close to the value calculated from the amino acid sequence of a monomer (30 kDa). Furthermore, the apparent hydrodynamic radius R h of the particles in this peak (25 Å) was close the R h value calculated from the crystal structure of a single TlpC LBD subunit (26 Å). In line with these results, analysis of the crystal packing using the PDBe PISA server (http://www.ebi.ac.uk/pdbe/pisa/) identified no quaternary structure and suggested that TlpC LBD is a monomer in the crystal.
SCIentIfIC RepORTs | 7: 14089 | DOI:10.1038/s41598-017-14372-2 Analysis of putative ligand-binding sites. We next examined the putative ligand binding sites, starting with the membrane-distal module. Inspection of the structure of the membrane-distal subdomain around the region implicated in binding of small-molecule ligands in other dCACHE-containing proteins revealed a well-defined groove that runs along the full length of the β-sheet and is flanked on one side by helix α3 and the stretch of amino acid residues connecting β2′ and α3, and on the other side, by the β3-β4-tongue ( Fig. 3a  and b). This cleft is lined by mostly aliphatic and small hydrophilic residues and has the following approximate dimensions: 30 Å in length, 11 Å in width, and 9 Å in depth. Structural comparisons show that the groove in the membrane-distal domain of TlpC is significantly larger than the small pocket present in Tlp3 and other dCACHE sensing domains that recognise their small-molecule ligands directly 23,24 . Superimposition of the membrane-distal subdomains of TlpC and Tlp3 over 121 C α atoms (r.m.s.d. of 2.9 Å) shows that helix α3 and the β3-β4-tongue in TlpC are positioned significantly further apart than the equivalent α-helix and β-tongue in Tlp3 (Fig. 3b). The cleft in TlpC appears too large to be a small-molecule-ligand binding site and could hypothetically fit a molecule of the size of a peptide, such as, for example, a loop or a terminal peptide of an-as-yet unidentified PBP.
We then analysed the molecular surface of the membrane-proximal subdomain of TlpC LBD using CASTp 54 with a probe radius of 1.4 Å. We detected a putative ligand-binding pocket with the surface area and solvent-accessible volume of 203 Å and 196 Å 3 , respectively (Fig. 3a). There was a clear electron density for a non-protein molecule bound in this pocket (Fig. 4). However, its shape did not match any of the components of the purification or crystallisation buffers, which suggested that the ligand trapped in the crystal could be a molecule that was present in the refolding mix or a product of proteolytic degradation of the sample.

Identification of lactate as a ligand for TlpC.
To identify the ligand captured by TlpC LBD, the protein was denatured to release the small molecules, and these were analysed by liquid chromatography-electrospray ionisation mass spectrometry (LC-ESI-MS). The negative ionisation mode MS data showed a small peak at m/z = 89.022 that was absent in the buffer control ( Supplementary Fig. 1). Within the experimental error, this   The (mFo -DFc) σ A -weighted electron density for lactate is shown in green. The map was calculated at 2.2 Å resolution and contoured at the 3.0 σ level. The lactate molecule is shown in all-atom ball-and-stick representation with C atoms coloured orange. The protein side chains that form direct contacts with lactate are shown in stick representation.
shape of the electron density in the membrane-proximal pocket. As lactate is the only one of these four compounds that is a natural metabolite produced during E. coli growth, we hypothesised that during refolding, the lactate was captured by the protein from the cell lysate.
Isothermal titration calorimetry (ITC) measurements confirmed that L-lactate binds exothermically to TlpC LBD with an apparent K D of 155 μM ± 5 μM (Fig. 5). The binding is driven by a favourable enthalpy change (∆H = −20 kcal mol −1 ) and is associated with a minor unfavourable entropy change (T∆S = −1.29 kcal mol −1 ). This binding appears specific to lactate because no significant heat release or absorption was observed with pyruvate, malate or oxaloacetate, that are chemically similar and metabolically exchangeable with lactate ( Supplementary Fig. 2).

Validation of lactate binding site in membrane-proximal module of TlpC dCACHE domain.
