Crystal structure of the cell wall anchor domain of MotB, a stator component of the bacterial flagellar motor: Implications for peptidoglycan recognition

Abstract

The stator ring of the bacterial flagellar motor is composed of the MotA and MotB proteins that act together to generate a turning force (torque) acting on the FliG ring of the rotor. The C-terminal domain of MotB (MotB-C) is believed to anchor the MotA/MotB complex to peptidoglycan (PG) of the cell wall. The first crystal structures of MotB-C and its complex with N-acetylmuramic acid (NAM) have been determined to 1.6- and 2.3-Å resolution, respectively. MotB-C is a dimer, both in solution and in the crystal. The two glycan chains of the PG ligand can be modeled as semirigid helices and docked into the grooves harboring the NAM molecules on the opposite faces of the dimer. The model suggests that a concave hydrophilic surface created upon edge-to-edge β-sheet dimerization and centered around the 2-fold axis of the dimer can accommodate the peptide cross-bridge linking the two sugar chains. Significant structural similarities were found between MotB-C and the PG-binding domains of reduction-modifiable protein M and peptidoglycan-associated lipoprotein exclude, suggesting that PG recognition by different outer membrane protein A-like proteins may be governed by very similar molecular mechanisms that evidently involve protein dimerization.

The bacterial flagellar motor is nature's rotary nanomachine that turns helical filaments, creating a propulsive force for bacteria to swim (1). The stator of this motor is formed by the membrane proteins MotA and MotB, which associate in complexes MotA₄MotB₂ (2) circumferentially arranged around the rotor (3). These complexes use the gradient of protons across the cytoplasmic membrane to generate the turning force (torque) applied to the FliG component of the rotor (4, 5). Previous biochemical and mutagenesis experiments provided evidence that the stator function involves protonation/deprotonation of a conserved aspartate residue in the transmembrane helix of MotB that is associated with a conformational change within the complex (6, 7). However, the detailed molecular mechanism by which the stator couples the ion flow to flagellar rotation is not yet understood, mainly owing to the lack of structural information for the stator and rotor-stator interface.

The MotB subunit has a short N-terminal cytoplasmic region, a hydrophobic α-helix spanning the cytoplasmic membrane, and a C-terminal periplasmic domain (8, 9). The C-terminal domain contains a motif LSX₂RAX₂VX₃L, conserved throughout the family of outer membrane protein A (OmpA)-like peptidoglycan (PG)-binding domains (10). It is therefore thought that the periplasmic domain of MotB binds to the highly cross-linked stress-bearing zone of the PG of the cell wall, thus immobilizing the stator ring of the bacterial flagellar motor. The functional importance of this domain has been highlighted by previous mutagenesis studies that mapped most of the mutations in the motB gene associated with a loss of motility in Escherichia coli to ≈160 C-terminal residues (11).

The Gram-negative pathogenic bacterium Helicobacter pylori associated with gastritis, gastric and duodenal ulcer, and gastric cancer (12, 13) possesses a single motB gene (14). Flagellar motility has been proven to be required during the initial colonization of the stomach (15, 16) and for attaining full infection levels in an animal model (17). In-frame deletion of the motB gene in H. pylori created flagellated nonmotile mutants with reduced ability to infect mice (17). H. pylori has meso-diaminopimelate (mDAP)-type PG with N-acetylglucosamine (NAG)–N-acetylmuramic acid (NAM)-l-Ala-d-Glu-(γ)-mDAP-d-Ala-d-Ala/Gly as the basic unit (Fig. 1 and ref. 18). NAG and NAM are connected in an alternating fashion by β-(1,4)-glycosidic bonds forming the glycan chains. Pentapeptide stems attached to NAM residues are cross-linked by transpeptidases in a tail-to-tail fashion, bridging the glycan chains to form a regular structure of the cell wall matrix.

Fig. 1.

Structure of the disaccharide pentamuropeptide unit of H. pylori PG.

