Mutations leading to altered CheA binding cluster on a face of CheY.

CheY is the response regulator of Escherichiacoli chemotaxis and is one of the best studied response regulators of the two-component signaling system. CheY can receive phosphate from the histidine kinase, CheA. Phospho-CheY interacts with the motor-switch complex to induce clockwise flagellar rotation, thus causing the cell to tumble. We used an enzyme-linked immunosorbent assay to study the direct interaction between the kinase, CheA, and the regulator, CheY. The products of random, suppressor, and site-specific cheY mutants were assayed for their ability to bind CheA. Nine mutants showed altered binding. We sequenced and mapped these point mutations on the crystal structure of CheY, and a high degree of spatial clustering was revealed, indicating that this region of CheY is involved in CheA binding. Interestingly, five of these altered binding mutants were previously defined as being involved in motor-switch binding interactions. This suggested a possible overlap between the motor-switch binding and CheA binding surfaces of CheY. Using CheY (Trp-58) fluorescence quenching, we determined the equilibrium dissociation constants of CheA(124-257) binding for these CheY mutants. The results from the fluorescence quenching are in close agreement with our initial enzyme-linked immunosorbent assay results. Therefore, we propose that the CheA and the motor binding surfaces on CheY partially overlap and that this overlap allows CheY to interact with either the CheA or the flagellar motor, depending on its signaling (phosphorylation) state.

Bacterial response to hostile environmental conditions is regulated by a complex network of interacting proteins, with the most predominant interactions being generated by members of two-component systems (1). In Escherichia coli chemotaxis, the interaction between the receptor-coupled histidine autokinase, CheA, and the response regulator, CheY, controls the bacterial response to chemical environmental changes (1)(2)(3)(4). In response to changes in the receptor's occupancy and adaptation, CheA autophosphorylates (5,6) and subsequently transfers its phosphate to either CheY or CheB (7,8,15). This phosphotransfer results from CheY's intrinsic autophosphorylation activity and is not due to catalysis by CheA (9,10). Studies indicate that phospho-CheY is the activated form that binds to a motor-switch complex, causing a clockwise flagellar rotation and a net change in the bacterial swimming direction (11)(12)(13).
Allosteric changes within a large complex containing a receptor dimer, a CheA dimer, and two copies of a small coupling protein, CheW, regulate CheA autophosphorylation and subsequent phosphotransfer to CheY (11,14,22,23). The autophosphorylation site, His-48, lies on the N terminus of CheA, and the CheY binding determinants lie between residues 124 and 257, commonly known as the P2 domain (17). However, the CheA binding site on CheY is not defined. Structural studies indicate that CheY is a single domain protein that folds into a (␤/␣) 5 topology, with five ␤-strands forming the hydrophobic core, surrounded by five ␣-helices (18,19). Three aspartate residues, Asp-12, -13, and -57 form the molecule's active site, with Asp-57 being the site where CheY receives the phosphogroup from CheA His-48 (20,21). Phosphorylated CheY loses its affinity for CheA (23) and shows high binding affinity for FliM, one of the motor-switch components (13).
It has been estimated that E. coli contains about 50 homologous two-component systems, which govern various cellular responses to stress (1,3). Since the active site of response regulators is highly conserved (1), CheA must be able to differentiate CheY from other potential response regulators. Very recently, an NMR study on the interaction between an Nterminal fragment of CheA (1-233) and CheY has defined the CheA (1-233) binding site on CheY by measuring the chemical shift changes in CheY upon CheA (1-233) addition (26). Residues identified here lie distinct from the active site on ␣-4, ␤-4, ␣-5, and ␤-5 of CheY.
In the present study, we have employed a CheY-CheA binding assay, based on the ELISA, 1 to characterize, in vitro, the interaction between full-length CheA and CheY and to screen CheY mutants for possible defects in CheA binding. The mutants that altered the CheA interaction had amino acid positions clustered on a face of the CheY protein, and this threedimensional clustering suggested that this region of CheY is involved in CheA binding.
