CheA Kinase and Chemoreceptor Interaction Surfaces on CheW*

Chemotactic responses of Escherichia coli to aspartic acid are initiated by a ternary protein complex composed of Tar (chemoreceptor), CheA (kinase), and CheW (a coupling protein that binds to both Tar and CheA and links their activities). We used a genetic selection based on the yeast two-hybrid assay to identify nine cheW point mutations that specifically disrupted CheW interaction with CheA but not with Tar. We sequenced these single point mutants and purified four of the mutant CheW proteins for detailed biochemical characterizations that demonstrated the weakened affinity of the mutant CheW proteins for CheA, but not for Tar. In the three-dimensional structure of CheW, the positions affected by these mutations cluster on one face of the protein, defining a potential binding interface for interaction of CheW with CheA. We used a similar two-hybrid approach to identify four mutation sites that disrupted CheW binding to Tar. Mapping of these “Tar-sensitive” mutation sites and those from previous suppressor analysis onto the structure of CheW defined an extended surface on a face of the protein that is adjacent to the CheA-binding surface and that may serve as an interface for CheW binding to Tar.

Escherichia coli motility involves short periods (1-2 s) of smooth swimming ("runs") punctuated by briefer episodes (0.1-0.2 s) of somersaulting ("tumbles") that result in random direction changes. As an E. coli cell swims about, its chemotaxis system monitors the levels of attractant and repellent chemicals encountered and modulates the turning probability to promote cell migration up concentration gradients of attractants and down gradients of repellents (1,2). Chemotactic movement results from communication between cell surface chemoreceptor proteins and components of the flagellar motors. Two complementary processes, excitation and adaptation, contribute to this control by modulating the intracellular level of a "tumble signal," phosphorylated CheY (3)(4)(5)(6).
The molecular machinery that mediates excitation and adaptation in this system has been defined and its mechanism of action elucidated in part: transmembrane receptor proteins (also called MCPs) 1 bind attractant and repellent ligands in the periplasm and modulate the autokinase activity of CheA (a protein histidine kinase) accordingly. Binding of repellent stimuli to MCPs is thought to stimulate CheA autokinase activity, whereas attractant stimuli inhibit this activity (7)(8)(9)(10)(11)(12). Autophosphorylated CheA can, in turn, donate its phosphoryl group to CheY (13)(14)(15), a modification that directs CheY to bind to the switch components of the flagellar motor and thereby promote tumbling (4). Another possible fate for the phosphoryl group of P-CheA is transfer to CheB (13,16,17); phosphorylation of CheB stimulates its methylesterase activity and therefore promotes sensory adaptation by adjusting the signaling properties of the MCPs via removal of methylester groups from the side chains of specific glutamate residues (16,17).
The events that enable MCP-mediated control of CheA autokinase activity are thought to occur within a ternary complex that includes MCPs, CheA, and CheW (9, 11, 18 -22). Understanding how this complex forms and how it mediates CheA regulation are, therefore, central to defining the molecular mechanisms of both excitation and adaptation in the chemotaxis system. Although many components of the bacterial chemotaxis system have been studied in considerable detail (for reviews, see Refs. 5,6,23,and 24), little is known about the specific biochemical contributions of CheW. This 167-amino acid protein (7) does not appear to have any enzymatic activity by itself; nor does it share sequence or structural similarities with motifs that do have catalytic activities. But CheW clearly plays a crucial role in the bacterial chemotaxis system, as indicated by the complete inability of cheW mutants to respond to any chemotactic stimuli (25). We are attempting to define how CheW contributes to chemotaxis signal transduction.
CheW has four known activities in vitro: (i) binding to CheA; (ii) binding to MCPs, such as Tar; (iii) promoting formation of CheA⅐CheW⅐MCP ternary complexes; and (iv) enabling MCPs to stimulate and/or inhibit CheA autokinase activity (9, 10, 12, 18 -22, 26, 27). Based on these observations, current models depict CheW as an "adapter protein" that serves to tether CheA to the MCPs (18,20,28). While such a simple adapter role would account for the known in vitro activities of CheW, it is important to emphasize that the relationships among these in vitro activities and their roles in vivo have not been clearly established. In particular, the functional importance of CheW's binding activities has not been clearly established by, for example, determining whether disrupting the CheW 7 CheA binding interaction affects formation of the CheA⅐CheW⅐MCP ternary complex and/or activation of CheA within this complex.
