Torque generation in the Salmonella flagellar motor is coupled to translocation of H+ ions through the proton-conducting channel of the Mot protein stator complex. The Mot complex is believed to be anchored to the peptidoglycan (PG) layer by the putative peptidoglycan-binding (PGB) domain of MotB. Proton translocation is activated only when the stator is installed into the motor. We report the crystal structure of a C-terminal periplasmic fragment of MotB (MotBC) that contains the PGB domain and includes the entire periplasmic region essential for motility. Structural and functional analyses indicate that the PGB domains must dimerize in order to form the proton-conducting channel. Drastic conformational changes in the N-terminal portion of MotBC are required both for PG binding and the proton channel activation.
The bacterial flagellum is a long, filamentous organelle that extends out from the cell body and rotates to drive bacterial motility. The flagellum is made of three major parts: the basal body, the hook and the filament. The basal body works as a reversible rotary motor, and the filament, typically 10–15 μm long, serves as a helical propeller. The hook is a universal joint that connects the motor with the filament to transmit torque regardless of the orientation of the filament. The flagellar motor is a membrane-embedded nanomachine powered by the electrochemical potential difference of hydrogen or sodium ions across the cytoplasmic membrane (Manson et al., 1977; Matsuura et al., 1977; Hirota and Imae, 1983; as review articles, see Berg, 2003; Minamino et al., 2008; Sowa and Berry, 2008; Terashima et al., 2008). In the Salmonella and Escherichia coli motor, torque is generated by rotor–stator interactions coupled with proton translocation through the channel formed within the stator. The Mot complex is composed of two cytoplasmic membrane proteins, MotA and MotB, which form a MotA4/MotB2 hetero-hexameric complex (Terashima et al., 2008). About a dozen stator complexes assemble around the rotor (Reid et al., 2006) and are believed to anchor to the peptidoglycan (PG) layer (Chun and Parkinson, 1988; De Mot and Vanderleyden, 1994; Muramoto and Macnab, 1998). When the basal body is isolated by solubilizing cell membranes with detergents, the stator complex is lost from the basal body due to its transient interaction to the rotor. Therefore, structural information on the rotor–stator interaction is limited. No structural information is available for the isolated MotA/MotB complex, either. Therefore, our present understanding on the molecular mechanisms of torque generation and energy conversion is limited.
MotA has four transmembrane segments and a large cytoplasmic loop, which contains conserved charged residues that are thought to interact with those of a rotor protein, FliG (Zhou et al., 1998a). MotB is composed of a small N-terminal cytoplasmic segment (residues 1–28), a single transmembrane helix (residues 29–50) and a large C-terminal periplasmic region (residues 51–309) (Muramoto and Macnab, 1998) (Fig. 1). The transmembrane helix, which contains a critical aspartate residue (Asp-33) required for proton translocation across the cell membrane, forms the proton channel together with two of the transmembrane segments of MotA (Blair and Berg, 1990; Stolz and Berg, 1991; Zhou et al., 1998b; Che et al., 2008). The periplasmic region of MotB contains a putative peptidoglycan-binding (PGB) motif (De Mot and Vanderleyden, 1994), which is believed to anchor the MotA/MotB complex to the PG layer around the rotor. Residues 149–269 of MotB show sequence similarity to other OmpA-like proteins (UniProt Accession NO. P55892). Most amino acid substitutions in motB that impair motility are within or adjacent to the PGB motif or within the transmembrane helix. Thus, anchoring the MotA/MotB complex to the PG layer is essential for the motor function (Blair et al., 1991; Togashi et al., 1997). Since the PG layer is separated from the outer surface of the hydrophobic core layer of the cytoplasmic membrane by about 100 Å (Matias et al., 2003; A. Okada et al., pers. comm.), the stator must have a relatively long extension between its transmembrane domain and its PG-binding domain.
One of the important features of stator function is the mechanism by which it assembles into the motor. To produce a fully functional motor, multiple stator units have to be incorporated at appropriate positions around the rotor and anchored there to generate torque on the rotor. However, abrupt, stepwise drops and restorations of the rotation speed of the motor have been observed even in steadily rotating motors, suggesting that the stators are replaced frequently (Block and Berg, 1984; Blair and Berg, 1988; Sowa et al., 2005). In agreement with these observations, recent fluorescent photo-bleaching studies showed rapid exchange and turnover of the stator complexes in the functioning motor, revealing that the stator association to, and dissociation from, the PG layer is highly dynamic (Leake et al., 2006).
