Structures of the ligand‐binding domain of Helicobacter pylori chemoreceptor TlpA

Abstract

Bacteria use chemoreceptor proteins to sense and navigate their chemical environments. The most common class of chemoreceptors are transmembrane proteins that sense chemical cues through binding of a small‐molecule ligand to a periplasmic domain, which modulates the receptor's ability to stimulate reversal of the cell's flagella motors. The prevalent gastric pathogen Helicobacter pylori uses such membrane‐bound chemoreceptors, called transducer‐like proteins (Tlp), to colonize and persist within the stomach. TlpA has been implicated in sensing arginine, bicarbonate, and acid, but no experimentally determined protein structures of TlpA were available to better understand ligand binding and signal transduction. Here, we report three crystal structures of the periplasmic portion of TlpA, which contains tandem PAS/Cache domains, similar to a recently published structure of the lactate‐sensing chemoreceptor TlpC from H. pylori. These structures are the first to show a tandem PAS/Cache‐form chemoreceptor in its native homo dimer oligomer, and we identify residues that are key contributers to the dimer interface. We performed sequence analyses to identify TlpA and TlpC homologs and used residue conservation among these homologs to implicate regions important for the general tandem PAS/Cache fold, and residues specific to TlpA function. Comparisons with TlpC show that despite high similarity across the general structure, TlpA lacks the residues required to bind lactate, and instead contains a pocket almost entirely hydrophobic in nature.

Short abstract

PDB Code(s): http://firstglance.jmol.org/fg.htm?mol=6dtm, http://firstglance.jmol.org/fg.htm?mol=6e0a, http://firstglance.jmol.org/fg.htm?mol=6o09

Abbreviations

AU

asymmetric unit

H. pylori

Helicobacter pylori

NCS

noncrystallographic symmetry

PDB

Protein Data Bank

rmsd

root mean square deviation

Tlp

transducer‐like proteins

Introduction

Helicobacter pylori is a gastric bacterial pathogen that colonizes approximately half the world's population,1 causing gastritis, stomach ulcers, and stomach cancer.2 H. pylori—as many bacteria—uses a sophisticated chemosensory pathway to sense chemical stimuli and direct motility through a phosphorylation cascade that modulates flagella rotation.3 The most prevalent class of bacterial chemoreceptors (~85%) are inner membrane‐bound and contain a variable periplasmic ligand‐binding domain, a trans‐membrane two helix bundle, and a cytosolic coiled‐coil that interfaces with the subsequent chemotaxis proteins in the signaling cascade, CheW and CheA.4 Signal transduction by membrane‐bound chemoreceptors is still an active area of research, but it is thought that ligand binding in the periplasmic region modulates a piston‐like movement or an order/disorder transition that governs whether the downstream transmembrane region promotes CheA autophosphorylation.3, 5, 6 H. pylori possesses three such membrane‐bound chemoreceptors: transducer‐like proteins (Tlp) A, B, and C, which several studies have concluded to be required for stomach colonization and full virulence, and are candidate targets for the design of novel therapeutics.7, 8, 9, 10

Valuable insight into H. pylori chemotaxis and chemoreceptor signaling has come from crystal structures of Tlp periplasmic (Tlp_p) fragments. Ligand‐bound structures of the TlpB_p were solved by our group in 2012,11 which revealed that the receptor uses a single PAS/Cache¹ domain to bind urea and mediate a chemoattractant response. Recently, the ligand‐bound structure of the TlpC_p was also determined, and found to contain tandem PAS/Cache domains.13 This study showed that TlpC binds and recognizes lactate as a chemoattractant. TlpA, which is conserved in all H. pylori strains, has been reported to mediate chemoattraction to arginine,14 and chemorepulsion from acid.15 ΔtlpA knockout strains display defects in colonizing the stomach antrum8 and compete poorly with wild type strains.16 TlpA_p is predicted to also contain tandem PAS/Cache domains based on sequence similarity to TlpC, but previously no structural data were available to provide insights into the sensing mechanism or ligand binding. Here, we report three crystal structures, determined to 2.3–2.5 Å resolution, of the periplasmic fragment of TlpA, the final member of the H. pylori membrane‐bound chemoreceptor repertoire to be structurally characterized.

