Self-assembly of receptor/signaling complexes in bacterial chemotaxis

Abstract

Escherichia coli chemotaxis is mediated by membrane receptor/histidine kinase signaling complexes. Fusing the cytoplasmic domain of the aspartate receptor, Tar, to a leucine zipper dimerization domain produces a hybrid, lzTar_C, that forms soluble complexes with CheA and CheW. The three-dimensional reconstruction of these complexes was different from that anticipated based solely on structures of the isolated components. We found that analogous complexes self-assembled with a monomeric cytoplasmic domain fragment of the serine receptor without the leucine zipper dimerization domain. These complexes have essentially the same size, composition, and architecture as those formed from lzTar_C. Thus, the organization of these receptor/signaling complexes is determined by conserved interactions between the constituent chemotaxis proteins and may represent the active form in vivo. To understand this structure in its cellular context, we propose a model involving parallel membrane segments in receptor-mediated CheA activation in vivo.

The Escherichia coli chemotaxis system provides a paradigm for understanding general principles of signal transduction and membrane receptor function. In E. coli, chemotaxis responses are mediated by changes in protein phosphorylation associated with the binding of ligands to membrane receptors at the cell surface. We are interested in defining the key protein–protein interactions that mediate signal transduction between the receptors and the protein kinase CheA, which modulates motor responses within the cytoplasm. Considerable evidence supports a model where CheA activity is regulated within large multimeric complexes formed between CheA, the α-helical cytoplasmic domain of the chemotaxis receptors, and an SH3 (Src homology 3)-like β-barrel protein, CheW. These components and their interactions are highly conserved among virtually all chemotactic bacterial and archaeal species (1, 2).

Atomic resolution structures of the cytoplasmic domain of the serine receptor (Tsr_C) (3), the cytoplasmic domain of another receptor, MCP_1143C (4), CheA (4–7), and CheW (4, 8) have been determined by x-ray crystallographic and NMR methods. Crystals of Tsr_C and MCP_1143C both contain long, antiparallel α-helical coiled-coil monomers that come together in parallel to form long, dimeric four-helix bundles (3, 4). The cytoplasmic domain of the aspartate receptor (Tar_C), which is 75% identical to Tsr_C, is predominantly an α-helical monomer in solution (9, 10).

The periplasmic aspartate-binding domain of the aspartate receptor (Tar) is dimeric both in solution and in crystals (11, 12), and the corresponding serine-binding domain of the serine receptor (Tsr) appears to have a similar structure (13). Cross-linking studies of Tar and Tsr in membranes are consistent with a dimeric organization, although recent results indicate that higher order assemblies are probably involved (14–17). Immuno-EM, immunofluorescence, and GFP-tagging studies have established that, within a cell, receptors, CheA, and CheW cluster together into large complexes that may contain thousands of subunits of each protein (18, 19). Formation of these assemblies depends on the presence of all three components. Complexes of Tar or Tsr, CheA, and CheW have also been formed from purified components in vitro (20, 21). The complexes are characterized by >100-fold increases in CheA-kinase activity that are subject to inhibition by addition of aspartate or serine (22, 23).

Soluble fragments of Tar_C or Tsr_C together with CheW have been shown to activate CheA (10, 24, 25). In the case of Tar_C, activation was greatly facilitated by the attachment of a leucine zipper dimerization domain to the Tar_C N terminus. The resulting lzTar_C hybrids formed mixtures of dimers and tetramers in solution (10, 25). Complexes of CheA and CheW with lzTar_C have been imaged by using transmission EM (TEM) (26–28). The structure of this assembly did not fit preconceived notions of receptor organization. It had generally been assumed that receptor dimers interacted with CheW monomers and CheA dimers to form 2:2:2 complexes (10, 29). The lzTar_C complexes are much larger structures, however, with a monomer subunit composition of ≈24 lzTar_C:6 CheW:4 CheA. The 12 lzTar_C four-helix bundles in the complex exhibited an unexpected barrel-like arrangement, with two sets of 6 parallel four-helix bundles emerging from a central region where CheA and CheW are bound and the leucine zipper domains at either end (27, 28).