Overproduction of the MotA/MotB complex does not affect cell growth, suggesting that the proton-conducting activity of the MotA/MotB complex is tightly coupled to its incorporation around the rotor (Stolz and Berg, 1991). MotA/MotB complexes diffusing in the cytoplasmic membrane must have a mechanism to suppress proton flow through the channel that would otherwise harm the cell. Deletion of residues 53 though 66, just after its single transmembrane segment of MotB, causes proton leakage and arrests cell growth upon overproduction of the mutant MotB protein together with wild-type MotA (Hosking et al., 2006), suggesting that this 14-residue segment acts as a plug to suppress proton leakage until the MotA/MotB complex is properly installed into the motor.
Earlier studies demonstrated that a MotB mutant lacking residues 51–100 (MotBΔL) can still form a stable, active stator and that a MotB mutant with an additional deletion of residues 271–309 is also functional, although unstable (Muramoto and Macnab, 1998). We therefore named the region containing residues 111–270 PEM (Periplasmic region Essential for Motility) (Fig. 1). The structure of PEM should provide essential information about the molecular mechanisms of assembly, anchoring and activation of the stator. In this study, we determined the crystal structures of C-terminal fragments of MotB (MotBC) that covers the entire PEM region. Based on the structure and subsequent structural-based mutational and biochemical studies, we propose a model for assembly coupled activation of the proton channel of the MotA/MotB complex.
Results and discussion
We recently showed that a C-terminal fragment of MotB consisting of residues 78–309 (MotBC1; Fig. 1) forms dimer in the periplasm and inhibits the motility of wild-type cells, suggesting that MotBC1 competes with the wild-type MotA/MotB complex for binding to stator attachment sites around the rotor (Kojima et al., 2008a). Since 1H-15N Hetero-nuclear Single Quantum Coherence (HSQC) spectra indicate that MotBC1 contains a highly mobile region composed of about 40 amino acid residues (data not shown), we constructed several N-terminally and/or C-terminally truncated MotBC1 variants for crystallization (Fig. 1). We succeeded in crystallizing MotBC2 (residues 99–276), MotBC6 (residues 88–291) and MotBC7 (residues 88–309) (Table S1). They were crystallized in three different space groups: Form I, P3221 for MotBC6 and MotBC7; Form II, C2221 for MotBC2; and Form III, P212121 for MotBC2. We determined the structures of Form I from MotBC6, Form II, and Form III at 2.0, 1.75 and 2.4 Å resolutions respectively (Table S1). We did not refine the structure of MotBC7 because the portion on which we could have built an atomic model was the same as that of MotBC6 (residues 108–282) and because the resolution was lower with MotBC7. Since Form I contained one molecule, and Forms II and III both contained two molecules, in the asymmetric unit, we built and refined five atomic models. Each model had missing segments due to disorder (Table S2) except for one in Form II that resolved the entire chain of MotBC2 (Fig. 2A). Since this model, refined at the highest resolution (1.75 Å), contains the largest number of residues (99–276) among the five models and covers PEM (111–270), and also because the other models adopt basically the same structure (Fig. 2B), we mainly describe the structure of MotBC2 Form II.
Overall structure of MotBC2
MotBC2 appears as a single-domain structure with a long N-terminal α-helix (α1) protruding from the domain (Fig. 2). The core of the domain has a typical OmpA-like structure, which adopts a β-α-β-α-β-β fold, and shows considerable structural similarities to other PGB domains, such as the C-terminal regions of PAL (Parsons et al., 2006), RmpM (Grizot and Buchanan, 2004) and MotY (Kojima et al., 2008b) (Fig. S1). The long α1 helix is connected to PGB core structure through α2, which is perpendicular to α1, and strand β1, which is connected to β2 in the PGB core β-sheet and extends the sheet in an antiparallel manner. Although the N-terminal region of MotBC2 shows a relatively low sequence similarity among MotB proteins from various bacterial species, secondary structures prediction by the PSIPRED server (Jones, 1999) suggests that α1, α2 and β1 are common structural elements (Fig. S2).
MotBC2 forms a dimer through an interaction between its PGB domains (Fig. 2C and D). The subunit interface is composed of α4 and β4 (Fig. 2C). Strand β4 forms hydrogen bonds with its counterpart in an antiparallel manner (Figs 2C and 3A), producing a large, twisted intersubunit β-sheet. The two α4 helices also align antiparallel. The combined effect of these interactions is to make the two PGB domains tightly packed. These interactions, and the subunit arrangement for dimer formation, are totally different from those of Helicobacter pylori MotBC (HpMotBC) (Roujeinikova, 2008) (Fig. S3), although the overall folds of the domain structures resemble each other.
