Organization of the Flagellar Switch Complex of Bacillus subtilis

Flagellar motility plays key roles in the survival of many bacteria and in the harmful action of many pathogens. Bacterial flagella rotate; the direction of flagellar rotation is controlled by a multisubunit protein complex termed the switch complex. This complex has been extensively studied in Gram-negative model species, but little is known about the complex in Bacillus subtilis or other Gram-positive species. Notably, the switch complex in Gram-positive species responds to its effector CheY-phosphate (CheY-P) by switching to CCW rotation, whereas in E. coli or Salmonella CheY-P acts in the opposite way, promoting CW rotation. In the work here, the architecture of the B. subtilis switch complex has been probed using cross-linking, protein interaction measurements, and mutational approaches. The results cast light on the organization of the complex and provide a framework for understanding the mechanism of flagellar direction control in B. subtilis and other Gram-positive species.

ABSTRACT

While the protein complex responsible for controlling the direction (clockwise [CW] or counterclockwise [CCW]) of flagellar rotation has been fairly well studied in Escherichia coli and Salmonella, less is known about the switch complex in Bacillus subtilis or other Gram-positive species. Two component proteins (FliG and FliM) are shared between E. coli and B. subtilis, but in place of the protein FliN found in E. coli, the B. subtilis complex contains the larger protein FliY. Notably, in B. subtilis the signaling protein CheY-phosphate induces a switch from CW to CCW rotation, opposite to its action in E. coli. Here, we have examined the architecture and function of the switch complex in B. subtilis using targeted cross-linking, bacterial two-hybrid protein interaction experiments, and characterization of mutant phenotypes. In major respects, the B. subtilis switch complex appears to be organized similarly to that in E. coli. The complex is organized around a ring built from the large middle domain of FliM; this ring supports an array of FliG subunits organized in a similar way to that of E. coli, with the FliG C-terminal domain functioning in the generation of torque via conserved charged residues. Key differences from E. coli involve the middle domain of FliY, which forms an additional, more outboard array, and the C-terminal domains of FliM and FliY, which are organized into both FliY homodimers and FliM heterodimers. Together, the results suggest that the CW and CCW conformational states are similar in the Gram-negative and Gram-positive switches but that CheY-phosphate drives oppositely directed movements in the two cases.

IMPORTANCE Flagellar motility plays key roles in the survival of many bacteria and in the harmful action of many pathogens. Bacterial flagella rotate; the direction of flagellar rotation is controlled by a multisubunit protein complex termed the switch complex. This complex has been extensively studied in Gram-negative model species, but little is known about the complex in Bacillus subtilis or other Gram-positive species. Notably, the switch complex in Gram-positive species responds to its effector CheY-phosphate (CheY-P) by switching to CCW rotation, whereas in E. coli or Salmonella CheY-P acts in the opposite way, promoting CW rotation. In the work here, the architecture of the B. subtilis switch complex has been probed using cross-linking, protein interaction measurements, and mutational approaches. The results cast light on the organization of the complex and provide a framework for understanding the mechanism of flagellar direction control in B. subtilis and other Gram-positive species.

INTRODUCTION

Bacterial flagella are rotating molecular machines powered by the cytoplasmic membrane ion gradient (1–5). They comprise a long helical filament functioning as a propeller, a curved, flexible segment in the role of a universal joint, and a basal body spanning the cell envelope that includes elements involved in generating torque (Fig. 1). Flagellar motors can turn in either direction. In enteric species such as Escherichia coli, counterclockwise (CCW) rotation leads cells to swim relatively smoothly in what is termed a run, whereas clockwise (CW) rotation induces rapid tumbling that reorients the cell. In the absence of stimuli, alternating runs and tumbles cause cells to move in a random walk. In the presence of chemical gradients, chemotactic signaling pathways regulate flagellar switching to make tumbles less frequent when cells are moving up a gradient of attractant or down a gradient of repellent, thus prolonging runs in a favorable direction (6–8).

FIG 1.

(A) Protein organization at the base of the flagellum in E. coli. The foundation element is the MS ring, formed from FliF, in the cytoplasmic membrane (c.m.). MotA and MotB, also in the membrane, form the stator (nonrotating part) of the motor. MotB attaches to peptidoglycan (not shown). Under high load, a motor can contain several (as many as 11 in E. coli [65]) stator complexes, but motors propelling cells in liquid (thus, under light load) contain only one or a few stator complexes (66). The 3-protein “switch complex” that regulates motor direction is mounted on the rotor. The arrangement shown is based on work with the proteins of E. coli. FliG functions most directly in rotation (28) and is positioned at the top, where it interacts with the cytoplasmic domain of MotA (29). FliM is at a middle level, and FliN is at the bottom. (B) Detail showing the organization of individual domains of the switch complex proteins of E. coli. The present study examined protein organization in the switch complex of B. subtilis, which contains FliG and FliM but utilizes the protein FliY (∼40 kDa) in place of FliN (∼15 kDa).

The structure of the flagellar basal body has been observed most clearly in electron microscopic reconstructions in Salmonella enterica (9). In this species, the basal body consists of a set of rings on a rod: the LP ring, which spans the lipopolysaccharide and peptidoglycan layers; the MS ring in and above the cytoplasmic membrane; and the C ring in the cytoplasm. The C ring coincides with the so-called “switch complex,” which functions closely in motor rotation and the control of CW/CCW direction switching (10). In Salmonella and E. coli, the switch complex is a drum-shaped structure about 45 nm in diameter and 15 nm high (Fig. 1) (9, 11, 12). The complex is significantly larger in Borrelia burgdorferi (13) and certain other species (14). In Salmonella or E. coli, the complex is built from the proteins FliG, FliM, and FliN. The organization of these proteins has been examined in several cross-linking and mutational studies that, together with electron microscopy (EM) reconstructions, support a model with FliG positioned at the top (i.e., nearest the membrane), FliM at a middle level, and FliN at the bottom (15–18). FliG contains three globular domains joined by helical linking segments (15, 19). The FliG N-terminal domain (FliG_N) secures the switch complex to the basal body, binding tightly to the protein FliF that forms the MS ring (20–23) (Fig. 1). The FliG middle domain (FliG_M) rests on the FliM subunits below (17, 24–26), while the C-terminal domain (FliG_C) rests on FliG_M of the adjacent subunit (15, 16, 19, 27). FliG_C is the part of the switch complex that functions most closely in the generation of torque (28). Charged residues on the upper part of FliG_C interact with charged residues in the stator protein MotA (29–31). FliM also contains three discrete regions. A fairly short N-terminal segment provides a site of interaction with the chemotaxis signaling molecule CheY-phosphate (CheY-P) (32–35). A larger (∼200-residue) middle domain forms the side wall of the C ring (18), and the C-terminal domain binds to FliN (36–38). FliN can form tetramers (39) and probably other multimers (40, 41). FliN and FliM_C together form the bottom of the C ring. In addition to its presumed structural role in the lower part of the complex, FliN contributes to the process of flagellar export by binding the export apparatus component FliH (42–44). In E. coli, FliN also functions in chemotactic regulation by providing the site where CheY-P molecules, once captured by FliM_N, bind to induce CW rotation (45).

Switch complex architecture has not been as extensively studied in Bacillus subtilis or other Gram-positive species. Purification methods suitable for detailed electron microscopic analysis have not been developed; thus, the overall structure and dimensions of the B. subtilis switch complex are unknown. The B. subtilis complex contains homologs of FliG and FliM but instead of FliN utilizes a larger protein called FliY, which displays sequence similarity to both FliN and FliM_M (46). The portion of FliY related to FliM_M is also homologous to CheC, another chemotaxis protein found in Gram-positive species (47). The FliY_M domain has phosphatase activity specific for CheY-P (48–50). Most remarkably, whereas CheY-P induces CW rotation (and thus tumbling) in Gram-negative species, in B. subtilis it has the opposite effect, promoting CCW rotation and smooth swimming (8).

In the present study, we have examined the organization of the proteins in the flagellar switch of B. subtilis, using in vivo cross-linking, measurements of protein-protein interaction by the bacterial adenylate cyclase two-hybrid (BACTH) method, and characterization of mutant phenotypes. A working model for the structure of the switch complex of B. subtilis is presented. The model rationalizes current data on the B. subtilis switch complex and provides an initial basis for understanding the mechanism of direction switching in this and other Gram-positive species.