In previous work (29), we followed a traditional mutational approach in an attempt to investigate the functional significance of CheW binding interactions with CheA and MCPs: we identified mutations that disrupted CheW's ability to support chemotaxis, and then we examined the effect of these mutations on the ability of the protein to bind to CheA and Tar. Each of the mutants we isolated following this approach was defective in formation of an activated ternary complex and exhibited defects in both binding interactions. While those results sug-gested that CheW's binding activities are indeed important for its role in chemotaxis signal transduction, they did not allow us to examine the effects of eliminating just one binding interaction at a time; nor did they allow us to identify what segments of CheW might be involved specifically in interactions with CheA and which with MCPs. To examine more rigorously the role(s) of CheW's binding activities and to begin defining potential interaction interfaces with CheA and Tar, we turned to the DITA approach of Inouye et al. (30) with the goal of identifying CheW mutants lacking just one of the two binding activities. Our results, in conjunction with the recently solved three-dimensional structure of CheW (31), indicate clustering of mutation sites that cause disruption of CheW 7 CheA interactions and therefore suggest the location of a possible CheA binding interface on CheW. In addition, our results help to further elucidate the CheW surface that appears to mediate contacts with MCPs, first defined by the genetic suppressor studies of Liu and Parkinson (26).
Strains and Plasmids-Saccharomyces cerevisiae strain YCJ4 and plasmids pLEX202PL, pAS-Snf1, and pLex-bicoid were kindly provided by Jeremy Thorner (University of California, Berkeley). Strains RP3098 (⌬(flhA-flhD) (37)) and D392 (⌬cheW (38,39)) are Escherichia coli K-12 derivatives kindly provided by J. S. Parkinson (University of Utah) and F. W. Dahlquist (University of Oregon), respectively. Strain D392 was used for swarm assays, whereas strain RP3098, lacking all chemotaxis genes, was used for purification of most of the proteins used in this study.
Plasmid pGAD:CheW encoded a CheW-Gal4AD fusion and was created using vector pGAD424 (CLONTECH). Wild type cheW was PCRamplified from pCW (27) using primers that generated EcoRI and SalI sites upstream and downstream, respectively, of cheW. The resulting PCR fragment was digested with EcoRI and SalI and then ligated into corresponding sites of pGAD424. Plasmid pLex:cTar encoding cTar-LexA fusion was created by ligating a fragment of tar encoding amino acids 257-553 into EcoRI and BamHI sites of plasmid pLex202PL. Plasmid pGBT9:CheA encoding CheA-Gal4BD fusion was created by inserting full-length cheA into vector pGBT9 (CLONTECH, Inc.) between SmaI and SalI sites.
The plasmid pCE (29) was used for overexpression and purification of CheW mutant proteins carrying a Met-His 6 -Cys extension at their N-terminal ends. Mutant cheW alleles were excised from corresponding pGAD:CheW plasmids and ligated into pCE using EcoRI and SalI sites. For use in the swarm assays, mutant cheW alleles were excised from pGAD:CheW vectors using AgeI and SalI and ligated into corresponding sites of pCnoW (29). All plasmids created in the course of this work were verified by dideoxy sequencing (performed at the University of Maryland Center for Agricultural Biotechnology).
Mutagenesis and Screening-Random cheW mutants were generated using an error-prone PCR procedure in which an unequal ratio of dNTPs and a high concentration of Mn 2ϩ created conditions favorable for nucleotide misincorporation by Taq polymerase (40). The PCRgenerated, mutagenized cheW DNA fragment was gel-purified and then transformed along with EcoRI/SalI-digested pGAD424 into YCJ4 strain carrying pGBT9:CheA and pLex:cTar (30). pGAD:CheW plasmid was restored in vivo in the resulting YCJ4 transformants by a gap repair mechanism (41). Half of the transformation mixture was plated on Yc medium lacking histidine, tryptophan, and leucine (YcϪhis,trp,leu) (selecting for pLex:cTar, pGBT9:CheA, and pGAD:CheW, respectively) and including 5-FOA. Colonies capable of growth in the presence of 5-FOA were transferred to Yc-his,trp,leu plates lacking uracil, and the results of growth on these plates were evaluated after 3 days at 30°C. The colonies that displayed a URA Ϫ phenotype were then restreaked on Yc-his,trp,leu plates; after a 2-day incubation at 30°C, these colonies were tested for ␤-galactosidase activity using a colony lift assay (42).
The second half of the transformation mixture was plated on Yc-his,trp,leu medium in which the amount of uracil was reduced from the normal level of 0.01 to 0.0008%; this selected for URA ϩ colonies (attempts to use medium completely lacking uracil were unsuccessful). Among the mutants growing on this type of medium, we then identified ␤-Gal Ϫ colonies as described above. pGAD:CheW from selected colonies was then isolated and retransformed into YCJ4 strain expressing cTar and CheA fusions. The resulting transformants were assayed for URA3 and lacZ expression to confirm the phenotype of the strains observed in the differential interaction trap assay. Nucleotide changes in these alleles were determined by dideoxy sequencing at the University of Maryland.