As a step toward elucidating the molecular basis for PG recognition by H. pylori MotB and investigating the possible role that its separate domains play in the oligomeric behavior and assembly of a single stator unit, the putative PG-binding domain of MotB (MotB-C) comprising the 132 C-terminal residues was expressed in E. coli, purified, and crystallized as described (19). This article reports determination of the crystal structures of MotB-C and its complex with NAM.

Results and Discussion

Overall Structure and Comparison to Other OmpA-Like PG-Binding Domains.

The crystal structure of recombinant H. pylori MotB-C (residues 125–256 plus an additional N-terminal GIDPFT fragment introduced by the cloning procedure) was determined to 1.6-Å resolution by using the multiple-wavelength dispersion (MAD) method with the ytterbium derivative. The asymmetric unit contains four molecules related by a 222-point group symmetry. The MotB-C structure is composed of four pairs of alternating β-strands and α-helices arranged in the topological order βαβαβαβα. Three parallel [β1 (residues 123–125), β2 (residues 156–162), and β3 (residues 200–206)] and one antiparallel [β4 (residues 226–234)] strands form a four-stranded β-sheet in the order 1423 (Fig. 2A). Helices α1 (residues 134–149), α2 (residues 175–194), and α4 (residues 235–250) pack against one face of the β-sheet, whereas the shorter helix α3 (residues 217–225) forms an N-terminal extension of strand β4. The loops connecting β-strands with α-helices at one end of the β-sheet (near the C termini of β2 and β3 and the N terminus of β4) are short and relatively rigid. The opposite end of the β-sheet expands into three longer petal-like loops, β1α1, β2α2, and β3β4. Analysis of the distribution of the main-chain temperature factor and average C_α rmsd for pairwise superpositions of the four molecules in the asymmetric unit [supporting information (SI) Fig. S1] identifies these loops as the most mobile elements in the structure.

Fig. 2.

The overall fold of H. pylori MotB-C and comparisons with other OmpA-like PG-binding domains. (A) Stereo representation of the MotB-C structure. The bound NAM molecule is shown in stick mode. The figure was prepared by using PyMOL (20). (B) Comparison of the structures (Upper) and the secondary structure topologies (Lower) of MotB-C from H. pylori, the periplasmic domains of PALs from E. coli and H. influenzae, and the PG-binding domain of RmpM from N. meningitidis. The conserved structural core is drawn in black and white in the topology schemes and colored cyan/magenta (helices/strands) in the 3D structure representation. Topological differences are highlighted with distinct colors. (C) Sequence alignment of the C-terminal domains of MotBs from H. pylori and E. coli and the OmpA-like domains shown in B. Conserved residues are highlighted in red. The positions of the aa substitutions associated with loss of function in E. coli MotB (11) are shown by black dots. Alignment was carried out by using the ClustalW server (25) and further refined based on structure superpositions. The figure was produced by using ESPript (26). (D) Stereoview of the superposition of the structures of MotB-C (blue), the periplasmic domains of PALs from E. coli (red), and H. influenzae (green) and the PG-binding domain of RmpM from N. meningitides (magenta). The 10 conserved residues are shown in black in the MotB-C structure.

In a comparison of the atomic coordinates of MotB-C against the structures deposited in the Protein Data Bank (21) using the DALI Server (22), significant similarities were found with PG-associated lipoproteins (PALs) from E. coli (PDB ID code 1OAP), Haemophilus influenzae (23), and Neisseria meningitides reduction-modifiable protein M (RmpM) (24) (Fig. 2B). MotB-C and the PG-binding domains of PALs and RmpM adopt a very similar fold, despite the limited sequence homology (<8% global sequence identity and <18% identity for pairwise comparisons of MotB-C with PALs and RmpM; see Fig. 2C). As shown in Fig. 2 B and D, the four proteins share a strongly conserved 100-residue structural core comprising a four-stranded β-sheet and three α-helices that pack against it in an almost identical manner. The structures of PALs from E. coli and H. influenzae and RmpM from N. meningitides can be superimposed on MotB-C over the 100 C_α atoms of the common core with an rmsd of 1.5, 2.1, and 1.9 Å, respectively. The sequence alignment of MotB-C, PALs, and RmpM identifies only 10 conserved residues (Fig. 2C). When their positions are highlighted on the structure of MotB-C (Fig. 2D), there is a clear prevalence of such residues in the core of the protein, except for the structurally important Gly-195 in loop α3β3.

