Active site interference and asymmetric activation in the chemotaxis protein histidine kinase CheA.

The histidine protein kinase CheA is a multidomain protein that mediates stimulus-response coupling in bacterial chemotaxis. We have previously shown that the purified protein exhibits an equilibrium between inactive monomer and active dimer (Surette, M., Levit, M., Liu, Y., Lukat, G., Ninfa, E., Ninfa, A., and Stock, J. (1996) J. Biol. Chem.271, 939-945). We report here a study of the kinetics of phosphorylation of the isolated phosphoacceptor domain of CheA catalyzed by the isolated catalytic domain of the protein. The reaction fits Michaelis-Menten kinetics (Km = 0.26 mM for ATP and 0.10 mM for phosphoacceptor domain; kobs = 17 min−1). The catalytic domain exhibits the same equilibrium between inactive monomers and active dimers as the full-length CheA protein. Thus, CheA dimerization is an intrinsic property of this domain, independent of any other portion of the molecule and is required for its catalytic activity. In equimolar mixtures of full-length CheA and catalytic domain, homodimers and heterodimers are formed in equal concentration, indicating that all of the determinants for the dimerization are localized entirely on the catalytic domain. An analysis of the kinetics of phosphorylation catalyzed by CheA-catalytic domain heterodimers indicates half of the sites reactivity. The rate of CheA phosphorylation within this heterodimer is over 5-fold greater than that observed in CheA homodimers. The dramatic increase in activity within this asymmetric dimer raises the possibility that CheA activation by receptors involves a mechanism that directs catalysis to one active site while preventing interference from the other.

Bacterial signal transduction networks that regulate motility and gene expression generally involve two central enzymatic components, histidine protein kinases and phosphoaccepting response regulators (1,2). Sensory information controls the rates of kinase autophosphorylation at specific histidine residues. These phosphoryl groups are subsequently transferred to a specific aspartate residue in the response regulator proteins, causing a conformational change in the regulator that leads to the generation of a response. The most intensively investigated signaling network of this type is the system that mediates chemotaxis responses in Escherichia coli and Salmonella typhimurium (3). The chemotaxis histidine kinase, CheA, and response regulator, CheY, that mediate chemotaxis in these species are found in all motile eubacterial and archaebacterial strains that have been examined and are clearly specialized for sensory motor regulation.
The CheA protein is composed of at least four functionally and structurally distinct domains (Fig. 1). An N-terminal phosphoaccepting domain of approximately 130 residues that contains the site of histidine autophosphorylation, termed the P1 or H domain, is connected to a domain of approximately 70 residues that binds CheY, termed the P2 or Y domain, which is in turn linked to a catalytic, C, domain of approximately 250 residues that binds ATP and catalyzes the phosphorylation of the N-terminal H domain. Finally, a distinct C-terminal regulatory, R, domain functions to mediate regulatory interactions between CheA and membrane receptor-transducer proteins. The H and Y domains have been isolated as independently folding units, and their structures have been determined by NMR methods (4 -7). Both are globular. The H domain is essentially an ␣-helical bundle with the phosphoaccepting histidine side chain extending from the solvent-exposed surface of one helix. The Y domain is an open faced ␣/␤ sandwich. The H and Y domains are attached to one another and to the C domain by flexible linker sequences, and it is apparent that the two globular domains are free to move in solution like balls on a chain. No structural information is available concerning the C and R domains, although it has been shown that the R domain can be cleaved away without dramatically affecting the activity of the CheA catalytic domain (8).
Purified CheA is in equilibrium between an inactive monomeric form and an active homodimer (K D ϭ 0.3 Ϯ 0.1 M) (9). Phosphorylation can occur in trans within the dimer, with the C domain of one subunit catalyzing the phosphorylation of the H domain of the other subunit (10,11). It has not been determined whether the inactivity of CheA monomers is due to the lack of a phosphoaccepting H domain in trans or due to an intrinsic requirement of C domain dimerization for catalytic activity.
To better understand the interactions of the H and C domains within CheA dimers we have undertaken a detailed analysis of the kinetics of phosphorylation of a fragment of CheA composed solely of H domain by a fragment of CheA that contains only the C domain. The results indicate that the C domain exists in an equilibrium between an active dimer and an inactive monomer with a K D essentially identical to the K D for dissociation of intact CheA dimers. Thus, the dimerization of CheA depends entirely on dimerization of the C domain, and C domain dimerization is required for kinase activity. An analysis of the kinase activities of heterodimers of CheA in association with fragments of CheA that contain the C domain indicates that CheA dimers exhibit half of the sites reactivity. Apparently, the interaction of an H domain with one of the active sites of a catalytic domain dimer precludes interaction of a second H domain with the other active site.

