The Structural Analysis of the Periplasmic Domain of Sinorhizobium meliloti Chemoreceptor McpZ Reveals a Novel Fold and Suggests a complex Mechanism of Transmembrane Signaling

Abstract

Chemotaxis is a fundamental process whereby bacteria seek out nutrient sources and avoid harmful chemicals. For the symbiotic soil bacterium Sinorhizobium meliloti, the chemotaxis system also plays an essential role in the interaction with its legume host. The chemotactic signaling cascade is initiated through interactions of an attractant or repellent compound with chemoreceptors or methyl-accepting chemotaxis proteins (MCPs). S. meliloti possesses eight chemoreceptors to mediate chemotaxis. Six of these receptors are transmembrane proteins with periplasmic ligand-binding domains (LBDs). The specific functions of McpW and McpZ are still unknown. Here, we report the crystal structure of the periplasmic domain of McpZ (McpZ_PD) at 2.7 Å resolution. McpZ_PD assumes a novel fold consisting of three concatenated four-helix bundle modules. Through phylogenetic analyses, we discovered that this helical tri-modular domain fold arose within the Rhizobiaceae family and is still evolving rapidly. The structure, offering a rare view of a ligand-free dimeric MCP-LBD, reveals a novel dimerization interface. Molecular Dynamics calculations suggest ligand binding will induce conformational changes that result in large horizontal helix movements within the membrane-proximal domains of the McpZ_PD dimer that are accompanied by a 5 Å vertical shift of the terminal helix toward the inner cell membrane. These results suggest a mechanism of transmembrane signaling for this family of MCPs that entails both piston-type and scissoring movements. The predicted movements terminate in a conformation that closely mirrors those observed in related ligand-bound MCP-LBDs.

Introduction

Chemotaxis is the process used by bacterial cells to seek out nutrient sources and eukaryotic hosts ^1–5. In addition, chemotaxis signaling pathways facilitate bacterial escape from environmental pollutants and other harmful chemicals, known as repellents ^6,7. Escherichia coli serves as the classical model system for elucidating the underlying molecular mechanisms of the chemotaxis system ^6,8. Universally, signaling is initiated by chemoreceptors or methyl-accepting chemotaxis proteins (MCPs), a group of bacterial receptors that detect environmental changes through specialized receiver domains ^6,9. Canonical MCPs are multidomain proteins containing a variable periplasmic sensory domain coupled to two conserved transmembrane helices and a large equally conserved cytoplasmic region ^4,8,9. Ligand binding regulates the autophosphorylation activity of the MCP-associated kinase CheA, which, in turn, controls phosphorylation of its cognate response regulator CheY ¹. Binding of CheY-P to the cytoplasmic base of the flagellar motor causes the peritrichous flagella to rotate in a clockwise (CW) direction ultimately resulting in a tumbling motion of the bacterial cells. Presence of an attractant causes CheA inhibition, reversal to counterclockwise (CCW) flagellar rotation, and smooth swimming toward the nutrient source. A complex adaptive system ensures that receptors do not become saturated as the bacterial cell migrates up the nutrient gradient ^4,10–12. Due to the complexity of chemotaxis systems and a scarcity of direct structural information for the transmembrane regions, many questions remain regarding the mechanism of signal transduction.

Bacterial chemotaxis also plays an important role in establishing the symbiotic relationship between the soil bacterium Sinorhizobium meliloti and its legume host alfalfa ^13,14. S. meliloti possesses eight chemoreceptors, sequentially named McpT to McpZ and IcpA (Internal Chemotaxis Protein A) ^15–17. McpZ is predicted to have an extraordinarily large and structurally distinctive periplasmic domain for which the ligand is unknown. To identify regions critical for ligand recognition and signal transduction, we characterized the crystal structure of the periplasmic domain of McpZ (McpZ_PD). The structure of McpZ_PD, solved at 2.7 Å resolution, revealed a dimeric protein with a novel helical trimodular (HTM) fold. Phylogenetic analysis suggests that this new receptor class emerged relatively recently in the Rhizobiaceae family. The HTM fold appears to have arisen through either internal gene duplication within a gene encoding a helical bimodular (HBM) receptor ¹⁸ or the insertion of an unrelated gene fragment into an HBM encoding sequence.

The distinctive dimeric structure of ligand-free McpZ_PD served as a starting point for molecular dynamics (MD) calculations to explore the conformational space of the protein and gain insights into the potential signaling mechanism. Intriguingly, the observed movements suggest McpZ seamlessly integrates elements of the two prevailing models for MCP cross-membrane signaling.