Overexpression and Purification of Proteins-Strains containing various vectors were grown in Luria broth at 37°C (30°C for CheY) to * This work was supported by National Institutes of Health Grant AI 18985. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. mid-exponential phase, induced appropriately for 4 -16 h, harvested and resuspended in 50 mM Tris (pH 7.9), and stored at Ϫ70°C until sonicated. Overexpressing lysates were obtained by thawing cells and sonicating them three to five times for 1 min and then centrifuging them at 30,000 ϫ g for 20 min. The supernatant fraction was used as a source of Che proteins. The wild-type CheY was overproduced from plasmid pRL22⌬Z in the E. coli-K12 strain CY15040 by inducing expression at 42°C (27). The mutant CheY proteins were overproduced from similar plasmid constructs (Table I) in SG1 strain (devoid of wild-type CheY), which was used to eliminate wild-type CheY contamination. All CheY proteins (wild-type and mutants) were purified as previously published (27). CheA was overproduced using the plasmid pDV4⌬EcoRV, and the CheA and CheW were overproduced using pDV4. CheA was purified as previously published (28). P2, a fragment of CheA, was overproduced from plasmid pTM22 in RP437 cells by inducing cells at mid-exponential phase with 100 M isopropyl-␤-Dthiogalactoside, and the protein was purified using a protocol by McEvoy et al. 2 In Vitro Mutagenesis of the cheY Gene-The plasmid-carrying cheY gene (pRL22⌬Z) was mutagenized using hydroxylamine mutagenesis (29). The mutagenized plasmid DNA was transformed into competent RP4079 cells, and transformants were plated by mixing with a tryptone swarm agar. After overnight incubation at 30°C, Che ϩ bacterial colonies appeared as miniswarms. Colonies exhibiting no swarm or minimal swarm were picked as mutants, and their mutant phenotype was confirmed by transforming a fresh batch of RP4079 competent cells with DNA isolated from each of the colonies.
DNA Sequencing-Double-stranded high copy number plasmid pRL22⌬Z, bearing putative cheY mutations, was DNA sequenced by a dideoxy chain termination method, using the protocol provided by the manufacturer of Sequenase, version 2.0 (U. S. Biochemical Corp.). Sequencing was facilitated by three appropriately situated, non-coding strand oligonucleotide primers that were specific for cheY.
Binding Assay-The binding assay was performed by coating microtiter plate wells with 100 l of 2-fold, serially diluted rabbit anti-CheY antibody in 50 mM Tris, pH 7.9 (starting with 240 ng antibody in 100 l of 50 mM Tris, pH 7.9). The plate was incubated overnight at 4°C, and the following day the wells were washed three times with 200 l of blocking buffer (buffer B, 1% dry skim milk and 0.1% sodium azide in 50 mM Tris, pH 7.9) and incubated overnight with buffer B to block any unoccupied sites. On the third day, the wells were washed three more times with buffer B, and 100 l of each protein (in 50 mM Tris, pH 7.9) was added in a sequential manner to react with the antibody, with each addition being followed by incubation at room temperature for 1 h and then washed three times with buffer B. The proteins used in the above step include 1) purified CheY or freshly prepared cell lysate with overexpressed CheY (50 -100 ng/l), 2) cell lysate with overexpressed CheA/ CheW (0.5-1.0 ng/l), 3) monoclonal anti-CheA antibody (CA1. 4.21, 1.0 ng/l), and 4) IgG-specific antibody linked to alkaline phosphatase (1:1000 dilution). Finally, the wells were washed five times with buffer B, and each well had 200 l of 37 mg of ortho-nitrophenyl phosphate (substrate) in 22 ml of glycine buffer (1 mM MgCl 2 , 0.1 mM ZnCl 2 , and 7.5 g of glycine in 1 liter adjusted to pH 10.0 with 1 M NaOH) added to it. After incubating for approximately 1 h at room temperature, absorbance readings were taken at 405 nm in an ELISA reader (Bio-tek Instruments, EL311).
Fluorescence Quenching-Equilibrium dissociation constants (K D ) of P2 (CheA 124 -257) binding for the wild-type as well as the mutant CheYs were determined according to Swanson et al. (26) by fluorescence quenching using a Perkin Elmer LS50 bioluminescence spectrophotometer. At a constant slit width of 10 nm, the Trp-58 of CheY was excited at 285 nm, and the emission was recorded at 346 nm. The P2 solution was added in small, measured increments, and the corresponding changes in the intrinsic fluorescence of CheY, as a function of P2 concentration, were monitored as a direct assay for P2 binding (26). In some experiments, at high concentrations of P2 (Ͼ20 M), the emission intensity, which until this point was approaching a minimum, began to increase again. Control experiments with P2 alone (no CheY) at concentrations over 20 M showed some absorbance at 285 nm, indicating the presence of some contaminating tryptophan in the P2 prep. More quantitative experiments (not described in detail) yielded a maximum emission of 500 ppm (Ͻ10% that of 1 M CheY) by 40 M P2. The corresponding data for CheY-P2 binding were adjusted accordingly.