To establish whether lactate binds to the membrane-distal or membrane-proximal module, we determined the crystal structure of TlpC dCACHE domain co-crystallised with 10 mM L-lactate. The co-crystals with lactate were isomorphous to the form A crystals grown with no lactate in the crystallisation mix. Superposition of the protein contents of the two asymmetric units based on the overlap of 767 C α atoms with an r.m.s.d. of 0.32 Å showed that, within the limit of the experimental error in the coordinates (0.33 Å for the co-crystal with lactate), their structures were essentially identical. Analysis of the electron density maps revealed no lactate binding sites other than the one in the membrane-proximal subdomain. This subdomain contained a lactate molecule bound in a very similar mode to that observed in the form A crystals grown with no added lactate (Fig. 4).
The lactate binding site is located in a pocket formed by residues F202, L210, N213, I218, L223, Y249, L252, S253, and Y285. Calculation of the accessible surface area (ASA) showed that lactate becomes almost completely shielded from the solvent upon binding to TlpC LBD, with 99.5% of its ASA buried by the protein. The carboxyl and hydroxyl groups of lactate form hydrogen bonds with the side chains of N213, Y249 and Y285, and with the main-chain amides of L252 and S253. The TlpC LBD/lactate complex is further stabilised by hydrophobic interactions between the methyl group of lactate and the side chains of F202, L210, I218 and L223 (Fig. 4).
To evaluate the contribution of individual amino acids to the lactate binding, N213, I218 and Y285 were individually replaced with alanine and Y249 with phenylalanine, and the effect of the single-amino acid substitutions was assessed by isotitration calorimetry. Comparison of the circular dichroism spectra of the variants with that of native TlpC LBD showed no significant differences, indicating that the amino acid substitutions did not alter the secondary structure (Fig. 6). ITC measurements demonstrated that each of the N213A, I218A, Y285A and Y249F substitutions abolished the binding of lactate to TlpC LBD ( Fig. 4 and Table 1). To further confirm that the membrane-distal subdomain does not bind lactate, the TlpC residues S104, Y151 and K153 -occupying the positions structurally equivalent to the ligand-binding residues Tyr118, Tyr167 and Thr170 in the membrane-distal subdomain of Tlp3 -were individually substituted with alanine. In contrast to the effect on the binding to the membrane-proximal domain, these substitutions only resulted in 2-3 fold reduction in the affinity to lactate, likely due to partial fold destabilisation rather than loss of interactions with the ligand.
H. pylori TlpC mediates positive chemotactic response to lactate. To test the physiological relevance of the observed specific interaction between TlpC and lactate, we assessed the lactate chemotactic response of H. pylori. Although there are isolates of the laboratory H. pylori strain 26695 that are motile 55 this strain is prone to motility loss and difficult to use in motility evaluation. We therefore used the human isolate pre-mouse SS1 (PMSS1) 56 , which displays a high level of reliable motility, and has been studied for chemotaxis responses in recent publications 28 . The TlpC ligand-binding domain from 26695 and PMSS1 are identical ( Supplementary  Fig. 3), so we reasoned both proteins would respond similarly to lactate. We assessed whether lactate is an H. pylori attractant or repellent using a swimming assay that enumerates flagellar-based bacterial reversals, a common read out for a chemotactic response 57 . In this assay, attractants cause decreased and repellents cause increased direction changes 26,27,32 . Wild-type H. pylori showed a significant response to 0.1 mM lactate in this assay, but lost the response at higher concentrations (Fig. 7).