Measuring CheW Affinities for CheA and Tar-These assays utilized wild-type and mutant CheW proteins that had been engineered to include an N-terminal His 6 -Cys tag, as described previously (29). The purified His 6 -Cys-CheW proteins were modified by covalently attaching fluorescein maleimide (Molecular Probes) to the single cysteine side chain. Fluorescently labeled His 6 -Cys-CheW (hereafter referred to as F*-CheW) was then used for binding assays. Binding of wild-type or mutant F*-CheW to CheA was monitored using fluorescence anisotropy as described previously (29). In short, these assays involved measuring the increase in fluorescence anisotropy ( ex ϭ 492 nm, em ϭ 517 nm) of the F*-CheW sample as increasing concentrations of CheA were added. These experiments were performed in TEND buffer (50 mM Tris-HCl, pH 7.5, 0.5 mM EDTA, 100 mM NaCl, 0.5 mM dithiothreitol) containing 10% polyethylene glycol 12,000. For each CheW variant, the magnitude of the observed anisotropy increase as a function of the CheA concentration was used to generate a binding isotherm that was fitted to a simple binding equation to generate an estimate of K d CheA , the dissociation constant for the F*-CheW⅐CheA complex. Details of this assay and analysis procedure have been published previously (29).
Assays of the binding of wild-type or mutant CheWs to Tar were performed using the "pull-down competition" approach detailed previously (29). These assays made use of inner membrane vesicles carrying overproduced levels of Tar (18). Such vesicles are easily pelleted by centrifugation and carry with them, into the pellet, F*-CheW in the expected amounts based on CheW binding to Tar. To assess binding of mutant CheW proteins binding to Tar, competition experiments were performed in which wild-type F*-CheW (final concentration 2 M) was premixed with a particular CheW mutant protein (0 -50 M). Then the inner membrane vesicles were added to this mixture, generating a final Tar concentration of 11 M in TEND buffer containing 1 mg ml Ϫ1 BSA. The resulting protein mixtures were allowed to equilibrate (in the dark, with gentle mixing) for 10 min at 25°C, and then the vesicles (and associated F*-CheW) were sedimented by centrifugation. The fluorescence emission intensity of the supernatants of these centrifuged samples were then measured ( ex 492 nm, em 517 nm). The data generated in these experiments showed the expected progressive increase in signal intensity (reflecting increasing levels of free F*-CheW) at increasing concentrations of competitor (unlabeled, mutant CheW). Analyses of these titration curves allowed us to define the K d of the complex formed by Tar with the mutant CheW (29).

Efficacy of the Differential Interaction Trap
Assay-We used the DITA of Inouye et al. (30) to quickly and easily observe CheW interactions with CheA and Tar. This approach is a three-hybrid modification of the yeast two-hybrid assay (43), and it involved simultaneous expression of three chimeric proteins, carried by three distinct plasmids, in S. cerevisiae strain YCJ4. The three chimeric fusions were as follows: CheW fused to the transcription activation domain of Gal4 (CheW-Gal4AD); CheA fused to the DNA-binding domain of Gal4 (CheA-Gal4BD); and the cytoplasmic domain of Tar (amino acids 257-553) fused to the DNA-binding protein LexA (cTar-LexA). As depicted in Fig. 1, interaction of CheW-Gal4AD with CheA-Gal4BD is expected to result in expression of the URA3 reporter gene in YCJ4, and interaction of CheW-Gal4AD with cTar-LexA is expected to drive expression of the lacZ reporter gene in this yeast strain. Our goal was to use this system to identify two distinct classes of cheW mutations: (i) those that diminished CheW binding to CheA, but not to Tar and (ii) those that decreased the affinity of CheW for Tar but not for CheA. Before proceeding with such a mutant search, we first exam-ined the specificity and sensitivity of the DITA two-hybrid interactions.
To assess the sensitivity of the DITA in our system, we evaluated the effects of six cheW mutations that we had previously shown to cause moderate to severe decreases in the affinity of CheW for CheA and Tar (29). Our results (Table I) indicate that expression of the URA3 reporter gene is sensitive to moderate (and large) changes of K d CheA but not to small affinity changes: a 6.5-fold increase in K d CheA gave rise to a Ura Ϫ phenotype, but a 3.3-fold increase did not. We observed somewhat greater sensitivity for altered expression of the lacZ reporter gene in response to changes in the affinity of CheW for Tar; a 3-fold increase in K d Tar was sufficient to generate a ␤-Gal Ϫ phenotype.