RESULTS

Organization of FliM_M.

In the switch complex of E. coli, many (34 or more) copies of the middle domain of FliM (FliM_M) are organized in an array that forms the side wall of the C ring, below FliG and above FliN (Fig. 1). Targeted cross-linking experiments in E. coli showed that the helix α1 edge of each FliM_M domain is near the helix α2′ edge of an adjacent FliM_M in the ring (18). Patterns of cross-linking through the α1 and α2′ edges were affected by attractant and repellent stimuli, suggesting some reorientation of the FliM_M domain upon CW/CCW direction switching (51). To determine whether FliM_M domains of B. subtilis interact with each other, we first used the bacterial adenylate cyclase two-hybrid (BACTH) method (52). Codons 45 to 229 of B. subtilis fliM were cloned into the BACTH vectors, and the four possible construct pairs (fusions of the adenylate cyclase domains to either the amino or carboxyl terminus of each partner) were examined, using MacConkey agar-lactose (Mac-Lac) plates followed by assays of β-galactosidase (β-gal) activity. One combination (FliM_M-pKT25 with pUT18-FliM_M) displayed strong color on Mac-Lac plates and gave β-gal activity of ∼3,000 Miller units (Fig. 2A), while negative controls showed negligible activity. The FliM_M domain of B. subtilis thus appears to interact with itself.

FIG 2.

Interactions between FliM_M domains. (A) FliM_M-FliM_M interaction in a bacterial adenylate cyclase two-hybrid (BACTH) experiment. β-Galactosidase activities are plotted for the construct pair that showed an interaction (using the pKT25 and pUT18 plasmids), together with corresponding negative controls. Representative results from a MacConkey agar-lactose plate are shown below the graph. (B) Disulfide cross-linking between Cys residues introduced on the α1 and α2′ edges of FliM_M. Positions of single- and double-Cys replacements are indicated. Oxidative cross-linking was induced using Cu-phenanthroline. Products were visualized on gels using antibody raised against E. coli FliM, which showed reactivity toward both FliM and FliY of B. subtilis. Cross-linked FliM multimers (reaching to hexamer or pentamer) are indicated by dots for the pairs that cross-linked most efficiently. (C) Relationship of adjacent FliM_M domains that can account for the observed cross-linking. Cysteine pairs are joined by lines with thickness proportional to cross-link yield.

To determine whether the relative orientation of the FliM_M domains in B. subtilis is similar to that deduced for E. coli, we carried out disulfide-cross-linking experiments with Cys residues introduced at several positions along the α1 and α2′ edges. Mutant fliM genes encoding the Cys replacements, under the control of an IPTG (isopropyl-β-d-thiogalactopyranoside)-regulated promoter, were integrated into the chromosome of a B. subtilis ΔfliM mutant. Cells were cultured to mid-log phase with induction by IPTG at a level (1.0 mM) that gave effective complementation in motility assays, and disulfide formation was induced using Cu-phenanthroline (Cu-phen), I₂, or H₂O₂. Products of cross-linking were characterized on immunoblots using a polyclonal antibody against E. coli FliM, which was able to detect both FliM and FliY of B. subtilis. Representative results with Cu-phen are shown in Fig. 2. Of 10 Cys pairs studied, seven (58/96, 58/187, 61/187, 65/92, 65/96, 65/97, and 65/187) gave cross-linked products upon oxidation with Cu-phen. Single-Cys controls showed relatively little cross-linking (Fig. 2B). Cross-linking was most efficient for the 58/187 and 65/187 Cys pairs, which gave products as large as pentamers. Cys pairs that failed to form disulfides (58/97, 61/96, and 61/97) were next tested with a bifunctional maleimide (bismaleimido-ethane [BMOE]) that permits cross-linking between more distant positions. The 61/96 and (to a lesser extent) 61/97 pairs were cross-linked by BMOE to form dimers and larger products (see Fig. S1 in the supplemental material). Cys replacements centered on the broader faces of FliM_M perpendicular to the α1 and α2′ edges (at positions 85 and 178) showed only weak cross-linking with Cu-phen (Fig. 2B). The pattern of cross-link yields is consistent with a FliM_M-FliM_M relationship similar to that determined previously in E. coli (18) (Fig. 2C).

Organization of FliY_M.

In place of the FliN protein occurring in E. coli, the switch complex of B. subtilis contains the larger protein FliY, which has N-terminal and middle domains homologous to those of FliM and a C-terminal domain very similar to that of FliN. The largest element unique to the Gram-positive switch complex is FliY_M, comprising about 200 residues. Crystal structures have been determined for both FliY_M and FliY_C of the hyperthermophile Thermotoga maritima (with FliY_C initially being termed FliN). (39, 49). The FliY_M domain has an overall fold similar to FliM_M and has similar dimensions except for an approximately 20% greater width between the α1 and α2′ edges. To test for self-interaction of the FliY_M domain, the BACTH system was employed with fusion constructs containing fliY codons 22 to 227 in each of the four possible combinations. One construct pair (FliY_M-pKT25 with pUT18-FliY_M, the combination that also detected a FliM_M-FliM_M interaction) showed strong color on Mac-Lac plates and β-galactosidase activities close to 3,000 Miller units. Negative controls displayed negligible activity (Fig. 3A). We infer that the FliY middle domain is involved in self-interaction.

FIG 3.

Interactions between FliY_M domains. (A) BACTH experiment demonstrating FliY_M-FliY_M interaction. β-Galactosidase activities are plotted for the construct pair that showed an interaction, along with corresponding controls. A representative Mac-Lac plate is shown below the graph. (B) Disulfide cross-linking between Cys residues introduced on the α1 and α2′ edges of FliY_M. Positions of Cys replacements are indicated; oxidative cross-linking was induced using Cu-phenanthroline. Products were visualized on gels using antibody raised against E. coli FliM. FliY-FliY cross-linking products (those requiring the presence of both Cys residues) are indicated by dots. For a discussion of bands observed in the single-Cys mutants, see the text. (C) Relationship of adjacent FliY_M domains that can account for the cross-linking. The thickness of lines reflects the largest multimer observed upon cross-linking (dimer, trimer, or tetramer).

To further explore FliY_M interactions, disulfide cross-linking was again used to provide constraints on the domain orientation. Mutant fliY genes encoding cysteine replacements at a number of positions on the α1 and α2′ edges of the domain were integrated into the chromosome of a ΔfliY strain, and oxidation was induced using Cu-phen. Products were characterized on anti-E. coli FliM immunoblots. Representative results are shown in Fig. 3B. All five Cys pairs examined yielded dimers or larger products on treatment with Cu-phen. Unlike the case for FliM_M, where the single-Cys controls showed little cross-linking, several of the single-Cys FliY_M proteins also formed dimers on treatment with Cu-phen. FliY has been shown to form homodimers held together through intertwined C-terminal domains (39). Cross-linking of the single-Cys FliY proteins might then occur between the FliY_M domains in a homodimer rather than between adjacent units in the switch complex. Consistent with this, several of the fliY single-Cys mutants still yielded some cross-linked dimer even when expressed in a ΔfliM ΔfliY strain that is unable to assemble switch complexes even with FliY reintegrated into the chromosome (see Fig. S2 in the supplemental material). Four Cys pairs (54/176, 77/155, 155/176, and 54/77) gave rise to cross-linked products that were not seen in the single-Cys controls or in the ΔfliM ΔfliY strain, including trimers and tetramers, and which are presumably due to cross-linking between adjacent FliY_M units in the complex. Together with the BACTH experiment, the cross-linking results indicate that multiple FliY_M domains are adjacent in the B. subtilis switch complex, approaching through regions (the α1 and α2′ edges) similar to those involved in the FliM_M-FliM_M interaction (Fig. 3C).

Next, two-hybrid experiments were used to test for interactions of FliY_M with other potential binding partners in the switch. FliY_M did not show detectable interactions with FliM_M (all 8 BACTH combinations tested), FliG_N (residues 1 to 102; all combinations tested), FliG_M (residues 103 to 188; all combinations tested), or FliG_C (residues 201 to 338; 6 combinations tested). While negative results of two-hybrid experiments do not definitively exclude interaction between tested proteins, we infer that the middle domain of FliY primarily mediates homomeric interactions.

