The CheYs of Rhodobacter sphaeroides*

The Escherichia coli two-component chemosensory pathway has been extensively studied, and its response regulator, CheY, has become a paradigm for response regulators. However, unlike E. coli, most chemotactic nonenteric bacteria have multiple CheY homologues. The roles and cellular localization of the CheYs in Rhodobacter sphaeroides were determined. Only two CheYs were required for chemotaxis, CheY6 and either CheY3 or CheY4. These CheYs were partially localized to either of the two chemotaxis signaling clusters, with the remaining protein delocalized. Interestingly, mutation of the CheY6 phosphorylatable aspartate to asparagine produced a stopped motor, caused by phosphorylation on alternative site Ser-83 by CheA. Extensive mutagenesis of E. coli CheY has identified a number of activating mutations, which have been extrapolated to other response regulators (D13K, Y106W, and I95V). Analogous mutations in R. sphaeroides CheYs did not cause activation. These results suggest that although the R. sphaeroides and E. coli CheYs are similar in that they require phosphorylation for activation, they may differ in both the nature of the phosphorylation-induced conformational change and their subsequent interactions with the flagellar motor. Caution should therefore be used when projecting from E. coli CheY onto novel response regulators.

Two-component signal transduction is widely used by bacteria to sense environmental conditions (reviewed in Ref. 1); an extensively studied example of this is bacterial chemotaxis (reviewed in Ref. 2). Two-component systems comprise histidine protein kinases and response regulators. Histidine protein kinases are homodimeric proteins in which each subunit transphosphorylates the other subunit on a histidine residue using ATP as the phosphodonor. The rate of this autophosphorylation reaction is controlled by sensory stimuli. Phosphorylated histidine protein kinases are able to transfer the phosphoryl group onto a conserved aspartate residue within the active site of their cognate response regulators. Phosphorylation of the response regulators changes their activity, allowing them to produce a response that is appropriate for the original stimulus. Over 4600 different response regulators have been identified by genome sequencing, with examples being found in plants, lower eukaryotes, and bacteria (3). The signaling modules within two-component pathways are highly conserved, yet some bacteria have more than 100 different two-component systems with apparently little cross-talk between them. This indicates that there must be a high degree of specificity between these homologous components.
In E. coli, a decrease in attractant concentration causes the chemoreceptors to signal via CheW, to increase the rate of autophosphorylation of the chemotaxis histidine protein kinase CheA (4,5). CheA-P phosphotransfers to the response regulators CheY and CheB (6,7). CheB-P is a chemoreceptor demethylase involved in adaptation (8). CheY-P interacts with the FliM component of the flagellar motor promoting a switch in the direction of flagellar rotation from counter-clockwise rotation to clockwise rotation, which causes the cell to change its swimming direction (9). Response regulators are believed to be in a dynamic equilibrium between their inactive and activated conformations. Phosphorylation stabilizes but is not a prerequisite for the activated conformation of the response regulator. Mutation is an alternative way of activating a response regulator; activating mutations stabilize the activated conformation of the response regulator. For example, the CheY(D13K,Y106W) mutant is nonphosphorylatable yet can still cause the motor to switch from counter-clockwise rotation to clockwise rotation (10).
There are numerous examples of single domain response regulators that do not function as CheYs (11)(12)(13). The Rhodobacter sphaeroides genome contains a total of 14 single domain response regulators (14); of these, 6 have been designated CheYs based on their level of sequence similarity to Escherichia coli CheY and their proximity to other chemotaxis loci (15).
R. sphaeroides has a single flagellum that rotates unidirectionally. Unlike E. coli, R. sphaeroides changes swimming direction by stopping the flagellar motor; during these stops the cell is randomly reoriented by Brownian motion (16). R. sphaeroides has multiple homologues of all of the E. coli chemotaxis genes except cheZ. Most of these are organized into three major chemotaxis operons (17). Of particular relevance to this study are the six cheY genes as follows: cheY 1 , cheY 2 , and cheY 5 are found in cheOp 1 ; cheY 3 is found in cheOp 2 ; cheY 6 is found in cheOp 3 ; and cheY 4 is found adjacent to a chemoreceptor gene at an unlinked locus (18). cheOp 2 and cheOp 3 have been shown to be essential for chemotaxis (17,19), whereas deletion of cheOp 1 has only minor effects upon chemotaxis under laboratory conditions (20). With the exception of the CheYs, the cellular localization of the R. sphaeroides chemotaxis proteins has been determined; the sensory transducing proteins encoded by cheOp 2 form a complex with the transmembrane chemoreceptors at the cell poles, whereas the proteins encoded by cheOp 3 localize with putative cytoplasmic chemoreceptors to a cytoplasmic cluster (21). The polar chemotaxis cluster contains CheA 2 , which has been shown in vitro to phosphorylate all six of the R. sphaeroides CheYs and both CheBs (22). In contrast, the cytoplasmic chemotaxis cluster contains the atypical kinase formed from CheA 3 and CheA 4 (subsequently referred to as CheA 3 A 4 ), which can only phosphotransfer to CheY 1 , CheY 6 , and CheB 2 . Phosphosignaling from both chemotaxis clusters has been shown to be essential for chemotaxis (23).
