INTRODUCTION

Top Abstract Introduction Materials & Methods Results Discussion References

FliG, FliM, and FliN are proteins of the bacterial flagellum that have multiple functions (25, 34, 35; reviewed in references 14, 15, and 24). All three proteins are essential for flagellar assembly, and all are involved in controlling motor switching between clockwise (CW) and counterclockwise (CCW) rotation. FliG also functions directly in torque generation by the motor (7, 12, 13). Genetic suppression studies by Yamaguchi et al. (34, 35) first provided evidence that FliG, FliM, and FliN function together in a complex, which has been termed the "switch complex." All three proteins were subsequently localized to the flagellar basal body by immunoelectron microscopy (4, 5, 8, 9, 36-38). Binding of FliM to FliN, FliG, and other copies of FliM was detected in experiments using the yeast two-hybrid system (16, 17). These and additional binding interactions were also observed in coprecipitation experiments using glutathione S-transferase (GST) fusion proteins (29). A FliM-FliN fusion protein can support assembly and also some motor function, consistent with the proposal that FliM and FliN function within the same complex (10). Most recently, Toker and Macnab (31) used affinity blotting to demonstrate binding of FliM to FliG, FliN, and phospho-CheY and to examine the effects of deletions in FliM upon each of these interactions. FliM may also interact weakly with MotA, a stator component that functions in transmembrane proton conduction (29), and with FliF, the protein that forms the membrane-embedded MS ring of the flagellar basal body (19).

The complex containing FliG, FliM, and FliN is thought to reside on the rotating part (the rotor) of the flagellar motor (4, 5, 16, 17, 27-29, 36-38). Among the three proteins, FliM has an especially large role in controlling the direction of motor rotation. Many point mutations in FliM affect CW-CCW switching (25), and binding studies using purified proteins showed that FliM can bind to phospho-CheY (2, 3, 32), the chemotactic signaling molecule that triggers switching to the CW direction (20, 33). FliM appears to have little if any direct role in torque generation per se. Although certain mutations of FliM can give a nonmotile, flagellated phenotype (25, 30), some motility is restored when either the mutant FliM protein or one of the other switch complex proteins is overexpressed (12, 30).

Here, we report the effects of several deletions and linker insertion mutations in FliM on the functions of the protein in flagellar assembly, motility, and switching and on its binding to FliN, FliG, CheY, and phospho-CheY. The results provide insight into the domain organization of FliM and identify segments of the protein involved in interactions with some of its partners. A model is proposed for the function and arrangement of FliM domains in the flagellar motor.

	MATERIALS AND METHODS

Top Abstract Introduction Materials & Methods Results Discussion References

Strains, media, and plasmids. The strains and plasmids used are listed in Table 1. Transformations, plasmid isolation, and DNA manipulations used standard procedures (23). Most linker insertions were made in plasmid pHT32, whose parent is pTM30 (18), a high-copy-number plasmid that expresses cloned genes from the tac promoter. Plasmid pHT32 was made by inserting a BamHI fragment encoding fliM, obtained by PCR amplification of the cloned fliM gene (27), into the unique BamHI site in pTM30. It encodes a translational fusion with the residues MLNDPH fused to the N terminus of FliM and complements the fliM null strain DFB190 (27) to wild-type motility on swarm plates when induced with 60 µM isopropyl--D-thiogalactopyranoside (IPTG). Plasmids expressing GST fusions to FliN and FliG have been described elsewhere (29). A plasmid expressing a GST-CheY fusion (pHT121) was made by replacing a BamHI segment of the GST-FliM expression vector pHT86 (29) with a BamHI segment encoding cheY, which was obtained from plasmid pHT111, a pTM30 derivative that encodes cheY. Plasmid pHT121 complemented a smooth-swimming, nonchemotactic cheY deletion strain (RP5232) to give frequent tumbles and good chemotaxis. The GST-only plasmid used for negative controls (pHT100) has been described elsewhere (29).