Dimerization through the PGB domain is essential for proper arrangement of the proton channel
To examine the significance of dimer formation, we mutated residues of MotBC6 on helix α4 at the dimer interface. Two of the changes introduce bulky side-chains that potentially disrupt the intersubunit interaction (A216W and D217W). The other two were substitutions that are known to impair motility (E213G and R223H) (Blair et al., 1991; Togashi et al., 1997) (Fig. 3A). MotBC6 fragments with each of these substitutions were monomeric in solution, as determined by size-exclusion column chromatography (Fig. 3B). MotBΔL proteins with these replacements were coexpressed with MotA in an E. coliΔmotAB strain, and their motility was assayed on soft agar plates (Fig. 3C). The mutations impaired motility either significantly or completely, suggesting that dimerization through the PGB domain is crucial to motor function.
How does disruption of dimerization of the PGB domains affect motility? Previous cross-linking studies demonstrated that the two transmembrane helices of the MotB dimer interact with each other in the MotA4/MotB2 complex (Braun and Blair, 2001; Braun et al., 2004), raising the possibility that periplasmic dimer formation influences the arrangement of the transmembrane helices. Since the F34C substitution in the transmembrane helix of MotB can cross-link the MotB dimer in E. coli (Braun and Blair, 2001), we introduced this replacement into the four mutant MotBΔL proteins and coexpressed them with MotA in the E. coliΔmotAB strain. All the four proteins significantly reduced amounts of cross-linked dimer products (Fig. 3D), indicating that dimerization of the periplasmic region is required for the proper arrangement of the transmembrane helices that form the proton channel.
A large conformational change is required to anchor the MotA/MotB complex to the PG layer
Superposition of the structure of PAL bound to the PG precursor (PDB ID code 2aiz) (Parsons et al., 2006) onto the MotBC2 structure allowed us to predict the PGB site of MotBC2 (Fig. 4A). It is located on the top surface of the MotBC2 dimer opposite to α1. Most mutations that confer non- or slow-motile phenotypes (Blair et al., 1991; Togashi et al., 1997) target this putative PGB site (Fig. S1), an observation that supports the idea that this region is essential for anchoring the stator unit to the cell wall.
Since MotBΔL can form a functional stator with MotA, the MotA/BΔL complex must be anchored to the PG layer around the rotor. The α1 helices of the MotBC2 dimer extend to the same side of the structure, and their N-termini are directly connected to the transmembrane dimer segments of MotBΔL. This situation strongly suggests that the PGB dimer of MotBΔL stands on the cytoplasmic membrane surface with the two long α1 helices as its legs and the two α4 helices on the upper surface, where the PGB sites are located [Fig. 4B, (i)]. However, the MotBC2 dimer is too small to reach the PG layer, since the distance between the surface of the hydrophobic core layer of the cytoplasmic membrane and that of the PG layer is about 100 Å, and the MotBC2 dimer is only about 50 Å tall. For the PGB sites on the top surface of MotBC2 to reach the PG layer, a large conformational change would be required. Because the PGB core forms a conserved, compact domain, the N-terminal region of PEM is not part of the PGB core. Thus, helices α1 and α2 and strand β1 are the most plausible candidates for being involved in the conformational change. If strand β1 detaches from the core β-sheet and extends colinearly with helices α1 and α2, the PGB domain would become long enough to reach the PG layer [Fig. 4B, (ii)].
The N-terminal region of PEM regulates proton translocation of the stator complex
We investigated the effects of several residue substitutions in the N-terminal region of PEM and found that L119P and L119E in α1 (Fig. 5A) affected cell growth without impairing motility (Fig. 5B). When MotBΔL mutants with either of these two changes were co-overexpressed with MotA in the ΔmotAB strain, motility was normal but growth was severely impaired. In contrast, no significant growth impairment was observed with full-length MotB and MotBΔL or with MotBΔL(L119P/E) without induction by IPTG (Fig. 5B). Cells expressing MotA/BΔL(L119P/E) formed chemotactic rings significantly larger than those formed by cells expressing MotA/BΔL under non-inducing condition (Fig. S4) and swam vigorously under an optical microscope, suggesting that the conformation of the MotA/BΔL(L119P/E) complex is more favourable than that of MotA/BΔL for efficient assembly into the motor. The L119P substitution does not seem to affect the arrangement of the transmembrane helices of the MotB dimer, based on the extent of disulphide cross-linking (Fig. 3D). Thus, the two changes at L119 must alter the conformation of the MotA/BΔL complex to allow proton translocation so that a massive proton influx causes growth impairment (Hosking et al., 2006).