Results and Discussion

The TlpA periplasmic region forms a tandem PAS/Cache homodimer

H. pylori TlpA is a 75 kDa protein, with residues ~28–299 corresponding to the periplasmic ligand‐sensing region. Crystal trials with the periplasmic fragment from H. pylori strain SS1 (NCBI:taxid102617, NCBI protein accession number WP_077231652.1) produced well‐diffracting crystals in three separate forms (Materials and Methods) with either a monomer (TlpA₁, 2.5 Å, PDB ID: http://firstglance.jmol.org/fg.htm?mol=6dtm), a dimer (TlpA₂, 2.4 Å, PDB ID: http://firstglance.jmol.org/fg.htm?mol=6e0a), or two dimers (TlpA₄, 2.3 Å, PDB ID: http://firstglance.jmol.org/fg.htm?mol=6e09) in the asymmetric unit for a total of seven independent chains (Table 1). These structures reveal that TlpA_p contains 6 α‐helices and 11 β‐strands with tandem PAS/Cache domains [Fig. 1(A)]; a distal domain (α2‐β5, residues 76–187) and a proximal domain (α4‐β11, residues 188–292) relative to the inner membrane. Despite crystallizing in different conditions and crystal forms, all chains show high similarity, with a rmsd of 0.48 Å across 188 Cα atoms and 1.5 Å rmsd between all 251 Cα atoms. Chemoreceptors form homo dimers with or without chemoeffector present,17 and indeed the TlpA₂ and TlpA₄ structures exhibit what we expect to be its native dimer arrangement [Fig. 1(B)]. To our knowledge, these are the first structures to capture the native homo dimer for a tandem PAS/Cache‐containing chemoreceptor. The dimer interface in total buries ~1030 Ų and consists mostly of residues from α1 and α2.18 Major contributors to this interaction are residues L38, N41, F52, E56, N63, I83, and L87, which each contribute >40 Ų buried surface area per residue. Interestingly, one structure was able to crystallize as a monomer, which suggests that the interactions promoting dimer formation, at least for the periplasmic region, are rather weak. The monomer crystallized in space group C2, but did not form a biologically relevant dimer across the twofold crystallographic axis, with that interface burying only 210 Ų (classified as nonbiologically relevant by the Eppic server19). An overlay of dimers from the TlpA₂ and TlpA₄ structures showed high similarity for the overlaid monomers (0.49 Å rmsd across 221 Cα atoms), but the corresponding dimer partners exhibited surprisingly large shifts, of up to 8 Å in several regions including α1, the β1–β2 loop, and, importantly, the C‐terminal region near β11 that connects to the transmembrane region [Fig. 1(C)]. While we cannot know if these specific shifts represent conformations that are physiologically relevant, it certainly indicates the structure is flexible enough to permit large structural changes, such as could be induced by binding of a ligand, to be transduced across the membrane.

Table 1.

Summary of Crystallographic Statistics

	TlpA1	TlpA2	TlpA4
	PDB code: http://firstglance.jmol.org/fg.htm?mol=6dtm	PDB code: http://firstglance.jmol.org/fg.htm?mol=6e0a	PDB code: http://firstglance.jmol.org/fg.htm?mol=6e09
Space group	C2	P21	P21 (SeMet)
Chains in the A.U.	1	2	4
Cell dimensions and angle (a, b, c, beta) (Å,°)	143.2, 67.8, 29.4, 94.1	72.8, 59.5, 77.7, 114.1	63.5, 72.4, 132.9, 93.3
Resolution (Å)a	29.3–2.1 (2.16–2.1)	70.0–2.43 (2.47–2.43)	15–2.30 (2.34–2.30)
Completeness (%)	97.3 (91.3)	99.7 (95.9)	99.9 (99.6)
Reflection measurements	116,878	83,485	1,233,407
Reflections (unique)	16,013	23,067	53,349
Average I/σ	10.3 (1.9)	31.3 (2.8)	13.2 (4.3)
R merge b	0.129 (0.781)	0.038 (0.177)	0.15 (1.75)
R pim	0.049 (0.397)	0.023 (0.132)	0.059 (0.703)
CC1/2 a	0.996 (0.788)	1.002 (0.943)	0.998 (0.948)
Crystallographic R‐factora	0.273 (0.30)c	0.197 (0.222)	0.189 (0.210)
R‐freeb	0.33c	0.256	0.246
Protein non‐H atoms (solvent atoms)	1940 (104)	4280 (174)	7903 (426)
Average B‐factors, protein atoms (Å2)	41	37	26
Average B‐factors, solvent	28	40	32
rms bond lengths (Å)	0.017	0.008	0.014
rms bond angles (°)	1.8	0.59	1.3