FliM linker insertions. Four 12-residue oligonucleotides, which are wholly or partially self-complementary, were used to make a series of linker insertion mutations in fliM. They are as follows: L40, 5'-GCTCCCGGGAGC-3' (SmaI linker); L41, 5'-GCCCGGGCACGT-3' (AatII to SmaI adapter); L42, 5'-GTACCCCCGGGG-3' (BsiWI to SmaI adapter); and L43, 5'-CCCCTCGAGGGG-3' (XhoI linker). In some cases, single proline codons rather than 12-residue linkers were inserted into fliM, as detailed below. Plasmids expressing the linker insertion mutant FliM proteins were named pMn, with n specifying the fliM codon after which the linker was inserted. Fifteen linker insertions, which fell into three groups according to the method of construction, were made. The first group includes pM38, pM60, pM144, pM258, pM267, and pM282, which were made by inserting one of the linkers into existing restriction sites in the fliM gene, resulting in the introduction of the nonnative residues specified in Table 2. The second group includes pM81, pM132, pM163, and pM227, which were made by inserting a single proline codon at the sites indicated. This modification was made either by inserting the triplet CCG or CCC into C/GG or /GGG sequences in the native fliM sequence (slashes indicate sites of insertion) or, in the case of pM163, by replacing nucleotide A at position 489 with the nucleotides CCCG. These mutations were made by using the Altered Sites procedure (Promega) on the fliM gene cloned in plasmid pHT41. These mutations generated SmaI sites (CCCGGG), which were confirmed by restriction digests. The proline insertion mutations were then transferred into pHT32 by exchange of a restriction fragment bordered by a BsiWI site in the fliM coding region and a HindIII site in the downstream polylinker. Each of these single-proline insertions disrupted FliM function, as judged by swarming assays, and so no further insertions were made. The third group includes pM16, pM111, pM212, pM241, and pM310. At these positions, except for pM241, single proline insertions were first made and subcloned into pHT32, as described above. In making the proline insertion at codon 16, a silent mutation was introduced in codon 16 (AATAAC). In the case of pM241, a SmaI site was introduced by changing codons 239 through 241 from TCG CGT AAT to TCC CGG GAT, which resulted in the substitution Asn241Asp. Swarming was not significantly affected by these proline insertions or by the AsnAsp mutation at residue 241, and so a 12-bp oligonucleotide (L43) was then inserted into each site, by using the SmaI site generated in the first step.

fliM deletion constructs. fliM deletions were constructed in either pHT32 or one of the pHT32-derived pMn plasmids. Plasmid pHT17 (encoding FliM_1-60) was made by deleting a segment extending from the PstI site in the upstream linker to an Eco47III site at codon 60. The PstI site was blunted with mung bean nuclease before ligation. Plasmid pHT67 (FliM_60-144) was made by deleting a 252-bp segment extending from the Eco47III site at codon 60 to an NruI site at codon 144. Plasmid pHT134 (FliM_145-241) was made from the Asn241Asp mutant plasmid, which contains an introduced SmaI site at codon 240, by deleting a 287-bp segment extending from the NruI site at nucleotide 431 to the SmaI site at nucleotide 718 and inserting in its place an 8-bp SalI linker (GGTCGACC) to restore the reading frame. Plasmid pHT126 (FliM_241-334) was also made from the Asn241Asp mutant plasmid, by deleting the segment between the SmaI site and an EcoRI site in the downstream polylinker. The EcoRI end was blunted with mung bean nuclease before ligation. Plasmid pHT127 (FliM_1-241) was made from the Asn241Asp mutant plasmid by deleting a segment extending from BamHI in the upstream polylinker to the SmaI site at nucleotide 718. The BamHI end was made blunt before ligation, by using T4 DNA polymerase and deoxynucleoside triphosphates. Plasmid pDFB81 (FliM_1-38) was made by first changing the BsiWI site at codon 38 in fliM to a BamHI site and then moving codons 38 to 334 of fliM into pTM30, by using the introduced BamHI site and a SmaI site in a downstream polylinker. Plasmid pMAM4 (FliM_145-334) was made by deleting a segment of pHT32 between the NruI site at nucleotide 431 and an EcoRV site in the downstream polylinker. Plasmid pPB3 (FliM_1-247) was constructed by inserting a segment of fliM encoding residues 248 through 334, obtained by PCR amplification of the cloned fliM gene (27) with primers that introduced an NdeI site at the 5' end and a BamHI site at the 3' end, into the T7 expression vector pAED4 (27).