In vitro assays have shown that all six R. sphaeroides CheYs are capable of binding to the FliM component of the flagellar motor, and that this binding is enhanced by CheY phosphorylation (24). If all six CheYs can bind to FliM, then an intriguing question is as follows: why does R. sphaeroides have six CheYs? The consequence of each of the CheYs binding to FliM is not known; the ability of a CheY to bind FliM in vitro does not necessarily imply that this interaction happens in vivo or that it mediates direction changing. In this study, we determined which of the six CheYs are important for controlling flagellar motor rotation. We identified the cellular localization of these important CheYs using a combination of CFP 2 /YFP tagging and immunogold electron microscopy. These CheYs were then targeted by site-directed mutagenesis with the aim of either preventing phosphorylation or stabilizing the activated conformation of the CheYs. One phosphorylation site mutant, CheY 6 (D56N), caused a stopped phenotype (cells were flagellate but nonmotile). This mutant was shown to be phosphorylated at an alternative site, and in vivo data suggest that phosphorylation at this alternative site mediates the stopped phenotype.

EXPERIMENTAL PROCEDURES
Plasmids and Strains-The plasmids and strains used in this study are shown in Table 1 and supplemental Tables I and II. E. coli strains were grown in LB medium at 37°C. R. sphaeroides strains were grown in succinate medium at 30°C either aerobically with shaking without illumination or photoheterotrophically in an anaerobic cabinet (Don Whitley Scientific) with illumination at 50 mol m Ϫ2 s Ϫ1 . Where required, antibiotics were used at concentrations of 100 g ml Ϫ1 for ampicillin and 25 g ml Ϫ1 for kanamycin, nalidixic acid, and tetracycline.
Molecular Genetic Techniques-All standard genetic techniques were performed as described (25). Pfu polymerase (Promega) was used for all PCRs. All primers were synthesized by Sigma-Genosys. The plasmid midi-prep kit (Qiagen) was used to prepare sequencing quality DNA, which was sequenced by the DNA sequencing service (Department of Biochemistry, Oxford University). DNA sequence was analyzed using Clone Manager version 7 (Scientific and Educational Software).
Site-directed Mutagenesis of cheY Genes in the R. sphaeroides Genome-Overlap extension PCR was used to generate the constructs for mutating the cheY genes; these constructs contained the mutated cheY genes plus 500 bp of flanking genomic sequence on each side. The most frequently used R. sphaeroides codon was used for each of the desired amino acid substitutions. These constructs were cloned into the allelic exchange suicide vector pK18mobsacB to generate the mutation plasmids (supplemental Table II). The mutation plasmids were checked by sequencing and then used to introduce the point mutations into the genome of R. sphaeroides by allelic exchange as described previously (Table 1) (19,26). Western blotting was used to confirm that the point mutations had no effect on protein expression levels.
Behavioral Analysis-Behavior was measured under both aerobic and photoheterotrophic growth conditions (Table 1  and supplemental Table I); cheY mutations that affected chemotaxis had similar effects under both sets of growth conditions. The swarm plate responses, the behavior of tethered cells in a flow chamber to the removal of propionate, and the photoresponses of R. sphaeroides strains were characterized as described previously (17). Nine data sets were obtained for swarm plate and photoresponse analysis. Three data sets that together contained at least 10 cells were obtained for each tethered cell analysis. Strains were described as chemotactic if swarm rings were visible on swarm plates. Swarm plates were also used to assess motility by reference to known nonmotile, nonchemotactic, and chemotactic strains; nonmotile cells form smaller colonies on swarm plates than motile but nonchemotactic cells which in turn form smaller colonies than chemotactic cells. Motility and free-swimming stopping frequency was assessed in photoheterotrophic cells at an OD 700 nm ϭ 0.6 by microscopic examination of free-swimming cells.
Fluorescence Microscopy-Differential interference contrast and fluorescence images of CFP/YFP fusion expressing R. sphaeroides strains were acquired as described previously (21). At least five fields of view each containing at least 50 cells from independent cultures were analyzed for each strain.
Immunogold Electron Microscopy-Aerobic cells at an OD 700 nm ϭ 0.6 (wild-type cells are motile at this optical density) were prepared for electron microscopy as described (27). Gold particles were scored according to their cellular position and their proximity to other particles as described (28). A total of 80 cell sections was analyzed for each strain.
Protein Purification-His-tagged CheA and CheY proteins were overexpressed and purified as described previously (22,23,29). Mutant CheY proteins were overexpressed and purified in the same way as the wild-type versions of the proteins. Protein purity and protein concentrations were measured as described (22). Purified proteins were stored at Ϫ20°C; activity remained constant for at least 1 year.