^aValues in parentheses indicate statistics for the highest resolution shell.

^b R _merge = Σ_iΣ_j(I _ij − <I>_i)/Σ_iΣ_j<I>_i, where I _ij is the amplitude of the jth observation of reflection i and <I>_i is the mean value of observations I _ij.

^a For definitions of these quantities, see Karplus, P. A. & Diederichs, K., Linking Crystallographic Model and Data Quality. Science 336, 1030–1033 (2012).

^b R‐factor = Σ||F _o| − |F _c||/Σ|F _o|, where F _o and F _c are the observed and calculated structure amplitudes, respectively.

^c Model refined to 2.5 Å only due to unidentified problems with the data (Materials & Methods).

Figure 1.

The TlpA periplasmic ligand‐binding portion is a tandem PAS/Cache dimer. (A) A stereo view of a TlpA_p monomer is shown with the distal and proximal PAS/Cache domains (relative to the inner membrane) colored light blue and pink, respectively. Secondary structure elements are labeled, and the putative proximal ligand‐binding pocket is noted with an Asterix (*). (B) The TlpA₂ structure is shown with residues buried at the interface highlighted based on buried surface area for Chain A on a green–yellow–red spectrum ranging from >0 Ų (green) to 85 Ų (red) buried. Residues contributing ≥40 Ų are depicted as sticks. Key secondary structural elements are noted, with (‘) denoting elements from Chain (B). (C) An overlay of the dimer from the TlpA₂ (white) and TlpA₄ structures shows that while one set of chains overlay almost identically (0.49 Å rmsd across 221 Cα atoms, not depicted), the partner chains exhibit significant structural shifts that may be relevant for signaling (yellow). The C‐terminal region near β11 (purple) that connects to the transmembrane region (noted as “TM”) and downstream CheA signaling interface is among several regions to shift ~8 Å. (D) Shown overlaid from all structures (7 chains, 0.27 Å rmsd across 99 Cα atoms) are the residues constituting the ligand‐binding pocket (shown as surface) present in the proximal PAS/Cache. Key contributors to the pocket are shown as sticks, with carbons white, nitrogens blue, and oxygens red. The pocket is almost entirely hydrophobic, with a small polar region contributed by Tyr228 and Tyr252. NCS‐averaged 2fo–fc electron density from the TlpA₄ structure for the unidentified ligand is shown as purple mesh contoured at 1.0 σ. Structures were overlaid using Chimera.35

Analysis of the proximal and distal PAS/Cache domains

The proximal PAS/Cache domain contains a putative ligand binding pocket of ~270 ų volume, and strong electron density is present in all three crystals for an unidentified bound ligand. The pocket is identical in all chains and is lined mostly by nonpolar side chains: F203 and I204 from β7, I224 from α5, A234 and T235 from α6, L263 from β10, and A285 and I287 from β11, with the only polar interactions contributed by Y228 from α5 and Y252 from β9 [Fig. 1(D)]. The electron density is the strongest and most clear in the TlpA₄ structure, and NCS‐averaging of the ligand density from the four chains suggests that the ligand may be a small, mostly hydrophobic molecule [Fig. 1(D)]. Although several candidate ligands (e.g., n‐butane and 1,2‐pentane diol) can be fit into the density, the identity of the ligand remains inconclusive and so we have left this site unmodeled. As no such molecule was present in the protein or crystallization buffers, we expect that the bound molecule was retained from the cell lysate and purification steps, as occurred for TlpB and TlpC.11, 13