MATERIALS AND METHODS
Strains and Plasmids-Derivatives of pQE12 (Qiagen) carrying insertions of cheA gene fragments encoding residues 1-138 (H domain) and 277-524 (C domain) flanked by BamHI and BglII sites were constructed by cloning polymerase chain reaction products amplified from the S. typhimurium cheA gene in pM04 (12). The primers used were primer Ia (5Ј-TTTGGATCCATGGATATTAGCGATTTT-3Ј) and primer Ib (5Ј-CCCAGATCTGACCGCCGGCGTGGTTTC-3Ј), and primer IIa (5Ј-TTTGGATCCGAGTCCACCAGCATTCGC-3Ј) and primer IIb (5Ј-CCCAGATCTCAGCGGCAGCAGAATACG-3Ј) for H domain-and C domain-encoding plasmids, respectively. The plasmids were transformed into E. coli strain M15/pREP4 (Qiagen). The final clones, pHD and pCD for the H domain and the C domain constructs, respectively, were confirmed by sequencing.
Expression and Purification of H Domain-M15/pREP4/pHD cells were grown at 37°C in 6 liters of LB, 100 g/ml ampicillin, 25 g/ml kanamycin to an A 600 of approximately 0.80. H domain expression was then induced by the addition of isopropyl-␤-D-thiogalactopyranoside to a final concentration of 1.0 mM. Incubation was continued for an additional 6 h, and cells were harvested, washed with 100 mM potassium phosphate, pH 7.0, resuspended in 100 mM potassium phosphate, 1.0 mM EDTA, 2.0 mM 1,4-dithiothreitol, 1.0 mM phenylmethylsulfonyl fluoride, 10% glycerol, pH 7.5, and frozen. After thawing, cells were sonicated, and the lysate was cleared by centrifugation at 105,000 ϫ g for 2 h. Ammonium sulfate was added to the supernatant to 37% of saturation. The pellet was dissolved in 50 mM sodium phosphate, 300 mM NaCl, pH 8.0, and loaded onto a 10-ml Ni-NTA column (Qiagen) equilibrated in the same buffer. After washing the column with this buffer, it was washed with 50 mM sodium phosphate, 300 mM NaCl, 10% glycerol, pH 6.0. The H domain was then eluted with a gradient of 0 -0.50 M imidazole in this buffer. Fractions that contained H domain were dialyzed against 50 mM sodium phosphate, 0.10 mM EDTA, 0.50 mM 1,4-dithiothreitol, pH 7.5, and loaded onto a 25-ml DEAE-Sephacel (Pharmacia) column equilibrated in the same buffer. The column was washed with buffer and eluted with a 150 ml 0 -0.50 M NaCl gradient. Fractions with H domain were dialyzed against 25 mM Tris-Cl, 25 mM NaCl, and 0.10 mM 1,4-dithiothreitol, pH 7.5, concentrated in a Centriprep-3 (Amicon), and stored at Ϫ20°C.
Expression and Purification of C Domain-M15/pREP4/pCD cells were grown and lysed essentially as described above. The lysate was dialyzed once against 5.0 mM sodium phosphate, 1.0 mM EDTA, 1.0 mM 1,4-dithiothreitol, pH 8.0, and loaded onto a 90-ml DEAE-650 M (Toyopearl) column equilibrated in 10 mM sodium phosphate, 1.0 mM EDTA, 1.0 mM 1,4-dithiothreitol, pH 8.0. The flow-through containing all of the C domain was dialyzed against 50 mM sodium phosphate, 300 mM NaCl and chromatographed on a Ni-NTA column as described above. The fractions of pure C domain were dialyzed twice against 50 mM Tris-Cl, 1.0 mM 1,4-dithiothreitol, 1.0 mM EDTA, pH 7.5, and finally against the same buffer with 0.10 mM EDTA. The protein was concentrated in a Centriprep-10 (Amicon) and stored at Ϫ20°C.