Mass Spectroscopy-CheY 6 (D56N)-P was prepared according to the above phosphotransfer method using CheA 3 A 4 as the kinase (nonradioactive ATP was used). Following electrophoresis, the gels were stained with Coomassie Blue, and the CheY 6 (D56N)-P bands were excised and washed with water for 30 min. Samples were digested with trypsin, reductively alkylated on cysteine residues, and prepared for mass spectroscopy according to standard methods (30). Samples were analyzed using a nano-liquid chromatography-MS/MS system consisting of a reversed phase high pressure liquid chromatography (Ultimate Plus TM (Dionex/LC Packings, The Netherlands)) coupled to an ion trap tandem mass spectrometer (HCTplus TM , Bruker Daltonics, Bremen, Germany) as described (31). Individual MS/MS data were searched against the Swiss-Prot data base (version 11.12.2005) using Mascot TM software (Matrixscience, London, UK). Identification of the phosphorylated residue Ser-83 of the CheY 6 (D56N)-P protein was based on the neutral loss (Ϫ97.98Da) observed in the tryptic peptide (M ϩ 3H) 3ϩ 608.05 Da, and a loss of Ϫ18 Da mass at this particular residue because of a conversion of serine to dehydroalanine at this position. Interpretation of the data was performed in accordance with published guidelines (32). 4 Are Essential for R. sphaeroides Chemotaxis-In-frame deletions of the cheY genes were generated to determine the minimal subset(s) of cheYs that could support chemotaxis. Deletion of all six R. sphaeroides cheY genes (strain JPA1337) abolished chemotaxis but not motility; the cells swam but failed to respond to chemosensory and photosensory stimuli. Chemotaxis was observed in strains lacking any one of cheY 1 , cheY 2 , cheY 3 , cheY 4 , and cheY 5 ( Table  1). Previously we have shown that cheY 6 is essential for chemotaxis (17); however, in this study we show that cheY 6 alone cannot support chemotaxis because the strain deleted for the other cheYs (JPA1025 (⌬cheY 1 , ⌬cheY 2 , ⌬cheY 3 , ⌬cheY 4 , and ⌬cheY 5 )) was nonchemotactic. Because cheY 6 alone is not sufficient for chemotaxis, a series of five strains were created with four cheYs deleted but retaining cheY 6 plus one other cheY. Of these five strains, only JPA433 (⌬cheY 1 , ⌬cheY 2 , ⌬cheY 3 , and ⌬cheY 5 ) and JPA434 (⌬cheY 1 , ⌬cheY 2 , ⌬cheY 4 , and ⌬cheY 5 ) were chemotactic ( Table 1), indicating that cheY 6 plus either of cheY 3 and cheY 4 can support chemotaxis. CheY 3 and CheY 4 appear to be functionally redundant for one another. A strain deleted for both cheY 3 and cheY 4 (JPA425) was nonchemotactic, suggesting that CheY 1 , CheY 2 , and CheY 5 are unable to partner CheY 6 in supporting chemotaxis. Thus, chemotaxis roles can only be assigned to three of the six R. sphaeroides CheYs: CheY 3 , CheY 4 , and CheY 6 ; these CheYs were therefore analyzed further.

CheY 6 and Either of CheY 3 or CheY
Localization of CheY 3 , CheY 4 , and CheY 6 -Strains were produced in which the cheY genes were replaced with genes encoding the corresponding CFP/YFP fusions. The functionality of the fusions was tested using swarm plates and in vitro phosphotransfer to the purified fusion proteins from their cognate kinases (supplemental Tables I and II). N-terminal fusions to both CheY 3 and CheY 4 were fully functional in both wild-type backgrounds and in backgrounds retaining cheY 6 but deleted for the four remaining cheYs. However, strains with either N-or C-terminal fusions to CheY 6 were nonchemotactic. Western blotting using an antibody against CheY 6 showed that the CheY 6 fusions were expressed at much lower levels than wild type (undetectable in the case of the N-terminal fusion), which could explain why strains containing these fusions were nonchemotactic. Only the N-terminal fusions to the three CheYs were successfully expressed and purified from E. coli, and these behaved like the wild-type proteins in the in vitro phosphotransfer assays (Table 2).
Fluorescence microscopy showed that the majority of YFP-CheY 3 was delocalized throughout the cytoplasm; no polar clusters were observed, although some protein was localized to cytoplasmic clusters (Fig. 1). In contrast, CFP-CheY 4 mainly localized to discrete regions at both the cell poles and in the cytoplasmic cluster with some diffuse fluorescence throughout the cytoplasm. The fluorescence from CheY 6 -YFP and CheY 6 (D56N)-YFP (a phosphorylation site mutant of CheY 6 ) was mainly localized to cytoplasmic clusters, although a substantial amount of diffuse fluorescence was seen throughout the cytoplasm; no polar clusters were visible. Because the CheY 6 fusions were nonfunctional, the localization of CheY 6 and CheY 6 (D56N) was also determined by immunogold electron microscopy ( Fig. 1). This method suggested that CheY 6 and CheY 6 (D56N) are distributed diffusely throughout the cytoplasm (72 and 75% of gold particles were cytoplasmic and unassociated, respectively).