The FliY_M array is in proximity to FliM_C.

We next sought to locate the FliY_M array relative to FliM. We introduced Cys residues at positions approximately centered on the broad faces of the middle domains of FliM (position 85 or 178) and FliY (position 69 or 165). Initial attempts to induce disulfide formation (using iodine, Cu-phenanthroline, or hydrogen peroxide) gave no cross-linking, indicating that the positions chosen on FliM_M and FliY_M are not in very close proximity. The more-permissive bifunctional reagent bismaleimido-hexane (BMH) was then used. A range of cross-linked products was seen even with no Cys introduced in FliM_M, indicating the involvement of one or both of the native FliM Cys residues (at positions 217 and 284 in the B. subtilis protein). Experiments were then repeated with one or the other native FliM Cys residue removed by mutation. With Cys217 removed, experiments again showed a diversity of bands, including the intended products (those arising from the introduced Cys residues) in only low yield but a substantial yield of slower-migrating products assignable to FliY dimer (Fig. 4, upper panel). Faster-migrating products were also observed in significant yield; experiments with the other native Cys removed (lower panel) indicated that the faster-migrating products are the result of cross-linking through Cys284, which is in FliM_C. The substantial yield of cross-linking through the native Cys residue FliM-284, greater than that occurring through either of the Cys residues introduced into FliM_M, points to a location for FliY_M in the vicinity of FliM_C.

FIG 4.

Cross-linking of Cys residues introduced in FliM_M and FliY_M. Cross-linking used the bifunctional reagent bis-maleimidohexane. Introduced Cys residues are indicated at the top; all Cys mutants displayed vigorous swimming in liquid culture. Upper panel, experiments using a mutant variant of FliM with the native Cys217 residue removed by mutation. Open arrowheads indicate relatively faint products assignable (on the basis of relative mobilities and a requirement for both introduced Cys replacements) to cross-linking between the introduced Cys residues in FliM_M and FliY_M. Filled arrowheads indicate products that arose in higher yield, which, based on comparison to the experiment in the lower panel, involve cross-linking to the native Cys284 of FliM. Additional bands (not labeled by arrowheads) indicate cross-linking of FliY to itself, most strongly through Cys165. Lower panel, experiments using a FliM variant with the native Cys284 residue removed by mutation. Open arrowheads indicate products assignable to the introduced Cys residues (occurring at the same positions as the similarly labeled fainter bands in the upper gel).

FliY and FliM C-terminal domains intertwine to form heterodimers.

In E. coli and Salmonella, the lower part of the switch complex is formed from FliN and FliM_C (36, 37). A crystal structure of the FliN homolog FliY_C from T. maritima shows the domain forming an intertwined homodimer (39). FliY of B. subtilis was previously shown to restore strong motility, but with an aberrantly CCW bias, to a Salmonella fliN mutant (46). We tested a FliY_C construct containing only the FliN-related region (residues 233 to 378) and found that it sufficed to restore strong, CCW-biased motility to an E. coli ΔfliN mutant (as assayed by microscopic examination in liquid cultures). This implies that B. subtilis FliY_C is folded similarly to E. coli FliN and can take the place of FliN in the E. coli switch complex to enable motor assembly and rotation.

FliM_C shows significant sequence similarity to FliY_C (39) and might therefore fold in a similar way, potentially forming intertwined FliM_C/FliY_C heterodimers. While the cross-linking experiments of Paul et al. (43) appeared to argue against such heterodimers, a recent crystallographic study by Notti and coworkers (41) provided clear evidence of intertwined FliM_C/FliN units in Salmonella. A mass spectrometric (MS) analysis of complexes formed by T. maritima FliM, FliY, and FliY fragments indicated the occurrence of FliM_C/FliY_C heterodimers (40) and further suggested that a short form of T. maritima FliY (the C-terminal domain, FliY_C) is produced from an alternative start site and can intertwine either with itself or with FliM_C (40).

To test for the presence of a short form of FliY in B. subtilis, whole-cell proteins were examined on immunoblots probed with anti-FliN antibody. Blots showed only the full-length protein (see Fig. S3 in the supplemental material). A C-terminal FliY construct comprising residues 233 to 378 gave a strong band at the expected position, indicating that the antibody was capable of detecting FliY_C. The absence of the band in cell lysates indicates that the B. subtilis switch contains only a small amount, if any, of shortened forms of FliY. We note, though, that when full-length FliY was heterologously expressed in E. coli, extended exposures showed a weak band at the position expected for FliY_C, indicating either translation from an internal start site or breakdown of the protein by an E. coli protease (Fig. S3). We conclude that the short form of FliY may be an artifact of heterologous expression and is likely not present or relevant in B. subtilis.

To begin probing the organization of FliY_C, we used BACTH experiments to look for interactions of the domain with itself or with FliM_C. FliY_C constructs contained residues 286 to 378, and FliM_C constructs included residues 241 to 332. Strongly interacting pairs of constructs were found for both FliY_C/FliY_C and FliY_C/FliM_C (Fig. 5A and B). β-Galactosidase activities were comparable for the FliY_C/FliY_C and FliY_C/FliM_C interactions and also showed a similar pattern of variation among the construct pairs (compare Fig. 5A and B). These results indicate that both FliY_C homodimers and FliY_C/FliM_C heterodimers can form, likely with similar geometries of interaction. In contrast, BACTH experiments with the FliM_C domain gave no evidence of homotypic FliM_C/FliM_C interaction (see Fig. S4 in the supplemental material).

FIG 5.

FliM_C-FliY_C and FliY_C-FliY_C interactions and presence of FliY homodimers in the B. subtilis switch complex. (A) BACTH experiments indicating a FliY_C-FliY_C interaction. Results are shown for the four possible construct combinations, with relevant negative controls on the right. Experiments used a minimum of six biological replicates. (B) BACTH experiments demonstrating FliM_C-FliY_C interaction, with relevant controls (those necessary in addition to the ones in panel A) on the right. (C) Homology model of the FliM_C/FliY_C heterodimer based on the structure of a Salmonella FliM_C/FliN complex (41), indicating positions of Cys replacements used to examine proximities. Iodine-induced cross-link yields are indicated by the thickness of the orange lines. The green dashed line joins a distant Cys pair included as a negative control. The FliM²⁶⁶/FliY³¹⁰ pair, indicated by the dashed orange line, showed reduced monomer intensities on blots and gave a cross-linked product observable only on heavier exposure (see Fig. S5 in the supplemental material) (D) Disulfide cross-linking between positions in FliM_C and FliY_C. Oxidation was with I₂; the blot used anti-E. coli FliM antibody. Positions of Cys residues are indicated at the top. (E) Localization of some FliY homodimer to the membrane fraction in cells that make flagella (ΔfliY) but not in nonflagellate (ΔfliMY) cells that cannot assemble switch complexes. Cross-linking was through Cys at position 310, with oxidation by I₂. The blot used anti-E. coli FliM antibody. FliY with the L310C mutation was expressed from a chromosomally integrated construct using the lac promoter.

To test the proposal that the FliM_C/FliY_C heterodimer is folded similarly to the FliY_C homodimer, for which the structure is known, we carried out disulfide-cross-linking experiments using five Cys pairs (FliM²⁶³/FliY³¹², FliM²⁶⁵/FliY³⁰⁹, FliM²⁶⁶/FliY³¹⁰, FliM³¹⁹/FliY³²⁸, and FliM²⁸⁴/FliY³⁶³) that should be in fairly close proximity in a FliM_C-FliY_C heterodimer modeled on the crystal structure of the FliY_C homodimer (39) (Fig. 5C). A pair that should be well separated (FliM³¹³/FliY³⁷⁶) was included as control. Mutant fliM and fliY genes encoding the pairs of Cys residues were integrated into the chromosome of a ΔfliM ΔfliY mutant. On treatment with I₂, four of the Cys pairs (FliM²⁶³/FliY³¹², FliM²⁶⁵/FliY³⁰⁹, FliM³¹⁹/FliY³²⁸, and FliM²⁸⁴/FliY³⁶³) gave a band at the position expected for a FliM/FliY heterodimer (Fig. 5D). The negative-control pair FliM³¹³/FliY³⁷⁶ gave only a very weak band that was also observed in the FliY³⁷⁶ single-Cys control. The FliM²⁶⁶/FliY³¹⁰ pair, although expected to be close, gave a comparatively weak band observable only with heavy exposures (see Fig. S5 in the supplemental material), which might reflect the fact that these residues are more buried than the others tested. Most single-Cys controls gave the expected negative result, but significant cross-linking was observed with single-Cys replacements in FliY at position 309, 310, or 376. The crystal structure shows that positions 309 and 310 are near “themselves” in the FliY_C homodimer (39). Cross-linking through Cys³⁷⁶ might involve adjacent dimeric units in the switch complex, as discussed further below. We note that the Cys³⁷⁶ mutant was nearly immotile, a possible further indication of a functionally relevant interdimer contact.