Kinase Kinetics-Rates of ATP hydrolysis were measured in 50 mM Tris-Cl, 100 mM potassium glutamate, 5.0 mM MgCl 2 , pH 7.5, at 23°C using the spectroscopic pyruvate kinase/lactate dehydrogenase coupled assay for ATPase activity as described previously (9). Molecular activities of CheA or C domain are expressed as apparent rate constants k obs (min Ϫ1 ), which correspond to the number of molecules of product (ADP or phospho-H domain) produced per min per monomer of CheA or C domain. In reactions catalyzed by heterodimers composed of CheA and C domain or of CheA and CheA S , k obs is calculated per dimer. The kinetics of H domain phosphorylation were also determined at 23°C in the same buffer used for the coupled ATPase assay with the indicated additions of [␥-32 P]ATP (Amersham Corp.) and other components. Reactions were initiated by the addition of ATP. At intervals, 10-l aliquots were mixed with 10 l of 0.010% SDS plus 100 mM EDTA, and aliquots of this mixture containing not more than 15 g of protein were spotted onto polyvinylidene difluoride membranes using a Schleicher & Schuell dot-blot filtration manifold. Membranes were washed with 20 mM Tris-Cl, 150 mM NaCl, 5.0 mM sodium pyrophosphate, 5.0% methanol, pH 8.0, and dried. Spots were excised, and levels of radioactivity were assayed in a liquid scintillation spectrometer.

RESULTS
Phosphoaccepting Activity of H Domain--It has previously been shown that a fragment of CheA containing only the C domain is able to phosphorylate a fragment containing only the H domain (13). This activity is at least 10-fold greater than the rate of H domain phosphorylation catalyzed by the intact CheA protein or by a naturally occurring variant of CheA termed CheA S produced from an alternative site of translational initiation corresponding to Met 98 in the full-length CheA protein (Fig. 2). Presumably, the H and/or Y domains in CheA and CheA S interfere with the ability of the free H domain to interact with the active site in the catalytic domain. As has been observed previously (13), the phospho-H domain acts as an efficient CheY phosphodonor in the complete absence of the CheY binding domain (Fig. 2, compare lanes 6 and 8).
At concentrations of purified C domain in the micromolar range the dependence of the rate of phosphorylation on H domain concentration exhibited Michaelis-Menten kinetics with an apparent K m for H domain of approximately 100 M (Fig. 3). The fact that the x axis intercept in the inset to Fig. 3, which corresponds to ϪK m , is the same at saturating and subsaturating ATP concentrations indicates that the affinity of C domain for H domain is unaffected by ATP binding. At saturating ATP, the estimated k obs for H domain phosphorylation is approximately 17 min Ϫ1 . This is almost twice the k obs for CheA autophosphorylation, approximately 9.0 min Ϫ1 , ob- tained under the same assay conditions (9). The apparent K m for ATP in the phosphorylation of H domain by C domain is approximately 0.26 mM, close to the K m of 0.33 mM obtained with CheA under similar conditions (9). This value is independent of the concentration of H domain (Fig. 4, inset).
These quantitative measures of rates of H domain phosphorylation were obtained in a coupled assay under steady state conditions where H domain phosphorylation was rate-limiting (i.e. sufficient CheY was added, 25 M, to ensure a relatively rapid dephosphorylation of phosphohistidine groups and subsequent hydrolysis of phospho-CheY). In the absence of CheY, the initial rate of ATP hydrolysis was the same as the steady state rate obtained in its presence (data not shown). In parallel experiments it was also shown that the initial rate of H domain phosphorylation measured by assaying production of 32 P-la-beled H domain directly was the same as the ATPase activity measured in the presence of CheY. To accurately assay the rate using radiolabeled ATP, it is necessary to directly measure the specific activity of [␥-32 P]ATP. Using thin layer chromatography (9), we have found that the portion of [␥-32 P]ATP in our commercially obtained material is generally only about 50% of the total radioactivity present. This type of contamination could explain previously reported differences between CheA autophosphorylation rates measured by 32 P labeling in the absence of CheY and ATPase activity measured in the presence of CheY (14).
C Domain Dimerization-At micromolar concentrations, purified C domain eluted during molecular sieve chromatography with an apparent molecular weight of 58,000. This is the value predicted for a globular dimer. Cross-linking studies performed at different concentrations of C domain indicated a K D of approximately 0.21 M (Fig. 5), essentially the same as that exhibited by homodimers of full-length CheA (9). This result indicates that the determinants for CheA dimerization are localized entirely to the C domain. If this were the case one would expect that in a mixture of equal concentrations of C domain and CheA there should be equal amounts of CheA homodimers, C domain homodimers, and CheA-C domain heterodimers. This prediction was confirmed by size exclusion chromatography of an equimolar mixture of CheA and C domain, where it was shown that 36% of the C domain eluted as a heterodimer with CheA while 64% eluted at a position corresponding to the C domain homodimer (Fig. 6).