In TlpA, the distal PAS/Cache domain may also bind ligands, as is the case for several receptors of similar structure, such as TlpQ from Pseudomonas aeruginosa (PDB ID: http://firstglance.jmol.org/fg.htm?mol=6fu4) and a family of bacterial histidine kinases.20 The TlpA proximal and distal PAS/Cache domains overlay within 2.8 Å for 63 Cα atoms, but no electron density suggestive of a bound ligand is apparent in the distal domain. We observed that the β3–α3 loop, which forms part of the distal pocket, was often disordered and residues 130–134 were not modeled in some structures. We speculate that this region could be involved in ligand binding and fold upon ligand recognition. TlpC was shown to bind lactate in its proximal PAS/Cache, and no ligand was identified for the distal PAS/Cache domain, but other chemoreceptor studies have found ligands at both distal21, 22, 23, 24 and proximal sites.13 There is also speculation that periplasmic binding proteins can directly interact with PAS/Cache domains in chemoreceptors, such as may be the case for TlpB and AibA/B from H. pylori,25 and so it is possible that the TlpA proximal site may bind a small hydrophbic ligand, whereas the distal site could recognize a periplasmic protein or peptide.

We attempted to identify the mystery ligand by both mass spectrometry and thermal shift assays to a diverse (>250) array of chemicals from BiOLOG Inc., including the previously reported chemoeffector arginine,14 but the results did not support binding by arginine or other tested small molecules (data not shown). We also attempted co‐crystallization of TlpA with arginine and soaking arginine into pregrown TlpA_p crystals, but the electron density maps did not show any evidence of arginine binding. A final attempt using isothermal titration calorimetry revealed no heat change upon addition of arginine to TlpA_p (data not shown). Arginine chemoattraction was reported for H. pylori strain 26695,14 but another group was unable to reproduce this result with H. pylori strain SS1,10 the strain used in this study. There exist a few sequence differences in the TlpA ligand binding domains between these two strains (96.3% identical overall), mostly in the distal PAS/Cache, but this is to be expected as that domain is more poorly conserved. It is possible these strain‐specific sequence differences could relate to apparent differences to the strains’ responses to arginine, although variability in chemotaxis assays could also account for this result. We note that the pocket in the proximal PAS/Cache is unsuited for binding arginine, as arginine is too large and is highly polar, suggesting unfavorable interactions within the largely hydrophobic pocket. Overall, our results provide no support for direct binding of arginine by TlpA.

Distribution and sequence conservation of TlpA homologs

TlpC is 40.2% identical to TlpA by sequence, and includes tandem PAS/Cache ligand‐binding domains.13 To identify evolutionary relationships and sequence conservation that might reveal sites important for function, we performed BLAST searches using the protein sequence of TlpA to retrieve ~1000 putative tandem PAS/Cache‐containing chemoreceptor homologues from the nonredundant sequence database, which included many TlpC sequences.26 We constructed a relatedness tree and found that TlpA and TlpC cluster distinctly, with 28.6% identical sites between TlpA‐like forms, and 49.1% identical sites between TlpC‐like forms, and 0.3% identical sites across all sequences [Fig. 2(A)]. Interestingly, we discovered that TlpA and TlpC are almost exclusively found in H. pylori and H. acinonychis species, and H. cetorum has TlpA, but not TlpC, arguing for a rather recent evolution of these proteins. A third group that contains representatives from the genera Campylobacter, Geobacter, Desulfospira, Pseuodomonas, and Pseudoalteromonas, which we refer to broadly as TlpA/C‐like, have predicted tandem PAS/Cache domains, but neither possesses the lactate‐binding residues of TlpC nor cluster well with TlpA [Fig. 2(A)].

Figure 2.