The arrangement of lzTar_C within these large assemblies (28) is different from the “trimer of dimers” arrangement of Tsr_C (3) or the “hedgerow of dimers” arrangement of MCP_1143C (4) that is seen in crystals in which the signaling domain hairpin regions are closely associated in the absence of CheW and CheA. As seen in these crystals, the unit of lzTar_C structure appears to be a four-helix bundle; however, in receptor/signaling complexes, the bundles splay apart from each other in the signaling domain hairpin region where CheW and CheA bind.

To rule out the possibility that the structure of lzTar_C/CheW/CheA (lzTar_CWA) complexes are a reflection of some special property of the lzTar_C hybrid receptor fragment, we have examined complexes formed with a Tsr_C protein that is essentially the same fragment whose structure has been determined by x-ray crystallography. Tsr_C does not have any dimerization domain or other elements that do not occur in the full-length Tsr receptor. We show here that although the Tsr_C protein is predominantly a monomer in solution, it is still capable of self-assembly into complexes with CheW and CheA. Moreover, these Tsr_C/CheW/CheA (Tsr_CWA) complexes have a similar size, shape, and organization to those formed with lzTar_C. Our results suggest that the structures of these ternary complexes may reflect inherent features of the interactions of these components in functional chemotaxis receptor/signaling complexes in vivo.

Results

The lzTar_C and Tsr_C proteins were compared by using gel filtration chromatography with inline multiangle light scattering to measure the absolute mass of protein eluents (Fig. 1). lzTar_C exhibits two elution peaks, one with a molecular mass corresponding to a dimer and the other with the expected mass of a tetramer. In contrast, Tsr_C eluted as a single peak with a mass corresponding to that of the monomer. Both lzTar_C and Tsr_C eluted with retention times corresponding to globular proteins with about twice the actual molecular mass, consistent with the extended α-helical conformation observed in crystals of Tsr_C (3) and MCP_1143C (4).

Fig. 1.

HPLC and multiangle light-scattering analysis of isolated lzTar_C and Tsr_C proteins. The left axis shows the molar mass of lzTar_C tetramers and dimers (○) and Tsr_C monomers (×). The right axis shows the absorbance at 280 nm of lzTar_C (solid line) and Tsr_C (dashed line). lzTar_C dimers have a predicted mass of 74.0 kDa; Tsr_C monomers are predicted to be 27.9 kDa.

Despite its monomeric state in solution, in the presence of CheW, Tsr_C activates CheA as effectively as lzTar_C. Whereas the specific activity of isolated CheA in solution under these conditions is 0.11 s⁻¹, the maximal activity in complexes with Tsr_C and CheW is 10.3 s⁻¹, which gives an activation factor of 94-fold, nearly identical to the 100-fold activation for CheA in complexes with lzTar_C and CheW (27). As reported previously with lzTar_C and another fragment of Tsr_C, this activation is highly dependent on receptor concentration (24, 26). For CheA activation as a function of receptor concentration, we observe a Hill coefficient of 1.7 and a midpoint of 40 μM for lzTar_C, as compared with a Hill coefficient of 3.5 and a midpoint of 54 μM for Tsr_C (data not shown).

Under conditions that are optimal for kinase activation, receptor/signaling complexes formed with lzTar_C or Tsr_C have a very similar hydrodynamic radius and protein composition as measured by gel filtration chromatography (Fig. 2). The active complexes were isolated by HPLC and denatured, and then their components were quantified in the presence of 6 M guanidine-HCl. This analysis indicated that lzTar_CWA and Tsr_CWA complexes had very similar subunit compositions (Fig. 2 Insets and Table 1). These subunit ratios are in good agreement with previous investigations, which used analysis of bands on SDS/PAGE gels and light scattering (26) or a three-dimensional EM reconstruction and fitting of molecular structures (28) to estimate the subunit composition. In both cases, four CheA monomers per complex were found, and we used this number as a known value in our analysis. In Table 1, we present the stoichiometry in terms of the number of CheW and receptor monomers per CheA monomer. We define this composition of one CheA subunit plus the listed number of CheW and receptor subunits as a “stoichiometric unit.” Using these calculated stoichiometries and the absolute molecular mass determined by multiangle light scattering, we find that the Tsr_CWA complexes each contain ≈4.2 stoichiometric units (Table 1) when the calculations are constrained to give 4.0 stoichiometric units per lzTar_CWA complex (data not shown).

