Conformational Barrier of CheY3 and Inability of CheY4 to Bind FliM Control the Flagellar Motor Action in Vibrio cholerae

Vibrio cholerae contains multiple copies of chemotaxis response regulator (VcCheY1–VcCheY4) whose functions are elusive yet. Although previous studies suggested that only VcCheY3 directly switches the flagellar rotation, the involvement of VcCheY4 in chemotaxis could not be ruled out. None of these studies, however, focused on the structure, mechanism of activation or molecular basis of FliM binding of the VcCheYs. From the crystal structures of Ca2+ and Mg2+ bound VcCheY3 we proposed the presence of a conformational barrier composed of the hydrophobic packing of W61, M88 and V106 and a unique hydrogen bond between T90 and Q97 in VcCheY3. Lesser fluorescence quenching and higher Km value of VcCheY3, compared to its mutants VcCheY3-Q97A and VcCheY3-Q97A/E100A supported our proposition. Furthermore, aforesaid biochemical data, in conjunction with the structure of VcCheY3-Q97A, indicated that the coupling of T90 and Q97 restricts the movement of T90 toward the active site reducing the stabilization of the bound phosphate and effectively promoting autodephosphorylation of VcCheY3. The structure of BeF3 − activated VcCheY3 insisted us to argue that elevated temperature and/or adequacy of phosphate pool might break the barrier of the free-state VcCheY3 and the conformational changes, required for FliM binding, occur upon phosphorylation. Structure of VcCheY4 has been solved in the free and sulfated states. VcCheY4sulf, containing a bound sulfate at the active site, appears to be more compact and stable with a longer α4 helix, shorter β4α4 loop and hydrogen bond between T82 and the sulfate compared to VcCheY4free. While pull down assay of VcCheYs with VcFliMNM showed that only activated VcCheY3 can interact with VcFliMNM and VcCheY4 cannot, a knowledge based docking explained the molecular mechanism of the interactions between VcCheY3 and VcFliM and identified the limitations of VcCheY4 to interact with VcFliM even in its phosphorylated state.


Introduction
Vibrio cholerae, the highly motile gram-negative bacterial pathogen that causes cholera, uses chemotaxis and motility to travel to its preferred intestinal niche to colonize [1]. Extensive studies on chemotaxis of Escherichia coli or Salmonella typhimurium showed that the ligand induced conformational change in methyl accepting chemotaxis protein (MCP) is sensed by the CheA-CheW complex eventually resulting autophosphorylation of the kinase CheA. Autophosphorylated CheA then donates phosphate to the response regulator CheY. Phosphorylated CheY interacts with the flagellar motor protein FliM and influence the direction of flagellar rotation from counter clock wise (CCW) to clock wise (CW) [2,3]. CCW rotation results smooth swimming and CW rotation causes the cell to tumble [4]. Because of the presence of a single polar flagellum, V. cholerae does not tumble as such but reverses direction briefly, allowing the bacterium to randomly reorient itself and swim in a new direction.
The genomes of a large number of bacterial species, including Vibrio cholerae, Pseudomonas aeruginosa, Rhodobacter spaeroides, Myxococcus xanthus, Borrelia burgdorferi, and Yersinia pestis, encode for multiple paralogues of the various chemotaxis genes and chemotaxis in these bacteria is more complex [5,6]. A recent genomic and bioinformatic analysis of over 450 bacteria indicates that more than 50% of the chemotaxis gene homologs have more than one copy of chemotaxis genes [5] and these genes are involved not only in flagellum-mediated chemotaxis but also in type-4 pilus-based motility [7,8], polysaccharide biosynthesis associated with pilusbased gliding motility [9] and flagellar morphogenesis [10]. In many cases, however, genetic analysis has not been successful in deciphering the function of these chemotaxis gene homologs [5,11].
The genome sequence of V. cholerae has three sets of Che protein and 45 MCP-like proteins [12]. Each set of che genes forms clusters where che cluster I (located on chromosome I) contains cheY1, cheA1, cheY2, cheR1, cheB1 and the putative gene cheW; cluster II of chromosome I contains cheW1, cheB2, cheA2, cheZ and cheY3, while cluster III of chromosome II contains cheB3, cheD, cheR3, cheW2, cheW3, cheA3 and cheY4.
So far, the molecular characterization of all four CheYs of V. cholerae (namely, VcCheY1-VcCheY4) is restricted to a few in vivo studies where some of the chemotaxis related genes are found to be involved in the virulence of V. Cholerae [13][14][15] [3]. Attempts to identify the V. cholerae cheY responsible for the flagellar motion showed that a deletion of cheY3 impairs chemotaxis [1] while insertional disruption and duplication of the cheY4 gene result in decreased and increased motility respectively [13]. Swarming assay and assessment of the swimming behaviour indicated that only VcCheY3 directly switches flagellar rotation, although this study could not rule out the involvement of VcCheY4 in the motor action [14]. Later, Bandyopadhaya and Chaudhuri (2009) showed that inactivation of cheY3 or cheY4 generates a less motile and less adherent mutant [15]. Sequence analysis of VcCheYs indicate that only 17% residues are identical among them which comprise the residues involved in binding of the divalent metal ion and stabilization of the phosphorylated intermediate ( Figure 1a). This implies that the basic machinery for the phosphorylation is intact for all four VcCheYs. Available literature, however, suggest that deletion of the cheY1 and cheY2 genes does not cause any defect in chemotaxis [14] and motility or adherence remains unaffected for the insertional mutants of cheY1 or cheY2 [15]. All these observations point to the fact that VcCheY3 and VcCheY4 are the key response regulators to control chemotaxis in V. cholerae.
Structures of CheY from different bacterial sources suggest that although all of these response regulators possess an overall (b/a) 5 fold, small differences in the amino acid sequence or point mutations lead to the subtle conformational variations that make each of these proteins unique in terms of their function [16][17][18]. Also, T87I and T87I/Y106W mutants of EcCheY were found to be phosphorylatable although these mutants were unable to generate clockwise rotation of the flagella [19]. In addition, both of these mutants had ,5-fold lower autodephosphorylation rates and the mutants were completely resistant to CheZ activity, indicating that an isoleucyl side-chain at position 87 renders EcCheY unable to perform its chemotactic functions [20].
VcCheY3 bears only 37% sequence identity with that of VcCheY4 ( Figure 1a) and so far, nothing is known about the structure, mechanism of activation or molecular basis of FliM binding for these two key response regulators, implicated in chemotaxis and virulence of V. cholerae. Here we report, the structures of VcCheY3 in Ca 2+ and Mg 2+ bound states, BeF 3 2 activated VcCheY3 (VcCheY3-BeF 3 2 ) and of the mutant VcCheY3-Q97A. Our structural observations identified a unique conformational barrier in VcCheY3 that controls its phosphorylation event. Implication of this barrier is established by fluorescence spectroscopic study on VcCheY3 and its mutants VcCheY3-Q97A, VcCheY3-Q97A/E100A and VcCheY3-D60A, comparison of their K m values and pull down assay with VcFliM NM . We have also reported the structures of VcCheY4 in free and sulfate bound states here and comparison of these structures helped us to argue that VcCheY4 has a strong tendency to be phosphorylated and the phosphorylated state would be more stable compared to its free state. While our pull down assay showed that only activated VcCheY3 can interact with VcFliM NM and VcCheY4 cannot, structure based docking explained the molecular mechanism of the interactions between VcCheY3 and VcFliM and identified the structural limitations of VcCheY4 to interact with VcFliM even in its phosphorylated state.