Material and Methods

Expression and purification of the McpZ_PD

The mcpZ_PD coding sequence, encompassing the codons for amino acids 39 to 424 of the original mcpZ gene, was expressed from plasmid pTYB11 in E. coli ER2566 to produce a fusion protein containing a self-cleaving N-terminal intein-chitin binding domain (intein-CBD) tag. Four liters of cell culture were grown at 37 °C in LB containing 100 μg of ampicillin/ml until they reached an OD₆₀₀ of 0.8. Gene expression was induced by adding 0.6 mM isopropyl-β-D-thiogalactopyranoside (IPTG) and continuing growth at 16 °C for an additional 16 hours. Cultures were then centrifuged, and cell pellets were collected. Cells were suspended in buffer A (20 mM Tris/HCl, 500 mM NaCl, 1 mM EDTA, and 10 % glycerol [pH 8.0]) supplemented with 10 mg/ml of DNase, 1 mM PMSF, and 1 × Halt™ Protease Inhibitor Cocktail (Thermo Fischer Scientific). Cells were lysed with a French pressure cell (SLM Aminco, Silver Spring, MD) at 16,000 lbs/inch², prior to clearing the lysate via centrifugation at 56,000 × g for 1 hour at 4 °C. The clear supernatant was filtered and loaded onto a chitin (New England BioLabs) affinity column. On-column cleavage of the intein-CBD-tag was performed by incubating the bound sample in cleavage buffer (buffer A supplemented with 50 mM DTT) for 72 hours at 4 °C. The protein was eluted with buffer A and concentrated using a 50 ml Amicon stirred cell (Millipore, Bedford, MA, U.S.A) with a 3 kDa MWCO regenerated cellulose membrane. The concentrated protein was further purified by loading the sample onto a HiPrep 26/60 Sephacryl S-200 HR column (GE Healthcare Life Sciences) that was pre-equilibrated in buffer B (100 mM HEPES, 100 mM NaCl, pH 7.0) or in buffer C (100 mM tricine, 150 mM NaCl, 1 mM EDTA, pH 8.0) for DSF and SEC-MALS experiments. Purity of the eluted protein was assessed via SDS-PAGE. Fractions containing only McpZ_PD were pooled and concentrated to 120 μM via a 50-ml Amicon stirred cell (Millipore, Bedford, MA, U.S.A) with a 3 kDa MWCO regenerated cellulose membrane and stored at -80 °C. The chromatogram for the elution from HiPrep 26/60 Sephacryl S-200 HR column and an SDS-PAGE of the elution peak are provided in Fig. S1a.

Differential scanning fluorimetry

For the differential scanning fluorimetry (DSF) assay, a 30 μl solution containing 50 μM McpZ_PD protein in buffer C and 2X SYPRO orange was pipetted into a 96-well optical plate. Thermal denaturation was performed in an ABI 7300 real-time PCR thermocycler (BioRad) by increasing the temperature from 10 to 95°C with a 30 s equilibration at each half degree Celsius. The melting temperature (T_m) was calculated by identifying the minimum of the negative first derivative of fluorescence intensity values. The described melting curve is provided in Fig. S1b.

Multi Angle Light Scattering Analysis

Size exclusion chromatography coupled with multi angle light scattering (SEC-MALS) was employed to determine the oligomeric state of McpZ_PD in solution. To this end, 40 μM McpZ_PD protein in buffer C was applied as a 100-μl injection to a Superdex 200 Increase 10/300 GL column (Cytiva) pre-equilibrated in the same buffer at a flow rate of 0.3 ml/min by an AKTA pure FPLC (Cytiva). The eluate was passed through an inline miniDAWN light scattering detector and Optilab differential refractive index (dRI) detector (Wyatt Technology Corporation). Calculations of the molecular mass from the intensity of scattered light and dRI measurements were performed using ASTRA 8.1 (Wyatt Technology Corporation). The SEC-MALS data are summarized in Fig. S1c.

Crystallization, X-ray diffraction data collection, structure determination, model building and refinement

Crystals of McpZ_PD were obtained through high-throughput screening of commercially available crystallization conditions using a sitting drop format. The hexagonal-shaped crystals used for diffraction data collection were obtained directly from the Morpheus II screen (Molecular Dimensions). The crystallization drop contained a 1:1 mixture of 0.12 mM McpZ_PD solution in buffer B and a crystallization screen condition composed of 2 mM Lanthanides (0.5 mM each Yttrium (III) chloride hexahydrate, Erbium(III) chloride hexahydrate, Terbium(III) chloride hexahydrate, and Ytterbium(III) chloride hexahydrate), 0.1 M MOPSO, Bis-Tris pH 6.5, and 31 % v/v Precipitant Mix 8 (10% w/v PEG 20000, 50% w/v Trimethylpropane, 2% w/v Non-detergent Sulfobetaine-195). Crystals grew over a 10-day period at room temperature. The crystals were transferred into a 2 μl drop of cryo-protectant obtained by mixing 70 μl mother liquor with 30 μl of a solution containing 20% ethylene glycol and 2M NDSB-201. Crystals were rapidly mounted and flash-frozen in liquid nitrogen.

Diffraction data collection was performed at the Advanced Light Source Beamline 4.2.2 at Lawrence Berkeley National Laboratory. Because the crystallization condition contains a mixture of Lanthanides several X-ray wavelengths were tested to optimize a potential anomalous signal to facilitate single anomalous dispersion (SAD) phasing. Ultimately, a wavelength of 1.0 Å was chosen to collect a 2.7 Å diffraction data set. Ten ordered heavy atoms in the crystal, all modelled as Yb(III), gave a strong anomalous signal that was used to generate a high-quality SAD-phased electron density map with the AutoSol routine of the PHENIX software package ¹⁹. The excellent quality of the initial electron density map permitted the building of almost the entire protein backbone using the AutoBuild routine of PHENIX. Iterative cycles of model building in WINCOOT ²⁰ and automated refinement with PHENIX were subsequently employed to construct and refine the final model. Diffraction data and refinement statistics are provided in Table 1. The final model has been deposited in the protein data bank with the PDB code 8F7N.

Table 1.