Flagellation, motility, and swarming. To determine the phenotypes of the fliM linker insertion mutants, the pMn plasmids were transformed into the fliM null strain DFB190. Staining and counting of flagella were carried out as described elsewhere (27). For swimming-speed measurements, overnight cultures were grown in TB and the appropriate antibiotic, diluted 100-fold into fresh TB containing various concentrations of IPTG, and then cultured for 4 h at 32°C. Cells swimming close to the coverslip were observed though a phase-contrast microscope and recorded on videotape. The paths of individual cells were measured by marking their positions on transparencies at intervals of 1/10 s (three video frames), by using a manual frame-advance feature of the recorder. Each cell was measured for about 1/2 s. Reported swimming speeds are averages for 48 cells.

GST fusion coprecipitation procedure. Coprecipitation experiments were carried out essentially as described elsewhere (29). These experiments employed an flhDC strain (RP3098) that expresses no flagellar proteins except those encoded on plasmids. In experiments to probe interactions of FliM with FliN or FliG, this strain was transformed with two plasmids, one that expresses a GST fusion protein and another that expresses FliM or its mutant variants. The transformants were cultured overnight in TB containing the appropriate antibiotics and 100 µM IPTG. Cells were harvested and lysed by sonication as described elsewhere (29). The following modifications were made to the earlier procedure in order to minimize proteolytic degradation of mutant FliM proteins. First, the 30-min incubation on ice prior to sonication was omitted, and the cells were instead lysed immediately after being resuspended. Second, the incubation of cell lysates with glutathione-Sepharose 4B beads was shortened from 30 to 15 min, and the elution step was shortened from 10 to 1 min. Finally, all steps were done either on ice or in a cold room.

Secondary structure prediction. Secondary-structure prediction of FliM used the neural network algorithm of Rost and Sander (21, 22), accessed via electronic mail to the web site maintained by the European Molecular Biology Laboratory in Heidelberg, Germany (http://www.embl-heidelberg.de/predictprotein/). Structure prediction was carried out by using all of the sequence data in an alignment of FliM proteins from Escherichia coli, Bacillus subtilis, Borrelia burgdorferi, and Caulobacter crescentus and also by using just the sequence from E. coli. Figure 5 displays the elements of secondary structure clearly predicted in both cases.

	RESULTS

Top Abstract Introduction Materials & Methods Results Discussion References

FliM linker-insertion mutants. The fliM gene was mutagenized by inserting short oligonucleotide linkers in 15 approximately regularly spaced locations. Each linker encoded from one to five nonnative residues and included at least one proline. The positions of the insertions and the residues introduced by the linkers are listed in Table 2. The linkers will presumably affect function significantly when they are inserted into interior segments of the protein that are important for folding or in exposed segments that contact other proteins. In segments that are not important for folding or function, insertions should have relatively minor effects. The effects of the linker insertions on FliM function were determined by expressing the mutant proteins from plasmids in the fliM null strain DFB190 and measuring numbers of flagella, rates of swarming in soft agar, and swimming speeds in liquid culture. The results are summarized in Table 2.