RESULTS
Detection of CheY-CheA Binding with ELISA-Based on ELISA, a sensitive and specific CheY-CheA binding assay was developed, which used a CheY-CheA complex as the antigen sandwiched between an anti-CheY antibody and an anti-CheA monoclonal antibody. This whole complex was detected by mouse-IgG-specific antibody conjugated with alkaline phosphatase (Fig. 1). Since the assay was done in microtiter plates,  it required a very small amount of the sample, and coating the microtiter wells with polyclonal rabbit anti-CheY antibody provided a simple means to affinity purify and concentrate the CheY from an overexpressed cell lysate (see "Materials and Methods"). The assay was performed using cell lysates overexpressing CheY and CheA/CheW. To measure background activity, lysates with either no CheY or a CheY deletion (⌬18 -28) were used. As seen in Fig. 2, the wells containing wild-type CheY produced stronger signals than the negative controls, with the signal to noise ratio being at least 2 to 1. The signal obtained from the negative controls was due to a cross-reactivity of the anti-CheY antibody to CheA, which could not be completely eliminated. The mouse monoclonal antibody to CheA and the class-specific anti-mouse antibody did not cross-react with CheY (data not shown).

Isolation and Screening of CheY Mutants Resulting in
Altered CheA Binding-65 random CheY mutants, generated by hydroxylamine mutagenesis, 6 CheY suppressors to motorswitch mutants, and 2 site-specific mutants, D13K and Y106W, were examined for CheA binding. 9 of these mutants exhibited altered binding properties, with 8 of the 9 showing decreased binding and 1 showing a slight increase in binding (Fig. 3). DNA sequencing revealed that all the random mutants, with altered CheA binding, carried a single-point mutation in the cheY gene (data not shown). The majority of the CheY mutants screened by ELISA demonstrated CheA binding similar to that of the wild-type CheY.
Mapping of Altered Binding CheY Mutants on the CheY Structure-Red atoms in Fig. 4 depict the positions on CheY where mutant CheY residues result in altered CheA binding. They cluster to a region, distinct from the active site on the CheY surface, and this high degree of clustering suggests that it is the CheA binding surface. The CheY mutants that do not alter CheA binding were not found on this proposed surface (Fig. 4, green atoms). This proposed CheA binding surface consists of solvent-accessible surfaces of the C terminus of ␤-4 (Thr-87), loop region between ␤-4 and ␣-4 (Ala-90, Glu-93), ␤-5 (Tyr-106, Val-108), the loop between ␤-5 and ␣-5 (Phe-111, Thr-112), and ␣-5 (Glu-117). Interestingly, several of these residues (Fig. 4, covered by stippling) were previously implicated (25) in motor-switch interactions. This identified a surface common to both interactions. All of these residues are surface located and solvent accessible. Their side chains, with the exceptions of Thr-87 and Phe-111, extend out to the surrounding solvent, toward a region where it can easily make contact with CheA. Asp-13 is not located on this proposed surface, and it seems possible that the altered CheA binding properties of the mutant D13K may be due to an altered conformation (see "Discussion").
The substitutions that affected binding also make sense chemically in the context of an altered protein-interaction surface. For example, the two glutamate-to-lysine substitutions (E93K, E117K) involve charge changes that could affect the electrostatic interactions between the interacting proteins. The other substitutions involve side group volume changes (A90V, Y106W, V108M) as well as the hydrophobicity changes (T87I, T112I). Among these residues, only Thr-87 and Phe-111 have their side chains directed in toward CheY's hydrophobic core, and in these cases, only the backbone portion of these residues seems to contribute to the proposed CheA binding surface, so the influence of the side chain may be more indirect.
Determination of Binding Affinities Using Trp-58 Fluorescence Quenching-Tryptophan fluorescence quenching was used to quantitate the CheY mutant affinities for the CheY binding region of CheA. The P2 region of CheA has been shown to contain the major binding determinants for CheY interaction (16,17). The P2 fragment is devoid of Trp and has been shown to quench the fluorescence of CheY, Trp-58, upon binding (26). Equilibrium dissociation constants (K D ) determined by fluorescence quenching of CheY (Table II) were found to be inversely correlated to the binding determined by ELISA (Fig. 3). For example, T87I showed slightly higher binding in the ELISA method and, as expected, 2-fold lower K D . Whereas, E93K, Y106W, V108M, F111V, T112I, and E117K showed decreased binding in the ELISA method and severalfold increased K D values (Table II). D13K and A90V show no changes in their P2 binding affinity as determined by fluorescence quenching. Although it is known that D13K is defective in Mg 2ϩ binding (31) and that Mg 2ϩ binding quenches the Trp fluorescence (30), assaying P2 binding to D13K may not be comparable to the  (18). Highlighted residues are as follows: red atoms represent residues where mutation affects CheA binding, and green atoms represent residues where mutation does not affect CheA binding. The ␣-carbon backbone of CheY is in white. White stippling depicts solvent-accessible surfaces of the residues that were implicated in motor-switch binding (25). The top and bottom pictures are two different views, 90°apart from each other. wild type or other mutants, it is also possible that the residues Ala-90 and Asp-13 may be critical for the intact CheA binding but not for the P2 binding. DISCUSSION The goal of this study was to define the CheA binding face of CheY. We screened cheY mutants for altered CheA interaction, using a modified version of the ELISA (32), which is a specific, economic, and sensitive assay. A technical problem with this method was a reduced signal to noise ratio caused by a high background. This was determined to be due to the cross-reactivity of the polyclonal anti-CheY antibody to CheA. Affinity purification of the CheY antibody on a CheY column reduced but did not eliminate this cross-reactivity. However, none of the other components nor antibodies demonstrated significant background.