Evolutionary conservation of TlpA. (A) A relatedness tree of 1024 TlpA homologues is shown with tandem PAS/Cache‐containing HpTlpA and HpTlpC noted as white circles. An unrooted relatedness tree was constructed using Geneious,36 with 66.4% pair‐wise sequence identity across all sequences. Six hundred and sixty‐five sequences clustered well with HpTlpA (pink, 95.6% pair‐wise identity), including homologs from H. acinonychis and H. cetorum. Two hundred and eighty‐nine sequences that clustered distinctly were identified as TlpCs (green, 95.3% pair‐wise identity) due to having close sequence similarity to HpTlpC and possessing the residues previously shown to be necessary for lactate binding.13 Seventy sequences that possess similar domain architecture but were disparate in sequence we group broadly here as “TlpA/C‐like” (orange). (B) Sequence conservation patterns from panel A are shown mapped onto Chain A of the TlpA₂ structure. Residues conserved in at least 90% of all sequences are shown in orange; positions conserved in at least 95% of the TlpA subgroup are shown in pink, except for residues that are also present in TlpC, which are shown in green. Residues that are not conserved are shown in white. Residues of the proximal PAS/Cache ligand binding pocket are shown as sticks in approximately the same orientation as in Figure 1(D). (C) An overlay of the TlpA_p (white) and TlpC_p (light green) monomers are shown, which overlay within 0.62 Å across 136 Cα atoms. (D) An overlay of the TlpA (white) and the lactate‐bound TlpC (light green) ligand binding pockets is shown, with atoms colored as in Figure 1(D). For each position, the residue identity and TlpA sequence number is given; for any differences in sequence with TlpC, the TlpC residue identity is stated after the sequence number, with dashes indicating a TlpC gap, and no change indicating that residue is also conserved in TlpC. Residue names are colored based on conservation as in (B). Key hydrogen bonds required for lactate binding are noted as light‐green dashes. For comparison the TlpA pocket is shown as a surface, behind the lactate.

We next utilized these sequences to highlight residues of potential importance for TlpA function by mapping conservation patterns onto the TlpA_p structure [Fig. 2(B)]. This analysis was made more difficult due to the fact that the TlpA and TlpC groups cluster tightly with little sequence diversity, and so we imposed rather strict cutoffs for our conservation analysis and grouped all residues into one of four categories: (1) positions conserved in ≥90% of all sequences (orange), (2) positions conserved only in ≥95% of TlpA‐like sequences that TlpC also contains (green), (3) positions conserved in only ≥95% of TlpA‐like sequences but absent from TlpC (pink), and (4) not conserved (white) [Fig. 2(B)]. This analysis revealed that the most highly conserved region in TlpA is the proximal PAS/Cache domain, with all residues lining the ligand‐binding pocket being either uniquely conserved in TlpA (I205, I224, A234, T235, Y228, A285) or conserved in all sequences (F203, L263, Y252) [Figs. 1(D) and 2(B)]. Not surprisingly, many residues of the core are conserved across all sequences, indicating they are important for the general fold of PAS/Cache domains, and not specific to TlpA function [Fig. 2(B)].

Comparison of TlpA and TlpC tandem PAS/Cache domains

An overlay of the TlpA_p and TlpC_p structures (PDB ID: http://firstglance.jmol.org/fg.htm?mol=5wbf) shows them to be highly similar, overlaying within 0.62 Å rmsd across 136 Cα atoms, and containing the same secondary structure features [Fig. 2(C)]. Between TlpA_p and TlpC_p, the only notable shifts occur at the N‐terminal α1 and α3 in the distal PAS/Cache [Fig. 2(C)], but these regions seem to be generally flexible as they are variable even across our three TlpA_p structures. Despite the tight global overlay and close sequence similarity, the proximal PAS/Cache pocket of TlpA does not possess the residues required for lactate binding [Fig. 2(D)]. TlpC has three key substitutions to facilitate lactate binding: I287 → Y (TlpA numbering), which hydrogen bonds to the lactate hydroxyl, a deletion in the β9‐β10 loop that brings two backbone NH groups close enough to donate hydrogen bonds to the lactate carbonyl, and Y228 → N, which would otherwise sterically clash with the lactate. These substitutions clearly show that TlpA is incapable of binding lactate and has evolved to bind some other ligand at this site. That we observe the generally nonpolar residues of the proximal TlpA PAS/Cache to be highly conserved provides further evidence this site is well suited for sensing a small hydrophobic ligand; however, further work will be required to test this prediction and determine whether ligand binding stimulates a chemoattraction or repulsion response in vivo.