Binding of NAM and Localization of the Putative PG Binding Site.

Cocrystallization of MotB-C with NAM yielded a crystal form different from that for free protein (Table S1 and Table S2). The crystal structure was determined by molecular replacement using the crystallographic tetramer of unbound MotB-C as a search model. Of 16 subunits in the asymmetric unit of the cocrystal with NAM, the electron density corresponding to the sugar molecule (Fig. 3A) was found in the two subunits with the lowest temperature factor.

Fig. 3.

Binding site for NAM. (A) (2mFo-DFc) sigmaA-weighted (27) electron density for NAM bound to MotB-C. The map was calculated at 2.3-Å resolution and contoured at 0.7-σ level. (B) Locations of the MotB-C residues structurally equivalent to the functionally important residues in E. coli MotB (named in parentheses): Ala-128 (Gly-164), Thr-163 (Thr-196), Asp-164 (Asp-197), Arg-172 (Glu-205), Ala-180 (Ser-214), Arg-183 (Arg-217), Met-188 (Arg-222), Gly-207 (Gly-240), Thr-209 (Ala-242), and Arg-226 (Arg-258). The positions of the C_α atoms of the residues with buried side chains are shown in blue, and surface-exposed chains are in red. (C) Stereoview of the molecular surface of MotB-C in the vicinity of the bound NAM molecule showing the putative glycan chain binding groove. The positions of residues Arg-172 and Thr-209 are highlighted. (D) View of the MotB-C dimer, as found in the crystal structure. The position of the bound NAM is shown for both monomers.

NAM binds in the middle of a shallow ≈25-Å-long groove formed by loops β2α2 and β3α3 at one end of the β-sheet (Fig. 3). Van der Waals contacts with the protein involve the peptide backbone of Ser-175, His-176, Tyr-177 and Gly-207 and side-chain atoms Asp-165(Cβ), Ser-175(Cβ), His-176(Nδ1), His-176(Cβ), Tyr-177(Cδ1), and Tyr-177(Cε1). The complex is further stabilized by hydrogen bonds between Tyr-177(N) and the carboxyl oxygen of NAM, Ser-208(Oγ) and NAM(O1), Ser-208(N) and the acetamido oxygen of NAM, and a water-mediated hydrogen bond between NAM(O6) and Leu-168(N). Fifty-eight percent of the accessible surface area (asa) of the sugar is buried upon this interaction. Superposition of the structures of free and NAM-bound MotB-C gave an rmsd of 0.25 Å for 132 Cα atoms, indicating that the overall protein structure does not undergo any significant conformational changes upon complex formation.

Blair et al. (11) reported substitutions in E. coli MotB that result in complete or partial loss of bacterial motility (11). Most of the structurally equivalent residues in H. pylori MotB-C have their side chains buried within the protein fold (Figs. 2C and 3B). Only substitutions Glu-205–Lys, Ala-242–Thr, and Ala-242–Val appear to involve surface-exposed side chains. The corresponding residues in H. pylori MotB-C, Arg-172 and Thr-209, reside on loops β2α2 and β3α3, respectively, flanking the opposite ends of the groove that accommodates NAM in the MotB-C/NAM crystal structure (Fig. 3 B and C). This finding strongly suggests that this groove forms part of the PG-binding surface in MotB.

MotB-C Dimerization in Solution and the Crystal.