Tar and other MCPs form homodimers (44,45) that appear to further interact and form oligomers (e.g. trimers of dimers (28,46)) that may represent the functional form of the chemoreceptors for in vivo signaling. Cytoplasmic fragments of Tar (cTar) and Tsr (cTsr) also form such complexes (12,47). However, it is not clear whether MCP dimerization/oligomerization is necessary for CheW binding. The dimerization status of the cTar-LexA fusion in our DITA experiments is also unclear. LexA has the ability to form dimers. Although the association constant for dimerization is relatively weak (2 ϫ 10 4 M Ϫ1 ) for LexA in isolation, this affinity is improved dramatically upon binding to the lexA operon (48,49). Thus, LexA dimerization might promote dimerization of our cTar-LexA fusion. Despite the potential complexities associated with dimerization or higher order oligomerization of cTar-LexA, our experiments indicated that the two-hybrid interaction between cTar-LexA and CheW-Gal4AD served as an accurate reporter of CheW 7 Tar interaction: mutations in CheW that affected in vitro binding to full-length Tar (in membrane vesicles) had the expected effect on the two-hybrid interaction.
Isolation of cheW Point Mutants Using DITA-We subjected the entire cheW gene to random mutagenesis using error-prone PCR (40); PCR products were introduced into appropriately cleaved plasmid pGAD424 (CLONTECH) and expressed as CheW-Gal4AD fusions in YCJ4 cells that already carried plasmids encoding CheA-Gal4BD and cTar-LexA. To isolate variants of CheW that exhibited weakened affinity for CheA, but not for Tar, we plated transformation mixtures onto medium containing 5-FOA, a nucleotide analog that inhibits growth of yeast cells expressing URA3 (50). Colonies that grew on these plates were then tested for ␤-galactosidase activity. Of ϳ3,000 5-FOA-resistant colonies, nine were found that maintained a ␤-Gal ϩ phenotype (reflecting maintenance of the CheW 7 Tar interaction). These Ura Ϫ ␤-Gal ϩ mutants (hereafter referred to as Class I) were subjected to DNA sequence analysis and retained for further analysis.
To identify variants of CheW that exhibited diminished affinity for Tar, but not for CheA, we first plated YCJ4 transformation mixtures onto medium depleted of uracil and then screened ϳ2,000 of these to identify four that were ␤-Gal Ϫ . These Ura ϩ ␤-Gal Ϫ mutants (hereafter referred to as Class II) were subjected to DNA sequence analysis and retained for further analysis.
The nucleotide sequences of three of the Class II mutants exhibited single point mutations, whereas the remaining Class II mutant and all nine Class I mutants had two or more nucleotide substitutions. For each of these multiple mutants, a single mutation was responsible for the observed DITA phenotype. This was determined by separating the mutations using convenient restriction sites and, in some instances, by recreating individual mutations using oligonucleotide-directed mutagenesis. The amino acid changes associated with the Class I and Class II mutations are summarized in Table II.
Biochemical Analysis of CheW Mutant Proteins-The DITA results suggested that specific mutations selectively inhibited CheW binding to CheA or Tar. This approach, while extremely convenient for initial identification of mutants, provided only an indirect and qualitative assessment of binding defects for the CheW variants. To support these results and quantify the  (29). Each value represents the average of at least two independent titration experiments Ϯ S.E.
expression was assessed as growth on selective medium lacking uracil. ϩ, growth after 3 days of incubation at 30°C; Ϫ, absence of growth under such conditions. b lacZ expression was assessed by a colony lift assay of ␤-galactosidase activity (42): ϩ, production of blue color when permeabilized cells were exposed to 5-bromo-4-chloro-3-indolyl-␤-D-galactopyranoside (Xgal); Ϫ, the absence of color change; Ϫ/ϩ, a faint blue color was observed in half of the assays performed with this mutant (the total number of trials was 4). effects of the mutations on CheW binding affinities, we performed in vitro binding titrations that utilized purified, fluorescein-labeled versions of the wild-type and mutant variants of CheW (29). To evaluate CheW binding affinity for CheA, we monitored the increase in fluorescence anisotropy exhibited by labeled CheW when it binds CheA (Fig. 2) (29). To assess CheW binding affinity for Tar, we used pull-down experiments in which membrane vesicles carrying high levels of Tar were used to sediment fluorescein-labeled CheW out of solution via centrifugation (Fig. 3) (18, 29). These experiments defined binding isotherms that we analyzed to estimate K d CheA (the dissociation constant for the CheW⅐CheA complex) and K d Tar (the dissociation constant for the CheW⅐Tar complex). The observed K d values (Table II) indicated that the DITA had successfully identified mutations that specifically diminished CheW affinity for CheA (without affecting its affinity for Tar); Class I mutants exhibited K d CheA values ranging from 5-to 12-fold higher than the wild-type K d CheA value, whereas K d Tar values for these mutants remained within 30% of the wild-type value. Class II mutants, as expected, exhibited significant increases in K d Tar (values ranged from 4 to 14 times the wild-type K d Tar value); however, this was accompanied by a moderate (ϳ3-fold) increase in K d CheA . Several factors could have contributed to this lack of specificity with the Class II mutants, such as an inadequately sensitive reporter system for CheW 7 CheA interactions and/or the nature of the Tar binding interface, a possibility that is considered in greater detail under "Discussion." Effect of CheW Mutations on Tar⅐CheW⅐CheA Ternary Complexes-In the presence of CheW and MCPs, CheA exhibits a dramatically enhanced autokinase activity (9,11,12,21,22,51,52). This activation of CheA appears to result from formation of an MCP⅐CheW⅐CheA ternary complex in which CheW serves as a "coupling factor." Little is known about how CheW accomplishes this receptor-kinase coupling or how it promotes formation of the ternary complex. As a first step toward improving our understanding of the functional role of CheW, we used our set of binding-defective CheW proteins to determine whether both of CheW's binding interactions are necessary for it to accomplish receptor-kinase coupling. We examined the ability of our mutant CheW variants to accomplish CheA activation by assaying CheA autokinase activity in mixtures containing purified CheW, CheA, the cytoplasmic domain of Tar (LZ-cTar), and CheY in the presence of an ATPase coupling system (34,51,52). These results (Fig. 4) indicate that each of our mutant CheW variants was less effective than the wildtype CheW in mediating MCP-CheA coupling in the concentration range tested. Although assays at higher CheW concentrations would have been informative for some of the mutant proteins, we were unable to pursue such assays because of concentration limitations for CheW, CheA, and CheY; adding higher CheW concentrations would have required corresponding decreases in the volumes of CheA, CheY, and/or ATPase coupling reagents, but we were unable to increase the stock concentrations of these proteins enough to accommodate the necessary changes.