Effect of C Domain Dimerization on Kinase Activity-We have previously shown that the autophosphorylation activity of CheA is completely dependent on its dimerization. One possible explanation for this is the requirement for an H domain in trans. Another possibility is that the C domain must be dimeric to be active. To test for the dependence of C domain dimerization on kinase activity, rates of H domain phosphorylation were examined as a function of C domain concentration. The kinase activity of the isolated C domain toward the H domain exhibits a concentration dependence in the submicromolar range that is consistent with an equilibrium between inactive monomers and active dimers. The apparent K D for C domain dissociation es- timated from the dependence of H domain phosphorylation activity on C domain concentration, 0.17 M, was essentially the same as the value estimated for the K D of C domain homodimers estimated from cross-linking studies, 0.21 M (compare Figs. 5 and 7). Thus, the catalytic activity of the C domain depends on its state of dimerization. This explains the requirement of dimerization for full-length CheA activity but does not exclude the possibility that there is also a requirement for H domain in trans. When rates of phosphorylation by CheA-C domain heterodimers were examined, however, dramatically different results were obtained.

Negative Cooperativity between Free H Domain Phosphorylation and Linked H Domain Autophosphorylation in Heterodimers Composed of CheA and C Domain-
When increasing concentrations of CheA where added to a fixed concentration of C domain, a dramatic inhibition of H domain phosphorylation was observed (Fig. 8). This decrease in H domain phosphorylation is presumed to be caused by the partitioning of the C domain from active C domain homodimers into relatively inactive CheA-C domain heterodimers. The solid curve in Fig. 8 shows the decrease in activity that would be predicted from this effect assuming the heterodimer to be completely inactive. From this result we conclude that the CheA-C domain heterodimer, like CheA homodimer, has a very low activity toward free H domain. Thus, the presence of the H and Y domains on one subunit is sufficient to effectively inhibit the interaction of both potential kinase active sites with free H domain. Moreover, in a complementary experiment high concentrations of H domain (400 M) caused a 30% inhibition of the rate of CheA phosphorylation in CheA-C domain heterodimers (Fig. 9). Free H domain had a much smaller inhibitory effect on CheA homodimer autophosphorylation, probably because the associated H domains compete with free H domain for accessibility at the site. These results indicate that in CheA dimers there is negative cooperativity between active sites. In other words, occupancy of one active site within a dimer by a substrate H domain precludes occupancy of the other active site.
The effect of increasing concentrations of C domain on the rate of CheA autophosphorylation was examined in the absence of free H domain (Fig. 10). The results indicate a dramatic stimulation due to the formation of CheA-C domain heterodimers with over 5-fold higher autophosphorylation activity than CheA homodimers. Similar results were obtained by adding increasing concentrations of CheA to a fixed concentration of C domain (Fig. 11). When the latter experiment was performed with a fixed concentration of CheA S , however, the results indicated that the CheA-CheA S heterodimers had only about a 20% greater autophosphorylation activity than CheA Aliquots of the fractions were subjected to 15% SDS-polyacrylamide gel electrophoresis followed by transfer to nitrocellulose (0.22 m, MSI). Proteins were detected by Western blotting using rabbit polysera against CheA. 125 I-labeled goat anti-rabbit antiserum (DuPont NEN) was used as secondary antibody, and the blots were quantitated using a Molecular Dynamics PhosphorImager. C domain that eluted in the time interval 8.25-9.25 min was assumed to be from C domain-CheA heterodimers, and C domain that eluted in the time interval 9.25-10.5 min was assumed to be from C domain homodimers. homodimers. This result is consistent with the observation that CheA S homodimers have an activity toward free H domain similar to that of CheA homodimers (Fig. 2). DISCUSSION The histidine kinase CheA functions in bacterial chemotaxis to integrate sensory receptor inputs and relay this information to response regulators in the cytoplasm. The protein occupies a central position in the chemotaxis signal transduction network, and it has been shown that CheA can interact directly with at least four other Che proteins, CheW, CheY, CheB, and CheZ, as well as with chemotaxis receptors such as Tar and Tsr (15)(16)(17)(18)(19). The complex regulatory functions of CheA are reflected in the complex multidomain structure of the protein, with the central histidine kinase or C domain being flanked by domains responsible for mediating interactions between the kinase domain and other signal transduction components (3,20). In order to better understand the mechanisms that regulate this core kinase activity, we have undertaken a detailed kinetic analysis of its activity in the absence of flanking regulatory regions. Using high concentrations of the isolated H domain as a phosphoaccepting substrate, we have shown that kinase activity fits Michaelis-Menten kinetics for a random order mechanism with apparent K m values for H domain and MgATP of approximately 0.10 and 0.26 mM, respectively and a k obs of 17 min Ϫ1 . The K m for MgATP is very close to that obtained for the autophosphorylation of CheA, and the k obs for the isolated C domain is almost 2-fold higher then that obtained with the CheA protein.