FliM deletion mutants. To obtain additional insight into the domain organization of FliM, we also constructed several deletion mutants. The choice of segments for deletion was based on the phenotypes of linker insertion mutants, available restriction sites, and sequence comparisons and predictions of secondary structure (see Fig. 5). Effects of the FliM deletions on function were assessed by measuring numbers of flagella, rates of swarming in soft agar, and motility of cells in liquid culture. The results are summarized in Table 3.All of the deletions tested gave a nonflagellate phenotype, with the exception of a 38-residue N-terminal deletion that gave a motile but nonchemotactic phenotype. Cells of the FliM_1-38 mutant swam smoothly, with few or no tumbles, indicating that their motors rotated with a strong CCW bias. The nonflagellate phenotype of most deletions was not caused by destabilization of the FliM protein, since most of the FliM fragments were stable enough to accumulate in cells. Stable FliM variants included a large C-terminal deletion (145-334) and a medium-sized N-terminal deletion (1-60). Some larger N-terminal deletions, which are not included in Table 3 but are described below, destabilized the protein so that it did not accumulate to a detectable level. A large N-terminal deletion of 241 residues left a fairly small (93-residue) C-terminal fragment that was stable under certain circumstances, as described below. Because this C-terminal fragment was marginally stable and its binding to other flagellar proteins proved especially interesting, we also constructed a plasmid that directs higher-level expression of a slightly smaller C-terminal fragment (residues 248 to 334). When overexpressed, this 87-residue C-terminal fragment accumulated in cells and was readily detectable on immunoblots. This fragment was used in some of the binding experiments described below.

Interactions with FliN or FliG. Insertions in FliM that disrupt function might disrupt interactions with other proteins. To test this possibility, we examined the binding of the mutant FliM proteins to other proteins with which FliM is known to interact. These experiments used GST fusions and coisolation assays developed previously to study binding interactions among the switch complex proteins (29). Because high-level overexpression of FliM might lead to nonspecific binding, we first used Coomassie blue-stained gels to determine whether FliM was highly overexpressed under the conditions used for binding experiments. Under the conditions typically used in binding experiments (induction of pHT32 or its variants with 100 µM IPTG), no band assignable to FliM was observed on gels of whole-cell proteins, indicating that FliM was not among the more abundant proteins present. FliM was readily detected on immunoblots of proteins from the same cultures (data not shown).

View larger version (19K): [in this window] [in a new window]   FIG. 1.   Coprecipitation of linker insertion mutant FliM proteins with GST-FliN (lanes labeled N) or GST-FliG (lanes labeled G). Positions of the linker insertions in FliM are indicated at the top. The experiment used the one-cell protocol (Materials and Methods). Coprecipitated material was analyzed on immunoblots probed with anti-FliM antiserum. Immunoblots of samples not exposed to the glutathione beads showed that all of the insertion mutant FliM proteins were present in cell lysates at levels comparable to that of the wild-type (w.t.) protein (data not shown).

View larger version (20K): [in this window] [in a new window]   FIG. 2.   Coprecipitation of FliM fragments with GST-FliN (lanes labeled N) and GST-FliG (lanes labeled G). The parts of FliM deleted are indicated at the top of each lane. (A) An initial set of FliM deletions, which together span the protein. The C-terminal fragment FliM242-334 (labeled 1-241 over the two rightmost lanes) was not stable except in the presence of FliN or GST-FliN, and so its failure to coprecipitate with GST-FliG is inconclusive. (B) Coprecipitation of an 87-residue C-terminal fragment of FliM with both GST-FliN and GST-FliG. This fragment accumulated in cells to detectable levels even in the absence of FliN (see text). (C) Coprecipitation of a 144-residue N-terminal fragment of FliM with GST-FliG. In negative-control experiments, neither FliM nor any of the FliM fragments was coprecipitated with GST alone (example control for a FliM fragment is shown here, and that for full-length FliM is shown in Fig. 4). w.t., wild type. Numbers to the left of each panel show molecular mass in kilodaltons.