Fig. 2.

HPLC two-dimensional gel filtration analysis of soluble receptor/signaling complexes. (A) lzTar_CWA. (B) Tsr_CWA. Complexes were purified by HPLC using a TSK-Gel G5000PW_XL column, and fractions were analyzed by using SDS/PAGE gels (not shown) to confirm that CheA, CheW, and lzTar_C or Tsr_C coelute in the leading peak at ≈14.5 min. Free CheA, lzTar_C, and CheW elute at ≈17.5, 19.2, and 20.4 min, respectively. Insets in A and B show chromatograms corresponding to the second dimension of analysis. Peak fractions containing purified complexes were run on a GF-250 column in 6 M guanidine-HCl. The CheA, receptor, and CheW peaks are resolved and were used to calculate the relative stoichiometry (Table 1).

Table 1.

Stoichiometry of receptor/signaling complexes

Components in active receptor/signaling complex	Subunit ratio of receptor:CheA	Subunit ratio of CheW:CheA	rms radius, nm	Mass of complex, MDa	Number of stoichiometric units
lzTarC/CheW/CheA	6.15 ± 0.10	1.02 ± 0.07	16.5	1.27	4.0
TsrC/CheW/CheA	5.31 ± 0.04	1.07 ± 0.07	15.3	1.00	4.2

We measured the stoichiometries for both varieties of complexes by using a two-dimensional HPLC analysis (see Fig. 2). The numbers are derived from the averages and standard deviations from 12 measurements (2 measurements each on two fractions from three separate HPLC separations). The results for the mass, rms radius, and number of stoichiometric units represent the average from two independent multiangle light-scattering data sets.

The structural similarity between complexes formed with either Tsr_C or lzTar_C was further confirmed by TEM. The Tsr_CWA and lzTar_CWA complexes have the same overall shape and organization (Fig. 3). Both are elongated and have higher density at the middle and ends. The radius of gyration for the lzTar_CWA complex is 13.0 nm when calculated from the three-dimensional reconstruction (28, 30). Comparison with the values in Table 1 suggests that the complex may have a more open or extended structure in solution as opposed to when it is fixed in uranyl acetate stain.

Fig. 3.

TEM analysis of receptor/signaling complexes. (A) TEM image of a field of lzTar_CWA complexes. (Scale bar, 100 nm.) (B) Two-dimensional average of 4,900 images of individual lzTar_CWA complexes (Upper) and the variance among those images (Lower). (Scale bar, 42 nm.) (C) TEM image of a field of Tsr_CWA complexes. (D) Two-dimensional average of 10,583 images of individual Tsr_CWA complexes (Upper) and the variance among those images (Lower). Both structures appear very similar in overall shape and organization, but the average and variance of the Tsr_CWA images have higher resolution features, suggesting that these complexes are somewhat more regular in structure.

At a concentration of 5 μM, lzTar_C forms few or no complexes in the presence of 10 μM CheA and CheW (10, 26), and a smaller Tsr cytoplasmic fragment that is similar to Tsr_C has been reported not to activate CheA at this concentration (24). Using the coupled system assay, we detected no activation of CheA by 5 μM Tsr_C. To confirm that no complexes form, a 5 μM concentration of either lzTar_C or Tsr_C was mixed with 10 μM CheA and CheW in RBIII buffer (see Materials and Methods for buffer composition) and analyzed by HPLC. As expected, no signal corresponding to the soluble complexes was observed at the elution time of ≈14.5 min. Running the corresponding fractions on the gel, no protein was detectable by Coomassie staining.

