Structure of Flagellar Motor Proteins in Complex Allows for Insights into Motor Structure and Switching

Background: The flagellar motor generates bidirectional rotation using the proteins FliG, FliM, and CheY and membrane-bound stator complexes. Results: The structures of portions of FliG and FliM in complex allow for insights into switching and structure. Conclusion: FliGMC and FliMM in a 1:1 complex may form the C-ring of the motor. Significance: Understanding motor structure is key to understanding its mechanism. The flagellar motor is one type of propulsion device of motile bacteria. The cytoplasmic ring (C-ring) of the motor interacts with the stator to generate torque in clockwise and counterclockwise directions. The C-ring is composed of three proteins, FliM, FliN, and FliG. Together they form the “switch complex” and regulate switching and torque generation. Here we report the crystal structure of the middle domain of FliM in complex with the middle and C-terminal domains of FliG that shows the interaction surface and orientations of the proteins. In the complex, FliG assumes a compact conformation in which the middle and C-terminal domains (FliGMC) collapse and stack together similarly to the recently published structure of a mutant of FliGMC with a clockwise rotational bias. This intramolecular stacking of the domains is distinct from the intermolecular stacking seen in other structures of FliG. We fit the complex structure into the three-dimensional reconstructions of the motor and propose that the cytoplasmic ring is assembled from 34 FliG and FliM molecules in a 1:1 fashion.

The flagellar motor is a rotary nano-turbine that drives motile bacteria. The motor generates torque through the free energy of an electrochemical gradient of extracellular cations flowing through the stator in the membrane to the cytoplasmic rotor components (1,2). The stator proteins MotA and MotB form a membrane channel (MotAB) that generates torque at the interaction site between MotAB and the rotor protein FliG (3,4). The rotation direction results from modulation of the FliG/MotAB interaction that is regulated by the "switch complex" composed of FliM, FliG, and FliN. Multiple copies of each protein component form the cytoplasmic ring (C-ring) 3 of the motor (5)(6)(7)(8). The structure of the C-ring has been visualized by cryo-EM tomography, and models of how the motor proteins fit into the C-ring have been published by several laboratories (24). X-ray crystallography has supplied the structure of parts of every protein that comprise the C-ring, yet an absolute arrangement has not emerged. The structure of the middle domain of FliM (FliM M ) has been solved alone (Protein Data Bank (PDB) 2HP7) and in complex (PDB 3SOH) with the middle domain of FliG (FliG M ). In each structure, alone or in complex, FliM M has assumed identical conformations, suggesting that the structure of FliM in the C-ring has been observed in the crystal structures.
The three domains of FliG, N-terminal, middle, and C-terminal, have also been crystallized, but in each case, the domains assume different orientations with respect to one another. The structures show the domains of FliG in two conformations either held apart by extended linker helices as seen in the 1LKV and 3HJL structures or collapsed atop each other as seen in the 3AJC structure. An interesting observation is that in all these structures, the middle and C-terminal domains of FliG form a pseudo-armadillo repeat motif (ARM) stack. The apparent ARM stack can form both intermolecularly (PDB 1LKV and 3HJL) or intramolecularly (PDB 3AJC), but the orientation and interactions of the two domains in all these structures are similar.
Here we report the 1.9 Å resolution crystal structure of the middle domain of FliM (FliM M ) in complex with the middle (FliG M ) and C-terminal (FliG C ) domains of FliG (FliG MC ) from Thermotoga maritima (see Fig. 1A). FliG MC assumes a compact conformation similar to the recently published FliG MC (PDB 3AJC) structure from the Namba laboratory (9). Our structure also contains an apparent intramolecular ARM stack as well as the same interface between FliG M and FliM M as seen in the 3SOH structure. Our structure was able to be modeled into the three-dimensional cryo-tomography reconstructions of the motor provided by DeRosier laboratory, allowing for a proposal of the three-dimensional arrangement of the motor structure to be formed.