A region from P52 to P65 just after the transmembrane segment of E. coli MotB (P53 to P66 in Salmonella MotB) has been proposed as a ‘plug’ for the proton channel when the MotA/B complex is not assembled (Hosking et al., 2006). Deletion of this plug region inhibits cell growth, but not motility, by acidifying the cytoplasm, suggesting that the plug regulates proton translocation depending on the assembly state of the stator. However, expression of MotBΔL, although it lacks the plug region, does not impair growth, indicating that some other region within PEM can also plug the proton channel. It could be that the whole of MotBC2 plugs the channel, as depicted in Fig. 4B, or that a specific arrangement of the transmembrane helices suppresses proton translocation because of the way it is attached to the MotBC2 dimer. The L119P/E mutations in helix α1 must destabilize the hydrophobic interactions with L149 in strand β2 and V183 and I187 after helix α3 (Fig. 5A). The loss of interactions of α1, α2 and β1 with the PGB core may allow MotBΔL to take an extended conformation (Fig. 4B) and to open the proton channel even when the MotA/BΔL complex is not assembled into the motor. The detachment of a strand from a β-sheet may appear rather drastic and unlikely. However, β1, α1 and α2 are not fully involved in the hydrophobic core of the PGB core domain, and such conformational changes involving peripheral elements have been observed in intermolecular β-sheets of many other protein complexes (Shin et al., 2003; Ranson et al., 2006).
Model for activation of the proton channel by association with the stator
Based on the structure of MotBC2 and the biochemical analyses described above, we propose that the activation of the proton channel of the MotA/B complex is coupled to the anchoring of the complex around the rotor. Initially, the MotA/B complex would diffuse through the cytoplasmic membrane. Even with the addition of the 60-residue linker to the transmembrane segment and the 37-residue C-terminal region, which are not present in MotBC2, the periplasmic domain would probably not contact the PG layer. When the MotA/B complex encounters a rotor, interactions between the cytoplasmic domain of MotA and FliG in the C-ring (Zhou et al., 1998a) could trigger conformational changes in the N-terminal region of PEM that open the proton channel and allow the PGB domain to anchor to the PG layer. Although this general type of mechanism has been proposed previously (Van Way et al., 2000; Hosking et al., 2006), our study reveals a possible structural basis for the conformational changes. Further studies, such as testing the effect of cross-linking the long α1 helix to the PGB core or the crystal structure analysis of a MotBC fragment with the L119P/E substitutions, which are currently underway, will be required to determine whether these events actually occur within the cell envelope.
Bacterial strains, plasmids and mutagenesis
Bacterial strains and plasmids used in this study are listed in Table S3. The plasmid containing in-frame deletion of MotB (pTSK30) was constructed as described by Toker et al. (1996). All the mutations in MotB were generated by the QuikChange site-directed mutagenesis method, as described previously (Kojima et al., 2008a). For the functional and cross-linking assays for each MotB mutant, we mutated motB(Δ51–100) in the plasmid pTSK30, and for the biochemical assays with the mutant MotBC6 fragment, mutations were introduced into the plasmid pTSK17.
Protein expression and purification
The genes encoding Salmonella typhimurium MotB residues 99–276 (MotBC2), 88–291 (MotBC6) and 88–309 (MotBC7) were PCR-cloned into the vector pET19b (Novagen) and overexpressed with a 5-histidine (for MotBC2) or 6-histidine (for MotBC6 and MotBC7) tag in E. coli strain BL21(DE3), as described previously (Kojima et al., 2008a). For overexpression of MotBC6 labelled with Se-Met (Se-Met MotBC6), we used B834(DE3) cells carrying the pLysS plasmid as the expression host. Cells were collected by centrifugation and suspended in buffer A (20 mM Tris-HCl pH 8.0, 150 mM NaCl) containing one tablet of Complete Protease Inhibitor cocktail (Roche Diagnostics). Cells were disrupted by French Press (Ohtake Works) and lysates were centrifuged at 100 000 g for 30 min. The soluble fraction was loaded onto a HisTrap column (GE Healthcare). His-tagged MotBC fragments were eluted by a linear 0–500 mM gradient of imidazole in buffer A. MotBC2 and MotBC7 were pure enough for crystallization after the HisTrap column. MotBC6 was further purified using a HiTrapQ column (GE Healthcare) with a linear gradient of 10–1000 mM NaCl in buffer B (20 mM Tris-HCl pH 8.0, 10 mM NaCl). Peak fractions were pooled and concentrated in buffer C (20 mM Tris-HCl pH 8.0, 100 mM NaCl) by ultrafiltration using an Amicon Ultra device (Millipore). Final concentrations of purified MotBC fragments used for crystallization were: 57 mg ml−1 (MotBC6), 18 mg ml−1 (MotBC2), 32 mg ml−1 (MotBC7) and 60 mg ml−1 (Se-Met MotBC6). Se-Met MotBC6 was purified by the same procedure as described MotBC6.