For the Class I mutants (Fig. 4A)

FIG. 3. Membrane vesicle pull-down assays of CheW binding to Tar.
A competition assay was used to monitor displacement of fluorescein-labeled wild type CheW by mutant CheW proteins. 2 M F*-CheW wt was added to 11 M Tar (in membrane vesicles) in the presence of increasing amounts of unlabeled wild type CheW (q), CheW mutants L158Q (‚), V87A (छ), E38D (Ⅺ), or BSA (E) in TEND buffer in the presence of 1 mg/ml BSA (added to minimize nonspecific association of CheW with membrane vesicles). After a 10-min incubation at 25°C, membrane vesicles and bound proteins were sedimented by centrifugation, and the fluorescence intensity remaining in the supernatant was measured. Lines represent best fits of the data to define the values of K d Tar . Details of the titration procedure and data analysis are described in a previous paper (29). binding contacts that mediate formation of the CheW⅐CheA complex (in the absence of MCP) are also involved in formation of the activated ternary complex.
Class II mutants also exhibited diminished coupling activities (Fig. 4B). However, attributing this decrease to a reduced affinity of CheW for Tar is complicated by the fact that these mutants also have K d CheA values that are ϳ3-fold higher than the wild-type value. For assessing the effects of an increase in K d Tar , we found it helpful to use the titration profile of mutant T46A (a Class I mutant) as a standard for comparison; T46A exhibits a K d CheA value that is ϳ6-fold higher than the wildtype value and a K d Tar value that differs from the wild-type value by only ϳ30%. Both of the Class II mutants that we studied in detail exhibited coupling abilities that were considerably worse than that of T46A. If the affinity of CheW for Tar had no effect on coupling ability, then we would have expected Class II mutants to perform at least as well as T46A and perhaps considerably better because they bound CheA somewhat better than did T46A. This was not the case: Class II mutants E38D and V87A both performed considerably worse than T46A in coupling assays. These results suggest that the binding contacts that mediate formation of the CheW⅐Tar complex (in the absence of CheA) are also involved in formation of the activated ternary complex. We also noted that, in contrast to the titration profiles observed with wild-type CheW and the Class I mutants, the Class II mutants generated distinctly sigmoidal titration profiles (Fig. 4B), perhaps reflecting cooperativity, a possibility that is considered further under "Discussion." In Vivo Complementation by CheW Mutant Proteins with Altered Binding Affinities-To investigate how well our mutant CheW proteins functioned in the context of the chemotaxis signaling pathway in intact cells, we transformed plasmids expressing each cheW allele into a ⌬cheW host strain and assessed the chemotactic abilities of the resulting transformants by examining their migration rates in semisolid "swarm" agar (Fig. 5). These results indicated that the mutant variants were surprisingly effective, supporting swarm rates that were close to that observed with wild-type CheW. These results indicate that in vivo activity of CheW is surprisingly insensitive to moderate changes in the affinity of CheW for CheA and MCPs.