Cloning, Overexpression and Purification
VcCheY3 and VcCheY4 were purified according to the previously described protocols [21,22]. Briefly, the genes encoding VcCheY3 and VcCheY4 were amplified from V. cholerae O395 genomic DNA and cloned into pET28a + vector. After transformation, cells were grown at 37uC until the optical density at 600 nm (OD 600 ) reached 0.4 to 0.6. Protein expression was induced by the addition of IPTG (isopropyl-D-thiogalactopyranoside) to a final concentration of 0.1 mM. The cells were harvested by centrifugation and the resuspended pellet was lysed by sonication in presence of PMSF. The cell lysate was then centrifuged (12000 g for 50 mins) at 4uC. The 66His tagged protein was isolated from the supernatant using Ni 2+ -NTA affinity chromatography (Qiagen) and were eluted with lysis buffer containing 150 mM imidazole. The eluted fractions were checked by 15% SDS-PAGE, pooled and dialyzed overnight against the thrombin clevage buffer (0.05 M Tris-HCl pH 8.0, 150 mM NaCl) and the 66His tag was cleaved with 1 U thrombin by overnight incubation at 4uC. The proteins were further purified by gel filtration chromatography using a Sephacryl S-100 (GE-Healthcare) column (7861.4 cm) pre-equilibrated with thrombin cleavage buffer containing 0.02% sodium azide at 4uC.
The gene encoding FliM NM (residue 1-250) was amplified from V. cholerae O395 genomic DNA and cloned into pET21b + vector with a C-terminal 66His-tag to get optimal expression level and solubility. The FliM NM protein was purified by growing cells in LB media to an optimal density 0.6-0.8 at 600 nm and induced with 1 mM IPTG. The cells were harvested after induction at 37uC for 3 h. Cell pellet was resuspended in lysis buffer containing 50 mM Tris-HCl pH 7.5, 250 mM NaCl, 1 mM PMSF and 10 mg lysozyme and lysed by sonication. After centrifugation (140006g, for 45 mins and at 4uC) FliM NM with C-terminal 66His-tag was isolated from the supernatant by using Ni 2+ -NTA agarose (Qiagen) and the protein was eluted with lysis buffer containing 200 mM Imidazole. After checking in 12% SDS-PAGE the eluted fractions were dialyzed against the lysis buffer.