Crystallization, data collection and structure determination
Crystals were obtained at 20°C using the sitting-drop vapour-diffusion method by mixing a 1 μl protein solution with a 1 μl reservoir solution. The protein fragments were crystallized into various forms under various conditions (Table S1). Crystals were soaked in a solution containing 90% (v/v) of the reservoir solution and 10% (v/v) MPD for a few seconds, and then immediately transferred into liquid nitrogen for freezing. All the X-ray diffraction data were collected under helium gas flow at −233°C using a synchrotron beamline BL41XU of SPring-8 (Harima, Japan). The data were processed with MOSFLM (Leslie, 1992) and scaled with SCALA (Collaborative Computational Project Number 4, 1994). Phase calculation was performed with SOLVE (Terwilliger and Berendzen, 1999) using the anomalous diffraction data from the Form I crystal of Se-Met MotBC6. The best electron-density map was obtained from MAD phases followed by density modification with DM (Collaborative Computational Project Number 4, 1994). The model was constructed with COOT (Emsley and Cowtan, 2004) and was refined against the Se-Met MotBC6 data to 2.0 Å using program cns (Brunger et al., 1998). A 5% fraction was excluded from the data for the R-free calculation. During the refinement process, iterative manual modifications were performed using ‘omit map’. Structures of the Form II crystal of MotBC2 and the Form III crystal of MotBC6 were determined by molecular replacement using MOLREP (Collaborative Computational Project Number 4, 1994). Data collection and refinement statistics are summarized in Tables S2 and S4.
The atomic co-ordinates and structure factors of MotBC6 Form I, MotBC2 Form II and MotBC2 Form III have been deposited with the accession codes 2zov, 2zvy and 2zvz respectively.
Analytical size-exclusion column chromatography
Analytical size-exclusion chromatography of purified MotBC6 fragments was performed with a Superdex 75 10/300 GL column (GE Healthcare), as described previously (Kojima et al., 2008a).
We used the E. coliΔmotAB strain RP6894 as the host to assess functional properties of mutant MotB proteins. RP6894 cells harbouring plasmid pTSK30 encoding motA and motB(Δ51–100) with or without motB mutations were grown overnight at 30°C in tryptone broth [1% (w/v) tryptone, 0.5% NaCl] containing 100 μg ml−1 ampicillin. A 1 μl aliquot of an overnight culture was spotted on a tryptone soft-agar plate containing ampicillin and 0.1 mM IPTG and incubated at 30°C for 8 h.
Growth was monitored as described (Zhou et al., 1998b). RP6894 cells harbouring respective plasmids were grown overnight at 30°C in tryptone broth containing 100 μg ml−1 ampicillin. The culture was then diluted 1:30 in 3 ml of the same medium and shaken at 30°C. IPTG was added to a final concentration of 0.1 mM after 3 h and A660 was measured every hour.
Cross-linking experiments were performed on whole cells of strain RP6894 harbouring pTSK30 plasmids encoding motA and motB(Δ51–100) with or without motB mutations, following the method described previously (Braun et al., 2004). Cells were grown at 30°C to log phase in tryptone broth containing 100 μg ml−1 ampicillin and 0.1 mM IPTG, harvested by centrifugation, re-suspended in 50 mM Tris-HCl, pH 8.0, containing 20 mM EDTA to an OD660 of 5.0. Disulphide cross-linking was induced with 100 μM iodine for 2 min at room temperature and then quenched with 10 mM NEM. Cross-linked products were analysed by SDS-PAGE under non-reducing conditions, followed by immunoblotting with polyclonal anti-MotBC antibody.
We thank T. Kato and K. Iwasaki for invaluable discussions, S. Tatematsu for technical assistance, N. Shimizu, M. Kawamoto, K. Hasegawa and M. Yamamoto at SPring-8 for technical help in use of beamlines. This work was supported in part by Grants-in-Aid for Scientific Research [to S.K. (18054010), K.I. (18074006), H.M. (18074003) and K.N. (16087207)] from the Ministry of Education, Science and Culture of Japan.