Membrane localization of FliY homodimers.

FliY homodimers might be incorporated into and important for the function of the B. subtilis switch complex or, alternatively, might provide a pool of excess FliY to ensure that all FliM_C domains can be paired with a FliY_C. To address this, we compared the localizations of FliY homodimers in cells that assemble flagella and cells that do not. Experiments used the Cys replacement at FliY position 310, which as noted is near the symmetry axis in the homodimer. Following oxidation with I₂, the membrane and soluble fractions were separated and examined for the presence of FliY monomer and dimer. Experiments were done in either a ΔfliY background where flagella are assembled (when the fliY construct has been integrated into its chromosome) or a ΔfliM ΔfliY background that remains nonflagellate (even with the fliY construct integrated). Cross-linked FliY dimers occurred in both the membrane and soluble fractions in the cells assembling flagella but were almost exclusively in the soluble fraction of the nonflagellate ΔfliM ΔfliY cells (Fig. 5E). The FliM-dependent membrane localization of FliY homodimers supports the proposal that FliY homodimers are present in the switch complex.

Other cross-links involving the FliM_C and FliY_C domains.

The switch complex presumably contains many copies of the FliY_C/FliY_C and FliY_C/FliM_C dimers, and if these are organized similarly to the corresponding part of the E. coli switch, adjacent dimeric units are expected to be in proximity. As noted above, the cross-linking observed in the FliY³⁷⁶ single-Cys control (Fig. 5) appears unlikely to take place between the two subunits in the dimer (Fig. 5) but might occur between adjacent dimeric units. Slightly heavier exposures showed two bands in the FliY³⁷⁶ Cys mutant; the lower of these disappeared when the native Cys284 of FliM was removed by mutation and is therefore assignable to a FliY³⁷⁶-FliM²⁸⁴ cross-link (Fig. 6). The remaining band, migrating with a somewhat higher apparent molecular weight (MW), is most readily explained as an interdimeric FliY³⁷⁶-FliY³⁷⁶ product. To see if nearby positions behaved similarly, additional Cys replacements were made, at position 366 in FliY (which is near position 376) and at residues 322 and 332 in FliM, which are in the positions corresponding to FliY³⁶⁶ and FliY³⁷⁶ in a homology model of the heterodimer (Fig. 6B). On treatment with Cu-phenanthroline, products assignable to FliM³³²-FliY³⁶⁶ and FliM³³²-FliY³⁷⁶ cross-linking were observed (Fig. 6). Like FliY³⁷⁶, these positions are distant from each other in the heterodimer, and their cross-linking is more plausibly explained as occurring between adjacent dimeric units.

FIG 6.

(A) Putative interdimer cross-linking through positions in FliM_C and FliY_C. Cys replacements were at positions that should be distant in either the FliY_C/FliM_C heterodimer or the FliY_C/FliY_C homodimer. Certain (unexpected) bands were found to involve the Cys natively present at position 284 in B. subtilis FliM; to assist with assignments, this Cys was removed by mutation for two experiments (as indicated). Cross-links involving the native Cys284 are indicated by open arrowheads. Two FliM-FliY cross-links that can be assigned to the introduced Cys residues are indicated by filled arrowheads. (B) Speculative arrangement of adjacent C-terminal domain dimers that could account for the observed cross-linking. Two FliM_C-FliY_C heterodimers are pictured; if the one on the left is replaced with a FliY^C homodimer, then two FliY³⁷⁶ residues come into proximity. Cross-linking between positions within a dimer would involve a significantly greater distance, as indicated by the dashed line.

Organization of FliG domains.

Crystal structures of FliG from Aquifex and Thermotoga show three globular domains joined by largely helical linking segments (19, 53). FliG is located in the upper (i.e., membrane-proximal) part of the switch complex, where it interacts with the membrane protein MotA of the stator (29). In the switch complex of E. coli, the FliG_N domains form a ring at a more-inward radius than the FliG_M/FliG_C array (9, 15, 20). To probe the organization of B. subtilis FliG_N, we introduced pairs of Cys residues in FliG_N at positions 46/83 or 47/102, homologous to the pairs 41/78 and 42/97 that cross-linked efficiently in E. coli (15). Genes encoding the Cys mutant proteins were integrated into the chromosome of a ΔfliG B. subtilis strain. The 46/83 mutant was motile but with an aberrantly smooth-swimming bias, whereas the 47/102 mutant migrated about as well as the wild type (wt) in soft agar (Table 1). On oxidation with I₂, both Cys pairs yielded multimers of FliG (Fig. 7A). The efficient cross-linking through these positions parallels what was observed in E. coli (15) and indicates a similar orientation of FliG_N domains in the two species.

TABLE 1.

Motility phenotypes of B. subtilis switch protein mutants

Protein and mutation(s)	Migration ratea	Commentb
FliM
R58W	1.11 ± 0.08
R58C	0.73 ± 0.03
D61W	0.10 ± 0.05	Smooth bias
D61C	0.32 ± 0.01	Smooth bias
R65W	0.88 ± 0.06
Y78C	0.75 ± 0.03
S85C	0.22 ± 0.01	Smooth bias
D87W	0.75 ± 0.06	Mild tumble bias
E92W	0.25 ± 0.07	Slow; smooth bias
E92C	0.44 ± 0.03	Smooth bias
R96C	1.18 ± 0.13
S97C	0.98 ± 0.05
S97W	0.00	Smooth bias
N100W	0.65 ± 0.16	Mild smooth bias
M133D	0.03 ± 0.05	Nonflagellate
E167W	0.12 ± 0.14	Slow, smooth
E178C	0.83 ± 0.06
E178W	0.81 ± 0.07
Q187C	0.75 ± 0.10
Q187W	0.58 ± 0.05	Tumble bias
F188C	1.26 ± 0.09
N215W	0.87 ± 0.05
T268W	0.82 ± 0.05
S269W	0.60 ± 0.05	Smooth bias
E270W	1.08 ± 0.06
L271D	0.00	Nonflagellate
C284W	0.91 ± 0.13
R58C/E92C	0.79 ± 0.06	Suppresses E92C
R58C/R96C	0.53 ± 0.09	Tumble bias
R58C/S97C	0.41 ± 0.07	Tumble bias
R58C/Q187C	0.05 ± 0.03	Smooth bias
D61C/E92C	0.01 ± 0.02	Smooth bias
D61C/R96C	0.24 ± 0.03	Smooth bias
D61C/S97C	0.67 ± 0.05	Smooth bias
D61C/Q187C	0.01 ± 0.02	Smooth bias
R65C/E92C	0.87 ± 0.11
R65C/R96C	0.85 ± 0.15
R65C/S97C	0.49 ± 0.05	Tumble bias
R65C/Q187C	0.54 ± 0.05	Tumble bias
FliY
I47C	0.01 ± 0.03	Smooth bias
I47D	0.00	Smooth bias
T54C	0.33 ± 0.05	Smooth bias
T54W	0.00	Smooth bias
T58W	0.62 ± 0.10	Mild smooth bias
S77C	0.64 ± 0.05	Mild smooth bias
S80W	1.87 ± 0.36
I119D	0.67 ± 0.05	Mild smooth bias
E132C	0.94 ± 0.12
I133D	0.39 ± 0.09	Smooth bias
I133W	1.62 ± 0.09
T155W	0.65 ± 0.07	Smooth bias
T155C	0.97 ± 0.20
G178W	2.08 ± 0.25
V189D	0.00	Nonflagellate
M298D	0.00	Nonflagellate
L310C	1.21 ± 0.13
I329D	0.00	Nonflagellate
I347D	1.42 ± 0.12
V353D	0.00	Nonflagellate
D366Cc	0.81 ± 0.03
L368D	0.71 ± 0.09	Mild smooth bias
E376Cc	0.00	Nonflagellate
I47C/S77C	0.00	Smooth bias
I119D/I133D	0.04 ± 0.05	Smooth bias
T54C/S77C	0.12 ± 0.07	Smooth bias
S77C/T155C	0.7 ± 0.11	Smooth bias
FliG
I20D	0.19 ± 0.08	Smooth bias
I39D	0.25 ± 0.11	Smooth bias
I46C	0.19 ± 0.08	Smooth bias
S47C	0.73 ± 0.18	Mild tumble bias
R83C	0.90 ± 0.14
L102C	1.22 ± 0.08
Q121C	1.01 ± 0.15
A162D	0.00	Nonflagellate
A162C	1.00 ± 0.17
R170C	0.90 ± 0.20
I219C	1.05 ± 0.07
L223D	0.02 ± 0.02	Nonflagellate
I46C/R83C	0.17 ± 0.06	Slow, smooth
S47C/L102C	1.05 ± 0.14
Q121C/R170C	1.02 ± 0.15
A162C/I219C	0.95 ± 0.04
R286E	0.00	Flagellate but immotile
E294K	0.00	Flagellate but immotile
Δ(173-175)	0.00	Nonflagellate