Site-directed Mutagenesis of CheY 3 , CheY 4 , and CheY 6 -To assess the contributions of CheY 3 , CheY 4 , and CheY 6 to the control of flagellar motor rotation, point mutations were made in the cheY genes in the R. sphaeroides genome (Table 1). These were in sites which in E. coli CheYs have been shown to either prevent phosphorylation or to cause activation. A variety of behavioral assays were performed on the mutants, and the in vitro phosphorylation of the CheYs was measured (Tables 1 and  2). For convenience, we have used the equivalent E. coli CheY residue numbers when referring collectively to mutations in the R. sphaeroides CheYs (Fig. 2). The phosphorylation site mutations D57A †3 and D57N † were made with the aim of preventing phosphorylation, whereas the D13K † , X95V † , and X106W † mutations were made with the aim of activating the CheYs. Each mutated cheY was introduced into an otherwise wild-type background. However, because CheY 3 and CheY 4 are redun-dant for one another, it is probable that the phenotype of a null point mutation in either one of these cheYs would be masked by the other cheY. Therefore, the cheY 3 and cheY 4 mutations were also introduced into strains deleted for cheY 1 , cheY 2 , cheY 4 , and cheY 5 and cheY 1 , cheY 2 , cheY 3 , and cheY 5 , respectively.
Mutation of Asp-13 and Xaa-106-Asp-13 † is strictly conserved in all CheYs and has been shown to be involved in Mg 2ϩ binding and catalysis. Mutation of this residue to lysine in E. coli prevents phosphorylation by CheA, yet increases the CW motor bias (33,34). Conversion of both Asp-13 to lysine and Tyr-106 to tryptophan further increases the CW motor bias. These mutations strongly favor the activated conformation of E. coli CheY, increasing the FliM binding affinity to that of CheY-P (10). The equivalent of the Asp-13 † residue is present in all R. sphaeroides CheYs (Fig. 2). However, CheY 3 and CheY 4 already have tryptophan instead of tyrosine at their equivalent of the 106 † position, and CheY 6 has valine at this position. Strains containing the cheY 3 (D10K) and cheY 4 (D10K) alleles in otherwise wild-type backgrounds were chemotactic; however, these alleles were unable to support chemotaxis in strains deleted for all of the other cheYs except cheY 6 . The cheY 6 (D10K) and cheY 6 (D10K,V104W) mutants were nonchemotactic. All of the strains expressing D13K † mutants of the R. sphaeroides cheYs exhibited wild-type stopping frequency (Table 1). In summary, strains containing the cheY 3 (D10K), cheY 4 (D10K), cheY 6 (D10K), or cheY 6 (D10K,V104W) alleles were phenocopies of their corresponding cheY deletion mutants (Table 1). This suggests that D13K † mutations make the CheY 3 , CheY 4 , and CheY 6 proteins nonfunctional. This contrasts with E. coli, where the CheY(D13K) mutant causes a very different phenotype (CW rotation) to the cheY deletion mutant (counter-clockwise rotation). The role of the residue at position 106 † was probed by mutating it to either phenylalanine (for CheY 3 , CheY 4 , and CheY 6 ) or to tryptophan (for CheY 6 only). None of these mutations had any effect on chemotaxis, stopping frequency, or in vitro phosphorylation (Tables 1 and 2). This suggests that unlike E. coli CheY, the R. sphaeroides CheYs are tolerant of substitutions at position 106 † .
Mutation of Residue 95-Another mutant that has been shown to activate E. coli CheY is the I95V mutation. This mutant is phosphorylatable and shows an increased affinity for FliM (35). CheY 3 and CheY 4 have a lysine residue at this position (Fig. 2). Interestingly, an E. coli CheY(I95K) mutant behaves like the ⌬cheY mutant and is unable to stimulate CW rotation of the flagellum (36,37). Mutation of the Lys-95 † residue in CheY 3 and CheY 4 to either valine or isoleucine had no detectable effects upon CheY in vitro phosphorylation, stopping frequency, or the ability of these CheYs to support chemotaxis (Table 1). These data suggest that the Lys-95 † residue is not important for the function of CheY 3 or CheY 4 .
Mutation of the Phosphorylatable Aspartate-The D57A † and D57N † mutations remove the phosphorylatable aspartate residue from CheY. These mutations have been widely used to prevent response regulator phosphorylation, allowing the in vivo properties of unphosphorylated response regulators to be studied (38). The D57A † mutants of CheY 3 , CheY 4 , and CheY 6 were all nonphosphorylatable in vitro and were in vivo phenocopies of their corresponding cheY deletion mutants (Table 1). This indicates that phosphorylation of these CheYs is required for chemotaxis.