Conclusion

Here we have described the first structures of the TlpA periplasmic ligand binding region and highlighted residues of importance by mapping sequence conservation onto the structure. Surprisingly, we found no evidence to support direct sensing of the previously reported chemoattractant arginine. The structures reveal TlpA contains a proximal PAS/Cache domain with a highly conserved hydrophobic pocket that is too small for arginine to bind, which was not previously recognized from sequence alone, and we speculate the protein recognizes an as of yet unidentified hydrophobic ligand. Based on our structural analyses, we identified residues that are major contributors to the homodimer interface, which represent positions that could be altered to engineer either more or less stable variants for future study. TlpA will require further structural and biochemical characterization to understand its sensing mechanism and definitively link its molecular function with in vivo observations.

Materials and Methods

TlpA_p cloning and protein purification

The periplasmic region of TlpA, TlpA_p, (residues 28–299 using UniProt27 numbering) was cloned from H. pylori SS1 gDNA (genome, NCBI:taxid102617, protein, NCBI protein accession number WP_077231652.1) using these primers: forward, 5′‐ATGTCCGGATCCGTGTCCTTAAACAGCAGGGTGAAAGAG‐3′ and reverse, 5′‐ATGTCCCTCGAGTTACGAGCCTACTTGCTCATAGACCTTGTCTTTTTC‐3′. Restriction sites BamHI and XhoI were used to subclone into a pBH vector that expressed TlpA_p with an N‐terminal 6×‐His tag under IPTG expression. JM109 DE3 E. coli was transformed with the TlpA_p pBH vector and expressed overnight at 18°C in Luria Bertani (LB) media with 1 mM IPTG. TlpA_p was purified using a standard Qiagen 6×‐His tagged protein protocol on Nickel‐NTA resin. Cells were diluted into lysis buffer containing 10 mM imidazole, 50 mM HEPES, and 300 mM NaCl pH 7.9 and lysed by sonication. The lysate was run over a gravity column pre‐equilibrated with lysis buffer, with 5 mL of Nickel‐NTA resin bed volume. The column was washed with lysis buffer to remove contaminant proteins, and then eluted with lysis buffer containing 300 mM imidazole. Prior to crystal trials, the native protein was dialyzed into 50 mM HEPES pH 7.5 and 150 mM NaCl, and the 6×‐His tag was cleaved via TEV protease, and concentrated to 12–15 mg/mL. Selenomethionine (SeMet)‐substituted TlpA_p was expressed from the same vector using methods from Van Duyne et al.28 and purified in the same manner, but the His‐tag was left uncleaved.

TlpA_p crystallization

Initial crystallization trials were performed using 96‐well crystal screens from Hampton Research, and we found TlpA_p crystallized in several conditions. We reproduced and optimized these initial hits using the hanging drop vapor diffusion method in 12‐well Linbro plates. The optimization screens had 1 mL of well solution with drops of TlpA_p (15–20 mg/mL) mixed 1:1 with well solution for a final volume of 4 μL. We were able to crystallize native TlpA_p in three different forms at room temperature: (1) as a monomer in 18% PEG 8000, 0.1 M CaOAc, 0.1 M Hepes pH 7.0 with paratone as a cryoprotectant, (2) as a dimer in 14.4% PEG 8000, 80 mM sodium cacodylate pH 6.5, 0.16 M ZnOAc, with 20% glycerol as a cryoprotectant, and (3) with two dimers (containing four chains) in 19% PEG 3350, 0.15 M NH₄NO₃, 0.15 M NaCl, 25 mM Hepes pH 7.5, 2 mM DTT with 20% ethylene glycol as a cryoprotectant. Crystals of SeMet‐substituted TlpA_p were obtained from drop #19 of the Hampton Research PEG/Ion screen, 0.2 M ammonium nitrate, 20% PEG 3350. Optimizing this crystal form for data collection required two rounds of microseeding using a horse‐tail hair (as recommended by Terese Bergfors in Protein Crystallography 29). The optimized condition was 0.15 M ammonium nitrate and 19% PEG 3350. The protein storage buffer for the SeMet protein was 25 mM Hepes pH 7.5, 0.15 M NaCl, 2 mM DTT. The crystals were equilibrated into a cryo mother liquor of 22% PEG 3350, 0.15 M ammonium nitrate, 25 mM Hepes pH 7.5, 0.15 M NaCl, 2 mM DTT, and 20% ethylene glycol.