^aRate of migration in soft (0.26% agar) tryptone plates, relative to that of the wild type. Values are means ± standard deviations for at least 6 biological replicates.

^bSwimming behaviors refer to unstimulated cells in liquid culture.

^cThese measurements used the ΔfliM ΔfliY strain with both fliM and fliY integrated at the amyE locus.

FIG 7.

Examples of efficient cross-linking involving FliG. In all panels, Cys pairs were at positions homologous to ones that cross-linked in high yield in E. coli (15, 16, 20, 54). Relevant domain interfaces are schematized in Fig. 9 and 10. (A) Cross-linking of Cys pairs at the FliG_N-FliG_N interface (20). Cross-linking used 1 mM I₂. Positions of FliG monomer and multimers are indicated by dots. (B) Cross-linking of a Cys pair in the region of the putative FliG_M-FliG_M contact (54). Cross-linking used 0.2 mM I₂. FliG monomer and multimers are indicated by dots. The relatively slow migration (apparent MW, ∼110 kDa) of FliG dimers cross-linked through these positions in E. coli was observed previously (54). (C) Cross-linking of a Cys on the top of FliG_M to Cys in the hydrophobic patch at the bottom of FliG_C. Cross-linking used 1 mM I₂ (15, 16).

In studies in E. coli, constraints on the orientation of FliG_M were likewise provided by cases of high-yield cross-linking. To determine whether FliG_M is oriented similarly in B. subtilis, we examined disulfide cross-linking with Cys residues introduced at positions 121 and 170, homologous to the pair 117/166 which cross-linked efficiently in E. coli (54). Treatment with I₂ gave rise to a ladder of products extending to pentamer (Fig. 7B), indicating a FliG_M-FliG_M relationship similar to that in E. coli.

In the switch complex of E. coli, FliG_M rests on FliM_M (24, 25) at an interface formed from the conserved motifs “GGXG” in FliM_M and “EHPQR” in FliG_M (17, 24–26). Residues occurring at this interface in crystal structures of the T. maritima FliM_M/FliG_M complex (24, 25) are almost perfectly conserved in B. subtilis, suggesting that the FliM_M/FliG_M interactions will be similar. To examine the FliM_M/FliG_M interaction in B. subtilis, we used the BACTH system with constructs containing FliM residues 45 to 229 (FliM_M) and FliG residues 103 to 188 (FliG_M). Of seven construct combinations tested, one (pKNT25-FliG_M/FliM_M-pUT18C) gave β-galactosidase activities well above control values (Fig. 8A). This FliG_M/FliM_M interaction was blocked by mutation (to Asp) of FliM residue Met133, which is at the FliM_M/FliG_M interface in the T. maritima crystal structure (24, 25). The FliG_M/FliM_M interaction appears to be essential for flagellar assembly; the M133D mutation prevented migration of cells in soft agar (Fig. 8B) while rendering cells nonflagellate as determined by staining (55).

FIG 8.

Interaction between FliG_M and FliM_M and effect of the M133D mutation in FliM. (A) BACTH results for the construct combination (of 7 tested) that showed a FliG_M-FliM_M interaction, and loss of the interaction in the M133D mutant. (B) Immotility of the M133D mutant in soft agar. Staining showed that cells of the mutant were nonflagellate.

Targeted cross-linking studies in E. coli showed that FliG_C stacks onto FliG_M of the adjacent subunit in the complex (15, 16). To test for this FliG_C/FliG_M interaction in B. subtilis, we examined disulfide cross-linking of a Cys pair (162/219) homologous to a pair (158/214) that cross-linked efficiently in E. coli (15). Products as large as pentamers were formed on oxidation with I₂ (Fig. 7C), indicating that a FliG_C/FliG_M stacking interaction occurs in B. subtilis similarly to how it occurs in E. coli.

Analysis of mutants.

(i) FliM_M domain. To assess the functional importance of various regions in the B. subtilis switch proteins, we examined the effects of point mutations. Function was assayed by rates of migration in soft agar and by light-microscopic examination of cells swimming in liquid. Immotile mutants were stained to determine whether cells were nonflagellate or flagellate but paralyzed. In addition to the Cys replacements made for cross-linking, a number of other replacements were made in various protein regions highlighted in previous studies in Salmonella or E. coli.

As described above, a mutation on the “top” of the FliM_M domain (M133D) weakened the interaction with FliG_M and prevented flagellar assembly. Several additional mutations in FliM_M, (many of them single- or double-Cys replacements made for the cross-linking experiments) were characterized. The results are shown on the structure in Fig. 9A and are summarized in Table 1. Three bulky (Trp) replacements on the broad faces of the domain (at positions 87, 178, and 215) had only minor effects, allowing migration at 75% of the wild-type rate or greater. Mutations on the edges of the domain had varied effects, in some cases allowing normal motility but most often causing slower chemotaxis as a result of aberrant run/tumble bias (Fig. 9A). Two double-Cys mutants (R58C/Q187C and D61C/Q187C) had defects much more severe than those caused by the individual mutations (and also formed disulfide cross-links [Fig. 2]). One double mutant (R58C/E92C) displayed significant intragenic suppression, migrating as well as the wild type in motility plates even though the E92C mutation by itself caused a fairly severe chemotaxis defect. The prevalence of FliM_M mutations that affect switching suggests that the FliM_M domain is important for regulating CW/CCW bias in B. subtilis, as is known to be the case in Salmonella (56) and E. coli (51).

FIG 9.

Effects of mutations in the switch complex proteins of B. subtilis. Positions of mutations are indicated by balls, which are colored according to phenotype: gray, close to wt function; yellow, impaired chemotaxis caused by excessive CW bias; blue, impaired chemotaxis caused by excessive CCW bias, with stronger color indicating a more severe defect; red, surface position where mutation causes a defect in flagellation; brown, buried (thus potentially structure-disrupting) mutations causing a defect in flagellation. At positions where two or more replacements were made, the more severe phenotype is shown. The amino acid replacements made and relative migration rates are summarized in Table 1. (A) Mutations in FliM_M. Two views of the domain, related by a 180-degree rotation, are shown. (B) Mutations in FliY_M. Two views are shown. Side chains shown as pink sticks and with numbers in magenta are residues at the phosphatase active sites. (C) Mutations in the FliY_C/FliM_C heterodimer. The structure shown is for a complex of the Salmonella FliY_C/FliM_C heterodimer with a segment of FliH (green), a protein involved in flagellar export (and thus in flagellar assembly) (41). FliY_C is in darker gray, with residue numbers on the left, in boldface. (D) Mutations in FliG_N. Two copies of FliG_N are shown together with their bound helical FliF_C segments (beige) (20). The purple dashed line traces the FliG_N-FliG_N interface. The arrangement of helical segments at the interface is a proposal based on previous cross-linking experiments in E. coli (20). Positions that cross-linked in the present study in B. subtilis (Fig. 7) are connected by orange lines. Mutations at three positions along the interface (positions 20, 39, and 46) caused significant motility impairments. (E) Mutations in the FliG middle and C-terminal domains. Two FliG_M domains and one FliG_C domain are shown. The stacking of FliG_C on a neighboring FliG_M is illustrated schematically. Orange lines join Cys pairs that cross-linked in B. subtilis in the present study and that were previously shown to cross-link in E. coli. Conserved charged groups at the top of FliG_C (cyan) gave an immotile but flagellate (Mot⁻) phenotype when mutated to residues of opposite charge.