Surprisingly, the D57N † mutants behaved differently to the D57A † and the cheY deletion mutants. Like the cheY 3 (D53A) and cheY 4 (D53A) mutations, the cheY 3 (D53N) and cheY 4 (D53N) mutations were not capable of supporting chemotaxis in strains retaining cheY 6 but lacking the remaining cheYs. However, unlike the D57A † mutations, the cheY 3 (D53N) and cheY 4 (D53N) mutations partially inhibited chemotaxis in otherwise wild-type backgrounds on swarm plates but behaved like the wild-type strain in the tethered cell and photoresponse assays (Table 1).
Unlike the cheY 6 deletion and the cheY 6 (D56A) mutant, the cheY 6 (D56N) mutant had a stopped phenotype, i.e. it did not swim in any of the behavioral assays. Electron microscopy revealed that this mutant was flagellate suggesting that the flagellar motor was stopped in this strain (data not shown). Unlike the E. coli flagellar motor, the R. sphaeroides flagellar  motor is a stop-start motor; it does not reverse, i.e. clockwise rotation in E. coli is replaced with a stop in R. sphaeroides. It is likely that the stopped phenotype of the cheY 6 (D56N) strain is analogous to the E. coli tumbly phenotype which is caused by exclusive clockwise rotation of the flagella. A strain containing cheY 6 (D56N) but lacking the remaining cheYs is also stopped, whereas a strain containing wild-type cheY 6 but lacking the remaining cheYs is motile, indicating that CheY 6 (D56N) is capable of causing the stopped phenotype in the absence of the other CheYs. Although the D57A † mutants of CheY 3 , CheY 4 , and CheY 6 were nonphosphorylatable, the D57N mutants were phosphorylatable by their cognate kinases (Table 2 and Fig. 3). This difference in phosphorylation between the D57A † and D57N † mutants might account for the different behavior of the D57A † and D57N † mutants. The phosphorylation of the D57N † mutants occurred on a much slower time scale than for the wild-type CheYs, suggesting that the mutant CheYs were being phosphorylated at a different site to the wild-type CheYs (Fig.  3). Even though the phosphotransfer reactions were slow for the D57N † mutants, the final phosphorylation levels of CheY 4 (D53N) and CheY 6 (D56N) were higher than those of the corresponding wild-type proteins, suggesting that the D57N † mutants dephosphorylate more slowly than the wild-type proteins. The E. coli phosphorylation site mutant, CheY(D57N) is phosphorylated at an alternative site, Ser-56, by CheA (39); CheY 3 and CheY 4 both have a threonine at this position, which is likely to be the alternative phosphorylation site for these proteins. However, CheY 6 has a nonphosphorylatable leucine residue at this position, and so the phosphorylation site of CheY 6 (D56N) must be elsewhere on the protein.
Phosphorylated CheY 6 (D56N) Stops the Flagellar Motor-The cheY 6 (D56N) mutant was the only stopped (flagellate, but nonmotile) mutant found in this study. Unlike wild-type cells, this mutant did not form an expanding colony in the swarm plate assays. However, after prolonged incubation of the swarm plates (4 days) flares could be seen emerging from the colony. Microscopic examination of the cells in these flares revealed that they had regained motility. Cells recovered from the flares were used to inoculate new swarm plates, where they formed  OCTOBER 27, 2006 • VOLUME 281 • NUMBER 43 expanding colonies with a diameter similar to that observed for motile but nonchemotactic strains (e.g. ⌬cheY 6 ). This recovery of motility appeared to be permanent suggesting that additional mutations had occurred that could suppress the stopped phenotype caused by the cheY 6 (D56N) mutation. A total of 10 suppressor strains were isolated from 10 different flares. The cheY 6 , fliM, fliN, cheA 4 , and cheA 3 loci in these suppressor strains were amplified by PCR and sequenced. All of the suppressor mutants retained the cheY 6 (D56N) mutation but had no further mutations in cheY 6 or in the fliM and fliN genes; instead all 10 of the suppressor mutations were found in either cheA 3 (six) or cheA 4 (four) (supplemental Table 3). All 10 mutations were different and were either deletions (eight) or insertions (two). Interestingly, two of the deletions were in-frame deletions of the kinase domain of CheA 4 , and one of the insertions had a repeated codon within the P1 domain of CheA 3 . Because CheA 3 and CheA 4 together form the kinase (CheA 3 A 4 ) that can phosphorylate CheY 6 (D56N), these data suggest that it is the phosphorylated form of CheY 6 (D56N) that is capable of stopping the R. sphaeroides flagellar motor.

CheYs of R. sphaeroides
To test whether CheY 6 (D56N) requires functional CheA 3 and CheA 4 to cause the stopped phenotype, the cheY 6 (D56N) allele was introduced into a cheA 3 (H51Q) mutant (lacks the phosphorylation site of CheA 3 ) and a cheA 4 (G220K) mutant (has an inactive kinase domain in CheA 4 ); CheA 3 (H51Q) and CheA 4 (G220K) do not participate in phosphorylation reactions in vitro nor do they support chemotaxis in vivo (23). Both of the product strains lacked the stopped phenotype, being motile but nonchemotactic, which is consistent with the hypothesis that CheY 6 (D56N) needs to be phosphorylated by CheA 3 A 4 to stop the motor.