To demonstrate that lzTar_C and Tsr_C are functionally interchangeable in the soluble complexes, we attempted to form complexes that contain a mix of the two. For these experiments, we relied on the fact that no complexes formed with a 5 μM concentration of receptor. If 5 μM lzTar_C or Tsr_C was added to a 45 μM concentration of the other and incubated with 10 μM CheA and CheW, complexes formed and CheA was activated. This activation was very similar to complexes formed with a 50 μM concentration of either lzTar_C or Tsr_C. Both types of receptor fragments coelute in the fractions that contain the receptor/signaling complexes and are readily visible on a Coomassie-stained gel of the corresponding fractions (Fig. 4), suggesting that individual complexes contained a mixture of lzTar_C and Tsr_C subunits.

Fig. 4.

HPLC and SDS/PAGE analysis of mixed receptor/signaling complexes. (A) Complexes containing 45 μM lzTar_C and 5 μM Tsr_C. (B) Complexes containing 45 μM Tsr_C and 5 μM lzTar_C. Gray lines indicate the breadth of each HPLC fraction that was run in the corresponding lane on the gel.

Discussion

We have characterized a fragment of Tsr, Tsr_C, that corresponds closely to the fragment whose structure was determined by x-ray crystallographic methods (3). The crystallized fragment contained Tsr residues 286–526, whereas our fragment contains residues 290–551. In the crystal structure of Tsr_C, a trimer of dimers was observed; in solution, however, we see that the Tsr_C protein is a monomer. The hydrodynamic properties of the Tsr_C monomers are consistent with the extended, antiparallel, coiled-coil α-helical monomeric unit seen in Tsr_C crystals and in MCP_1143C (4) crystals. These monomers presumably come together during the crystallization process to form the assemblies of dimers seen in the crystal structures.

Immuno-EM labeling and three-dimensional reconstruction of lzTar_CWA complexes previously established that CheA and CheW are in the central region of the barrel-like receptor/signaling complexes (27, 28), whereas the N and C termini of the receptors are at the ends. Examination of the corresponding two-dimensional average of the TEM images of the Tsr_CWA complexes suggests an essentially identical overall organization (Fig. 3). The length of the Tsr_CWA complex is somewhat shorter, consistent with the lack of density at the ends where the leucine zippers are present in the lzTar_CWA complexes. The Tsr_CWA complexes appear to be better ordered overall, with a more structured average and variance map.

Our finding of 4.2 stoichiometric units versus 4.0 per receptor/signaling complex is probably not a significant difference, given the uncertainties in the calculation. We thus conclude that both types of complexes contain four CheA subunits and four CheW subunits per complex. The difference in the number of receptor subunits (Table 1) is outside of the range of variation observed and suggests that the Tsr_CWA complexes have, on average, 21 or 22 receptor subunits, rather than the 24 we observe in the lzTar_CWA complexes. This result is consistent with the lower stability we observe for the Tsr_CWA complexes when they are diluted or otherwise perturbed, and it suggests that the Tsr_C receptor fragments are in equilibrium between the monomer form seen in solution (Fig. 1) and a dimer or multimer in the complex. In contrast, because lzTar_C is always dimeric, a full complement of receptor fragments is always observed in the complex.

We have previously shown that, in lzTar_CWA complexes, the arrangement of dimeric four-helix bundles differs substantially from what is seen in the crystal structure of Tsr_C (28). Here, we show that analogous receptor/signaling complexes formed with Tsr_C have essentially the same structure as those formed with lzTar_C. The organization and shape of these complexes therefore appears to be determined by the nature of the conserved interactions among the receptor signaling domains, CheW, and CheA rather than by any particular property of the soluble receptor fragments that are used. This structure is also consistent with a recent analysis of the in vivo FRET data. The analysis suggests that the receptors form isolated, strongly coupled clusters rather than interacting with many receptors in an extended array (31). In addition, the stoichiometries in the soluble receptor/signaling complexes are consistent with the recent crystal structure of the CheA–CheW complex elucidated by Park et al. (4) and are similar to the stoichiometries proposed in their model of the receptor/signaling complexes.