To determine the oligomeric state of MotB-C in solution, multiangle laser light scattering (MALLS) analysis coupled to gel filtration chromatography was carried out in a set of four different buffers with the salt concentration and pH identical to those in the crystallization mix. MotB-C eluted as a single symmetrical peak with a polydispersity index value close to 1 in all four buffers, indicating that the eluted particles were homogenous with respect to the molar mass. The weight average molecular weight values ranged between 33.8 and 34.8 kDa, which is approximately twice the molecular mass calculated from the amino acid sequence (15.9 kDa).

Analysis of the packing of monomers in the crystal lattice in two different space groups identified an obvious dimer with 2-fold symmetry and approximate dimensions of 50 × 40 × 30 Å (Fig. 3D). β-Sheets of the two subunits associate via strands β3 in an antiparallel edge-to-edge manner. The dimer interface is formed by residues from helix α2, strand β3, and turn α2β3. About 7.8% (594 Ų) of the subunit asa is buried upon dimerization, which is at the low end of the range found for other dimeric proteins (28). Calculation of the Stokes radius from the crystal structure of the MotB-C dimer gives the value of 2.7 nm, which is very close to the value derived from the MALLS analysis in solution (2.5 nm in all tested conditions). These results demonstrate that H. pylori MotB-C exists as a dimer in solution, in line with the previous reports of the dimeric behavior for the periplasmic domain of Salmonella MotB (29) and other OmpA-like PG-binding domains, such as that of N. meningitidis RmpM (24) and E. coli PAL (30). The MotB-C dimer is ≈70 Å long at its maximum dimension and thus can insert into the pores of the PG matrix, the smallest diameter of which is estimated to be 70 Å (31).

Model for PG Binding by a MotB-C Dimer.

The C1–O1 and C4–O4 bonds of NAM bound in the middle of the long groove point to its opposite ends, marked by the positions of Arg-172 and Thr-209 (Fig. 3 B and C). This geometry suggests an obvious direction of the binding of the (NAM-NAG)_n chain of the natural PG ligand. Previous NMR studies of the cell wall fragment in solution revealed that the glycan chain forms a relatively rigid right-handed helix with a periodicity of three for the NAG–NAM repeat (31). The model for a nine-residue segment of the glycan chain was constructed by consecutive addition of sugar rings adhering to the experimentally determined average values for glycosidic torsion angles (31). The glycan fragment was then manually docked into the groove on the MotB-C surface by overlapping one of its NAM residues with the crystallographically observed NAM followed by minor readjustments of the torsion angles, to remove steric clashes with the protein atoms, and several cycles of structure idealization as implemented in REFMAC (27).

Analysis of this model (Fig. 4A) suggests that the groove formed by loops Asp-165–Tyr-177 and Ser-208–Asp-216 may interact with as many as seven consecutive residues of the glycan chain, buring a total of 540 Ų of the protein surface at the interface. Calculations of the surface area of the sugar rings shielded from the solvent upon this interaction give the values of 27%, 12%, 39%, 58%, 25%, 37% and 40% for the rings −3, −2, −1, 0, +1, +2 and +3, respectively (numbering from the nonreducing to the reducing end with 0 corresponding to the NAM position observed in the crystal), suggesting good surface complementaritiy between the protein and the ligand in the model. Only two of the six hydrogen bonds between the protein and the glycan involve protein side chains (Asp-165 and Ser-208); the rest are formed with main-chain atoms Leu-168(O), Lys-174(O), His-176(N), and Asn-215(N). Van der Waals contacts with protein involve the main-chain peptides in segments Asp-165–Tyr-177 and Ser-208–Asp-216 and the protein side-chain atoms Pro-167(Cβ), Lys-174(Cγ, Cδ), Ser-175(Cβ), His-176(Nδ1), His-176(Cβ), Tyr-177(Cδ1), Tyr-177(Cε1), Tyr-177(OH), Pro-211(Cγ), and Pro-214(Cβ).