CheY 6 (D56N) Is Phosphorylated at Ser-83-The E. coli phosphorylation site mutant, CheY(D56N), is phosphorylated at Ser-56 by CheA; however, R. sphaeroides CheY 6 has a nonphosphorylatable leucine residue at this position. The phosphorylation site was determined by mass spectroscopy. Phosphorylated CheY 6 (D56N) was prepared as described under "Experimental Procedures." Following digestion with trypsin, the resulting peptides were analyzed using MS/MS spectroscopy. The peptide ICML(S*)SV-AVSGSPHAAR was found to be phosphorylated at the serine residue highlighted with an asterisk, which corresponds to residue Ser-83 in CheY 6 (D56N) (Fig. 4).
In Vitro Phosphorylation of Ser-83 Mutants of CheY 6 -The predicted phosphorylation site of CheY 6 -(D56N) is Ser-83. This residue was mutated in both wild-type CheY 6 and in CheY 6 (D56N). Three different substitutions were made as follows: for S83D, aspartate substitutions have been successfully used in other proteins to mimic phosphoserine (40,41); for S83T, this is a conservative substitution, which should be phosphorylatable; and for S83A, this mutant should be nonphosphorylatable. Phosphotransfer from CheA 3 -P to the CheY 6 mutants was measured under multiple turnover conditions in the presence of CheA 4 and excess ATP, allowing CheA 3 phosphorylation to continue throughout the course of the reactions. The progress of these phosphotransfer reactions after 30 min is shown in Fig. 5; phosphotransfer occurred in reactions where a decrease in CheA 3 -P levels was accompanied by an increase in CheY 6 -P levels.
Phosphotransfer occurred to a similar extent for the CheY 6 (D56N) and CheY 6 (D56N,S83T) mutants (Fig. 5). However, no phosphotransfer was observed to the CheY 6 (D56N,S83D) mutant, and only a small amount of phosphotransfer was seen to the CheY 6 (D56N,S83A) mutant (there was ϳ200-fold less CheY 6 (D56N,S83A)-P than CheY 6 (D56N)-P). In E. coli, the asparagine 57 residue of the CheY(D57N) mutant has been shown to spontaneously deamidate to aspartate on a slow time scale (42). If the CheY 6 (D56N,S83A) mutant underwent a similar deamidation reaction, then the product of this deamidation would be CheY 6 (S83A), which could be phosphorylated at Asp-56 by CheA 3 -P. Deamidation and subsequent phosphorylation of a small fraction of the CheY 6 (D56N,S83A) mutant could account for the residual phosphotransfer seen for this mutant. Presumably, the CheY 6 (D56N,S83D) mutant does not deamidate, and consequently no phosphorylation was observed. These results are again consistent with Ser-83 being the phosphorylation site of CheY 6 (D56N).
Phosphotransfer from CheA 3 -P to wild-type CheY 6 and subsequent dephosphorylation of CheY 6 -P was complete within 30 s, i.e. after 30 s no CheA 3 -P or CheY 6 -P was detectable (Fig.  5). Phosphotransfer to the CheY 6 (S83D) mutant and its subsequent dephosphorylation was much slower than wild type with high levels of CheA 3 -P and CheY 6 (S83D)-P remaining after 30 min. Phosphotransfer to CheY 6 (S83T) and CheY 6 (S83A) was much faster than for CheY 6 (S83D) but still slower than wild type because detectable levels of CheA 3 -P remained after 30 min. These results show that Ser-83 is an important residue in CheY 6 where mutations affect both the rate of phosphotransfer from CheA 3 -P and the rate of CheY 6 -P dephosphorylation.
In Vivo Behavior of the CheY 6 Ser-83 Mutants-The wildtype cheY 6 gene in the R. sphaeroides genome was replaced with the genes encoding the D56N and/or Ser-83 mutants analyzed above. The chemotactic behavior of the resulting strains was analyzed. Unlike the stopped cheY 6 (D56N) mutant, the strains containing the nonphosphorylatable CheY mutants CheY 6 (D56N,S83D) and CheY 6 (D56N,S83A) were motile but nonchemotactic, i.e. they behaved like the cheY 6 deletion. Interestingly, cells containing the phosphorylatable mutant, CheY 6 (D56N,S83T) showed severely reduced motility where the cells were predominantly stopped with occasional very brief periods of movement. The CheY 6 (D56N,S83T) mutant therefore has a slightly less severe motility defect than the CheY 6 (D56N) mutant, which could indicate that the precise geometric configuration of the alternative phosphorylation site is critical for causing the stopped phenotype. These data are consistent with the hypothesis that phosphorylation of Ser-83 is required for CheY 6 (D56N) to cause the stopped phenotype.