To account for possible chemotactic effects due to pH change upon sodium L-lactate treatment of BB10, the pH of the media with and without treatment was assessed. While treatment with 10 mM HCl decreased the pH by more than 1.5 pH units, treatment with any concentration of sodium L-lactate only decreased the BB10 pH by less than 0.05 pH units. Furthermore, the pH difference between the highest and lowest amount of sodium L-lactate (10 mM and 0.1 mM) was only ~0.01. This analysis suggests that sodium L-lactate did not substantially change the medium pH, and thus any effect due to pH change upon sodium L-lactate treatment were likely negligible compared to chemotactic effects due to sodium L-lactate itself. We next examined whether TlpC was required for this lactate chemotaxis response. We generated an isogenic null mutant strain lacking tlpC (∆tlpC). tlpC is part of a single gene operon 58,59 , and thus mutations are unlikely to have polar effects. Consistent with this idea of specificity, the ∆tlpC mutant retained chemotaxis responses to signals that act through other chemoreceptors: the known repellent HCl that acts through TlpB, TlpA, and TlpD, as well as the attractant dipyridyl that acts through TlpD 27,28,30 (Fig. 7). These responses confirmed that the ∆tlpC mutant was not generally chemotaxis defective. Deletion of tlpC, however, abolished the chemotactic response to lactate at all tested concentrations (Fig. 7). Additionally, while ∆tlpC PMSS1 H. pylori displayed fewer reversals on average compared to WT PMSS1, this difference in basal reversal frequency was not significant (Fig. 7). This data thus supports that TlpC is the chemoreceptor for lactate in H. pylori.

Discussion
dCACHE domains of bacterial chemoreceptors and histidine kinases consist of two subdomains, membrane-distal and membrane-proximal, each of which could, in principle, harbour a binding site for a small signal or regulatory molecule. In all previously characterised dCACHE domains, including LBDs of Tlp3 from C. jejuni 23 22,45 , and Mlp24 from V. cholerae 60 , direct sensing involved binding of the signal molecule to the membrane-distal subdomain. Our analysis of the structural basis of lactate recognition by H. pylori chemoreceptor TlpC changes this paradigm regarding the mechanism of sensing by dCACHE domain by providing the first example where direct sensing of the signal molecule is mediated by the ligand binding to the membrane-proximal, rather than membrane-distal, subdomain.
This result has important implications for the conceptual framework of dCACHE-mediated sensing and signal transmission across the membrane. As this and previous studies showed, the membrane-distal and membrane-proximal modules of dCACHE are intimately associated with each other, and are therefore structurally and dynamically coupled 23,24 . For example, we previously demonstrated using X-ray crystallography that, upon binding to an attractant, the dCACHE membrane-distal subdomain of C. jejuni Tlp3 closes around the ligand and loses its tight association with the membrane-proximal domain, which, as a result, adopts a more open conformation 23 . The structural coupling of the membrane-distal and membrane-proximal subdomains is  Table 1. Thermodynamic parameters of lactate binding to TlpC LBD and its variants derived from ITC measurements.

Figure 7.
Lactate triggers a TlpC-dependent attractant response in H. pylori. Cultures of isogenic wild-type (WT) and ΔtlpC (∆TlpC) H. pylori PMSS1 were grown in BB10 overnight, treated with various concentrations of sodium L-lactate or control compounds as indicated, and then immediately filmed. Direction changes were enumerated over a 3 second period in at least 200 cells per treatment in two biological replicates. 10 mM HCl or 50 μM dipyridyl serve as known repellent and attractant response controls, respectively. Error bars represent the standard error of the mean. *p < 0.05; **p < 0.01, comparisons performed using a two-way ANOVA, followed by Tukey's pairwise comparisons (α = 0.05). There were no significant differences in the basal behaviour between wild type and its tlpC mutant.