Fig. 4.

Docking of a glycan chain fragment into the groove of MotB-C accommodating NAM. (A) Stereoview illustrating the interactions between the docked nine-residue (NAM-NAG) fragment, shown in black lines, and protein loops Asp-165–Tyr-177 (drawn with carbons in brown) and Ser-208–Asp-216 (drawn with carbons in cyan) of the MotB-C monomer. (B) Stereo drawing showing the direction of the peptide stems (yellow arrows) attached to the (0) and (+2) NAM residues of the glycan chain, with respect to the molecular surface of MotB-C. The side chains of Arg-172, Thr-209, Asp-165, and Arg-221 are shown in red in a ball and stick representation.

The peptide stem attached to the (+2) NAM residue of the modeled PG fragment would bind between loop β2α2 and helix α3, possibly forming hydrogen bonds with the side chains of Arg-221 and Asp-165 (Fig. 4B). Arg-221 is strongly conserved in MotBs (Fig. S2) whereas Asp-165 can be conservatively substituted by Asn. A topologically similar binding site for the peptide moiety of PG had been previously identified in the region formed by loops β1α1, β2α2, and helix α3 of H. influenzae PAL (23).

The two symmetry-related glycan-chain binding sites of the MotB-C dimer are located on opposite faces, separated by ≈50 Å (Figs. 3D and 5), which implies that MotB can bind to PG in the mode that engages two glycan strands simultaneously. The peptide stem attached to residue (0) NAM of the bound natural PG ligand would extend over and approximately parallel to strand β3, toward the 2-fold axis of the MotB-C dimer (Figs. 4B and 5). Modeling suggests that the tails of the two peptide stems extending from NAM (0) of the two glycan chains bound at opposite ends of the dimer can be close enough to form a cross-link. Thus, the hydrophilic concave surface formed by strands β2 and β3, and centered around the 2-fold axis of the dimer, may provide a binding site for the peptide cross-bridge, with the pseudo 2-fold symmetry of the latter matching the symmetry of the dimer (Figs. 3D and 5). The fact that the surface distance between the NAM carboxyl groups at the opposite ends of the dimer (43 Å) is very close to the calculated length for the peptide bridge in the extended conformation (43.6 Å; ref. 32) supports this view. The structure therefore suggests that MotB-C may form extensive interactions with both the glycan and the peptide moieties of PG with a total of ≈2,400 Ų of the protein asa shielded upon binding to PG. The presented structural analysis provides a useful foundation for more systematic biochemical and genetic studies.

Fig. 5.

Molecular surface of the MotB-C dimer colored according to the average main-chain rmsd for the superimposition of the structures of MotB-C and the PG-binding domains of PALs and RmpM (see Fig. 2D). The color gradient from blue, through white, to red corresponds to the rmsd range from 0 to ≥4 Å. The modeled glycan chains are shown in ball representation, and the hypothetical position of the peptide cross-bridge linking the (0) NAM residues is shown in yellow.

Implications for PG Binding by Other OmpA-Like Domains.