Binding of FliM to CheY and phospho-CheY. Phospho-CheY is the signaling molecule that triggers switching of the motor to CW rotation. Binding of purified FliM to phospho-CheY has been demonstrated previously in coisolation assays in which CheY was covalently linked to Sepharose beads (32) and by chemical cross-linking (2, 3). These studies showed that binding was significantly stronger in the presence of agents that phosphorylate CheY (acetyl phosphate and Mg²⁺) than in the absence of these agents.

View larger version (32K): [in this window] [in a new window]   FIG. 3.   Coprecipitation of FliM with GST-CheY in the presence or absence of agents that phosphorylate CheY, with FliM from two sources. (Left lanes) Experiment using FliM that was purified as described elsewhere (19, 32). Purified FliM was added to a suspension of RP3098 cells (which express no flagellar proteins) to give a final FliM level similar to that in other binding experiments. These FliM-supplemented cells were mixed with cells expressing GST-CheY, the cells were lysed, and a binding experiment was performed by the two-cell protocol. (Right lanes) A binding experiment done in the same way, except that the FliM was expressed within the RP3098 cells, and the samples were not supplemented with purified FliM.

View larger version (38K): [in this window] [in a new window]   FIG. 4.   Effects of deletions or linker insertions in FliM on binding to CheY in the presence or absence of the phosphorylating agent acetyl phosphate (all experiments contained Mg2+). (A) Coprecipitation of FliM deletion-mutant proteins with GST-CheY. The FliM deletions are indicated at the top. For the experiment using wild-type FliM protein, the GST-only negative control is also shown (first lane in panel A); all other lanes used the GST-CheY fusion protein. Negative controls for each of the FliM fragments showed that none was coprecipitated with GST alone (data not shown). Blots were typically exposed to film for 5 min after addition of the chemiluminescence reagents, but with the 1-38 and 1-60 mutants, the binding was somewhat weaker in the lanes without acetyl phosphate and significantly weaker in the lanes with acetyl phosphate, so overnight exposure was used. (The densitometry results in Table 3 used uniform exposures and a range of protein concentrations.) (B) Coprecipitation of FliM linker insertion proteins with GST-CheY. Positions of the linker insertions are indicated at the top. w.t., wild type. Numbers to the left of each panel show molecular mass (m.w.) in kilodaltons.

Sequence conservation and predicted secondary structure of FliM. The fliM genes from several species have been cloned and sequenced. Figure 5 displays patterns of sequence conservation as determined by an alignment of FliM sequences from four species and shows the principal elements of secondary structure predicted by a neural net algorithm (21, 22). These analyses give clues to the overall organization of the protein and will provide a useful framework for discussing the results presented here and in other studies of sequence-function relationships in FliM (2, 25, 30, 31). Features of interest are as follows. A short segment near the N terminus (residues 8 to 15) is well conserved and is predicted to be mainly -helical. The bulk of the protein is predicted to be organized into domains with mixed  and  secondary structure, joined by sizable segments that are poorly conserved and predicted to have nonregular secondary structure and which might be interdomain linkers (residues 17 to 34, 136 to 145, and 237 to 247). Adjacent to one of these putative linkers is a short segment (residues 132 to 135) that contains two invariant Gly residues and a third Gly that is present in all species but at slightly different positions. The most C-terminal part of the protein, from about residue 287 to the end, is relatively poorly conserved.

View larger version (27K): [in this window] [in a new window]   FIG. 5.   Patterns of sequence conservation, predicted secondary structures, and effects of linker insertions in FliM. The alignment used FliM sequences from E. coli, B. subtilis, B. burdorferi, and C. crescentus. The sequence from Agrobacterium tumefaciens is also known, but it is quite different from all of the others and was not included. Outlined and shaded boxes indicate residues that are identical in the four sequences, and nonoutlined, lightly shaded bars indicate positions where residues with hydrophobic character are found in all four sequences. Secondary structures were predicted by the neural net algorithm of Rost and Sander (21, 22), by using the information from all of the sequences. The -helices are represented by coiled lines, the -strands are represented by zigzag lines, and segments of nonregular secondary structure are represented by straight lines. Segments where predictions were ambiguous are left blank. Inverted triangles indicate positions and phenotypes of linker insertion mutations: solid, nonflagellate; stippled, flagellate but most cells nonmotile; striped, motile but nonswarming because of a strong CCW bias; and open, close to wild-type swarming. (See Table 2 for exact phenotypes and sequences of the insertions.)