Previously reported in vivo cross-linking studies are consistent with the clusters of dimers that are seen in the lzTar_CWA complexes (16, 17), and EM analyses of cell sections have shown evidence of membrane folds at the cell poles that could support the type of receptor signaling domain-to-signaling domain juxtaposition seen in our soluble receptor/signaling complexes (32). Given that the sequences of the receptor cytoplasmic domains, CheA, and CheW are highly conserved in virtually all chemotactic prokaryotes, it seems that this organization of the receptor complexes, a tight complex with two sets of signaling domains oriented toward one another, could reflect the structure of active receptor/CheW/CheA complexes in vivo.

Within the higher order assemblies of membrane receptors, CheW, and CheA that mediate chemotaxis in vivo, only a small fraction of CheA is in the activated state. Estimates of the cellular stoichiometry in vivo are ≈1.2 CheY proteins per CheA monomer (33). Previous studies have estimated that, in wild-type cells under steady-state conditions, ≈30% of CheY is phosphorylated (34). The specific activity of activated CheA in vitro is much higher than the in vitro rate of CheZ-mediated CheY dephosphorylation (21, 35). Although accurate measurements of the latter are difficult, it is almost certain that specific activity of activated CheA is at least 10-fold higher than that of CheZ. Based on these values, at most 4% of the CheA in the cell is fully activated by participation in a receptor/signaling complex. Fig. 5 presents a schematic model with only a small fraction of receptors and CheA participating in a complex that contains the receptor signaling domains from two opposing segments of membrane. In other regions, CheA and CheW may be bound to the receptors, but CheA is not activated. This schematic is similar in appearance to some of the membrane invaginations that have been observed in sections of wild-type cells by TEM (32).

Fig. 5.

A model for membrane receptor/signaling complexes in vivo in the receptor cluster at the pole of an E. coli cell. Receptors transition from a nonactivating (R) to a CheA-activating (T) complex through an invagination of the membrane that permits the formation of complexes similar to those seen for the soluble receptor/signaling complexes (Fig. 3). Receptors in the R state also participate in signaling by serving as substrates for ligand-specific methylation and demethylation.

The response time of E. coli cells to stimuli is generally found to be ≈50–100 ms (36, 37). Although large changes in membrane architecture may not occur on this time scale, local movements could permit additional CheA and receptor proteins near the edge of an active receptor/signaling complex to join into it and thus increase the net CheA activity. Additionally, conformational changes in the receptors seem to be able to inactivate CheA without requiring dissociation (24, 29, 38). The receptors outside the active receptor/signaling complexes may still play an important role in cellular memory and adaptation, because these receptors would be substrates for attractant-mediated methylation and demethylation by CheR and CheB (39–42). Over longer time scales, the changing methylation states of the receptors could determine which receptors participate in the active receptor/signaling complexes.

A requirement for interactions between receptors in opposing membranes would imply that a curvature of the membrane is involved in CheA activation (Fig. 5). From this, one would expect that changes in membrane properties would affect CheA activity. For example, CheA is activated in vivo by alcohols that tend to make the membrane more fluid (43), as well as by increases in osmolarity that cause plasmolysis with attendant membrane invaginations (44).

In summary, active, soluble receptor/signaling complexes self-assemble from the monomeric Tsr_C protein, CheW, and CheA. These complexes have essentially the same size, composition, and architecture as those formed from lzTar_C. We conclude that the organization of these receptor/signaling complexes is determined by conserved interactions between the constituent chemotaxis proteins and may genuinely represent the architecture of membrane receptor/signaling complexes in vivo, and we propose a model for the formation and regulation of these complexes.

Materials and Methods

Protein Expression and Purification.

CheA, CheW, CheY, and lzTar_C (all Q Tar 257–553 fused to an N-terminal leucine zipper domain) were purified as described in refs. 10, 26, and 27. Calculated extinction coefficients at 278 nm (M⁻¹·cm⁻¹) are 7,000 for lzTar_C, 15,527 for CheA, and 5,600 for CheW (45).