Mutagenesis
VcCheY3-D60A, VcCheY3-Q97A and VcCheY3-Q97A/E100A were prepared by two-step PCR and verified by commercial sequencing. All the mutant proteins were purified using the same protocol described for the wild type protein.
FliM NM -CheY Interaction through Nickel Pull-down Assay 50 ml of Ni 2+ -NTA slurry (Qiagen) was washed three times with binding buffer containing 10 mM imidazole, 150 mM NaCl, 5 mM MgCl 2 , 0.15% Tween 20 and 50 mM Tris-Cl (pH 7.5) and the resin was then incubated with 0.1 ml purified FliM NM -His protein in a concentration of 0.2 mg/ml at 25uC for 20 mins with gentle shaking. The beads were then washed for three times with the binding buffer before adding VcCheY3, VcCheY3-Q97A, VcCheY3-Q97A/E100A, VcCheY3-D60A or VcCheY4. For activation, respective protein was pre-incubated for 20 mins with BeF 3 -(100 mM). The mixture was then added in the FliM NM -His bound Ni 2+ -NTA resin maintaining 1:1 molar ratio and incubated for another 10 mins at 25uC. The beads were washed three times with the buffer and then resuspended in 25 ml of 46SDS-PAGE gel loading buffer and were subjected to SDS-PAGE analysis and Coomassie blue staining.

Fluorescence Spectroscopy
Fluorescence measurement was carried out using a spectrofluorometer, Hitachi F-7000. Fluorescence was measured at an excitation wavelength of 295 nm and an emission wavelength of 340 nm with slit widths of 2.5 nm and 5 nm for excitation and emission, respectively. All reactions were carried out at 25uC. Equilibrium titrations of VcCheY3, VcCheY3-Q97A/E100A, VcCheY3-Q97A and VcCheY3-D60A were carried out with acetyl phosphate (acP) and beryllium fluoride (BeF 3 2 ). The reactions in presence of acP were performed in a buffer containing 20 mM sodium phosphate (pH 7.5), 50 mM NaCl, and 2 mM MgCl 2 and the same were 50 mM Tris-Cl (pH 7.5), 150 mM NaCl and 5 mM MgCl 2 for BeF 3 2 . For all proteins the final concentrations were 1 mM, BeF 3 2 concentrations varies from 0 to 400 mM and the concentrations varies from 0 to 6 mM for acP. The fluorescence values were corrected for dilution. Km was determined as described previously by Lukat et al (1992) [23]. Acetyl phosphate and BeF 3 2 concentrations were plotted versus (I o 2 I)/ (I 2 I inf ), where I o is initial fluorescence intensity, I is the intensity at the corresponding acetyl phosphate concentration, and I inf is the intensity at the saturating concentration. From the plot, the reciprocal of the slope of the line corresponds to the K m value. According to proposed reaction scheme [23,24], shown as follows, K m = Ks. k3/k2.

CheYzR*P / ?
Where Ks is the equilibrium dissociation constant between CheY and acetyl phosphate (the phosphor-donor, R,P) and k2 and k3 are the phosphorylation and dephosphorylation rate constants, respectively.
If it is assumed that the observed quenching is a direct effect of the reduced quantum yield of phospho-CheY relative to that of CheY, the steady-state fluorescence at a given concentration of phospho-donor may be related to the kinetic parameters of the reaction scheme (Eq.1), where (I o 2I)/(I2I inf ) = ([R,P]k2)/(k3K S ). All experimental data points were fitted by linear fit analysis using Microsoft EXCEL and Origin 8.

Crystallization and Data Collection
Crystallization data of VcCheY3 [21] and VcCheY4 [22] have been published earlier. Briefly, crystals of VcCheY3 that grew in low-salt condition using 5% (w/v) PEG 6000 in 0.1 M Tris-HCl pH 8.0 as precipitant, belong to space group R3 and diffracted to a resolution of 1.67 Å . Crystals of VcCheY3 were also grown in the presence of Mg 2+ in a similar condition which diffracted up to 2.2 Å . VcCheY4 crystals grew in AMS at two different pH conditions. In the high-pH condition, hexagonal-shaped crystals were obtained using 0.8 M ammonium sulfate, 0.1 M Bicine pH 9.0, 4% glycerol as precipitant. In the low-pH condition, cubeshaped crystals were obtained using 0.8 M ammonium sulfate, 0.1 M citrate, 4% glycerol as precipitant. The low-pH and high-pH condition crystals were diffracted upto 1.67 Å and 1.9 Å with the space group C2 and P3 2  For data collection, crystals were fished out from the crystallization drops using nylon loop, briefly soaked in cryoprotectant solution and flash-cooled in a stream of nitrogen (Oxford Cryosystems) at 100 K. The diffraction data sets were collected using a MAR Research image-plate detector of diameter 345 mm and Cu K a radiation generated by a Bruker-Nonius FR591 rotating-anode generator equipped with Osmic Max Flux confocal optics and operated at 50 kV and 70 mA. Data were processed and scaled using AUTOMAR (http://www.marresearch.com/ automar/run.html). Data-collection and processing statistics are given in Table 1.