	DISCUSSION

Top Abstract Introduction Materials & Methods Results Discussion References

The phenotypes of fliM linker insertion mutations. Four linker insertions in FliM, at residues 16, 111, 240, and 310, allowed nearly normal function. This result can be rationalized in terms of the predicted secondary structures and patterns of sequence conservation (open inverted triangles in Fig. 5). Residue 16 is at the N-terminal boundary of a segment that is poorly conserved and predicted to have nonregular secondary structure. This region might be a linker between the well-conserved N-terminal segment (residues 8 to 15) and the rest of the protein. Residue 111 is in a segment predicted to be a loop between two -strands. This loop is evidently not on a functionally important surface of the protein. Residue 240 is near the middle of a segment (residues 237 to 247) with several properties suggestive of an interdomain linkerpoor conservation, polar character, and nonregular secondary structure. Residue 310 is in a segment that is relatively poorly conserved and is predicted to have nonregular secondary structure.

A C-terminal FliN-binding domain. A number of observations indicate that a ca. 90-residue C-terminal domain of FliM is necessary and sufficient for binding FliN. C-terminal FliM fragments accumulated in cells when they were coexpressed with FliN (or GST-FliN) or when they were highly overexpressed, but not otherwise. These FliM fragments bound FliN in coprecipitation experiments, whereas a large N-terminal fragment (residues 1 to 240) did not. Binding to FliN was abolished by two adjacent linker insertions in the C-terminal domain of FliM, at residues 267 and 282. As noted, the C-terminal domain of FliM is joined to the rest of the protein by a segment (residues 237 to 247) that might function as an interdomain linker.

View larger version (63K): [in this window] [in a new window]   FIG. 6.   Sequence alignments of segments of FliN with segments in the C-terminal domain of FliM, for species where both sequences are known (B. subtilis, Treponema pallidum, B. burgdorferi, C. crescentus, and E. coli). (The sequences are also known for Salmonella but are not significantly different from those of E. coli.) In B. subtilis and T. pallidum, the FliN homolog is called FliY and is a much larger protein that shows close homology to FliM in a short segment near the N terminus (1) (see the text). Residue numbers are not given for FliM from T. pallidum because the entire sequence is not known. Darkly shaded boxes indicate residues identical in FliM and FliN (or its homolog FliY) from the same species. Lightly shaded bars indicate positions where a sizable hydrophobic residue is found in both proteins from all species. Arrows indicate positions of the linker insertions in FliM that disrupted its binding to FliN.

Site(s) of FliG binding. Both deletions and insertions in FliM had surprisingly small effects on its binding to FliG (Tables 2 and 3), suggesting that this interaction might involve multiple, noncontiguous segments of the protein. This prediction was confirmed in experiments with the N-terminal fragment FliM_1-144 and the C-terminal fragment FliM_248-334, both of which were coprecipitated with GST-FliG (Fig. 2). Our conclusions concerning the FliG-binding site thus differ from those of Toker and Macnab (31), whose affinity blot study suggested that determinants for FliG binding are located in the middle of the protein, in the segment between residues 140 and 220. We cannot rule out binding of the middle part of FliM to FliG, because fragments lacking large segments from the N terminus were unstable (data not shown), but our results suggest that the middle of FliM is not the only part involved in binding FliG.