Under conditions of CheW overproduction, some differences were observed with several of the mutants. Elevated levels of wild-type CheW resulted in severe inhibition of chemotactic ability (Fig. 5), a phenomenon that has been reported previously (14,53). Class II mutants lacked this inhibitory ability, whereas one of the Class I mutants (L158Q) appeared to have enhanced inhibitory ability (Fig. 5B and data not shown). L158Q exhibited an affinity for Tar that is slightly higher (30%) than that of wild-type CheW. In previous work, we reported a similar enhancement of inhibition with cheW154ocr, a mutant that exhibited a 3-fold increased affinity for Tar compared with  (21,34). A, CheA activation by wild type CheW (q) and by CheW mutant proteins that are defective in binding interactions with CheA: T46A (Ⅺ), V45L (छ), and K56I (‚). B, CheA activation by wild type CheW (q) and CheW mutant proteins that were initially identified as being defective in binding interactions with Tar: V87A (Ⅺ) and E38D (‚). A control reaction carried in the absence of LZ-cTar (E) is shown in both panels. Lines connecting the data points were added to help the reader to distinguish among data sets and have no theoretical significance.

FIG. 5. Complementation abilities of mutant CheW variants.
Swarm plate assays were carried out using E. coli strain D392 (⌬cheW) expressing wild type CheW (top gray line), no CheW (bottom gray line), and the following CheW mutant proteins: T46A (‚), V45L (E), and K56I (Ⅺ) (A); L158Q (q) (B); and V87A (Ⅺ) or E38D (q) (C). Transformants were inoculated into 0.3% tryptone agar plates containing 0 -50 M isopropyl-␤-D-thiogalactoside (IPTG) and 100 g ml Ϫ1 ampicillin. The diameters of the outer edges of swarm colonies were measured after 16 h of incubation at 30°C. Lines connecting the data points were added to help the reader to distinguish among data sets and have no theoretical significance. The reported swarm diameters are averages for at least two independent experiments. Deviations in these values from one experiment to another were small, such that the error bars for most measurements were smaller than the size of the symbols used in the figure. In a few cases, the error bars were large enough to be shown. the wild-type CheW (29). These results indicate a correlation between the affinity of CheW for MCPs and its effectiveness as an inhibitor of chemotaxis at elevated levels of expression; mutants with significantly diminished affinity for Tar are ineffective as inhibitors, whereas mutants with even moderately enhanced affinity for Tar are more effective inhibitors. DISCUSSION Rationale, Advantages, and Limitations of the DITA Approach for Analysis of CheW Binding-To identify CheW mutants lacking just one of the two binding activities, we applied the DITA approach of Inouye et al. (30). One advantage of this selection/screen is that it allowed us to focus specifically on the protein-protein interactions of interest in a heterologous reporter system without possible interference from other components of the chemotaxis system or from unknown in vivo activities of CheW. In addition, this approach selected against several types of mutations that were uninteresting for our purposes, including those that caused large scale perturbations of the three-dimensional structure (e.g. mutants with disrupted folding patterns), nonsense mutations, and mutations that caused significant changes in CheW stability, turnover, or expression level.
This approach led to identification of several mutations (Class I) that diminished CheW affinity for CheA without simultaneously affecting its interaction with Tar. The existence of such mutants and their location on the three-dimensional structure of CheW (see below) suggests that the binding interfaces for CheA and Tar involve distinct faces of the CheW surface. The DITA approach was less successful in identifying mutations that specifically influenced CheW 7 MCP binding interactions; each of our Class II mutants exhibited a moderately decreased ability to bind CheA in addition to the anticipated decrease in affinity for Tar. Nonetheless, we were able to isolate one mutant (E38D) that exhibited a large defect in Tar binding and a considerably smaller defect in CheA binding. Our two classes of mutants allowed us to address several basic questions about the functional role(s) of CheW's binding interactions. These questions are considered below.

Mutations That Affect CheW Pairwise Binding Interactions Have Corresponding Effects on Its Receptor-Kinase Coupling
Activity-Titration of CheA with increasing concentrations of CheW gives rise to a simple hyperbolic binding isotherm; there are no indications of multiple affinities of binding sites or cooperativity. Similar results are observed for titrations that monitor binding of wild-type CheW to MCPs (18,29,52). We isolated mutants that exhibit diminished binding affinities in these simple pairwise interactions, and these mutants allowed us to address a simple question: Do these mutations also affect the ability of CheW to promote formation of the CheA⅐CheW⅐MCP ternary complex that is thought to mediate key events in chemotaxis signal transduction? At first glance, this question might seem trivial and its answer obvious. However, it is important to emphasize the potential complexities of CheW binding interactions in CheA⅐CheW⅐MCP complexes. In contrast to the apparent simplicity of CheW pairwise binding interactions, when CheW is placed in mixtures of CheA and MCPs, there appear to be multiple competing binding equilibria that result in formation of several distinct ternary complexes. These different versions of "the ternary complex" have different stoichiometries, different kinase activities, and different stabilities (21). This complexity suggests that the component proteins can interact with one another in a variety of different ways, some of which lead to a ternary complex in which CheA autokinase activity is very high (the "activated ternary complex") and some of which lead to a ternary complex in which CheA is essentially inactive (the "inactive ternary complex"). In view of this complexity, we deemed it important to investigate the relationship between the pairwise interactions and those that mediate formation of the activated ternary complex.