The MotB-C regions proposed to interact with PG in this study show the least main-chain deviation in the superposition of the structures of MotB-C and the PG-binding domains of PALs and RmpM (as highlighted by the blue color in Fig. 5; also see Fig. 2D). The solvent-exposed part of the β-sheet, the shallow groove formed by loops β2α2 and β3α3, and the adjoining surface pocket between loop β2α2 and helix α3 are topologically and spatially preserved in the compared structures. This finding suggests that PG recognition by different OmpA-like proteins may be governed by very similar molecular mechanisms that evidently involve protein dimerization. However, the lack of conservation of surface residues in these proteins begs a question: how is high affinity to PG preserved without any apparent strict evolutionary constraints on the nature of the side chains at the protein–PG interface? Analysis of the model for the glycan chain binding to MotB-C (Fig. 4A) reveals that most of the hydrogen bonds and a large proportion of the van der Waals contacts with the glycan are made by the protein main-chain atoms. Further examination shows that, of the eight side chains found in the walls of the sugar-chain binding groove, six (Asp-165, Pro-167, Lys-174, Ser-175, Tyr-177, and Pro-214) are positioned tangentially to the glycan chain. Therefore, variation of the side-chain length at these locations would not introduce steric clashes with the ligand or alter the width of the glycan-chain binding groove, consistent with significant sequence variability at these positions in OmpA-like domains (Fig. 2C and Fig. S2). The MotB-C side chains that point radially toward the glycan chain in the model are His-176, Pro-211, and Ser-208. His-176 is conservatively substituted by Asn in other OmpA-like proteins. Pro-211 is conserved in MotBs and is conservatively substituted by Ala in some of OmpA-like PG-binding domains. Ser-208 (predicted to form a hydrogen bond with the glycan chain) is substituted by a polar residue (Asp, Glu, Lys) in many homologous proteins. These structural features provide an explanation for the great sequence variability of the surface residues in OmpA-like proteins that fine-tuned themselves to recognize species-specific PG (33).

Materials and Methods

Crystallization and Data Collection.

Native and ytterbium derivative crystals of MotB-C from H. pylori strain 26695 were obtained as described (19). The MotB-C/NAM crystal complex was produced by cocrystallization with 10 mM NAM under similar conditions. X-ray diffraction data for the native cryo-cooled crystal and the NAM complex were collected to 1.6- and 2.3-Å resolution, respectively, using the Swiss Light Source (PX06). A MAD experiment was performed on a single crystal of Yb derivative at 100 K on European Synchrotron Radiation Facility beamline BM16. All of the data were processed and scaled by using programs MOSFLM (34) and SCALA (27) (see Table S1).

Structure Determination.

The locations of the two Yb sites for the derivative were found by using ShelxD (35). Refinement of heavy-atom parameters and calculation of phases were carried out with MLPHARE (27). An improved map was obtained by combined solvent flattening, histogram matching, and four-fold noncrystallographic symmetry averaging, as implemented in DM (36). The first model was built manually into a 3.0-Å resolution electron density map by using Coot (37). After refinement with REFMAC (27) using the native data set to 1.6-Å resolution, phases from this model and the MAD phases were combined in SIGMAA (38) to generate an improved map. A complete model of MotB-C, except for the four C-terminal residues, was built from this map in a semiautomatic fashion by using Coot and ARP/WARP (39). The structure of the MotB-C/NAM complex was solved by molecular replacement using MOLREP (27) with a crystallographic tetramer observed in the native crystal as a search model.

Gel Filtration Chromatography and MALLS Analysis.

The absolute hydrated molecular mass of MotB-C in solution was determined by MALLS analysis coupled to gel filtration chromatography. The sample containing 400 μg of MotB-C at a concentration of 8–10 mg/ml (similar to the concentration used in crystallization) was loaded onto a Superdex 200 24/30 gel filtration column (Amersham Pharmacia) equilibrated with buffer 1 [50 mM sodium phosphate (pH 6.4), 200 mM NaCl], 2 [buffer 1 + 200 mM (NH₄)₂SO₄], 3 [100 mM Tris (pH 6.4), 200 mM NaCl], and 4 [buffer 3 + 200 mM (NH₄)₂SO₄] flowing at 0.5 ml/min. The eluant was passed through an in-line DAWN EOS 18 laser photometer (λ = 688 nm) and an Optilab rEX refractometer (Wyatt Technologies). For calculations of the molecular mass and hydrodynamic radius, the light-scattered intensity and the refractive index were analyzed (Table S3). Theoretical calculations of the hydrodynamic radius from the crystal structure were carried out as described by Garcia de la Torre et al. (40) by using HYDROPRO version 7.C (http://leonardo.fcu.um.es/macromol).