X-ray diffraction data collection and refinement statistics

Space Group	I 41 2 2 (#98)
Unit Cell: a, b, c (Å)	184.3 184.3 113.4
Unit Cell: α, β, γ (º)	90, 90, 90
Resolution range (Å)	46.6 –2.7 (2.8–2.7)*
Total reflections	794,884 (78182)*
Unique reflections	27,023 (2669)*
Multiplicity	29.4 (27.3)*
Completeness (%)	99.9 (99.9)*
I/σ(I)	22 (1.5)*
Rmerge	0.14 (3.64)*
CC1/2 value	1.0 (0.7)*
Refinement statistics
Resolution (Å)	46.6 –2.7
Rwork/Rfree	0.244/0.260
Root mean square deviation bonds (Å)	0.010
Root mean square deviation angles (º)	1.2
Mean B factor (Å2)	104
Ramachandran Plot
Favored Region (%)	94.5
Outliers (%)	0

^*Outer shell statistics are provided in parentheses.

Molecular Dynamics Simulations

The ligand-free crystal structure of McpZ_PD lacked a complete sidechain for E189 which was replaced with a complete amino acid using PyMOL 2.5.4 ²¹. Fifteen rotamers were calculated with the best having an RMSD of 0.024 (4 to 4 atoms), and a strain of 22.07.

The GROMACS 2020.4 software suite ^22,23 was used for all MD simulations. The CHARMM36m forcefield and the CHARMM-modified TIP3P water model ^24–26 were applied. The system was built in a cubic box (4882 nm³) with a minimum solute-box distance of 1 nm. The system contained 150 mM KCl (445 K atoms and 441 CL atoms) to achieve a net neutral charge and to mimic physiological conditions. Residues and peptide termini were protonated according to their canonical states at pH 7.

To remove structural constraints imposed by crystal packing contacts, energy minimization was performed with the steepest descent algorithm with a maximum force constraint of 1000 kJ/mol*nm. Three replicate systems were individually equilibrated at constant volume and temperature (NVT), then constant pressure and temperature (NPT), while restraining heavy atoms. NVT equilibration used the modified Berendsen temperature coupling method²⁷ at a temperature of 298 K for 100 ps. Next, isothermal-isobaric conditions of 298K and 1 bar were applied for 100 ps using the Berendsen pressure coupling method.

Atom restraints were removed, and periodic boundary conditions were applied to all MD simulations. A short-range cutoff of 1.2 nm was applied to all nonbonded interactions. Long-range interactions were calculated using the particle mesh Ewald (PME) method ^28,29 using cubic interpolation and a Fourier grid spacing of 0.16 nm. An integration timestep of 2 fs was used along with the fourth order P-LINCS algorithm ³⁰ to constrain all bonds. The modified Berendsen temperature coupling method, Parrinello-Rahman pressure coupling method ^31,32, at 298 K and 1 bar respectively, and the Verlet cutoff scheme were used for all production MD simulations. Periodic boundary conditions were employed. Van der Waals forces were computed with the Lennard-Jones equation and smoothly switched to zero from 1.0 – 1.2 nm. After the initial production MD simulation, the three replicates were each extended to 500 ns, for a total of 1.5 μs of simulation time. Initial starting structures, dominant morphology structure files from simulation, and parameter files can be found on Open Science Framework (https://osf.io/82n73/) in Research Projects. Trajectories were analyzed with root-mean-square deviation (RMSD), root-mean-square fluctuation (RMSF), protein secondary structure (DSSP), and a principal component analysis (PCA), by using the GROMACS analysis package ²² and in-house scripts. A clustering analysis was conducted over the last 300 ns of the simulation using the Gromos algorithm with a rms cutoff of 0.2 nm. Quantification of protein movement was calculated from the trajectory of the protein and positional data. Displacement, pitch angle, and roll angle measurements were obtained for each helix, each four-helix bundle, and each four-helix bundle pair. Calculations were made using in-house python scripts and python scripts from https://github.com/speleo3/pymol-psico/blob/master/psico/orientation.py. The results of the MD simulations are summarized in supplementary Figures S2.–S4, Supplementary Tables 1–3, and two animations M1 and M2.

Bioinformatics

BLAST searches ³³ against NCBI sequence databases (RefSeq and Clustered nr) were performed with default parameters. Multiple sequence alignment was constructed using the MAFFT v7 L-INS-i algorithm ³⁴. HHpred search ³⁵ with default parameters was performed against profile models from Pfam and PDB databases. Taxonomy information was retrieved from the Genome Taxonomy Database ³⁶, release 202, and genome trees were generated with AnnoTree ³⁷.