consistent with the finding that signalling across the membrane presumably can be triggered by direct ligand binding to either subdomain -the membrane-proximal subdomain, as in TlpC, or membrane-distal subdomain as in Tlp3, Mlp37, McpB or McpC. Furthermore, one cannot eliminate the possibility that different ligands may signal through the same receptor, with some binding to the membrane-distal and others to the membrane-proximal subdomain. Indeed, all membrane-proximal subdomains of dCACHE sensing domains characterised to date contain a putative small ligand-binding pocket, including those that sense ligands with the membrane-distal domain 23,24,43,51 . For example, the structural study of LBD of C. jejuni chemoreceptor Tlp1 revealed an acetate ion bound to the dCACHE membrane-proximal module 24 . Acetate has not yet been found to trigger a chemotaxis response in C. jejuni 61 . However, it has been shown to induce either a positive or negative chemotactic response in other species 18,[62][63][64] . Apart from implicating the membrane-proximal, rather than membrane-distal, subdomain in direct sensing of a small molecule ligand, our crystallographic analysis revealed one more difference between the dCACHE domain of TlpC and that of other structurally characterised chemoreceptors of this type -the presence of a long groove in the membrane-distal subdomain, instead of a small well-defined pocket 23,43,51 . This groove might represent a putative binding site for a larger molecule, such as an intermediate ligand-binding protein from the periplasmic binding protein (PBP) family. PBP-mediated sensing is used for chemotaxis in other bacteria, where some PBPs have a function of a primary chemoreceptor that recognises and binds a small molecule in the periplasm, and, in the ligand-bound form, associates with its cognate, membrane-bound transducer-like protein, initiating the signal 65 . There are at least six putative PBPs encoded in the H. pylori genome which have been shown to bind autoinducer 2 66 and nickel 67 , and proposed to bind other compounds including peptides 68 , molybdenum, amino acids, and iron 69 . Whilst the membrane-proximal domain of TlpC mediates direct sensing of lactate, its membrane-distal domain -the shape of which does not imply small-molecule binding -may partner with a PBP to sense some other ligand.
Our analysis of the chemotactic behaviour of wild-type H. pylori showed that lactate induced an attractant response in a concentration-dependent manner, and that this response was drastically reduced in a chemotactically competent isogenic ΔtlpC mutant, demonstrating that TlpC is the primary chemoreceptor for lactate in H. pylori. Within the tested range of concentrations of lactate, the response was strongest at 0.1 mM, detectable at 1.0 mM, and not detectable at 10 mM. Putting the observed TlpC-dependent chemotactic behaviour towards L-lactate in the context of the receptor-ligand interactions, we note that the order of the optimal concentration at which lactate is sensed by H. pylori as attractant (0.1 mM) is the same as the order of the dissociation constant K D (0.155 mM) for its binding to the TlpC dCACHE domain. L-lactate, secreted by gastric mucous cells, reaches the concentration of 0.3-1 mM in gastric juice 70,71 . Presumably lactate forms a gradient with its highest amount at the source, the cells, but the stomach distribution of lactate is not known. Lactate is known to promote H. pylori growth in the stomach 72 and in media that is lacking dextrose, suggesting it can serve as either a carbon or energy source, or both 73 . Metabolically, lactate can be generated by lactate dehydrogenase (LDH) from pyruvic acid as the end product of glycolysis. However, LDH can also catalyse the reverse reaction, converting lactate into pyruvate 74,75 . Thus, if exogenous lactate was imported into H. pylori, it could be oxidised into pyruvate, which would then enter the tricarboxylic acid cycle. Alternatively, lactate can donate electrons to NADH and, in turn, to the electron transport chain to enhance proton motive force and bacterial energy levels 76 . In H. pylori, the proteins necessary for the import and utilisation of lactate have been identified 69,73 , including two lactate permeases and two LDHs 73 . At least one of the lactate utilisation genes has been shown to be required for stomach colonisation, supporting the importance of this process in vivo 77 .
TlpC mutants have mouse stomach colonisation defects but only when competing with wild type 29 . This phenotype is consistent with the idea that either lactate is limiting and wild type utilises it more efficiently, or that wild type follows a lactate gradient and occupies key niches before the tlpC mutant can get there. Thus, the in vivo fitness defect observed with isogenic TlpC mutants is consistent with an inability to efficiently access and catabolise lactate in vivo 29 .