(ii) FliY_M domain.

Most of about two dozen FliY_M mutations studied caused significant impairments in chemotaxis plates (Fig. 9B; Table 1). Exceptions that supported migration at 75% of the wild-type rate or greater were near the center of one of the broad faces (position 204), near the α2′ edge (positions 80 and 178), or near the α1 edge but fairly conservative in character (T155C). Several mutations near the α1 edge (I47C, I47D, T54C, T54W, T58W, and T155W) caused aberrantly smooth swimming. These replacements are in positions near the phosphatase active sites (Fig. 9) where they might hinder the interaction of FliY with CheY-P. The most severe defects occurred with replacements of residue 47, immediately adjacent to the phosphatase active-site residue N46 (49, 50). Decreased phosphatase action and resulting hyperphosphorylation of CheY would be expected to give a smooth-swimming phenotype in B. subtilis. Two surface-exposed hydrophobic residues at the top of the domain, I119 and I133, also proved to be important for regulating CW/CCW bias: individual replacements with aspartate reduced chemotactic migration significantly, and an I119D/I133D double replacement prevented chemotaxis entirely, again as a result of excessively smooth swimming (Table 1). These residues, particularly I133, are also positioned where they might affect the interaction with CheY-P. Flagellar assembly was prevented by just one of the mutations examined in FliY_M; this was V189D, at a position that is buried and likely important for folding of the domain.

(iii) FliM and FliY C-terminal domains.

A few nonconservative point mutations in FliM_C were studied and were mostly found to have only minor effects. Exceptions were L271D, which prevented flagellar assembly, and L295E, which caused delayed-onset motility, indicating that some step(s) in flagellar assembly is made less favorable. Leu271 is partially buried and might be expected to contribute to stable folding. Leu295 is in the region involved in binding the flagellar-export protein FliH (Fig. 9C).

In a small collection of FliY_C mutations, about half were tolerated by the criterion of 75% function. A few mutations caused a complete loss of motility, in all cases by preventing flagellar assembly. These included I329D in a partially buried residue that could be important for stable folding and V353D in a surface region that has been implicated, in E. coli and Salmonella in binding to the flagellar-export protein FliH (Fig. 9C) (41, 42, 44, 57). A third, M298D, is in a segment near the beginning of the domain whose actual position in the complex is uncertain. Unlike the nearby L295E mutation in FliM that retarded flagellar assembly, the V353D mutation gave a nonflagellate phenotype at all stages of growth. The corresponding mutation in E. coli (the V111D replacement in FliN) gave a nonflagellate phenotype and had been shown to weaken binding to FliH (57). As noted above, the FliY³⁷⁶ Cys replacement also caused a severe motility impairment, for reasons that are unclear but which might involve interactions between adjacent dimeric units.

(iv) FliG.

Mutational studies in E. coli identified surface regions of FliG that are essential for switch complex assembly (15, 16, 19, 53). Phenotypes of mutations on the corresponding surfaces of B. subtilis FliG support the proposal that it is organized similarly. Motility was significantly impaired by three mutations in FliG_N (I20D, I39D, and I46C) (Fig. 9D) at positions implicated in the FliG_N-FliG_N interface in the present experiments (Fig. 6) and in previous experiments in E. coli (15). Motility and flagellar assembly were prevented by the replacements A162D in FliG_M and L223D in FliG_C, which lie at the FliG_C/FliG_M interface (Fig. 9E) identified in previous mutational and cross-linking experiments in E. coli (15, 16, 19, 53).

Conserved charged residues in a helix at the top of the FliG_C domain are known to be important specifically for motor rotation, as opposed to flagellar assembly, in both E. coli (29, 58) and Vibrio (30, 31). In E. coli, individual alanine replacements of a conserved basic or acidic residue were tolerated, while a double replacement gave an immotile but flagellate phenotype (58). In B. subtilis FliG, charge-neutralizing mutations at the corresponding positions (Arg286 and Glu294) were tolerated, both singly and together. Charge-reversing mutations at either position gave a nearly immotile phenotype, with only a very small fraction (fewer than 1%) of cells showing any movement, while cells remained normally flagellated. A double mutation reversing both charges gave a fully immotile phenotype. Thus, in B. subtilis, as in E. coli, this region on FliG is important for motor rotation.

In Salmonella, a three-residue deletion in FliG (Δ169-171), near the beginning of the helix that links FliG_M to FliG_C, was found to give strong CW motor bias (59). In B. subtilis, where the default motor direction is already CW, the corresponding deletion (Δ173-175) gave nonflagellate phenotype (Table 1). This resembles what was found in Vibrio alginolyticus, where the corresponding deletion also gave a nonflagellate phenotype (60).

DISCUSSION

Because most studies of flagellar ultrastructure have been undertaken in E. coli and Salmonella, information on the organization of the switch complex in Gram-positive species has been limited. The switch complex of B. subtilis can serve as a model for many Gram-positive species and might also provide clues to switch complex architecture in other species, such as Helicobacter pylori and certain spirochetes, whose switches contain a FliY-like protein. Efforts to examine the B. subtilis switch complex by electron microscopy have been hampered by the difficulty of purifying basal bodies in an intact form. Here, we used biochemical and mutational approaches that do not rely on isolation of basal bodies and that previously provided useful constraints on switch protein organization in E. coli.

Given the overall similarity in components, it might be supposed that general features of protein organization should be similar in the switch complexes of B. subtilis and E. coli. In major respects, this appears to be true. The FliM_M domains are organized similarly (Fig. 2); the FliM_M domains are positioned under the FliG_M domains, as in E. coli (Fig. 7); multiple copies of both FliG_N and FliG_M are organized into arrays (presumably rings) in which the adjacent subunits interact through the regions analogous to those identified in E. coli (Fig. 6A and B); and the FliG_C domains rest, as they do in E. coli, on the adjacent FliG_M domains in the array (Fig. 6C). Finally, as in E. coli, conserved charged residues at the top of FliG_C function specifically in motor rotation, giving a flagellate but paralyzed phenotype when mutated (Table 1).

A major difference from E. coli is the presence in the B. subtilis switch complex of an additional array containing the FliY_M domains. Evidence for an assembly with multiple FliY_M domains came from BACTH experiments that showed a FliY_M-FliY_M interaction (Fig. 3A) and from cross-linking experiments indicating proximity of the α1 and α2′ edges of adjacent FliY_M domains (Fig. 3). If the FliY_M array is presumed to be a ring, then the larger size of FliY_M relative to FliM_M in the relevant dimension (an ∼20% greater distance between the α1 and α2′ edges) would suggest that it lies at greater radius than the FliM_M ring. The cross-linking experiments do not locate the FliY_M array precisely but point to a position low in the complex, in proximity to FliM_C. This location is supported by tomographic studies of Zhao and coworkers, who found a FliY-related feature in this region of the basal bodies of Leptospira interrogans (61).

The FliG subunits in the B. subtilis switch appear not only to be organized similarly to those in E. coli but to be supported in a similar way, through interactions of the EHPQR motif of FliG_M with the GGXG motif at the top of FliM_M. The actual picture might be more complex, however, given the occurrence of both FliY homodimers and FliM/FliY heterodimers in the structure. A reasonable (though not the only) possibility is that the FliY homodimers occupy positions quasiequivalent to those of the FliM/FliY homodimers, in which case some FliY_M domains would be interspersed in the otherwise-FliM_M array. (An alternative proposal would be that the FliY homodimers form their own rings, giving rise to a larger switch complex.) Because FliY_M appears not to interact with FliG, the presence of FliY_M domains would require some adjustments to the model in the region of the FliG-FliM interface, as discussed further below.