Results

McpZ_PD assumes a novel helical tri-modular fold

The crystal structure of McpZ_PD was determined via single-wavelength anomalous dispersion (SAD) phasing using the strong anomalous signal generated by heavy metal ions contained in the crystallization solution that crystallized with the protein. The asymmetric unit of the crystal was composed of a single McpZ_PD molecule. The final refined model of this molecule encompasses amino acids 48 to 410 of McpZ (Fig. 1). The nine amino-terminal and fourteen carboxy-terminal residues of the crystallized construct did not yield interpretable electron density and are presumed to be structurally flexible. McpZ_PD is non-globular in shape, with a long dimension of about 137 Å. The protein is entirely helical, consisting of eight alpha helices, which form three concatenated FHBs. The membrane-proximal and the central FHB are vertically aligned. Yet, the central FHB is rotated by about 152º in the clockwise direction relative to the membrane-proximal FHB. Helices H1, H2, H3, and H8 form the membrane-proximal FHB. Helices H3, and H8 form the rigid linker to the central FHB where they pair with helices H6, and H7. Leading out to the membrane-distal FHB, helices H3 and H6 are strongly kinked at residues N168 and L291, respectively. The distal bundle is completed by the short helices H4 and H5 with the latter assuming a notably curved shape (Fig. 1a.). Overall, the central and membrane-distal FHBs are arranged at an approximate 57º angle. Helix H8 forms part of the proximal and distal FHBs, while the ninety-residue long helix H3 provides the rigid conduit that links all three subdomains. We performed a search of the Protein Data Bank (PDB) for structurally related proteins and discovered that there are a number of proteins that align well with the combined membrane-proximal and central subdomains. However, no other known structure contains a third FHB. Therefore, the presence of a helical tri-modular (HTM) domain places McpZ into a new receptor family.

Figure 1. Structure of McpZ_PD.

a. Cartoon depiction of McpZ_PD. A single McpZ_PD molecule forms the asymmetric unit of the crystal. The three four-helix bundle subdomains are labeled according to their relative positions to the inner membrane of the cell. The length of the molecule and the angle between the central and distal FHBs are displayed. The figure was generated using Pymol ⁶². b. Topology diagram for McpZ_PD. c. Correlation between primary and secondary structure of McpZ_PD.

Phylogenetic evidence for Rhizobial origins of the HTM domain

A BLAST search with default parameters against the NCBI Reference proteins (RefSeq) database was performed using a sequence corresponding to the periplasmic domain of McpZ (RefSeq accession WP_014529036.1) flanked by predicted transmembrane regions (residues 16 to 448). A total of 1,245 unique sequences were identified (Dataset S1: all similar sequences) and of those 1,169 matched the query sequence for more than 90% of its length (Dataset S2: all similar sequences of similar length), indicating the presence of a homologous domain of the same size. To generate a taxonomically balanced, representative dataset of these sequences, we performed a search using the same query against the NCBI nr clustered database, where sequences are clustered at 90% identity and 90% length using the MMseqs2 algorithm³⁸. A total of 238 clusters were identified and 181 of these clusters contained sequences matching the query sequence over more than 90% of its length. Sequences representing each cluster (automatically generated by the nr clustered database) were downloaded and used as Dataset S3: representative similar sequences of similar length).

Phyletic distribution of HTM domains was first analyzed by assessing NCBI taxonomy of BLAST hits identified in searches against the RefSeq database. This analysis showed that essentially all hits in Dataset S1 were from alphaproteobacteria, specifically members of the orders Rhizobiales and Hyphomicrobiales. Then, we used a more robust genome phylogeny based taxonomic scheme³⁶ with a substantially revised taxonomy for alphaproteobacteria. This analysis showed that all identified HTM domains originated from a single branch of a single bacterial family – Rhizobiaceae – in the order Rhizobiales (Fig. 2.a.). The only exception was one sequence (RefSeq accession WP_226920380.1) from a betaproteobacterium Kinneretia sp. XES5 from the order Burkholderiales, which is likely indicative of a horizontal gene transfer event. Thus, we conclude that the HTM domain family originated in the common ancestor of the evolutionarily youngest (farthest from root of the genome tree: Fig. 2a.) group of the family Rhizobiaceae.

Figure 2. Phylogeny and sequence conservation of McpZ.

a. Distribution of HTM-containing chemoreceptors in the family Rhizobiaceae. The presence of HTM is marked by red circles on terminal branches of the Rhizobiaceae family genome tree. Branches shown in blue identify genera where chemotaxis systems are present. The likely origin of the HTM domain is marked by a red circle next to the longest internal branch. Numbers in brackets show the number of genomes in each clade. Asterisk indicates that only one genome is available in a given clade. These symbols are automatically generated by AnnoTree b. McpZ_PD colored according to sequence conservation. c. Packing of a single layer of McpZ_PD molecules inside the crystal. A single symmetric dimer is highlighted. d. Cartoon depiction of a McpZ_PD dimer. Molecules are colored according to sequence conservation using the same approach as in 2 b. The large highly conserved patch between the central and membrane-distal FHBs is now buried at the dimer interface. Figures b. and d. were generated with the Consurf server ⁶³ using the subset of McpZ homolog sequences also used in fig. 2a. The conservation scores were generated using the default Bayesian method.