EXPERIMENTAL PROCEDURES
FliG (residues 115-327) and FliM (residues 46 -230) were cloned into the pJY5 vector. Proteins were expressed using BL-21 Rosetta cells. The cell pellets were suspended in 25 mM Tris, pH 7.5, 400 mM NaCl, 10 mM imidazole, 1 mM EDTA and lysed with a French pressure cell. The lysate was heat shocked at 80°C for 15 min followed by centrifugation at 13 rpm in a Beck-man JA-10 rotor. The clarified lysate was passed through a nickel affinity column and eluted with 200 mM imidazole. The eluant was cleaved with tobacco etch virus protease in 50 mM Tris, pH 8.0. Following cleavage, the proteins were passed for a second time through a nickel affinity column. The flowthrough was collected and exchanged into 10 mM Tris, pH 7.9, 10 mM NaCl. The proteins were mixed in a 1:1 ratio of 75 M each and crystallized in 50 mM calcium acetate, 0.1 M sodium cacodylate, pH 6.0, and 25% v/v 2-methyl-2,4-pentanediol at 20°C in a hanging drop plate. The crystals were frozen in liquid nitrogen after being dipped into cryo-buffer containing 0.1 M Tris, pH 8.8, 8% PEG 8000, 35% glycerol. The data sets were collected at beamline 5.0.2 at the Advanced Light Source in Berkeley, CA. Images were processed and scaled using HKL2000 (10). Because the data were highly anisotropic, ellipsoidal truncation and anisotropic scaling were performed using the UCLA Molecular Biology Institute diffraction anisotropy server (11). The structure was solved using T. maritima FliG MC (PDB 1LKV) and T. maritima FliM (PDB 2HP7) coordinates as a template for molecular replacement, refined using PHENIX (12). The structure was manually built in COOT (13).

RESULTS AND DISCUSSION
The FliG MC -FliM M complex (Fig. 1A) was generated by mixing truncated forms of FliG (residues 115-327) and FliM (residues 46 -230) in an equimolar ratio. Optimized crystals displayed diffraction with Bragg spacing of 1.9 Å. The structure was solved by molecular replacement using the previously solved structures of FliG (PDB 1LKV) and FliM (PDB 2HP7) as search models (7,14). The complex was refined to a resolution of 1.9 Å with an R free of 24.7% and an R work of 21.7%. Complete refinement statistics are shown in Table 1. The overall architecture of the complex is cylindrical with a vertical height of ϳ100 Å and a diameter of ϳ25 Å, with a contact surface area between FliM and FliG of 703 Å 2 (15,16). Four of the C-terminal residues of FliM and seven residues of a loop in FliG were not observed in the electron density and were presumed disordered. The unstructured region of FliG does not contain conserved residues, suggesting that this is a variable region among species. Furthermore, the missing residues precede the conserved Gly-Gly hinge located between the domains of FliG and may be unstructured due to the inherent flexibility in this region of the protein (17).
Upon binding to FliG MC , FliM M does not undergo any significant conformational changes, with an r.m.s deviation of 0.6 Å as compared with the FliM M 2.0 Å resolution structure (PDB 2HP7) or the 3.5 Å resolution structure of FliM M bound to FliG M both previously solved by the Blair and Crane laboratories (7,18). FliM M consists of three ␤-strands (␤1-␤3) and three ␣-helices (␣1-␣3) that form a pseudo-symmetric ␣/␤/␣ three-layered sandwich. The highly conserved GGXG sequence motif, located in the loop between helices ␣3-␣1Ј, is involved in FliG binding and confirmed by both the 3SOH and this structure (6,19). In the 2HP7 structure, the two glycine residues of the GGXG motif are resolved, but the following four residues (PGEN) are disordered (7). Upon binding to FliG M , the loop in FliM M surrounding the GGXG motif becomes ordered and stabilized and is observable in both the 3SOH and this structure. The FliG M /FliM M interaction surface is positioned directly above the secondary binding site of CheY located on FliM M that was identified by NMR experiments (20). When the response regulator CheY is phosphorylated (CheY-P), it is able to bind to FliM to signal the switch in rotation direction. The proximity of the CheY-P binding site to the FliG M /FliM M interface could explain how the switch signal is propagated from FliM through FliG toward the MotAB complex.
FliG MC in complex with FliM M is predominantly ␣-helical consisting of 14 helices arranged in an extended structure (Fig.  1A). A right-handed super-helix is generated from the stacking  of pseudo-armadillo repeats from helices 1-4 (ARM M ) and helices 6 -8 (ARM C ) (Fig. 1B), mimicking the stacking observed in eukaryotic ARM motifs (21,22). The ARM M /ARM C stacking results in the burial of 14 hydrophobic residues forming a contact area of 745 Å 2 (15,16). Helix MC lies below ARM M and interacts with helices 1-2 of FliG MC as well as parts of FliM M . The FliM M loop between helices ␣1Ј and ␣3 and residues within the helices themselves interact with ARM M and Helix MC of FliG, forming a network of bonding interactions. Fig. 1C shows the packing of hydrophobic residues and hydrogen bonding between the two proteins. The C-terminal domain (Helices C1-6 ) is located above the intramolecular ARM stack with the charged Helix C5 positioned on top of the domain, free to interact with MotAB (3). The C-terminal domain of FliG is principally connected to the ARM stack through the loop between helices 8 and C1, suggesting that changes in the position of FliG C relative to the ARM stack may be a possible component of motor switching.