Biochemical reagents were from standard suppliers, including Sigma–Aldrich (St. Louis, MO), Fisher (Pittsburgh, PA), and Roche (Indianapolis, IN). All restriction enzymes and DNA purification reagents were from New England Biolabs (Ipswich, MA) and Qiagen (Valencia, CA), respectively.

A cytoplasmic fragment of Tsr (PW029; all Q Tsr 290–551), referred to in this report as Tsr_C, was purified by using the expression plasmid pPW029. Plasmid pPW029 was constructed in analogy to plasmid pPA90 (24) by using the vector pCJ30 (46), with an insert from plasmid pHSE5.tsr_QQQQ (47). pHSE5.tsr_QQQQ contains the E. coli tsr gene with all of the methylation sites mutated to glutamine. Plasmid pHSE5.tsr_QQQQ was digested with HindIII and partially digested with Sau3AI. A DNA fragment of the desired size was agarose gel-purified and ligated into the BamHI-HindIII sites of pCJ30. The insert was verified by restriction digestion and automated DNA sequencing.

The Tsr_C protein was purified in a fashion similar to lzTar_C (10). Briefly, DH5α cells with pPW029 were grown in Terrific Broth (48) with 50 μg/ml ampicillin at 37°C to OD₆₀₀ = 0.8, induced with 1 mM isopropyl β-d-thiogalactoside, grown for 5 h more, harvested, washed with 0.1 M NaPO₄ (pH 7), and frozen. All further steps were carried out at 4°C or below. Cells were thawed and resuspended in 0.1 M NaPO₄, pH 7/0.15 mg/ml lysozyme/1× protease inhibitor mixture (Calbiochem, San Diego, CA)/0.5 mM EDTA. The cells were sonicated three times for 30 s on ice mixed with NaCl. Insoluble material was pelleted at 24,000 rpm in an SW-28 rotor (Beckman Coulter, Fullerton, CA). Protein in the supernatant was precipitated twice with 25% ammonium sulfate, collected by centrifugation at 14,000 rpm in an SLC-1500 rotor (Sorvall, Asheville, NC), and then desalted by using a G-25 column (Amersham Biosciences, Piscataway, NJ). Using a FPLC system (Pharmacia, Uppsala, Sweden), protein was loaded onto a Source-30Q column (Amersham Biosciences) in 25 mM Tris/25 mM NaCl/0.5 mM EDTA, pH 8, and eluted with a linear gradient of 0–400 mM NaCl in the same buffer. The Tsr_C protein eluted as a broad peak. The Tsr_C-containing fractions were precipitated with 40% ammonium sulfate, resuspended, and run on a Sephacryl S-300 column (Pharmacia) in 25 mM Tris/50 mM NaCl/0.5 mM EDTA, pH 7.5. Fractions containing pure Tsr_C protein were concentrated by using a Centriprep YM-10 concentrator (Millipore, Bedford, MA) to a final concentration of 582 μM, based on a calculated ε at 280 nm of 5,690 M⁻¹·cm⁻¹ (45). At 278 nm, the ε is 5,600. Concentrated protein was 0.22-μm filtered by using Spin-X filters (Corning, Corning, NY), snap-frozen in liquid nitrogen, and stored at −80°C. Mass spectrometry of the purified protein gave a single peak at a mass of 27,925 Da, which agrees with the predicted mass of 27,928 Da to within the estimated accuracy of the instrument. Mass spectra were acquired at the mass spectrometry facility of the Princeton University Department of Chemistry by using a model 1100 HPLC (Hewlett Packard, Palo Alto, CA) equipped with a Luna C-8 reverse-phase column (Phenomenex, Torrance, CA) and a mass-selective detector electrospray mass spectrometer (Hewlett Packard).

HPLC Characterization of lzTar_C and Tsr_C.

All HPLC analyses were performed with a model 1100 HPLC (Agilent Technologies, Palo Alto, CA). Samples of lzTar_C and Tsr_C were analyzed by using a Superose-12 HR 10/300 column (Pharmacia) in 50 mM KPO₄/200 mM NaCl/0.01% NaN₃, pH 7.8, connected to a UV detector inline with a miniDawn multiangle light-scattering detector (Wyatt Technology, Santa Barbara, CA). Data were collected and analyzed by using ASTRA 4.73 software (Wyatt Technology) to estimate the absolute mass of each protein by using the extinction coefficients at 278 nm, a value of 0.165 for the differential refractive index (dn/dc), and 1.330 (water) for the solvent refractive index. Gel filtration standards from BioRad (Hercules, CA) and purified CheW were used to determine the relationship between retention time and log(molecular mass).