The binding site for CheY. Small N-terminal deletions in FliM weakened its binding to CheY and reversed the effect of CheY phosphorylation so that binding was weaker in the presence of phosphorylating agents. FliM with a 38-residue N-terminal deletion conferred a smooth-swimming phenotype, indicating that motor switching is impaired. These results suggest that residues near the N terminus of FliM are important for binding to CheY, and particularly for binding to phospho-CheY. This conclusion agrees with the recent study of Bren and Eisenbach (2), who used peptides in competition experiments to show that the first 16 residues of FliM contain determinants for binding phospho-CheY. Evidence of CheY binding to the N-terminal part of FliM was also obtained in the deletion study of Toker and Macnab (31). A linker inserted after residue 16 did not interfere with flagellar assembly, motor rotation, chemotaxis, or CheY binding. This finding implies that residues immediately C-terminal to residue 16 are not important for binding to CheY or for any conformational changes that accompany motor switching.

Assembly of FliM into the flagellum. A model for the arrangement of FliM and the other switch complex proteins in the flagellum is presented in Fig. 7. Key features of the hypothesis are as follows. FliN and the C-terminal domain of FliM bind to each other and together form the bulk of the C ring (5, 38). FliN and FliM alternate in a regular pattern, which is suggested to be three FliN molecules per FliM molecule on the basis of present estimates of subunit stoichiometry (ca. 110 FliN and 35 FliM molecules per flagellum [38]). Ratios of 4:1 or 2:1 are also possible and would not alter the essentials of the model. The C ring contacts another ring formed from FliG. Because FliG is present in 25 to 50 copies per flagellum (37), whereas FliM and FliN together total more than 100 copies, the FliG subunits most likely contact only some of the subunits in the C ring. Both N-terminal and C-terminal domains of FliM are pictured binding to FliG. The C ring does not contribute directly to the site of torque generation, but it is important for positioning the C-terminal domain of FliG there and for ensuring that directional switching occurs synchronously in all parts of the rotor (Fig. 7C to E). Binding to phospho-CheY is also suggested to involve multiple domains of FliM. This binding might induce a relative movement of FliM domains, and of the FliG domain(s) to which they are attached, causing changes at the rotor-stator interface that lead to CW rotation (Fig. 7C to E).

View larger version (65K): [in this window] [in a new window]   FIG. 7.   Model of the arrangement of the FliG, FliM, and FliN proteins in the flagellar motor. (A) The FliG-FliM-FliN assembly mounted on the MS ring, as viewed from the cytoplasm. Subunit stoichiometries are approximate and are based on immunoblots of proteins in isolated flagellar structures (37, 38); FliN-FliM stoichiometries of 2:1 or 4:1 are also possible. The location of FliN in the C ring is based on immunoelectron microscopy (5, 38). Results of the present study suggest that FliN and the C-terminal domain of FliM occupy similar positions in the structure. The C-terminal domain of FliG is placed at the rotor-stator interface on the basis of mutational studies of FliG and of the stator proteins MotA and MotB (6, 7, 12, 13, 39, 40). (B) Side view of the FliG-FliM-FliN assembly. For clarity, only a subset of the proteins is shown. The cytoplasm is toward the top, and the periplasm is toward the bottom. The MS ring and MotA-MotB complexes are located in the cytoplasmic membrane, which is not pictured. (C to E) Hypotheses for the movements that might be triggered by binding of phospho-CheY to FliM, to cause switching to the CW direction of motor rotation. In each case, switching is suggested to involve a change in the position or orientation of the FliG C-terminal domain, relative to the stator and/or other parts of the rotor. (C) Binding of phospho-CheY might cause a domain of FliM, and the attached domain of FliG, to move up or down in a direction parallel to the rotation axis of the motor. (D) Phospho-CheY might induce subunits of the C ring to tilt, causing attached domains of FliG to move tangentially relative to other components of the rotor. Here and in panel E, the view is rotated 90° relative to that in panel C. (E) Phospho-CheY might induce tilting of both the C-ring subunits and the attached FliG domains, changing the angular orientation of the FliG domains.