To identify a protein domain ancestral to HTM, we performed two types of analyses. First, the sequence corresponding to McpZ_PD (residues 38–426) was used in a HHpred search against profile models of the Pfam and PDB databases. The best matches were to the HBM, Helical bimodular sensor domain from Pfam (Pfam accession number PF16591.8; Probability = 98.56%) and to the structure of the ligand-binding domain of McpS chemoreceptor from Pseudomonas putida (PDB ID: 2YFA; Probability = 98.51%), which is a founding sequence for the HBM domain family. Both best hits aligned to the N-terminal half of the McpZ periplasmic domain (Fig. S5). The same models also match to the C-terminal half of the McpZ periplasmic domain, but with slightly lower probability scores (96.21% and 95.94%, correspondingly) and N- and C-terminal matching profiles overlapped significantly (Fig. S6). This suggests that (i) the HBM was a likely ancestor of the HTM and (ii) an internal duplication of HBM or insertion of an unrelated FHB region into HBM gave birth to the HTM. To further verify this, we aligned HBM sequences with closest BLAST hits that had less than 90% length match to the query (found in Dataset S1, but not in Dataset S2). These closely related sequences had much shorter periplasmic regions (~260 compared to nearly 400 amino acids of McpZ_PD), while all HTM sequences had a large insertion in the middle of the periplasmic region, corresponding to the membrane-distal FHB in the HTM (Fig. S7). To determine the domain family of the shorter (~260 aa) periplasmic regions identified in the original BLAST search, we performed a HHpred search against Pfam and PDB profiles. The closest such sequence, methyl-accepting chemotaxis protein from Rhizobium sp. PP-WC-1G-195 (RefSeq accession WP_245423295.1; 45.89% identity to McpZ in the periplasmic region, E value = 2e-27), matched closely to Pfam HBM (Probability = 99.61%) and to sensor kinase TorS from Vibrio parahaemolyticus (PDB code 3O1I; Probability = 99.58%) over the entire length of its predicted periplasmic region. This strongly suggests that the HBM domain is the closest relative and therefore the likely ancestor of the newly discovered HTM family.

Crystal packing, sequence conservation patterns, SEC-MALS, and structural homology point toward a symmetric McpZ_PD dimer

Canonical MCPs are dimeric, but the isolated periplasmic domains often only dimerize in the presence of their cognate ligand. However, even though there is only a single molecule in the asymmetric unit of the crystal, our packing analysis suggested the presence of a symmetric McpZPD dimer. Application of the two-fold axis generates an intertwined McpZ_PD dimer with an extensive interface. Using AreaIMOL from the CCP4 program suite ³⁹, we computed a combined buried surface area of 6,300 Ų at the dimer interface. Examination of the sequence conservation patterns obtained from the multiple sequence alignments of McpZ homologs (Fig. S7) revealed two highly conserved regions at the predicted dimerization interface. The first region is located within the membrane-proximal FHB module. The second area containing numerous conserved surface-exposed residues maps to the connecting region between the central and distal modules (Fig. 2.d.& S8). The notion that the observed dimer constitutes a biologically relevant unit is further supported by the comparison of McpZ_PD with its closest known structural homologs, the periplasmic domains of the MCP McpS from Pseudomonas putida KT2440, McpS_PD, (PDB code 3O1I) ⁴⁰ and of the signaling histidine kinase TorS from Vibrio parahaemolyticus, TorS_PD, (PDB code 2YFA) ⁴¹. The alpha-carbons of 219 residues of McpZ_PD and McpS_PD could be superimposed with an RMSD of 2.8 Å (Fig. 3.a & S9.), while the superposition of a McpZ_PD monomer with a TorS_PD molecule yielded an RMSD value of 3.5 Å for the 254 aligned alpha-carbons (Fig. 3b & S10) ⁴². Notably, both McpS_PD and TorS_PD form dimers in their crystal structures with very similar interfaces within the membrane-proximal FHB modules, where each protein uses a pair of helices to create an extensive interface (Fig. 5.a). As observed in the former two structures, the membrane-proximal and central FHBs of McpZ_PD form a similar interface in the symmetry-generated dimer. However, the McpZ_PD dimer shows important differences to the former two structures in the alignment of the membrane-proximal FHBs (Fig. 5.b &c). Dimerization is solely mediated by the two H1 helices to form a comparatively small interface. A large number of the residues in H1 is conserved, some of these mediate dimerization, such as N49, L52, S56, and K59. However, this conformation also leaves the conserved hydrophobic residues L55 and Y62 from H1 as well as W401 and A408 from H8 either partially or completely solvent exposed (Fig. S11). Using SEC-MALS, we confirmed that McpZ_PD forms a dimer in solution, and as such this dimer is likely the principal building block of the crystal (Fig. S1.C).

Figure 3. Structural comparison of McpZ_PD with McpS_PD and TorS_PD.

a. Left panel. Superposition of McpZ_PD and McpS_PD. Right panel. Conservation-colored surface depiction of the region in McpZ_PD, which is equivalent to the succinate binding pocket in the membrane-proximal FHB of McpS_PD. Conserved residues are labeled and their side chains are displayed. The succinate is faintly shown to mark the putative ligand binding pocket. b. Left panel. Superposition of McpZ_PD and TorS_PD. Right panel. TorT molecule (grey) modeled with a McpZ_PD dimer (blue and green). The model was created by superimposing a 1:1 TorS-TorT complex onto the McpZ_PD dimer. The TorS_PD molecules were then hidden to mark potential binding sites for periplasmic binding proteins on McpZ_PD.

Figure 5. Mechanistic implications of the distinctive dimerization interface of McpZ_PD in the membrane-proximal FHB region.

a. Dimerization interfaces of the ligand bound McpS_PD, PcaY_PD (PDB code: 6S3B), and TorS_PD viewed from the perspective of the inner membrane. N- and C-terminal helices, which connect to the transmembrane helices in the full-length proteins are color-coded blue and red, respectively. Numbers indicate the distances between the helices in Å b. MD-predicted scissoring motion. Left, N-terminal and C-terminal helices of the McpZ_PD dimer viewed from the perspective of the inner membrane. Right, MD calculations predict significant horizontal helical shifts of N- and C-terminal helices of the McpZ_PD dimer. We propose that the shifts occur upon ligand binding. Numbers indicate the distances between the helices in Å. c. MD-predicted piston-type motion. Left, side view of the membrane-proximal FHB of the McpZ_PD dimer. The indicated dihedral angle measures the H8-H1-H1-H8 angle between the termini of these helices. Right, MD calculations predict up toa 5.6 Å vertical shift of H8 toward the membrane of the McpZ_PD dimer, whereas H1 remains in the same plane. We propose that the shift occurs upon ligand binding. The vertical motion is also manifested in a nearly sixty degree shift of the H8-H1-H1-H8 dihedral angle between the termini of these helices. Angle values are averaged across the three replicates.