Ser-83 was mutated in an otherwise wild-type CheY 6 to assess the role of this residue in wild-type CheY 6 . The cheY 6 (S83T) and cheY 6 (S83A) mutants both supported wildtype levels of chemotaxis, whereas the cheY 6 (S83D) mutant phenocopied the cheY 6 deletion, i.e. was motile but nonchemotactic. The lack of chemotaxis of the strain expressing the CheY 6 (S83D) protein can be explained by the reduced rates of phosphotransfer and dephosphorylation of this mutant protein. However, the normal behavior of the cheY 6 (S83A) mutant indicates that phosphorylation of Ser-83 is not essential for wild-type chemotaxis.

DISCUSSION
Prior to the sequencing of its genome, it was believed that R. sphaeroides had only four CheYs (CheY 1 , CheY 2 , CheY 3 , and CheY 4 ) (18). A genetic study of the roles of these CheYs was performed; signaling events that could not be explained by these CheYs were attributed, by the process of elimination, to the newly discovered CheY protein CheY 5 (43). However, there was yet another undiscovered CheY protein, CheY 6 (17). In this study we demonstrate that CheY 6 plus either of CheY 3 and CheY 4 are essential for chemotaxis in R. sphaeroides. No chemotaxis roles could be found for the cheOp 1 -encoded CheY 1 , CheY 2 , or CheY 5 .
Why would an organism have six CheYs when two are sufficient? One possibility is that all six CheYs are used in the wildtype strain but the robustness of the R. sphaeroides chemotaxis system allows the system to cope even when some of those CheYs are removed. Alternatively, because the chemotaxis genes in cheOp 1 are expressed at very low levels under the conditions used in our assays (43), they may have little or no role in chemotaxis; however, it is possible that under alternative growth conditions, the CheYs encoded by cheOp 1 may be more highly expressed and may adopt a more significant role in chemotaxis. A further possibility is that cheOp 1 and the CheYs encoded within it are not involved in chemotaxis but instead control some other cellular process; there is a precedent for this in the ␣-subdivision of proteobacteria, where Rhodospirillum centenum uses chemotaxis-like systems to control both flagella biosynthesis and cyst development (44). Why are a minimum of two CheYs required for R. sphaeroides chemotaxis when many bacteria, e.g. E. coli, Bacillus subtilis, and Vibrio cholerae, can manage with only one (45,46)? Sinorhizobium meliloti and R. sphaeroides lack the CheZ, CheC, CheX, and FliY signal terminating CheY phosphatases that have been described in other bacteria (47)(48)(49). Like R. sphaeroides, S. meliloti also requires two CheYs for normal chemotaxis; S. meliloti CheY2-P can bind to the FliM component of the flagellar motor and bring about direction changing, whereas CheY1 cannot bind to the motor and acts as a signal terminating phosphate sink (50). Could R. sphaeroides be using a similar system? All six R. sphaeroides CheYs can bind to FliM in vitro, and this binding is enhanced by CheY phosphorylation (24). If all six CheYs bind FliM in vivo, then this would suggest that any CheY acting as a phosphate sink would be capable of binding to FliM. However, the binding of a CheY to FliM may not result in a change in flagellar rotation. Some of the R. sphaeroides CheYs may be unable to stop flagellar rotation, but they might compete for binding to FliM with CheYs that can stop flagellar motor rotation. This competition could be central to integrating the signals from both the cytoplasmic and polar chemoreceptor clusters.
In E. coli, the predominant polar localization of CheY is dependent upon CheA (51). CheA 3 and CheA 4 are localized to the cytoplasmic chemotaxis cluster, whereas CheA 2 is localized to the polar chemoreceptor cluster (21). Phosphosignals from CheA 2 , CheA 3 , and CheA 4 have been shown to be essential for chemotaxis. CheY 6 is phosphorylated by both CheA 2 and CheA 3 A 4 (22,23). CheY 3 and CheY 4 are phosphorylated by CheA 2 but not by CheA 3 A 4 . If the R. sphaeroides CheYs localized to the sites of their cognate kinases, then CheY 6 would be expected to be found at both the cytoplasmic and polar clusters, whereas CheY 3 and CheY 4 would be found only at the cell poles. Experimental measurements of CheY localization showed that significant proportions of all of these CheYs were delocalized, consistent with the requirement of CheYs to be mobile signal carriers allowing communication between the signaling clusters and the flagellar motor. CheY 6 -YFP did localize to the cytoplasmic chemotaxis cluster; however, unlike YFP-CheY 3 and YFP-CheY 4 , this fusion protein was nonfunctional. Immunogold electron microscopy did not confirm this cytoplasmic clustering of CheY 6 and instead indicated that the majority of CheY 6 was delocalized throughout the cytoplasm. The presence of the YFP tag on CheY 6 -YFP may have increased the affinity of CheY 6 for the cytoplasmic cluster possibly by stabilizing a conformation of CheY 6 that associates with the cluster. CheY 4 but not CheY 3 localized to the polar chemoreceptor cluster along with its cognate kinase, CheA 2 . Surprisingly both CheY 3 and CheY 4 also localized to the cytoplasmic chemotaxis cluster even though no known cognate kinase for these proteins is found there. This suggests that CheY 3 and CheY 4 may interact with a component of the cytoplasmic chemotaxis cluster; we are currently attempting to identify which component of this cluster is the binding partner for CheY 3 and CheY 4 .