Our results clearly indicate that mutations that diminish the affinity of the pairwise interactions have a corresponding effect on the ability of CheW to promote formation of the activated ternary complex. In theory, the differences observed between the wild-type and mutant titration profiles (Fig. 4) could reflect either diminished levels of activated ternary complex or diminished activity of CheA within this complex. We did not pursue experiments to distinguish between these two possibilities; however, it is notable that the shapes of the profiles and the observed magnitudes of the maximal activities match fairly closely with those predicted by computer simulations generated by assuming diminished complex formation. 2 These results suggest that CheW coupling activity involves both of its binding interactions (i.e. CheW must interact with both CheA and Tar to enable formation of the active complex). Moreover, these findings are consistent with the idea that the binding contacts that mediate pairwise CheW 7 CheA and CheW 7 Tar interactions are closely related to (perhaps even identical to) the contacts that are required for formation of the activated ternary complex.
Two of our mutant CheW proteins (V87A and E38D) generated a sigmoidal relationship between the concentration of CheW and the levels of activated complex, suggesting that activated complex formation is a cooperative process, at least with these mutants (see Fig. 4B). We noted a similar effect with the mutant CheW154ocr in previous work (29). CheW154ocr and the Class II mutants analyzed here had altered affinity for receptors. However, pairwise titration experiments using these mutants and Tar gave rise to simple hyperbolic binding curves; only in the presence of CheA (in the kinase activation assays) did we observe the apparent cooperativity. Cooperativity in regulation of CheA activity by receptors has been reported previously (12,21,22,54) and may be linked to the phenomenon of receptor clustering (19,45). Our results suggest that CheW might contribute to this phenomenon through its binding interactions with receptors. Binding of CheW to an MCP dimer might, for example, generate or uncover additional contact sites that promote formation of a complex involving CheW, CheA, and a cluster of MCPs. These "revealed sites" might include MCP positions that become available for interaction with CheA, similar sites on CheA that become available for interaction with MCPs, or possibly some additional sites on CheW itself that allow it to act as a nucleator of higher-order macromolecular aggregates. It is puzzling that we observed apparent cooperativity with several CheW mutant variants but not with other mutants and not with wild-type CheW. Further detailed analysis of these mutants may provide insight into the molecular mechanism by which CheW promotes formation of activated ternary complexes.
In Vivo Chemotactic Signaling Tolerates Significant Decreases in CheW Binding Affinities-We examined the ability of our binding-defective CheW variants to support chemotaxis in vivo. When examining the results of these complementation experiments, it is useful to consider two mutants as representative test cases: K56I for examining the effect of diminishing the affinity of CheW for CheA and E38D for examining the effect of diminishing the affinity of CheW for Tar. The K56I mutant exhibits an affinity for CheA that is weaker than that of wild-type CheW by a factor of ϳ12, but its affinity for Tar is essentially the same as wild type. The E38D mutant has a diminished affinity for Tar by a factor of 14, whereas its ability to bind CheA is influenced to a lesser extent (ϳ3-fold). Both of these mutants exhibited diminished abilities to promote formation of the activated ternary complex in vitro, as discussed above. Therefore, we were surprised to find that these and other such mutants performed quite well when tested by in vivo complementation (swarm) assays; the mutant proteins supported swarming rates that were comparable with that supported by wild-type CheW. One conceivable, albeit extreme, interpretation of these findings would be that in vitro binding, ternary complex formation, and CheW coupling activity are in vitro artifacts that have little to do with CheW activity in vivo. However, in previous work we demonstrated that several cheW null alleles (isolated by virtue of their Che Ϫ phenotype in swarm assays) encode variants of the CheW protein that have severely disrupted binding activities and that completely lack coupling activity in vitro. This observation and the dramatic increase in CheA activity that results from ternary complex formation (9,21) argue against the extreme interpretation suggested above. A second possibility is that the chemotaxis signaling pathway is able to make due with a severely diminished level of activated ternary complex. Several recent publications have demonstrated that the signaling circuitry of the chemotaxis system is a "robust" network that can maintain responsiveness and sensitivity despite large fluctuations in the activities and/or concentrations of key signaling components (55)(56)(57). In previous work (58), we observed that CheA active site mutants having autokinase activities as low as 6% of the wild-type activity can support chemotaxis (as measured by swarm assays). Thus, the 10-or 20-fold reduction in the amount of activated ternary complex expected for our E38D and K56I mutants might be accommodated by this robust network. Such a situation would explain why most cheW null alleles (isolated by screening for a Che Ϫ phenotype (29)) have very large effects on both K d CheA and K d Tar ; less drastic effects on CheW binding would have no phenotype in swarm assays and would never be identified for further analysis. Yet a third possible explanation of the surprising efficacy of our weakly binding CheW mutants is that in vivo conditions compensate in some other way for the poor binding abilities of the mutant CheW proteins, perhaps as a result of macromolecular crowding in the cytoplasm (59). Receptor clustering (19,60) or adjustments of MCP methylation levels (55,61,62) might also contribute to the ability of the in vivo chemotaxis system to overcome defects in CheW binding affinity. However, our qualitative examination of MCP methylation patterns in cells expressing these cheW alleles did not reveal any clear differences from methylation patterns observed in cells expressing wildtype cheW, leading us to doubt the contribution of methylation as a compensating factor (results not shown).