Small ligand binding pockets and a putative docking site for periplasmic binding proteins are partially conserved in McpZ_PD

A well-established DSF assay using Biolog MicroPlates™ (PM1, PM2A, PM3B, PM5 and PM6) containing an array of carbon sources, nitrogen sources, nutrient supplements, and peptides failed to reveal any ligands for McpZ_PD ^43–45. We also tested the polyamines cadaverine, putrescine, spermine, and spermidine, because a cluster of genes encoding a spermidine/putrescine ABC transporter is located downstream of the mcpZ gene but no binding has been observed (data not shown). Using the crystal structure, we therefore sought to identify potential ligand binding sites in the McpZ_PD through comparative analysis. The membrane-proximal and central FHBs of McpZ_PD align well with the two modules of McpS_PD (Fig. 3.a.). McpS_PD has two ligand binding pockets, one within each FHB ⁴⁰. The membrane-proximal FHB binds succinate and malate, whereas acetate is recognized by the second FHB. Superposition of the two McpS FHBs demonstrated that the same regions are associated with ligand binding in the two modules⁴⁰. The succinate/malate binding site contains three arginine and several other polar amino acids but only a small number of hydrophobic residues. The structurally equivalent region of McpZ_PD revealed a number of conserved residues, however these are non-polar such as M65 and L69. The well-preserved residues Y62 and I397 also line the pocket, while two moderately preserved asparagine residues, N66 and N398, lend some polarity to the putative ligand binding site (Fig. 3.a & S6.). Overall, if this pocket is indeed involved in ligand binding, the electrostatic properties of the ligand should be very different from succinate. The acetate binding pocket in the second FHB of McpS_PD is also characterized by the presence of three arginines that facilitate acetate binding (Fig. S9.). Only one of these arginines, R125, is conserved in the central FHB of McpZ_PD. R139 reaches into the putative ligand binding pocket, however, this residue is not conserved within the central FHBs of McpZ homologs.

In addition to binding to ligands directly, MCPs are known to detect various nutrient sources indirectly through the mediation of ligand bound Periplasmic Binding Proteins (PBPs). There is no known crystal structure of an MCP-PBP complex in the HBM family of receptors, however, the histidine kinase TorS has been co-crystallized with its cognate trimethylamine-N-oxide (TMAO) bound PBP TorT ⁴¹. The TorT binding site is located at the far end of the membrane-distal FHB and extends across the TorS dimer. The PBP binding site is highly conserved among TorS homologs⁴¹. When the TorS_PD and McpZ_PD structures are superimposed through their distal and central FHBs, respectively, TorT is positioned at the top of the central McpZ_PD near the kink that leads to the third FHB (Fig. 3.b & S10). The PBP appears to fit just underneath the membrane-distal FHB suggesting a potential role of this region in PBP recruitment. However, overall sequence conservation of the putative PBP binding pocket is low among McpZ homologs, whereas the equivalent sections are highly conserved in TorS. Thus, there is no strong supporting evidence for the presence of a PBP docking site at this location. Collectively, the comparative analysis of McpZ_PD with McpS_PD and TorS_PD in conjunction with sequence conservation patterns only supports the presence of a ligand binding pocket in the membrane-proximal FHB of McpZ_PD.

Molecular Dynamics simulations of McpZ_PD suggest vertical and horizontal displacements throughout the dimer