Structure Determination and Refinement
The structures of wild type VcCheY3, VcCheY4, VcCheY3-Q97A and activated VcCheY3 (VcCheY3-BeF 3 2 ) were solved by molecular replacement using MOLREP of CCP4 suite [25]. Packing considerations indicated the presence of one molecule in the asymmetric unit for all the structures.
The wild type VcCheY3 structure in its Ca 2+ bound form was solved by using the coordinates of the Salmonella CheY (PDBID: 2 CHE) as template. The structure was refined by alternating cycles of model building and refinement using 'O' and CNS [26,27] to a final R cryst and R free values of 20.2% and 22.9% respectively. The poly-ala model of VcCheY3 was used as search model for VcCheY4 (low pH) and the refined structure of VcCheY4 (low pH) was used as search model to determine the structure high pH CheY4 (VcCheY4 free ). Low pH VcCheY4 (VcCheY4 sulf ) was refined to R cryst 21.8% and R free 24.6% and VcCheY4 free was refined to R cryst 22.5% and R free 26.0%. VcCheY3-BeF 3 2 structure was solved by using the coordinates of E. coli activated CheY i.e. EcCheY-BeF 3 2 (PDB code: 1F4V) as the search model. Strong electron density of beryllofluoride was found close to the active-site residue D60. The structure was refined upto R cryst of 23.1% and R free of 24.3% by several rounds of refinements and manual rebuilding by using the programs CNS [27] and COOT [28], respectively. The structure of VcCheY3-Q97A was solved using VcCheY3 as template and refined by the similar protocol to R cryst of 22.5% and R free of 25.2%. The structure of Mg 2+ bound VcCheY3 was also solved using VcCheY3 (Ca 2+ bound) as template after removing the coordinates of Ca 2+ and waters and refined by the similar protocol to R cryst of 20.0% and R free of 22.5%. Details of the refinement parameters for all the structures along with the geometric parameters determined by PROCHECK [29] are given in Table 2.

Calculation of Normalized B Factor
Since VcCheY4 free and VcCheY4 sulf crystals grew in different space groups and their diffraction resolutions are different, to compare their B factors we have plotted their normalized B-factor or B9-factor. Crystallographic B-factors of proteins determined even at high resolutions show large variations from one structure to another but the B-factors expressed in units of standard deviation about their mean value (normalized B-factor or B9factor) shows consistent behaviour [30][31][32]

Overall Structures of VcCheY3 and VcCheY4
As expected, both VcCheY3 and VcCheY4 possess (b/a) 5 fold (Figure 1 b, c) typical of the response regulators. Structure of VcCheY3 in free state superposes on S. typhimurium CheY (StCheY; PDB code: 2 CHE) with a root mean square deviation (rmsd) of 0.4 Å (for 108 Ca atoms) (Figure 1b). VcCheY4 was crystallized in two different states; one is in free state with no ligand attached (VcCheY4 free ) and another with a sulfate and a Ca 2+ ion bound at the active site (VcCheY4 sulf ). Interestingly, when VcCheY4 free is superposed on VcCheY4 sulf significant differences are observed at the active site, together with helix a4, b4a4 loop and b3a3 loop (Figure 1c). Since VcCheY4 free and VcCheY4 sulf were crystallized in different space groups, we have checked the probable influence of the crystal packing on the observed structural differences. Our packing analysis suggests that, in either case, these regions are rather loosely packed and their conformations are not influenced by crystal packing. VcCheY4, in either state, is significantly different from that of VcCheY3 ( Figure 1b) and superposition of VcCheY4 free and VcCheY4 sulf on VcCheY3 produces rmsd values of 1.4 Å and 1.2 Å respectively. VcCheY4, in either state, differs from VcCheY3 mainly in the a1, a5, a4 regions and in the b3a3 loop ( Figure 1b). It is to be noted that a1 and a5 were implicated previously in CheA and FliM binding respectively [33].
We have solved the structures of VcCheY3 in Ca 2+ and Mg 2+ bound states to the resolutions of 1.67 Å ( Figure S1a) and 2.2 Å (Figure S1b) respectively. The location of the Ca 2+ (or Mg 2+ ) ion in VcCheY3 is similar to that of Mg 2+ ion in StCheY. The Ca 2+ (or Mg 2+ ) of VcCheY3 is heptacoordinated where four coordinations occur with protein atoms and three with water molecules (Figure 2a). In contrast to that, the Mg 2+ of StCheY is hexacoordinated. Although D12 of StCheY is not coordinated to Mg 2+ , D15 of VcCheY3 that corresponds to D12 of StCheY, coordinates with the metal ion ( Figure 2a). Except this residue the disposition of the side chains of the other residues that coordinate with the metal ion are more or less similar in these structures (Figure 2a). The average coordination distance between Ca 2+ and the protein atoms is about 2.4 Å while this is of about 2.1 Å in case of Mg 2+ which is due to the size difference of the ions.