In two of the four FliG structures (PDB 1LKV and 3HJL), the domains are held separate by extended linker helices (Fig. 2, C and D) (14,23). Brown et al. (14) suggested that Helix MC is stabilized through crystal contacts and postulated that this region would be flexible in vivo if the crystal packing represented nonphysiological interactions. Furthermore, Van Way et al. (17) used bioinformatic analysis in conjunction with mutational experiments to show that the conserved Gly-Gly pair, at the end of Helix MC , acts as a flexible hinge between the two domains of FliG MC . The 3AJC structure (Fig. 2B) shows Helix MC in a new position as compared with the 1LKV and 3HJL structures. The new position of Helix MC was likely a product of the crystal packing facilitated by domain mobility through the Gly-Gly pair (9). In the FliG MC -FliM M complex, Helix MC is packed against FliM M beneath ARM M as opposed to being pointed away from the ARM stack as in the 3AJC structure. The multiple positions of Helix MC observed in the structures suggest a degree of freedom that is likely limited when FliG M binds to FliM M , allowing Helix MC to pack against FliM M . In the 3AJC structure, the position of Helix MC contacts the neighboring molecule, and movement of Helix MC in response to CheY-P has been suggested to play a role in switching (9,24). Helix MC in our complex structure is not in contact with neighboring molecules as it is in the 1LKV and 3AJC structures. This suggests that it may play a role in switching but not through the communication of multiple FliG monomers.
The majority of the C-ring models argue that the ARM stacking observed in the FliG structures is a central structural feature  (14,23). This results in an ARM C -ARM Mϩ1 stack. In the FliG MC -FliM M complex, the ARM stacking is observed within a single FliG protomer (Fig. 3A). The intramolecular ARM stacking is also assumed in the 3AJC structure (Fig. 3B), although the residues connecting the domains are not observed (9). We propose that ARM stacking is driven by the requirement to bury the hydrophobic residues that comprise the ARM domains, and the intramolecular stack is driven by the binding to FliM as seen in the complex. Determining whether the ARM stacking is intra-or intermolecular in the C-ring requires further experimentation, but in the co-crystal, the presence of FliM apparently drove the ARM domains to stack within a single FliG MC monomer, suggesting that the in vivo interaction could be intramolecular.
The intramolecular ARM stacking we observed in the crystal may represent a preassembly conformation of FliG. In complex with FliM, FliG in the intramolecular stacking configuration may be a precursor building block of the C-ring, and as they are assembled into the motor, FliG may undergo a transformation to the intermolecular interaction. The in vivo stacking may be represented by either the stacking we observed in the co-complex structure or the intermolecular stacking observed in the other FliG structures. If the intermolecular stacking proposal were correct, the rearrangement from cytoplasmic intramolecular stacking to the motor assembled intermolecular stacking would require the 745 Å 2 of contact surface between the ARM domains to be broken and reassembled during the transition. This transition would have to overcome the energetics of exposing the 14 buried hydrophobic residues to reassemble the ARM stacks in the motor in the exact way they were previously, but now in the intermolecular stacked arrangement. Further experimentation on intact isolated motors will be necessary to unambiguously determine the arrangement of FliG inside the C-ring.
FliG and FliM have been extensively studied by mutational analysis and biophysical techniques to determine the interac-   (9). C, ARM stacking observed in the 1LKV crystal (14). D, ARM stacking observed in the 3HJL crystal (23). tion site or sites (17,19,20,25,26). FliG M contains a conserved EHPQ motif that when mutated disrupts the binding to FliM M . These same residues were shown to be in contact in the FliG M -FliM M co-crystal (PDB 3SOH) and in the complex presented here. The interaction surface between FliM M and FliG M has also been analyzed by NMR and chemical cross-linking. In NMR experiments, the resonances assigned to the EHP(Q/R) motif broaden when labeled FliG M is titrated with FliM M , suggesting that the surfaces seen in the crystal interact in solution as well (20). Cross-linking experiments were designed from the 3SOH structure and showed that FliG M is able to cross-link to FliM M in vivo, confirming that these faces are in contact in the assembled motor.