Interestingly, several H. pylori lab strains appear to lack TlpC protein expression 27 . One of these strains, G27, has a single base indel that creates a frameshift and results in loss of TlpC expression 27 . Another, B128, also has a tlpC frameshift but its gerbil-selected daughter strain, 7.13, has regained TlpC expression 27 . These findings suggest that TlpC is not required for lab growth. The stomach may provide different selective pressures such that TlpC expression is an advantage. In support of this idea, four clinical H. pylori isolates analysed all expressed TlpC 27  Our observation that a higher concentration of lactate (10 mM) did not elicit a positive chemotactic response in H. pylori is in agreement with the reports that, at a concentration of 10 mM or above, lactate has an inhibitory effect on H. pylori growth 72,80 . The anti-H. pylori activity of high levels of exogenous lactate was first observed in co-cultures with lactic acid bacteria (LAB) [81][82][83] . Lin et al. 84 demonstrated that short chain fatty acids (SCFA) (acetic, formic, propionic, butyric and lactic acid) secreted by LAB, and the associated low pH values reduce H. pylori viability, with lactic acid exhibiting the strongest inhibitory effect out of all tested SCFA 84,85 . Although the exact mechanism by which high levels of lactic acid exert anti-H. pylori activity remains unclear, it is likely a combination of its inhibitory effect on H. pylori urease activity and the reduced ability of H. pylori to survive at low pH in the absence of urea 80,84-87 .
Full-length chemoreceptors function as trimers of dimers 88 . Although the degree and mechanism of the contribution dCACHE LBDs make to oligomerisation in vivo remains to be established, previous crystallographic studies on the dCACHE domains of C. jejuni Tlp3 23 , M. mazei HK1 s -Z3 43 and V. cholerae DctB 52 suggested that they likely dimerise through their stalk helix, with the twofold axis approximately perpendicular to the membrane plane. The dimerisation forces between isolated dCACHE domains are weak, as all domains of this type characterised so far, including that of H. pylori TlpC (this study), Tlp1 and Tlp3 from C. jejuni 23,24 , CtaA and CtaB from Pseudomonas fluorescens 89,90 , VfcA from Vibrio fischeri 91 , and PctA from P. putida 92 , are monomeric in solution. Our analysis showed that the isolated recombinant dCACHE LBD of TlpC is monomeric in the crystal as well. However, the observed structural similarity between LBDs of TlpC, Tlp3, HK1 s -Z3 and DctB allows for the possibility that, in the context of the membrane-embedded full-length receptor, TlpC LBD may also dimerise through its stalk helix.
In conclusion, this study reports the first example of the dCACHE type chemoreceptor that directly senses its ligand via its membrane-proximal subdomain, and that H. pylori seeks out lactate using chemotaxis. It adds to the mounting evidence that dCACHE sensing domains have evolved to recognise their ligands via several different direct and indirect mechanisms that may utilise either the membrane-distal, or the membrane-proximal, subdomain, or both. This raises an intriguing question about whether, despite this diversity, different dCACHE sensing domains share a common mechanism of signal transduction across the membrane.

Methods
Site-directed mutagenesis, protein expression, and purification. The expression vectors for single-point variants of TlpC LBD in which S104, Y151, K153, N213, I218 or Y285 were replaced by alanine, and Y249 by phenylalanine, were prepared from a TlpC-expressing plasmid described previously 93 . This plasmid expresses codon-optimised H. pylori 26695 TlpC LBD consisting of amino acid residues 34-297 (Fig. 1a). Mutants were created using the QuikChange Mutagenesis Kit (Stratagene). TlpC LBD and its variants were expressed and purified following the previously published procedure 93 . Crystallisation, data collection and structure determination. Form A crystals of TlpC LBD were obtained as described 93 . The crystals grew in space group C2 (Table 2) and contained three monomers in the asymmetric unit. Co-crystallisation with 10 mM sodium L-lactate under similar conditions produced the crystals of the TlpC LBD/lactate complex that were isomorphous with form A crystals ( Table 2). Two platinum derivatives were obtained by soaking the TlpC LBD crystals in either potassium tetrachloroplatinate (1 mM) or potassium hexachloroplatinate (1 mM). The derivative crystals belonged to space group P321 (form B) with a monomer in the asymmetric unit (Table 2). Native X-ray diffraction data (λ = 0.95 Å) and SAD data for the derivatives (λ = 1.07 Å) were collected on the MX1 and MX2 beamlines of the Australian Synchrotron (AS) 94 . All data were processed with iMOSFLM 95 and scaled with AIMLESS 96 from the CCP4 software suite 97 ( Table 2).