A major role of B. subtilis FliY_M is to catalyze dephosphorylation of CheY-P at two distinct active sites (49, 50). If the phosphatase active sites are accessible (i.e., are not occluded by the adjacent FliY_M domains), they should act to intercept and inactivate CheY-P molecules diffusing to the switch until the rate of CheY-P arrival exceeds the enzymatic capacity of the array. The benefit of a saturable opposing process in enhancing switch-like (nonlinear) behavior in signaling processes has been noted (62). Further, the proximity of FliY_M to FliM_M (and the possible presence of some FliY_M domains within the FliM_M array) means that switching might alter the accessibility of the phosphatase active sites to provide an additional feedback mechanism controlling the duration of CW versus CCW intervals.

Certain species, including the previously mentioned Leptospirillum and others such as H. pylori, contain variants of FliY that lack the Glu and Asn residues (49) associated with phosphatase activity. If FliY forms an additional ring within the switch complex as we suggest, then in those species it might serve a structural rather than a regulatory role, reinforcing the complex to allow it to function in the face of greater load presented by rotation in a viscous mucosal environment (H. pylori) or in the periplasm (Leptospirillum).

Other features distinctive to the Gram-positive switch occur at the bottom of the complex, which we propose is formed from both FliM_C/FliY_C heterodimers and FliY_C/FliY_C homodimers. Although we previously argued against FliM_C/FliN heterodimers (homologous to FliM_C/FliY_C heterodimers) in the switch complex of E. coli (43), a more recent structural study indicates that they are likely present in the switch complex of the very closely related Salmonella (41). This structural work together with recent mass spectrometric results (40) provides a basis for credible, structurally grounded hypotheses for the organization of the lower parts of the flagellar switch and injectisome. These models involve either heterodimers formed from a large component and a small component (e.g., FliM and FliN) or intramolecular pseudodimers formed from two FliN-like regions (in Spa33), together with homodimers of FliN or a FliN-sized protein produced from an alternative translational start site (40, 41). Blots of whole-cell proteins did not indicate significant levels of a short (FliN-like) form of FliY in B. subtilis cells, however, so we presently favor a model based on just full-length FliY for the B. subtilis switch. This has implications for the architecture of the lower part of the complex. In the models proposed for the Salmonella switch and Shigella injectisome (40, 41), the presence of numerous copies of the small unit (FliN in the flagellar switch or the C-terminal domain of Spa33 in the injectisome) allows for a continuous, coil-like organization that could not occur in the B. subtilis switch unless the numbers of full-length FliY and/or FliM subunits were doubled, with the accompanying formation of additional large rings. In the absence of ultrastructural data, we cannot rule this out, but we suggest that it is a less-likely alternative.

Though probably not forming a continuous coil-like structure like that in Salmonella, the dimers of C-terminal domains in the lower part of the B. subtilis switch complex would be in close proximity. Cross-linking experiments with Cys at surface locations in the C-terminal domains indicate that regions around residues 366 and 376 in FliY are fairly near the corresponding regions (residues 322 and 332) in FliM (Fig. 6). Significant cross-linking of some single-Cys mutants indicates that this C-terminal part of the protein is fairly dynamic, however, so the present results do not provide tight constraints on the relative positioning of the dimeric units. Regarding the function of C-terminal domains, the mutational results indicate that the domains are essential for flagellar assembly and the phenotypes of certain mutations in surface residues suggest that the region implicated in FliH binding in E. coli and Salmonella is likely to perform this function in B. subtilis as well (Fig. 9C).

The BACTH results and the effects of the FliM M133D mutation (Fig. 8), as well as a high degree of sequence conservation in the relevant residues, indicate that FliG_M interacts with FliM_M in B. subtilis similarly to their interaction in E. coli. BACTH experiments with all construct pairs suggest that FliG_M does not interact significantly with FliY_M. As noted above, the presence of some FliY_M domains in the FliM_M array would then alter the pattern of interaction with FliG. FliG subunits at these positions could rest on the occasional FliY_M domain just as they do on FliM_M, but without stabilizing interactions strong enough to register in the BACTH experiments. Alternatively, the FliG subunits at these positions might adopt an extended conformation that allows them to circumvent the FliY_M sites (illustrated in Fig. 10B). This model is similar to a recent proposal for FliG organization in the E. coli switch, where stretching of some FliG subunits was invoked for a different reason, namely to accommodate the mismatch in FliG versus FliM subunit numbers (15).

FIG 10.

(A) Model for subunit organization in the B. subtilis switch complex. FliM is magenta, FliY is cyan, and FliG is different shades of green (N-terminal domain darkest). In the upper panel, the view is from above (looking down through the membrane). The structure shown is based on rings with 34 subunits, but the actual number of subunits and the overall size of the complex are unknown. FliG_M and FliG_C domains have been removed from the lower left quadrant to show features below. A feature of note is the hypothesized presence of some FliY_M domains within the inner, predominantly FliM_M array. FliG domains are labeled in the plan view; FliM and FliY domains are labeled in the side view below. The N termini of FliM and FliY are shown as extending away from the proteins; these are believed to contain helices that bind to CheY-P, joined by flexible linkers to the middle domains. (B) An alternative arrangement of FliG that would not require FliG_C to bind to FliY_M even if some FliY_M domains are present in the otherwise FliM_M-based inner ring. In this alternative arrangement, FliG subunits at the FliY_M-containing positions are extended, possibly through unfolding of the FliG_M-FliG_C linking helix, to allow the C-terminal domain to bypass the FliY_M domain and rest on the FliM_M domain one position over. The arrangement is similar to one proposed in E. coli to accommodate FliG-FliM copy number mismatch (15).

The flagellate-but-immotile phenotypes of mutants with mutations in conserved charged residues of FliG_C (the only instances of a Mot⁻ phenotype observed in the present study) indicate that the charge-bearing ridge at the top of B. subtilis FliG_C functions specifically in rotation, as in E. coli. Given the close involvement of FliG_C in rotation, which apparently arises from direct interaction with the stator (29), movements in FliG_C are likely responsible for the reversal in torque that occurs on direction switching. In Salmonella, the helix joining the middle and C-terminal domains of FliG was found to be important for switching, with a 3-residue deletion near the beginning of the segment causing an especially strong CW bias (59). The corresponding deletion gave a nonflagellate phenotype in B. subtilis. We suggest that in B. subtilis, where the CW conformation of the switch is already the more stable resting state, shortening of the helix forces some structural element(s) beyond this stable CW arrangement into a state that does not allow the complex to assemble.

Given the evidently similar organizations of FliG in B. subtilis and E. coli (Fig. 7) (15) and the similar involvement of FliG_C charged residues in rotation, we propose that the CW and CCW switch states are similar in the two species. The opposite action of CheY-P in B. subtilis could then be accounted for by postulating that (i) the CW state is intrinsically more stable than the CCW state (which is the case) and (ii) the binding of CheY-P induces similar, except oppositely directed, movements in the two cases. A well-conserved segment near the N terminus of FliM is present in both species and is likely to function in the initial capture of CheY-P. In E. coli, CheY-P then interacts with a second site, on FliN, to trigger switching (45). The site(s) of action of CheY-P in B. subtilis has not been directly identified, but nuclear magnetic resonance (NMR) experiments in T. maritima (63), whose switch components resemble those of B. subtilis, point to a site in FliM_M rather than FliN. The opposite actions of CheY-P in the two species might then arise from its interactions with different regions of the switch.

A model for the B. subtilis switch complex developed from the present results is shown in Fig. 10. A subunit copy number of 34 has been assumed for the major components, matching the typical number in E. coli, but the actual numbers and the size of the complex are not known. Essential features are as already described: FliG is at the top (nearest the membrane) and is organized similarly to those in E. coli and Salmonella; FliM_M domains, possibly interspersed with some FliY_M, form an array at the middle level and an array of FliY_M domains forms a second major ring at a larger radius; and the bottom of the ring is formed from FliY_C/FliM_C heterodimers and FliY_C/FliY_C homodimers. An alternative “stretched” arrangement of a FliG subunit, enabling accommodation of the inner FliY_M domains in a different way, is shown in the Fig. 10B.

In conclusion, the present study provides a foundation for understanding protein organization in the flagellar switch complex of a Gram-positive bacterium. The components critical for rotation are organized similarly in B. subtilis and E. coli. We propose that the CW and CCW states of the switch are similar in Gram-negative and Gram-positive species and that the opposite effects of CheY-P are due to its interactions with different regions of the switch to induce oppositely directed conformational changes. Switching likely involves significant movements in both FliM and FliG in both Gram-negative and Gram-positive species. The exact nature of these movements remains to be determined.