Because the structure offered a rare view of a ligand-free dimeric MCP receptor domain, we sought to explore the conformational space the protein could sample by using molecular dynamics (MD) simulations. Initially, we attempted to create a larger MCP receptor model that encompassed the transmembrane helices using both AlphaFold ⁴⁶ and Robetta ⁴⁷. Individual chains of these models were extremely similar to each other but had a root-mean-square deviation (RMSD) of more than 5 Å from the crystal structure. Superposition of the McpZ_PD dimer with an AlphaFold-generated dimer gave an RMSD of more than 7.9 Å, indicating a lack of confidence in the predicted dimer interface (not shown). Therefore, we decided, instead, to pursue MD simulations utilizing the experimentally determined structure in order to probe structural shifts that might occur at the dimer interface. This work is based on the observation from single molecule studies that ligand-free proteins dynamically cycle through many conformations including those that represent ligand-bound states, albeit at lower frequency. Three independent MD simulations runs were performed for 500 ns each, for a total sampling time of 1.5 μs RMSD calculations of the protein backbone indicated that the structure was mostly unchanging after initial equilibration (Fig. S2) and clustering algorithms applied to the last 300 ns of simulation time provided dominant morphologies that represented 85–88% of simulation time for each replicate (Fig. 4.a). Secondary structure was also unchanged for the duration of the simulation as compared to the starting structure (Fig. S3), suggesting that the resolved structure had a favorable secondary and tertiary structure arrangement that was maintained when simulated in solvent. The three replicates converged on very similar clusters RMSD values between 1.5 to 2.2 Å. The observed trends were the same for all clusters but the degree whereby certain motions occurred did vary. The comparison of the starting structure with the dominant morphologies from simulations is represented in Fig. 4.a to highlight movements. During simulation, conformational changes are initiated at the hinge region between the central and distal FHBs. The change, signified by a straightening of the bend angle of H3 from ~120⁰ to ~130⁰ (Fig. 4.b), causes a downward shift of H7 in the central FHB. This motion, in turn, forces a straightening of helix H8 and a piston-type vertical motion of the C-terminus of H8 towards the cell membrane (Fig. 5.c). In addition to this vertical shift, we also observed a horizontal twisting motion within the membrane-proximal FHB, which leads to a compaction of the McpZ_PD dimerization interface (Fig. 5.b). Rotation and displacement of FHB cross-chain pairs are generally coupled and symmetric (Supplementary Tables 1–3). The large displacements of the FHBs are driven by the rotation of individual helices rather than their displacement. (Supplementary Tables 1 & 2). The shorter displacement of the central FHBs in the dimer indicates that they rotate about their center of mass during the hinging process (Table 2). The bending angles of H3 and H8 in the starting and end structures are shown Fig. 4.b. Two animations provided with the supplementary material visualize the overall predicted motion trajectories for the McpZ_PD dimer (S.M1) and the horizontal movements predicted to occur within the membrane proximal FHB (S.M2). These structural changes make intriguing predictions about a possible mechanism of transmembrane signaling which are explored further in the discussion section.

Figure 4. Results of MD calculations.

a. Superposition of the most prevalent conformations obtained the MD calculations with the original structure of the McpZ_PD dimer. RMSD values and percent of simulation time the dimers are observed in these prevalent conformations are indicated below. The final conformations of the three replicate runs are very similar with RMSD values ranging between 1.5 to 2.2 Å. b. Kink angles of helices 1 and 8 of McpZ. Comparison of initial structure (left)and the MD endpoint conformations (right). The hinge angle between the central of distal FHBs straightens, which in turn causes a straightening of the kinked H8 helix. The net results are a ~55⁰ clockwise rotation of the membrane proximal FHBs around and a ~5 Å downward shift of H8.

Discussion

E. coli, the classical model system for studying chemotaxis, utilizes only five chemoreceptors ⁶, however, genome sequence data suggest that many bacteria have vastly larger MCP repertoires at their disposal. Pseudomonas species, for example, may encode up to 37 MCPs⁹. Based on their molecular architecture MCPs may be classified into several families, with the overwhelming majority falling into the superfamilies PAS, CACHE, or FHB ⁹. Receptors in the FHB family have so far been classed into four sub-categories. More than 70% of FHB receptors contain just a single FHB, while the remainder falls into the TarH, CHASE3, and HBM families ^9,12. Our structural and phylogenetic analyses have revealed that the periplasmic domain of McpZ (McpZ_PD) belongs to the new HTM family within the FHB superfamily that contains over 1,100 identifiable members. The HTM family genes are still rapidly evolving, especially within the central and membrane-distal FHBs, perhaps suggesting that many of the receptors contained within this family have yet to settle on distinctive functions for these two submodules.

HBM-type receptors such as P. putida McpS and McpQ are most closely related to McpZ. Although McpS has been shown to utilize both FHB modules for signal sensing, its primary response is mediated by the membrane-proximal module ⁴⁰. McpS recognizes dicarboxylates intermediates of the tricarboxylic acid (TCA) cycle such as succinate and malate using a binding pocket that involves residues from both protomers within the McpS_PD dimer ⁴⁰. The McpS paralog McpQ preferentially binds citrate and citrate-metal-ion complexes using a comparable mechanism ⁴⁸. Reflecting the ionic nature of the ligand, the associated binding pocket of McpS contains three arginine residues and two additional polar amino acids. In contrast, although the structurally equivalent region is enriched in conserved residues among McpZ homologs, the pocket is much more hydrophobic, suggesting that a TCA cycle intermediate is not an McpZ ligand, a notion that is supported by our unsuccessful ligand screening effort, which included these compounds. Because TorS objectively constitutes the closest structural relative of McpZ_PD in the PDB, we also strongly considered the possibility that McpZ signal sensing is indirect through the mediation of a PBP, which has been observed for numerous other MCPs ⁴⁹. There is experimental evidence that the HBM containing CtpL from P. aeruginosa senses inorganic phosphate through interactions with the PBP PstS, however, the location of their interface is not known ⁵⁰. Therefore, we used the superposition of McpZ_PD with TorS_PD in a TorS_PD-TorT complex to identify a putative PBP docking site in McpZ_PD. Indeed, this comparison suggests that a PBP could bind just between the central and membrane-distal FHBs, however, while the TorS-TorT interface is strongly conserved in TorS, no such conservation pattern was observed for McpZ_PD.