Identification of a Conformational Barrier Towards Activation of VcCheY3
In StCheY or EcCheY, upon phosphorylation at D57, a series of structural changes occur near the active site. T87 along with b4a4 loop moves toward the active site and stabilizes the bound phosphate through hydrogen bonding. Y106 of b5 executes an 'inward' movement (shown in line in Figure 2b) with minimal conformational adjustments of W58 and M85 and that inward movement of Y106 is essential for the binding of FliM at a4-b5-a5 face of CheY. K109 and the Mg 2+ contribute to stabilize phosphorylated D57 [34]. In the free state StCheY, W58 stays more on the surface (with x1 of 174u, x2 of -101u) and M85 side chain adopts such a x1 value (2155u) that together these residues leave a preformed cavity for the 'inward' positioning of Y106 upon activation ( Figure 2b).
D60 is the site of phosphorylation in VcCheY3 as it corresponds to D57 of StCheY (Figure 1a). Both in the Mg 2+ and Ca 2+ bound free state structures of VcCheY3, the side chain of W61 (that corresponds to W58 of StCheY) is observed in a conformation, substantially different from that of StCheY (Figure 2c, 2d). In the free state structure of VcCheY3, the side chain of W61 buries unusually deeply with a x1 of 2135u and x2 of 2133u (Figure 2c). Y109 stays in its 'out' position and the side chain of M88 (with x1 of 64u, x2 of 175u) stays between W61 and Y109, packing snugly with W61, Y109 and V106 through hydrophobic interactions (Figure 2c). This packing essentially fills up the pocket, required for the 'inward' positioning of Y109 upon activation (Figure 2c, 2d). Moreover, in this inactivated structure of VcCheY3, the crucial T90 of b4a4 loop (that corresponds to T87 of StCheY), which stabilizes the bound phosphate on D60 upon activation, is hydrogen bonded with Q97 (Figure 2c). To the best of our knowledge, this kind of interaction involving the Thr of b4a4 loop was not observed so far in any other response regulator. In VcCheY3, T90 and Q97 are oriented in such a fashion that together they form a capping on the aforesaid hydrophobic packing and at the same time block the 'out to in' trajectory of Y109 (Figure 2d). Additionally, the side chain carboxylate group of E100 (which is Ala in EcCheY or StCheY) forms a water mediated hydrogen bond with NE1 of W61 (Figure 2c; Figure S3a, S3b). Therefore, the hydrophobic packing of W61 with M88 and V106, together with the hydrogen bond between T90 and Q97 and the water mediated interaction between W61 and E100 seem to make a conformational barrier that may affect the process of activation in VcCheY3.

Comparison of Phosphorylation Events Through Fluorescence Spectroscopy
To investigate the contribution of the proposed 'conformational barrier' of VcCheY3 towards its activation, we prepared three mutants VcCheY3-Q97A, VcCheY3-Q97A/E100A and VcCheY3-D60A. Since W61 is within the Forster distance of D60, tryptophan quenching study was performed with VcCheY3 and its mutants to monitor the phosphorylation event using acetyl phosphate (acP) as substrate. Interestingly, VcCheY3 showed very low quenching (Figure 3a) indicating that phosphorylation at D60 does not induce any conformational change in W61 and W61 remains buried even after the treatment with acP. VcCheY3-Q97A and VcCheY3-Q97A/E100A, on the other hand, showed considerable quenching in the presence of acP (Figure 3a), suggesting that in the absence of the hydrogen bond between T90 and Q97 (and also in absence of E100), conformational alteration of W61 may take place more easily and it can move toward the surface of the molecule. As expected, quenching is almost negligible for the nonphosphorylatable analog VcCheY3-D60A (Figure 3a). Based on these experiments we have calculated the K m (K m = Ks. k3/k2) values where a higher K m value implies a decrease in the binding affinity between CheY and the phosphodonor (i.e. larger Ks), a slower rate of phosphorylation of bound CheY (i.e. smaller k2) or a faster rate of autodephosphorylation (i.e. larger k3) [35]. K m value, obtained by us, was the highest for VcCheY3 (6.460.45 mM) followed by VcCheY3-Q97A (2.360.4 mM) and VcCheY3-Q97A/ E100A (2.060.2 mM) (Figure 3b) which are in accordance with our structural observations.

Structure of VcCheY3-Q97A
To investigate whether the hydrogen bond between T90 and Q97 affects the hydrophobic packing of W61, M88 and V106, we have solved the structure of VcCheY3-Q97A. As expected, the overall structure of VcCheY3-Q97A is almost identical to that of VcCheY3 and the Mg 2+ ion bound at the active site occupies the equivalent position to that of Mg 2+ (or Ca 2+ ) of VcCheY3 (Figure 4a). Interestingly, even in the absence of the hydrogen bond between T90 and Q97, the conformation and packing of W61, M88 and V106 are found to be unaltered with respect to the wild type VcCheY3 (Figure 4b). However, the water mediated hydrogen bond between W61 and E100 is not seen in this mutant. E100 is slightly reoriented here and has moved toward the CD1 atom of the adjacent I69 (Figure 4b). These observations, coupled with the quenching results, point to the fact that although the hydrophobic packing of W61, M88 and V106 is independent of the hydrogen bond between T90 and Q97 in free state, in the Quenching data using acP (Figure 3a) clearly indicate that obtaining of stable VcCheY3-P for crystallographic study is not possible. Since BeF 3 2 readily forms persistent activated complexes with many response regulators, regardless of the half-lives of their phosphorylated states, this is regularly used to structurally mimic the phosphorylated state of the response regulators [36]. Fluorescence quenching experiment for VcCheY3 and its mutants, performed in the presence of BeF 3 2 , showed approximately 30 fold lowering of the K m values (219.060.6 mM, 110.062.1 mM, and 96.461.4 mM for VcCheY3, VcCheY3-Q97A and VcCheY3-Q97A/E100A respectively) compared to that of acP (Figure 3c). Thus, to visualize the structural changes in VcCheY3 upon phosphorylation, we have activated VcCheY3 using BeF 3 2 and solved the structure of VcCheY3-BeF 3 2 to 2.1 Å . The active site of VcCheY3-BeF 3 2 largely resembles to that of EcCheY-BeF 3 2 (PDB code: 1F4V) (Figure 4c). In VcCheY3-BeF 3 2 , BeF 3 2 is covalently linked with D60 and Mg 2+ is properly poised to interact with BeF 3 2 (Figure 4c, 4d). To stabilize the bound BeF 3 2 , the side chain of K112 reorients and T90 along with the b4a4 loop moves toward the active site with a conformational change, hallmark for the activation of this type of CheYs (Figure 4d). The hydrogen bond between T90 and Q97 is abolished and Q97 side chain moves away from T90 (Figure 4e). Breaking the hydrophobic packing with M88, the side chain of W61 moves toward the surface (with x1 of 2166u, x2 of 234u) acquiring a conformation similar to that observed in EcCheY-BeF 3 2 (Figure 4e). Under that situation, M88 occupies the space left by W61 and creates a pocket, sufficient to accommodate the 'in' position of Y109 which is essential for FliM binding (Figure 4e).