The two isomorphous SAD data sets were used to locate the platinum sites and calculate the phases for the form B crystals with Autosol 98 from the PHENIX software suite 99 . The resulting phase set (overall figure of merit of 0.30 for data between 51.4 and 3.3 Å) was used to generate an initial partial model for a TlpC LBD monomer with AutoBuild (PHENIX). This model was used for phasing the form A data for the native crystal and the co-crystal with lactate by molecular replacement. Refinement statistics and stereochemistry are given in Table 3. For both models, all the non-glycine residues lie in permitted regions of the Ramachandran plot, with 97% of these in the most favoured regions.
SEC-MALS analysis. The hydrated molecular mass and hydrodynamic radius of TlpC LBD in solution were determined by SEC-MALS. Protein was dialysed against buffer A containing 100 mM Tris-HCl pH 8.0 and 150 mM NaCl, and concentrated to 3 mg ml −1 . A 100 µl sample was loaded onto a WTC-030S5 SEC column (Wyatt Technology Corporation) pre-equilibrated with the same buffer flowing at 0.4 ml min −1 . The eluate was passed through an inline DAWN HELEOS light scattering detector, an Optilab T-rEX differential refractive index  detector and a quasi-elastic light scattering detector (WyattQELS, Wyatt Technology Corporation). The experiment was repeated in the presence of 10 mM sodium L-lactate. Calculations of the molecular mass and hydrodynamic radius from the intensity of the scattered light and refractive index were performed using ASTRA 6.0 (Wyatt) ( Table 4). Theoretical calculations of the hydrodynamic radius from the crystal structure were carried out using HYDROPRO version 10 100 . Chloramphenicol-resistant mutants were selected using 10 μg/ml chloramphenicol on Columbia Horse Blood Agar as previously described 29 . Mutation of tlpC was confirmed by PCR and western blot.   Chemotaxis assay. Swimming behaviour assays were done with H. pylori PMSS1 strains described above grown in Brucella broth (BD BBL/Fisher) with 10% FBS (Life Technologies) (BB10), with shaking, at 37 °C, under microaerobic conditions of 5% O 2 , 10% CO 2 , balance N 2 . Overnight cultures (~OD 600 0.25-0.5) were diluted to an OD 600 of 0.1 in fresh BB10, and then incubated as above until an OD 600 of 0.12-0.15 was reached. Motile, OD 600 0.12-0.15, cultures were treated with sodium L-lactate (0.1 mM, 1 mM, 10 mM) or an equal volume of H 2 O as an untreated control. As a repellent control, 10 mM HCL was used as done previously 30,32,36 . As an attractant control, 50 μM dipyridyl was used as done previously 27 . Dipyridyl results in fewer directions changes, an attractant response, dependent on chemotaxis in general and TlpD specifically. Dipyridyl induces and attractant response as it counters reactive oxygen species via chelation of iron 27 . The pH of BB10 upon treatment was independently assessed using a Denver Instruments pH meter. Cultures were filmed immediately after ligand addition at 400x magnification using a Hamamatsu Digital Camera C4742-95 with the μManager software (Version 1.4.22), mounted on a Nikon Eclipse E600 phase contrast microscope 103 (Supplementary videos 1-12). Videos were relabeled to blind the observer to the strain identity. For each sample, >100 3-s-long bacterial tracks from two independent cultures were analysed manually to identify stops followed by direction changes and to calculate the average number of direction changes in 3 s. Statistical analysis of the data for treated versus untreated samples was performed using a Student's t-test.

LC-ESI-MS analysis.
PDB submission codes. The atomic coordinates and structure factors of the TlpC LBD/lactate complex obtained at 2.2 Å resolution have been deposited in the Protein Data Bank (http://www.rcsb.org) under accession code 5wbf.