Formation of Soluble Receptor/Signaling Complexes.

Ternary complexes of lzTar_C or Tsr_C with CheW and CheA were generated as described in refs. 26 and 27. Proteins were mixed in RBIII buffer (25 mM Tris·HCl/50 mM potassium glutamate/25 mM NaCl/5 mM MgCl₂/5% DMSO/10% glycerol, pH 7.5), the final concentration of potassium glutamate was adjusted to 50 mM, and DTT was added to a concentration of 10 mM. A range of concentrations of lzTar_C and Tsr_C were tested in the presence of 10 μM CheA and 10 μM CheW. Mixtures were incubated at 30°C overnight. After formation, complexes were stored at room temperature (≈25°C). For TEM analysis, Tsr_CWA complexes were assembled instead by using a final concentration of 300 mM KPO₄ (pH 7.5), 5 mM MgCl₂, and 10 mM DTT or 5 mM neutralized Tris(2-carboxyethyl)phosphine (TCEP). In this buffer, complexes formed as efficiently as in RBIII, but the absence of DMSO and glycerol resulted in improved TEM sample preparation.

HPLC Analysis of Receptor/Signaling Complexes.

lzTar_CWA and Tsr_CWA complexes were analyzed as described in refs. 26 and 27 by using a TSK-Gel G5000PW_XL column (30 × 0.78 cm) with a TSK-Gel G4000PW guard column (Tosoh Bioscience, Tokyo, Japan) using a flow rate of 0.5 ml/min of RBIII (with additional MgCl₂ up to 10 mM). Receptor/signaling complexes with final receptor concentrations of 60 μM lzTar_C and 80 μM Tsr_C were used for the determination of the complex stoichiometry and absolute molecular mass. Fractions were collected across the elution peak corresponding to receptor/signaling complexes and were analyzed by using 15% SDS/PAGE gels and Coomassie staining.

For determination of the stoichiometry of proteins in the receptor/signaling complexes, two-dimensional HPLC was used. Using the TSK-Gel G5000PW_XL column, two 0.4-min fractions beginning at elution times of 14.1 and 14.5 min were collected for each sample, and a final concentration of 2.5 mM neutralized Tris(2-carboxyethyl)phosphine (TCEP) was added to each. The resulting protein mixture was precipitated by using a final concentration of 10% trichloroacetic acid at 0°C, centrifuging at 12,400 × g for 15 min, washing twice with −20°C acetone, and allowing to air dry at 25°C. Pellets were resuspended in 6 M guanidine-HCl/20 mM NaPO₄/1.5 mM TCEP, pH 6.5, and loaded onto a Zorbax GF-250 (4.6 × 250 mm) column (Agilent Technologies) running at 0.25 ml/min in the same buffer. Absorbance was monitored at 278 nm, and protein concentrations were determined by using extinction coefficients at 278 nm (45). The area under each peak of the chromatogram was divided by the extinction coefficient of the corresponding protein to determine the number of moles of protein in the peak, and then the numbers for CheW and lzTar_C or Tsr_C were divided by the number for CheA to find the stoichiometry.