Free and Sulfated Structures of VcCheY4
Although the overall structures of VcCheY4 free and VcCheY4 sulf are similar, substantial conformational differences are observed between these two, especially around the active site, in helix a4 and b4a4 loop. A Ca 2+ ion is located at the active site of VcCheY4 sulf which coordinates with D9, D52 and main chain carbonyl oxygen of N54 with an average coordination distance of 2.4 Å (Figure 5a; Figure S2a). A tetrahedral positive electron density was observed in the active-site pocket of VcCheY4 sulf during refinement which was interpreted as a sulfate ion because VcCheY4 was crystallized using ammonium sulfate as precipitant ( Figure S2a). In contrast to that, neither a metal ion nor a sulfate ion was observed at the active site of VcCheY4 free although both of these components were added during crystallization (Figure 5a; Figure S2b). Absence of the divalent metal ion do not cause any change in the side chain conformation of D9 and D52 compared to VcCheY4 sulf , but the carbonyl oxygen of N54 points away from the metal binding side (Figure 5a). As a result, the b3a3 loop of VcCheY4 free takes a different conformation and moves about 3 Å away from the active site (Figure 5a). In VcCheY4 free , helix a4 is shorter and b4a4 loop is unusually longer compared to those of VcCheY4 sulf (Figure 1c). Electron density around the b4a4 loop of VcCheY4 free is shown in the Figure S2c. The plot of B9-values indicated that the crystallographic B-factor of the b4a4 loop is much lower in VcCheY4 sulf compared to that of VcCheY4 free (Figure 5b). In VcCheY4 sulf part of the b4a4 loop is stabilized and adopts a helical structure effectively extending the length of a4 ( Figure 5c) and overall, the VcCheY4 sulf structure seems to be more compact compared to VcCheY4 free .
The location of the sulfate ion at the active site of VcCheY4 sulf is somewhat similar to BeF 3 2 of VcCheY3-BeF 3 2 (Figure 5a). T82 and K104, which are well known to stabilize the phosphoryl group in the other reported CheY structures, stabilize the sulfate ion in VcCheY4 sulf through hydrogen bonding. A movement of about 2 Å towards the active site occurs for T82 along with the b4a4 loop ( Figure 5a). Interestingly, in VcCheY4 sulf , an additional hydrogen is generated between T82 and K89 (K89 corresponds to Q97 of VcCheY3) which might further contribute to the compactness of a4 in VcCheY4 sulf (Figure 5a, 5c).
The crucial residue at b5 that acquires 'in' position upon activation is a Trp (W101) in case of VcCheY4 and in both the structures of VcCheY4 the side chain of W101 acquired 'in' position. In fact, this is the first structure of a naturally occurring CheY where Trp at this crucial position is observed to spontaneously occupy 'in' position, even without activation. In this case, W101 fits in a hydrophobic pocket made of V53, F61 and M80 (Figure 5c) and apart from making a hydrogen bond with T82, the hydrophobic part of K89 packs with W101 further contributing to the stability of VcCheY4 sulf .