Use of this assay enabled us to identify nine point mutants, D13K, T87I, A90V, E93K, Y106W, V108M, F111V, T112I, and E117K, which showed altered CheA binding. The most striking result is their location on the CheY molecule and the fact that they clearly cluster on the face of CheY. This clustering becomes functionally more significant, since many of these residues (90, 108, 111, 112, and 117) were previously identified by genetic suppression analysis to be involved in motor-switch interaction (25). The motor-switch suppressor E27K was the only putative position that interacts with the motor switch that did not alter CheA binding. Another study, using NMR spectroscopy to examine the interaction of CheA1-233 with CheY (26), came to the same conclusion that an overlap exists between the motor switch and the P2 binding surfaces on CheY. Interestingly, another study indicated that a number of residues, showing chemical shift changes upon phosphorylation, lie in this region (24). Combining these results with the observations that phospho-CheY does not stay in complex with CheA (23) but binds to FliM with a higher affinity than apo-CheY (13), we propose that 1) the overlap region in the unphosphorylated state contributes to the CheA binding surface, 2) phosphorylation alters the topology of this overlap region, and 3) the phosphorylation-induced changes may be responsible for both CheA releasing CheY and CheY's increased affinity for the flagellar motor. This kind of overlap enables CheY to interact with several proteins with the specificity for each interaction being governed by its signaling state.
Although the region where our mutants clustered is similar to the region identified by the P2-CheY NMR study (26), some residues were unique to each study. In our study, the mutagenesis may not have been saturated, despite the recurrence of T112I, and since we screened for total non-chemotaxis, it is possible that we might have missed some mutants that had reduced binding and partial chemotaxis function. In the NMR study, chemical shift changes in the backbone amide residues were measured, and only those chemical shift changes greater than 60 Hz were considered significant. It is possible that some residues, indicated in this study, may affect the binding through their solvent-accessible side chains but do not result in ⌬Hz greater than 60. The side chain of Phe-111 is buried inside CheY's hydrophobic core (18) and may seem to be an exception, but NMR data show that Phe-111 shifts upon adding CheA(1-233), although the observed chemical shift change is less than 60 Hz. Since CheA has been shown to bind to CheY as a dimer (23), some of the residues we identified, which are not indicated by NMR studies, may be critical for CheA dimer binding.
The D13K mutant is a dominant tumbly mutation, also found to be defective in CheA binding. One possible explanation is that this mutant has acquired a conformation resembling phospho-CheY, and hence it is defective in CheA binding.
Welch et al. (13) found that it binds to FliM with higher affinity than the apo-CheY, and this is consistent with the possibility of a conformational similarity between this mutant and the phospho-CheY.
Our results identify mutations that can alter the CheY-CheA binding, and it is possible that some of our mutations may have introduced a structural change that sterically disrupted binding rather than remove an interaction contributing to the binding. Also, a mutation can alter CheA binding, either by specifically changing the interaction surface or by nonspecifically causing a change in folding. Since all of the mutants we characterized were overproduced in the stable form and can be phosphorylated by CheA (except D13K (data not shown)), the probability of any major folding defects occurring is reduced. Also, the crystal structures of T87I and Y106W 3 do not show altered folding (33). The most striking evidence is derived from the clustering of these mutations on one face of CheY molecule. We also mapped the positions of five cheY mutants not displaying altered CheA binding, and these mutations are not found on our proposed CheA binding face (green atoms, Fig. 4).
It is clear that the CheA binding face of CheY that our study identified lies distinct from the active site of the molecule. Recognition by CheA, away from the highly conserved active site of CheY (3), may be a way of acquiring specificity for this interaction. On the other hand, evidence for an overlap between CheA binding face and motor-switch binding face makes the structural aspects of CheY more interesting. While more work will be needed for the absolute determination of the region of overlap between the two faces, it will be interesting to know the region of CheY involved in the binding of CheZ, the only protein besides CheA and motor-switch proteins, that is known to interact with CheY (34). This will provide a better understanding of the structural aspects of CheY's activity and the regulation of chemotaxis.