Blair and co-workers (7,18) have established that mutations in the conserved hydrophobic patch in ARM C also affect binding to FliM and have performed chemical cross-linking of intact cells to determine whether FliG C is in contact with FliM M in vivo (6,19). Their cross-linking results show that the hydrophobic patch of FliG C binds to the GGXG surface on FliM M as FliG M . In previous work (20), we used NMR to monitor the interactions of FliG C and FliM M . This showed chemical shifts in the resonances corresponding to the Helices C1-6 region of FliG C and not in the ARM C region, suggesting that the crosslinking results may not be present in solution. Composition gradient multiangle light scattering was also performed to measure the affinity of FliG C and FliM M , resulting in a K d of 580 M. (data not shown). This low affinity suggests that the interaction would probably not survive EM extraction. A model for the C-ring presented by Blair and co-workers (7,18) imagines that a portion of the FliG molecules in the C-ring interacts with two FliM M domains, one via the FliG M domain and the other via the FliG C domain. The remaining FliG molecules interact with FliM M via the FliG C domain only. DeRosier and co-workers (27) suggest that this unusual arrangement allows for matching the 26 -34-fold stoichiometry difference inferred from averaging of the EM reconstructions of the flagellar motor. The yields of the cross-links they observe between these domains are relatively low, suggesting that these interactions may reflect the behavior of a subpopulation of both FliM and FliG.
The FliG MC -FliM M complex was modeled into the high-resolution maps of the motor derived from three-dimensional electron micrograph reconstructions of a clockwise locked motor from Salmonella typhimurium (27,28). The C-ring from the three-dimensional reconstructions consists of two parts, an inner lobe with 26-fold symmetry and an outer lobe with 34-fold symmetry. The inner lobe of the C-ring disappears in three-dimensional reconstructions of a motor deleted for FliG N (29). This suggests that FliG N is only present in the inner lobe and that the outer lobe contains FliG MC , FliM M , and FliN that is mirrored by our construction of the C-ring.
The outer lobe is divided into three parts consisting of an upper, middle, and lower section. The upper section faces the membrane and is bridged to the middle section by a thin strip of density. Below the middle section is a ring that is believed to contain the FliN tetramer and the C terminus of FliM. The FliG MC -FliM M complex structure was docked into the density by the program VEDA with a correlation coefficient of 66.6 and an R-factor of 58.0. The alignment positioned the ridge of charged residues in FliG C toward the membrane in position to interact with the MotAB membrane channel (Fig. 4A). FliG M aligns nicely into density between the upper and middle sections, interacting with both FliM M and FliG. The orientation of FliM M in the middle lobe places the FliM M -FliM M subunit interfaces in close proximity, as identified previously by disulfide cross-linking experiments (7). Once we were satisfied with the position of our complex, the FliN tetramer was docked into the lower section of density. An expanded view of the docking is shown in Fig. 4C. The orientation of FliN to FliM was estimated from cross-linking experimental data and then refined by the VEDA fitting macro. Our x-ray structure and EM model fitting places the FliG C domain, FliM, and FliN subunits similarly to the model proposed by Lee et al. (15) without invoking the crystal packing ARM C -ARM Mϩ1 notion they proposed. Furthermore, we observed a partial melting of Helix MC , suggesting that Helix NM might be destabilized as well in the complex. This has major implications with regard to some of the rigid body movements proposed for these proteins during the flagellar motions (9, 18, 23). The symmetry of the C-ring observed in the three-dimensional reconstructions shows a shift from 34-fold to 26-fold when going from the outer to inner rings, respectively (Fig. 4B). The modeling of the C-ring presented here shows FliG and FliM in a 1:1 stoichiometry, which implies that the mismatch in the symmetry is managed in the connection from the outer to inner rings. We postulate that only 26 out of the 34 of the FliG N domains are able to bind to the 26 FliF copies located in the inner ring. The eight unbound FliG N domains would most likely be located unbound inside the inner space of the C-ring and would not participate in torque transfer from the outer to inner rings. This would allow for the motor to reconcile the symmetry mismatch in the simplest method possible. The symmetry mismatch may be a dynamic part of motor assembly. The three-dimensional reconstructions showed that the C-ring varies between 31-, 33-, and 34-fold symmetries, and the free FliG N domains may explain the ability of the C-ring to assume different copy numbers by displacing FliG-FliM complexes when necessary.