The molecular masses and rms radii of the lzTar_CWA and Tsr_CWA complexes were determined by HPLC by using the inline miniDawn multiangle light-scattering detector and the TSK-Gel G5000PW_XL column with the ASTRA software. Analyses were performed on data centered on each elution peak, from 14.1 to 14.9 min for the lzTar_CWA complex and from 14.2 to 15.0 min for the Tsr_CWA complex. The concentrations of the lzTar_CWA and Tsr_CWA complexes were estimated from UV absorption by using the extinction coefficients for the receptor/signaling complexes, determined as follows. The extinction coefficient of each complex was calculated from the extinction coefficients of the component proteins by using the stoichiometry that was determined as described above (Table 1). Similar to previous analyses, the calculated coefficient for the complex was corrected to account for the difference between the actual UV absorbance of the proteins in the lzTar_CWA complex in the elution buffer and the predicted UV absorbance of the individual protein components in 6 M guanidine-HCl (26). The correction factor was adjusted so that the analysis returned the known value of 4.0 CheA subunits per complex. This same correction was applied to the calculated extinction coefficient of the Tsr_CWA complex. We used a value of 0.165 for dn/dc, and the solvent refractive index of the RBIII elution buffer was measured to be 1.358 by using a model 33-45-58 refractometer (Bausch & Lomb, Rochester, NY).

Receptor/signaling complexes formed with mixtures of lzTar_C and Tsr_C were also analyzed by using the TSK-Gel G5000PW_XL column in RBIII (with additional MgCl₂ to a concentration of 10 mM). Fractions were collected every 0.4 min from 13.5 to 21.5 min and analyzed by using 15% SDS/PAGE gels with Coomassie staining. Complexes were formed at 30°C with a 5 μM concentration of either lzTar_C or Tsr_C, a 0 or 45 μM concentration of the other receptor fragment type, and a 10 μM concentration each of CheW and CheA in RBIII. Gels were scanned with a Scanmaker 4900 flat-bed scanner (Microtek, Carson, CA).

Kinase Activity Measurements.

Kinase activity of CheA alone and in complexes was determined by using a spectrophotometric coupled ATPase assay in RBIII buffer, as described in ref. 26. Data were collected at 20°C by using a DU-530 spectrophotometer (Beckman Coulter) with a temperature-controlled cuvette holder.

EM.

Microscopy of the lzTar_CWA complexes is described in ref. 28. Tsr_CWA complexes were diluted up to 20-fold in 50 mM Tris/250 mM KCl/5 mM MgCl₂, pH 7.5, as necessary to give a suitable density of particles on the grids. Five-microliter aliquots of the diluted complexes were immediately applied to 400 mesh copper microscope grids covered by continuous carbon foils, stained with 2% uranyl acetate, washed with deionized water, stained again with 2% uranyl acetate, blotted, and allowed to dry. TEM images at ×60,000 magnification were recorded with doses of <20 electrons per Ų on SO163 film (Eastman Kodak, Rochester, NY) by using a CM12 microscope (Philips Electronic Instruments, Mahwah, NJ) equipped with a model 626 cryostage and a model 651 anticontaminator (Gatan, Pleasanton, CA), both cooled to ≈−170°C. Micrographs were digitized by using an SCAI scanner (Carl Zeiss, Oberkochen, Germany) at 7 μm/pixel. Pixels were averaged 2 by 2 to give 14-μm pixels corresponding to 2.33 Å/pixel in the specimen.

Image Analysis.

The defocus of the individual micrographs was determined by using CTFFIND2 (49). Images with appreciable astigmatism were not used. The micrographs were phase-corrected for the contrast transfer function (CTF) by reversing the phase flip imposed by the CTF. Images of single particles were manually selected by using Ximdisp (MRC package) (50) and were windowed from the phase-corrected TEM micrographs by using SPIDER (51). Images were treated to remove any ramp of density and then were normalized, setting the average to zero and the standard deviation to a constant. The particles were iteratively aligned in two dimensions by using SPIDER. Particles were aligned first to an average of the lzTar_CWA complex (28) and then were aligned in subsequent rounds to the averages of the aligned Tsr_CWA images.

The radius of gyration for the lzTar_CWA complex was calculated from the surface map corresponding to the previously reported three-dimensional reconstruction by using the program HYDROMIC (30).

Abbreviations

Tsr

serine receptor

Tsr_C

cytoplasmic domain of Tsr

Tar

aspartate receptor

Tar_C

cytoplasmic domain of Tar

TEM

transmission EM

lzTar_CWA

lzTar_C/CheW/CheA

Tsr_CWA

Tsr_C/CheW/CheA