MATERIALS AND METHODS

Media.

Liquid cultures of B. subtilis and E. coli strains were grown at 37°C in LB medium (10 g tryptone, 5 g NaCl, and 5 g yeast extract per liter). Plates contained LB and 1.5% Bacto agar. Antibiotics were used at the following concentrations: ampicillin, 100 μg/ml; kanamycin, 50 μg/ml; spectinomycin (Spec), 100 μg/ml; and erythromycin, 1 μg/ml. IPTG was used at 0.5 mM in both assays of motility and two-hybrid experiments. Soft-agar plates for motility assays contained LB and 0.27% Bacto agar. Plates were poured and left at room temperature overnight before use.

Strains.

Strains and plasmids are listed in Tables S1 and S2 in the supplemental material. To generate the ΔfliMY strain, 1-kbp regions upstream of fliM and downstream of fliY were amplified, combined using PCR overlap extension, and cloned into the KpnI and SalI sites of the integration plasmid pMiniMAD, which contains a temperature-sensitive origin of replication and an erythromycin resistance cassette, to make pEW241. The plasmid was transformed into E. coli strain TG1, which creates concatemers enabling more ready integration into the B. subtilis chromosome, and then purified and transformed into B. subtilis strain PY79 with erythromycin selection at the restrictive temperature (37°C) to promote single-crossover recombination. The plasmid was evicted by culturing in LB (3 ml) overnight at 22°C without selection. The culture was serially diluted and grown on LB plates at 37°C overnight without selection. Multiple colonies were patched onto LB plates with and without erythromycin to identify colonies that had lost the plasmid. Chromosomal DNA was isolated from candidate colonies and screened for loss of fliM and fliY by PCR of genomic DNA.

Assay of motility and flagellation.

Motility plates containing 0.5 mM IPTG were spotted with 3 μl of cells from cultures grown in LB overnight at 37°C. Plates were incubated at 37°C. The diameter of expanding colonies was measured at regular intervals, and plots of diameter versus time were fitted to a line to obtain rates. Reported rates are relative to those of wild-type controls measured in the same experiment. For assays of motility in liquid medium, overnight cultures were diluted 100-fold into LB with 0.5 mM IPTG and grown for 4 h at 37°C, to mid-log phase. Cells were scored visually under a microscope and compared to a similarly grown control strain containing the wild-type gene. For immotile mutants, the extent of flagellation was determined by microscopic examination of cells stained using a wet-mount procedure (55).

Cloning and mutagenesis of B. subtilis switch complex genes.

The B. subtilis fliG, fliM, and fliY genes were PCR amplified from genomic DNA of strain 3610 and cloned into the expression vector pKG116, a gift from J. S. Parkinson. The cloned genes and adjacent Shine-Dalgarno sequence were amplified from these constructs and cloned into the IPTG-inducible plasmid pDR111 for integration into the amyE locus of the B. subtilis chromosome. Site-directed mutations were made using the QuikChange procedure (Agilent) and confirmed by dideoxy sequencing (by University of Utah core facilities or by Genewiz, LLC).

BACTH experiments.

Studies of protein-protein interaction by the two-hybrid method (bacterial adenylate cyclase two-hybrid [BACTH] experiments) used the kit from Euromedex. BBL MacConkey agar was purchased from Becton, Dickinson and Company. MacConkey agar plates were spotted with 2 μl of overnight cultures and incubated for 16 h at 32°C. β-Galactosidase assays were adapted from those described by Zhang and Bremer (64). The optical density at 600 nm (OD₆₀₀) was measured to estimate the density of 5-fold dilutions of overnight cultures, and then 20 μl culture was mixed with 80 μl permeabilization solution (0.8 mg/ml hexadecyltrimethylammonium bromide, 0.4 mg/ml sodium deoxycholate, 100 mM Na₂HPO₄, 20 mM KCl, 2 mM MgSO₄, and 5.4 μl/ml β-mercaptoethanol) and incubated at 32°C for 30 min with shaking. Six hundred microliters of substrate solution (60 mM Na₂HPO₄, 40 mM NaH₂PO₄, 2.7 μl/ml β-mercaptoethanol, and 1 mg/ml o-nitrophenyl- β- d-galactopyranoside) was added, and the samples were incubated at 32°C. Upon development of moderately strong yellow color (or after 1 h for samples that displayed low activity), the reaction was terminated by addition of 700 μl of 1 M Na₂CO₃. Samples were centrifuged to remove cells, and then the OD₄₂₀ was measured. Activities are given in Miller units: 1,000 × OD₄₂₀/(culture OD₆₀₀ × volume used × reaction time in minutes).

Disulfide cross-linking.

Overnight cultures of cells in LB with the appropriate antibiotic were diluted 100-fold into LB with 1 mM IPTG and grown at 37°C for 4 h. Cells were pelleted and resuspended in 1× phosphate-buffered saline (PBS) (137 mM NaCl, 2.7 mM KCl, 4.3 mM Na₂HPO₄, 1.47 mM KH₂PO₄, pH 7.4), and Cu-phenanthroline cross-linking solution was added to final concentrations of 4 mM Cu²⁺ and 16 mM phenanthroline. Samples were gently mixed by rotation at ∼60 rpm for 5 min at room temperature, and then the reaction was stopped by addition of N-ethylmaleimide (NEM) and EDTA to final concentrations of 20 mM and 60 mM, respectively. Cells were pelleted (13,000 × g, 1 min), resuspended in 200 μl of lysis buffer (10 mM Tris-HCl [pH 8.0], 0.5 mM EDTA [pH 8.0], 1 mg/ml lysozyme), and incubated at 37°C for 30 min. Samples were diluted by mixing with 300 μl PBS, and 160 μl was taken, mixed with 40 μl 5× SDS nonreducing buffer, and boiled for 10 min. Samples were then sonicated (Branson model 450 or 250 instrument, 30 pulses, power 3, duty cycle 50%). Cross-linked products were resolved by SDS-PAGE and visualized by immunoblotting, using anti-FliM or anti-FliN antibodies at a dilution of 1:2,500 or 1:1,000, respectively.

BMH and BMOE cross-linking.

Cultures were treated as described above for oxidative cross-linking. After resuspension in PBS, 0.2 mM cross-linker (bismaleimido-hexane [BMH] or bismaleimido-ethane [BMOE]) dissolved in dimethyl sulfoxide (DMSO) was added, and the samples were incubated at room temperature for 1 h. The reaction was quenched by addition of β-mercaptoethanol (2.5 μl) for 5 min at room temperature. Cells were pelleted, suspended in lysis buffer for 30 min as described above, sonicated, pelleted, and mixed with 40 μl 5× SDS reducing sample buffer.

Cell fractionation for localization of FliY.

Cultures were grown overnight from single colonies at 37°C in LB-Spec medium and then diluted 100-fold in 50 ml LB-Spec and cultured for 4 h. Cells were collected by centrifugation (10,000 × g, 15 min, 4°C), resuspended in 10 ml PBS, and divided into 10 1-ml aliquots, half of which were cross-linked using Cu-phen and half of which were used as non-cross-linked controls. Cross-linking proceeded for 5 min at room temperature and was quenched by addition of 30 mM NEM and 100 mM EDTA. Cells were collected by centrifugation and frozen (−80°C) for later use. Cell pellets were thawed, combined in 7 ml chilled lysis buffer (50 mM Tris [pH 7.5], 150 mM NaCl, 5 mM MgCl₂, 10% [vol/vol] glycerol, 1 mM phenylmethylsulfonyl fluoride [PMSF], 50 μg/ml lysozyme), and incubated for 2 h at 4°C. Cells were lysed using three passes through a French pressure cell (125 MPa), with a 5-min cooling step between each pass. Cell debris was collected by centrifugation (12,000 × g, 15 min, 4°C), and then cytosolic and membrane fractions were separated by ultracentrifugation (SW50 rotor at 45,000 rpm [average relative centrifugal force {RCF_av} = 189,000 × g], 90 min, 4°C). Membranes were resuspended in 500 ml nonreducing dye. Supernatant protein was precipitated using trichloroacetic acid (10%), collected by centrifugation, washed with chilled acetone, and resuspended in nonreducing dye. Samples were resolved by 10% SDS-PAGE, and FliM and FliY were detected by immunoblotting using antibody against E. coli FliM.