Mutation Sites Define Possible Interaction Interfaces for CheW Binding to CheA and MCPs-One of the goals of this study was to define the segments of CheW that mediate its binding to CheA and to MCPs. The results discussed above suggest that the binding surfaces used in pairwise CheW⅐CheA and CheW⅐Tar complexes are also utilized in the activated ternary complex, so identifying sites that are required for pairwise binding interactions is a viable approach for defining sites that play important roles in the protein-protein interactions of the activated ternary complex. The three-dimensional structure of CheW from Thermotoga maritima has recently been determined by NMR methods in the Dahlquist laboratory (31). This structure indicates that CheW consists of two fivestranded ␤-barrels surrounding a hydrophobic core (Fig. 6). The T. maritima protein has an amino acid sequence that is quite similar to that of the E. coli protein (63); therefore, it seems likely that they have similar structures. Such similarity has been observed in the structures of another component of the chemotaxis system (CheY) from E. coli and T. maritima (64). Mapping our mutation sites onto the structure of T. maritima CheW allowed us to visualize the relative orientation, in three-dimensional space, of the amino acid positions altered by our Class I and Class II mutations. The most striking result from such analysis is the clustering of the Class I mutation sites on one face of CheW (Fig. 6A). In particular, these sites cluster along the solvent-exposed edges of ␤-strands 4 and 5, an extended loop that connects ␤ 3 to ␤ 4 , and the C-terminal ␣-helix. We propose that this surface of CheW serves as a binding interface that mediates interactions between CheW and CheA in pairwise combinations as well as in the activated ternary complex. The same surface of CheW was proposed to mediate its interaction with CheA based on NMR chemical shift changes observed in T. maritima CheW in the presence of CheA (31). Possible CheA 7 CheW interaction sites have also been proposed as a result of computer models generated by assuming that CheW adopts a three-dimensional structure resembling that of the C-terminal regulatory domain of CheA (28,65). Our results support the prediction of Bilwes et al. (65) that the ␤ 3 -␤ 4 region of CheW (corresponding to ␤ 10 and ␤ 11 of CheA) mediates CheW 7 CheA binding interactions.
We were successful in identifying only four Class II mutants. With such a small sample size, "clustering" of these on the three-dimensional structure of CheW (Fig. 6B) is not necessarily a surprising or insightful finding. However, there are numerous cheW mutations that were identified by Liu and Parkinson (26) as suppressors of MCP mutations. Presumably, these mutations affect positions that participate in CheW-MCP contacts. Mapping these mutation sites and our Class II sites onto the structure of T. maritima CheW defines an extended surface adjacent to the putative CheA binding site of CheW (Fig. 6). We suggest that this extended surface mediates CheW 7 MCP binding in both the pairwise CheW 7 MCP complexes as well as in the activated ternary complex.
Our approach identified mutations that altered CheW 7 CheA or CheW 7 Tar binding. Several different types of changes in protein structure might underlie the observed binding defects: removal of binding contacts, steric disruption of binding contacts, nonspecific changes in global structure, and/or protein folding. In this regard, it is important to note that the Class I mutants exhibited normal affinities for Tar, so it seems unlikely that the binding phenotype of these mutants (diminished affinity for CheA) arose from nonspecific alterations of CheW structure. The results generated with the Class II mutants are less clear cut; these mutations had the expected effects on Tar binding affinity but also caused small decreases in the affinity of CheW for CheA. The Class II mutations affect ␤-strands located on a common surface of CheW and might cause subtle structural changes in the orientations of these strands, changes that could be propagated to the CheA-binding interface of CheW. In this regard, it is interesting to note that the C-terminal ␣-helix of CheW and ␤-strand 10 (connected to the N-terminal end of this helix) together form a structure that runs along the entire length of the CheW molecule and contacts most of the other secondary structural elements. This might provide a physical link that conveys structural perturbations (such as those caused by mutations) from the MCP-binding interface of CheW to its CheA-binding interface. Deleting most of this helix (in CheW154ocr) has the interesting effect of enhancing the affinity of CheW for Tar by a factor of 3 (compared with wild-type CheW) (29). Perhaps, in wild-type CheW, this helix mediates communication between the two binding interfaces of the protein.