The mutations that have been shown to activate E. coli CheY did not appear to activate any of the R. sphaeroides CheYs. Instead, strains expressing the mutant R. sphaeroides cheY(D13K † ) alleles behaved like the corresponding deletion strains, whereas those expressing the cheY(X95I,X95IV † ) and cheY(X106W,X106W Y † ) alleles behaved like wild type. These results may suggest that although the R. sphaeroides CheYs and E. coli CheY are similar in that they require phosphorylation for activation, they may differ either in the subsequent conformational change caused by that phosphorylation or in their interaction with the flagellar motor. Structural studies on S. meliloti CheY 2 also suggest that the phosphorylation-induced conformational change that occurs upon activation of E. coli CheY may not be universally applicable (52).
In vivo studies of the signaling role of the dephosphorylated forms of the CheYs using strains where the wild-type cheY gene had been replaced with a phosphorylation site mutant had unexpected results. The two mutations commonly used to prevent phosphorylation of response regulators, D57A † and D57N † , caused different phenotypes. The D57N † mutants could be phosphorylated at an alternative site by their cognate kinase, whereas the D57A † mutants could not be phosphorylated. This difference was most pronounced in the case of CheY 6 , where the cheY 6 (D56A) allele produced a motile but nonchemotactic strain, whereas the cheY 6 (D56N) allele produced a stopped strain, i.e. one that was flagellate but did not swim, suggesting that CheY 6 (D56N) could bind to and stop the flagellar motor. CheY 6 (D56N) can be phosphorylated at Ser-83 by the cognate kinase for CheY 6 , CheA 3 A 4 . Furthermore, the stopped phenotype of the cheY 6 (D56N) mutant requires functional CheA 3 and CheA 4 , suggesting that CheY 6 (D56N) needs to be phosphorylated at Ser-83 to stop the flagellar motor. Therefore, although chemotaxis requires a minimum of two In the control reaction an equal volume of buffer was added instead of the response regulators. Phosphotransfer is indicated by the appearance of phosphorylated response regulator and/or a reduction in the amount of CheA 3 -P. The chemotaxis and motility of strains bearing the mutant cheY 6 alleles in otherwise wild-type backgrounds are summarized below the phosphorimage. A ߜ indicates wild-type levels of either motility or chemotaxis, and a ϫ indicates a complete lack of either motility or chemotaxis. The cheY 6 (D56N,S83T) allele gives poorly motile cells, indicated by (ϫ/ߜ).
What is the role of the alternative phosphorylation site (Ser-83) in wild-type CheY 6 ? There is no evidence suggesting that alternative site phosphorylation occurs in wild-type CheY 6 . Even if it did, the slow phosphorylation kinetics are inconsistent with the ability of R. sphaeroides to respond to a stimulus in less than 1 s (53). Phosphorylation of Ser-83 is not essential for wild-type chemotaxis, because the cheY 6 (S83A) mutant supports wild-type levels of chemotaxis. Despite this, substitutions of Ser-83 altered both the rate of phosphotransfer from the kinase and the rate of response regulator dephosphorylation. This was most obvious for the CheY 6 (S83D) mutant which failed to support chemotaxis. Ser-83 must therefore either be an important residue in the CheY 6 active site, or Ser-83 substitutions must affect its conformation.
A previous study on E. coli CheY has shown that the CheY(D57N) mutant is partially activated by phosphorylation of Ser-56 (39). Phosphorylation of mutants where the primary phosphorylation site has been replaced with asparagine has also been detected for the response regulators FixJ from S. meliloti and NtrC from E. coli (54,55). No cases of response regulator phosphorylation where the primary phosphorylation site was mutated to alanine have been reported. A likely explanation for this has been proposed (39); the carbonyl group of the phosphorylatable aspartate is required to chelate the Mg 2ϩ ion that is essential for catalysis (56); asparagine mutants retain this carbonyl group, whereas alanine mutants do not.
In many studies of response regulator action either the equivalent of the D57A † or D57N † mutation is made to facilitate the study of the signaling role of the dephosphorylated response regulator. Our results and those of other researchers suggest that caution should be used when interpreting the results of analogues of the D57N mutation. We have demonstrated that even if no phosphorylatable residue is present at position 56 † , it is still possible for the D57N † mutant of the response regulator to be phosphorylated and to be active. Investigators should confirm, rather than assume, that D57N † mutants of response regulators are not subject to alternative site phosphorylation. Other mutations that activated E. coli CheY did not appear to activate the R. sphaeroides CheYs, suggesting that not all CheYs undergo identical conformational changes upon phosphorylation. E. coli CheY is often used as a model for response regulators. We have shown that even among CheYs, there is not necessarily conservation of critical residues and that activating/ inactivating point mutations do not automatically translate from one response regulator to another.