TlpA_p structure determination

The TlpA_p construct readily crystallizes in at least three useful crystal forms, and diffraction data were collected from three representative single crystals. Ultimately all structures were solved using a combination of SIRAS and molecular replacement techniques (Table 1). As the amino acid sequence for the periplasmic region contains 10 methionine residues, SeMet‐substituted protein was prepared and crystallized. Diffraction data were collected from a single frozen monoclinic crystal (100 K) at wavelength 0.97 Å, to resolution 2.1 Å at beamline 34‐ID at the Advanced Photon Source. The Autosol component of the Phenix program package30, 31 identified 33 Se sites and produced an initial electron density map with figure of merit 0.34. The Autobuild component subsequently produced a partially traced initial model, containing several disconnected segments consisting of 774 amino acid residues of the ~1040 amino acids in the final model. Several rounds of manual model rebuilding using Coot32 and model refinement using Phenix led to the final atomic model described in Table 1. These crystals were found to contain four chains in the asymmetric unit, forming two head‐to‐head dimers. Here, we distinguish our three structures by the number of chains observed in the asymmetric unit, and so we refer to this structure as “TlpA₄.” We believe the dimers, but not the interdimer interactions, are physiologically relevant as chemoreceptors are well known to form homo dimers and trimers of homo dimers, but the orientation required for the latter is not observed in this structure.

Diffraction data for TlpA₁ was collected on a Bruker diffractometer from a single native crystal at 100 K and data were processed with the Bruker package. Diffraction data for TlpA₂ were collected on an R‐Axis IV (Molecular Structure Corp.) and processed with the HKL2000 program suite.33 Using the TlpA₄ structure as a search model, the molecular replacement component in the Phenix package was used to solve TlpA₁ and TlpA₂, which were found to contain a monomer and dimer of TlpA in the asymmetric unit, respectively. On the basis of various tests (omit maps, etc.), we are confident of the basic correctness of the TlpA₁ structure; however, the electron density map is rather poor, and resulted in a model with high crystallographic R‐factors. This data set was collected from a single crystal that was never reproduced, and may have been partially twinned. Despite the structure's modest quality, we have included it in this article because it shows the construct can exist as a monomer, and its overall features are similar to the TlpA₂ and TlpA₄ structures. Final statistics for all models are presented in Table 1.

TlpA homolog and sequence conservation analysis

BLAST searches of the TlpA periplasmic domain using default threshold values were performed on May 31, 2018, to identify 907 pylori and 191 non‐pylori putative TlpA homologs. Sequences were aligned with MUSCLE34 and manually curated to retain only sequences possessing a tandem PAS/Cache domain that are also chemoreceptors, identified by the universally conserved CheA/CheW interface. Interestingly, only Helicobacter species such as H. pylori, H. acinonychis, H. cetorum, H. felis, and H. bizzozeronii were found to possess a TlpA, TlpC, or TlpA/C‐like homolog, and further inquiries into other species such as H. hepaticus did not reveal any additional putative homologs at the standard threshold values.

Accession Numbers (PDB IDs)

The crystal structures from this work have been deposited to the Protein Data Bank with the following identifiers: TlpA₁: 6dtm; TlpA₂: 6e0a; TlpA₄: 6e09.