Molecular Mechanism of FliM Binding in V. Cholerae
To investigate the binding ability of VcCheY3 and VcCheY4 with VcFliM, we performed an in-vitro pull down assay. VcFliM NM (a construct having the N-terminal and the middle domain of VcFliM with a C-terminal 66His-tag) was immobilized on Ni-NTA resin, which was then incubated with VcCheY3, VcCheY3-Q97A, VcCheY3-Q97A/E100A, VcCheY3-D60A and VcCheY4, individually, in presence of Mg 2+ but with or without BeF 3 2 . Our results showed that while the activated VcCheY3, VcCheY3-Q97A and VcCheY3-Q97A/E100A can interact with VcFliM NM , VcCheY3-D60A and VcCheY4 do not show any significant interaction with VcFliM NM even in presence of BeF 3 2 and Mg 2+ (Figure 6a). VcCheY3-D60A was used as the negative control, and the experiment performed with BeF 3 2 and without VcFliM NM quantified the basal level of adherence of VcCheYs in Ni-NTA agarose during experiment.
To identify the structural features of VcCheY3 and VcCheY4, responsible for the difference in affinity towards VcFliM, it was necessary to critically analyse their FliM binding surface. To start with, we prepared a model of the N-terminal 16 peptide of VcFliM N by 3D-JIGSAW and VcFliM N , thus prepared, was docked at the FliM binding face of VcCheY3-BeF 3 2 and VcCheY4 sulf . The FliM N part of the coordinates of EcFliM N -EcCheY-BeF 3 2 complex structure (PDB code: 1F4V) were used as a template to prepare the model of VcFliM N and EcFliM N -EcCheY structure as a whole was used as template for the docking. The resulting models of VcFliM N -VcCheY3-BeF 3 2 or VcFliM N -VcCheY4 sulf were then analysed to identify the structural determinants responsible for the differential FliM binding of VcCheY3 and VcCheY4. VcFliM N is observed to fit properly at the a4-b5-a5 cleft of VcCheY3-BeF 3 2 with considerable number of hydrogen bonds and hydrophobic interactions (Figure 6b, 6c) which are comparable with those of EcFliM N -EcCheY-BeF 3 2 (Table 3). In contrast to that, the probable interactions of VcCheY4 sulf with VcFliM N are inadequate (Figure 6d, 6e; Table 3). The FliM binding face of VcCheY4 sulf is not compatible enough for VcFliM. In VcCheY4 sulf , the space between a4 and a5 is ,2 Å wider compared to that of VcCheY3-BeF 3 2 which might cause a loose fit of VcFliM N at a4-b5-a5 face of VcCheY4 sulf . Residues T2 and D3 of VcFliM N are found to interact with VcCheY3-BeF 3 2 , but no such interaction is possible with VcCheY4 sulf (Figure 6c, 6e). Furthermore, in VcCheY3-BeF 3 2 , K122 of a5 is poised to form a salt bridge with D12 of VcFliM N , corresponding residue of VcCheY4 sulf is T114 which is spatially away from D12 of VcFliM N and naturally no interaction is expected between this pair (Figure 6e). As a result, the overall interactions between VcFliM N and VcCheY4 sulf are reduced significantly (Figure 6e, Table 3) supporting the observation of the pull down assay (Figure 6a).