There are few examples where both dimeric apo- and holo-forms of MCPs have been structurally characterized to reveal conformational changes associated with ligand binding. The classic paradigm for how force is generated through ligand binding was established using E. coli Tar and Tsr chemoreceptors as model systems ^51,52. The overall consensus is that chemoreceptors use a universal signaling mechanism characterized by a subtle piston-like motion exerted by the periplasmic domain onto the second transmembrane helix ^53–55. Within the periplasmic domains, the structural changes associated with ligand binding are small conformational shifts within each protomer of the dimer. This model fits well with the observation that chemoreceptor arrays are fairly static in the absence and presence of a stimulant ^6,56. However, two recent studies suggest that alternative means of force generation are possible. The periplasmic domain of the chemoreceptor MCP2201 was shown to be dimeric in the absence of a ligand but forms trimers in the presence of citrate ⁵⁷. Ligand binding is also associated with conformational changes, but the dimer-trimer transition appears to be the more significant signaling event ⁵⁸. Recently, the FHB receptor PcaY_PD from P. putida was crystallized both in the absence and presence of several ligands ⁵⁹. Ligand binding caused a shift of the protomers parallel to the dimerization interface, which results in a scissoring motion rather than a piston-type movement. In addition, although the ligand-bound dimer of PcaY_PD is overall more symmetric than the apo-protein, ligand binding consistently caused the partial unfolding of the very C-terminal end of H4 in one of the protomers. This relaxation may also have significance for signal transduction but has yet to be explored.

Interestingly, independent of their starting points, the ligand-bound structures of McpS_PD, TorS_PD, and PcaY_PD all display similar arrangements of their N-terminal and C-terminal helices (Fig. 5.a), suggesting related signaling mechanisms. All three exhibit a four-helix interface with very similar spacing (Fig. 5.a.). The present structure of McpZ_PD offers another example of a ligand-free dimeric ligand binding domain of an MCP. In this context, the striking differences between the dimerization interfaces of McpZ_PD on one hand and those observed in TorS_PD, McpS_PD, and PcaY_PD on the other hand may have mechanistic implications. These differences are particularly pronounced within the membrane-proximal FHBs, where all contacts between the two McpZ_PD protomers are mediated by the two H1-helices. Three residues pair directly with their symmetry-related counterparts, namely N49, L52, and S56, while two hydrogen-bonded K59-N60 pairs complete the interface between the two protomers. N49, L52, S56, and K59 are part of a cluster of well-conserved residues in helix H1, spanning from S49 to L69 (Fig. S11). The C-terminal H8 helix of McpZ_PD, is stacked against H1 but, unlike in TorS_PD, McpS_PD, and PcaY_PD, does not participate in dimer interactions (Fig. 5.b & S11). The trajectories of our molecular dynamics simulations converge on a second conformation for the McpZ_PD dimer that closely resembles those of the ligand-bound conformations observed in the other three receptors. In this conformation, H1 and H8 from the two protomers are evenly spaced forming a four-helix interface. This structural rearrangement requires a large horizontal movement of the four helices, which supports a scissoring-type motion at the membrane. Remarkably, the collective movements also suggest a vertical shift of the H8 helices towards the cell membrane by up to 5.6 Å. H1 on the other hand undergoes no vertical movement (Fig. 5.c). This observation is consistent with the piston-type movements that have been classically associated with MCP signaling. Therefore, the overall findings of the present study suggest a mechanism of transmembrane signaling for FHB MCPs that seamlessly integrates the two prevailing models for transmembrane signaling. Although we cannot dismiss the possibility that the observed changes are exaggerated due to the absence of the transmembrane helices in the studied structure, such integrated mechanism has been previously reported for two-component signaling histidine kinases NarX and NarQ ^60,61. In these systems the amplitudes of the ligand-induced changes are smaller but the overall trends are the same.

Conclusion

In summary, the reported crystal structure of the periplasmic domain of S. meliloti McpZ represents the first example of a new receptor subfamily with a helical tri-modular periplasmic domain. Sequence conservation patterns identified strong conservation only at the dimerization interface and within the membrane-proximal FHB, which suggests that McpZ homologs are still rapidly evolving. The ligand-free McpZ_PD structure revealed a novel dimerization interface. Supported by MD calculations, we propose that this conformation of the dimer is broadly representative of the ligand-free state for multiple members of this receptor family. We further hypothesize that ligand binding triggers signal transduction by inducing both scissoring and piston-type motions in the membrane-proximal FHB.

Supplementary Material

Supinfo2

Dataset S1. 1245 unique sequences obtained from BLAST search with default parameters against the NCBI Reference proteins (RefSeq).

Dataset S2. 1169 matched sequence matching McpZ for at least 90% of its length.

Dataset S3. 238 sequences representing clusters with 181 of these matching the query sequence over more than 90% of its length.

Supinfo4

Movie M2. View of MD predicted movement trajectories from the perspective of the inner membrane.

Supinfo5

Figure S1. McpZ_PD purification data, melting curve, and SEC-MALS analysis data.

Figures S2-S4 Summary of MD results.

Figure S5. Snapshot of the HHpred graphical output.

Figure S6. Multiple sequence alignment of representative chemoreceptors with closely related HTM and HBM domains.

Figure S7. Multiple sequence alignment of HTM domains and flanking TM regions from the representative dataset

Figure S8. Schematic detailing the interactions at the McpZ_PD dimerization interface

Figure S9. Superposition of the McpZPD dimer and McpSPD

Figure S10. Superposition of the McpZ_PD dimer and TorS_PD-TorT heterotetramer

Figure S11. Close-up view of the McpZPD dimerization interface in the membrane-proximal FHB.

Tables 1-3 Detailed breakdown of structural changes predicted through the MD simulation.

Supinfo3

Movie M1. Side-view of the MD predicted movement trajectories.