We have previously shown that the intact kinase CheA, is in equilibrium between an inactive monomer and an active dimer with a K D of 0.3 Ϯ 0.1 M (9). It has previously been shown that in CheA and homologous histidine kinases the phosphotransfer reaction can occur in trans with the catalytic domain of one subunit within a dimer catalyzing the phosphorylation of a histidine in the other subunit (10, 21, 22). The inactivity of monomeric CheA could therefore be attributed solely to an inability of monomeric CheA to autophosphorylate its associated phosphoaccepting domain. We show here, however, that the requirement for CheA dimerization is an intrinsic property of the C domain alone, independent of any requirement for transphosphorylation of associated H domains. Moreover, all of the determinants for dimerization seem to reside within the C domain, since the K D values for C domain dimer dissociation are essentially identical to the dissociation constants of CheA dimers or of CheA-C domain heterodimers. From the activities of heterodimers of different mutant CheA proteins it has previously been proposed that residues from both subunits within a dimer participate in formation of the active site (23). The dimerization requirement for CheA activity is consistent with this hypothesis.
The k obs for CheA homodimer autophosphorylation is 9.0 min Ϫ1 , the value obtained for the C domain homodimer is 17 min Ϫ1 , and that for the CheA-C domain heterodimer is approximately 60 min Ϫ1 . The relatively low activity of CheA homodimers is presumably due to interference between the two associated H domains. The associated CheY binding and regulatory (Y and R) domains could contribute to this interference. CheA-CheA S heterodimers exhibit a k obs of 12 min Ϫ1 . Thus, despite the fact that more than half of the H domain is missing in CheA S , the remaining H domain and the intact Y and R domains still interfere with kinase activity to almost the same extent as in CheA homodimers. The dramatically higher rate observed with the heterodimer compared with the C domain homodimer indicates that the associated H domain in CheA is better positioned to serve as a substrate than free H domain.
Assuming that CheA dimers are symmetric, there must be two active sites/dimer. The fact that C domain activity follows Michaelis-Menten kinetics indicates that the two sites are independent or that the protein exhibits half of the sites reactivity. The apparent interference between the two H domains in CheA homodimers suggests that H domain binding at one site may preclude the binding and phosphorylation of an H domain at the opposing site, i.e. CheA dimers may exhibit half of the sites reactivity. This mechanism is supported both by the fact that H domain is poorly phosphorylated by the heterodimer and by the inhibitory effects of high concentrations of H domain on heterodimer autophosphorylation. It has previously been shown that both H domains within a CheA homodimer are phosphorylated with kinetics that fit a single exponential function (14). This is not inconsistent with negative cooperativity or half of the sites reactivity, however, since the two sites could act sequentially to phosphorylate the two associated H domains.
CheA functions in chemotaxis in a ternary complex with CheW and receptor, and within this complex CheA autophosphorylation rates can be increased over 100-fold compared with values obtained with pure CheA under comparable conditions (24). One of our primary goals in studying the kinetics of purified CheA is to understand the mechanism of this activation. In evaluating the rates of various CheA constructs compared with rates obtained with CheA-CheW-receptor complexes it is necessary to compare absolute rates rather than -fold activation. In experiments where activation has been measured within complexes, the reported degree of activation depends both on CheA activity within the complex, which is generally optimized, and on the activity of pure CheA under the same conditions, which is generally far from optimal. Measurements of CheA autophosphorylation under conditions of CheA-CheW-receptor complex formation give values for k obs of up to 70 min Ϫ1 (25). This value is close to the k obs of approximately 60 min Ϫ1 that we have obtained for autophosphorylation within the C domain-CheA heterodimer. Because of incomplete complex formation, the k obs of 70 min Ϫ1 is undoubtedly an underestimate. Nevertheless, the relatively high rate of autophosphorylation within the heterodimer suggests that the mechanism of CheA activation within the ternary complex may involve an asymmetric mechanism that restricts the relative positioning of the domains associated with the C domain and/or causes a conformational change in the C domain to favor catalysis by one active site without interference from the other.