Discussion
Unlike E. coli two-component chemosensory pathway that relies on a single copy of response regulator CheY, V. cholerae possesses four CheY homologues. Occurrence of multiple CheYs is not unusual in bacteria as these are also found in R. sphaeroides and B. Burgdorferi [5]. Recent studies have demonstrated that multiple copies of CheY play specific roles in the chemotactic signal transduction mechanisms. As for example, among the three CheYs of B. burgdorferi only CheY3 directly regulates motor action while the other two cannot bind to the motor and act as signal terminating phosphate sink [37]. Similarly, in R. sphaeroides only CheY6 can change the direction of the flagellar motor, although the others bind FliM probably to regulate the level of the phosphodonor [38,39]. An intriguing question, therefore, arises about the role of multiple copies of CheY in V. Cholerae, especially of VcCheY3 and VcCheY4.
Together, phosphorylation at the active site Asp, hallmark movement of the Thr and the b4a4 loop toward the active site to stabilize the bound phosphate, 'in' positioning of the crucial hydrophobic residue of b5 and FliM binding at the a4-b5-a5 face to reverse the flagellar motion constitute the general mode of action of the chemotactic response regulators. In EcCheY or StCheY, a preformed pocket was seen to accommodate the 'in' position of the crucial b5 residue Y106 upon activation (Figure 2b). In contrast to that, in VcCheY3, this pocket is preoccupied by the hydrophobic packing of W61, M88 and V106 (Figure 2c). A unique hydrogen bond between T90 and Q97 additionally restricts the outward movement of W61, which is necessary to make a pocket for the 'in' positioning of Y109. This hydrogen bond also obstructs the movement of T90 toward the active site essentially hindering the stabilization of the phosphoryl group by T90. VcCheY3 shows minimum quenching in the presence of acP which further support the hindered movement of W61 upon phosphorylation at D60 (Figure 3a). VcCheY3-Q97A and VcCheY3-Q97A/E100A, on the other hand, show considerable quenching in the presence of acP indicating that in the absence of the hydrogen bond between T90 and Q97, W61 can easily be reoriented toward solvent and T90 can move toward the active site to stabilize the phosphoryl group.
Higher K m value of VcCheY3 compared to its mutants VcCheY3-Q97A and VcCheY3-Q97A/E100A further establishes the hindrance caused by the hydrogen bond between T90 and Q97 in stabilizing the acyl phosphate on D60. The lower K m values of VcCheY3-Q97A and VcCheY3-Q97A/E100A are due to the loss of the coupling between T90 and Q97 which facilitates the movement of T90 toward the active site and stabilize the acyl phosphate. A comparison of the K m value of VcCheY3 with the CheYs from Helicobacter pylori or E. coli shows that the K m of VcCheY3 is also higher than that of HpCheY1 (1.0760.31 mM) and EcCheY (3.260.4 mM). As mentioned by Lam et al. (2010), K m increases with the increase in the ionic strength of the buffer used in the experiment [24]. While 200 mM salt was used in the experiment of EcCheY, only 50 mM salt was used for HpCheY1and VcCheY3 (and its mutants). Since our experimental condition is same as that of HpCheY1, we can clearly infer that the K m value of VcCheY3 is about six fold higher than that of HpCheY1.  As mentioned earlier, a higher K m (K m = Ks. k3/k2) implies a decrease in the binding affinity between CheY and the phosphodonor (larger Ks), a slower rate of phosphorylation of CheY (smaller k2) or a faster rate of autodephosphorylation (larger k3) [35]. The high K m value of VcCheY3 implies that either its phosphorylation occurs slowly or it has a higher rate of autodephosphorylation. Based on the swarming assay and swimming behaviour Hyakutake et al, (2005) reported that only the VcCheY3 directly switches the flagellar rotation [14]. Our pull down assay shows that VcCheY3 and its mutants VcCheY3-Q97A and VcCheY3-Q97A/E100A bind VcFliM NM efficiently in the presence of BeF 3 2 and Mg 2+ . Docking results suggest that VcFliM N can fit properly at the a4-b5-a5 face of the activated VcCheY3 with significant number of hydrogen bonding and hydrophobic interactions (Figure 6a, 6b; Table 3). Moreover, sequence comparison of VcCheY3 with EcCheY or StCheY denotes that the crucial residues implicated in binding the kinase CheA are conserved in VcCheY3 (Figure 1a). These observations indicate that although VcCheY3 has all the requisites for the phosphorylation, stabilization of the acyl phosphate is hindered due to the obstructed movement of T90 towards the active site. Lesser stabilization of the bound phosphate might be implicated in enhanced autodephosphorylation (larger k3) for VcCheY3, effectively causing lower rate of activation which is reflected in its higher K m value. The conformational barrier of VcCheY3, therefore, acts as a molecular switch to control the level of VcCheY3-P. Elevated temperature and/or adequacy of phosphate pool might break the barrier of the free-state VcCheY3 and flip it to the phosphorylated state for FliM binding.
Two distinct conformations, differing at helix a4 and the crucial b4a4 loop, are observed for VcCheY4. Among these two structures, VcCheY4 sulf possesses a bound sulfate ion near the active site which occupies a position similar to the BeF 3 2 of StCheY-BeF 3 2 and VcCheY3-BeF 3 2 (Figure 5a). A bound sulfate ion was also observed in HpCheY1 structure (PDB code: 3GWG) where that sulfate ion caused conformational changes similar to the activated structure [24]. However, in HpCheY1, along with the conventional conformational changes, an unusual orientation of D53 was observed [24]. In VcCheY4 sulf , the sulfate ion did not alter the side chain conformation of catalytic D52 but stayed very close (,2.5 Å ) to it (Figure 5a). Since VcCheY4 sulf was crystallized at pH 4.0, at this pH D52 might be protonated allowing the sulfate ion to come to its close vicinity. In VcCheY4 sulf , the sulfate ion is properly coordinated with the Ca 2+ ion and is stabilized through the interactions with T82 and K104 (Figure 5a). Considering the compactness of the VcCheY4 sulf structure having a shorter b4a4 loop with low B-factors, long a4 helix, movement of T82 and b4a4 loop to stabilize the sulfate ion and additional hydrogen bond between T82 and K89, it can be said that VcCheY4 has a strong tendency to be phosphorylated in the presence of a divalent metal ion and the phosphorylated state is more stable compared to its free state.
Despite the fact that the crucial b5 residue W101 of VcCheY4 consistently acquires 'in' position, VcCheY4 fails to interact with VcFliM NM (Figure 6a). Through mutagenesis and structurefunction studies Matsumura and collaborators showed that substitution of Y106 of EcCheY with tryptophan (Y106W) produces a phosphorylation-dependent, hyperactive mutant that generates mainly clockwise rotational bias upon interacting with FliM [40]. In contrast to that, despite the consistent 'in' position of W101, VcCheY4 does not interact with VcFliM, as the N terminal part of VcFliM does not fit at the a4-b5-a5 face of VcCheY4 because of their spatial and electrostatic incompatibility (Table 3, Figure 6e). This apparent contradiction suggest that FliM binding by CheY is not just influenced by the 'in' positioning of the b5 hydrophoc residue but the spatial and electrostatic compatibility of the a4-b5-a5 face of CheY with the N-terminal part of FliM plays a vital role in this process. Since, CheZ and FliM share a common face of CheY for binding with similar mode of interactions [41], VcCheY4 is expected not to interact efficiently with CheZ as well. This observation corroborates with the fact that no cheZ is found in the cluster III where cheY4 is located. Since VcCheY4 can be phosphorylated but cannot bind FliM and probably not CheZ as well, VcCheY4 might act as phosphate sink or it might induce the expression of some other genes upon phosphorylation which can indirectly modulate flagellar action and/or virulence.
VcCheY4 was seen to slightly enhance the spreading of an E. coli cheZ mutant in semisolid agar and based on that Hyakutake et al proposed that VcCheY4 can affect chemotaxis by removing a phosphoryl group from VcCheY3 [14]. Our observations intend us to hypothesise that if a phosphate pull is shared by VcCheY3 and VcCheY4 then VcCheY4 can cause a phosphate depleted situation for VcCheY3, as phosphorylated state of VcCheY4 is more stable compared to its unphosphorylated state, which is other way round for VcCheY3. Alternatively, in a phosphate depleted situation, additional energy might help phosphorylated VcCheY4 to release the phosphoryl group through conformation dependent autodephosphorylation, as proposed by Pazy et al., 2009 [42] based on their observations of the mutant EcCheY.  Supporting Information S1.

Supporting Information
(PDF)