Receptor density balances signal stimulation and attenuation in membrane-assembled complexes of bacterial chemotaxis signaling proteins

Abstract

All cells possess transmembrane signaling systems that function in the environment of the lipid bilayer. In the Escherichia coli chemotaxis pathway, the binding of attractants to a two-dimensional array of receptors and signaling proteins simultaneously inhibits an associated kinase and stimulates receptor methylation—a slower process that restores kinase activity. These two opposing effects lead to robust adaptation toward stimuli through a physical mechanism that is not understood. Here, we provide evidence of a counterbalancing influence exerted by receptor density on kinase stimulation and receptor methylation. Receptor signaling complexes were reconstituted over a range of defined surface concentrations by using a template-directed assembly method, and the kinase and receptor methylation activities were measured. Kinase activity and methylation rates were both found to vary significantly with surface concentration—yet in opposite ways: samples prepared at high surface densities stimulated kinase activity more effectively than low-density samples, whereas lower surface densities produced greater methylation rates than higher densities. FRET experiments demonstrated that the cooperative change in kinase activity coincided with a change in the arrangement of the membrane-associated receptor domains. The counterbalancing influence of density on receptor methylation and kinase stimulation leads naturally to a model for signal regulation that is compatible with the known logic of the E. coli pathway. Density-dependent mechanisms are likely to be general and may operate when two or more membrane-related processes are influenced differently by the two-dimensional concentration of pathway elements.

The signaling pathway that mediates chemotaxis in Escherichia coli—like many systems—consists of transmembrane and membrane-associated proteins that function in the two-dimensional (2D) space of the lipid bilayer, where clustering, allostery, and cooperative interactions may all contribute to the regulation of signaling (1, 2). Studies of E. coli have shown that chemoreceptors, the adaptor protein, CheW (W), and the histidine kinase, CheA (A), form 2D arrays that are remarkable for the large number of receptors involved and their location at the cell poles (3–6). The ligand-binding domains of the homologous chemoreceptors endow the array with responsiveness toward specific attractant molecules; the conserved cytoplasmic domain (c-domain) conveys signals generated by ligand binding. In addition, the sensitivities of receptors toward cognate ligands are adjusted by the reversible methylation of a few conserved glutamic acid residues in the c-domain (7). Temporal comparisons are made between current and recent past chemoeffector concentrations by pathway proteins that propagate and terminate signals, which serves to bias the random-walk swimming behavior of E. coli in chemical gradients (8). Given its remarkable properties of sensitivity and wide dynamic range, an effort is being made to elucidate the structure and function of the receptor array.

Progress toward knowing the physical arrangement and action of the array has been made by identifying pairwise and higher-order interactions among receptors and signaling proteins (9–14). The x-ray structures of receptor c-domains, in vivo site-directed cross-linking, and electron microscopic analyses of soluble protein complexes have provided key information. Based on these data, a trimer-of-dimers organization of the receptors in E. coli is regarded by many to play a relevant role, because in vivo cross-linking and mutational data are compatible with this arrangement (9–11, 14). However, the situation is complicated by the fact that c-domain dimers in large self-assembled complexes with A and W are not arranged as trimer-of-dimers (15); yet these complexes stimulate CheA activity. Moreover, a “hedgerow” arrangement of dimers was found in crystals of a c-domain derived from a Thermotoga maritima chemoreceptor; this arrangement is also suggested to have functional relevance (13). Finally, the size and diversity of the microbial chemoreceptor superfamily gives credence to the idea that the E. coli paradigm may not represent all of its members (7, 16). The precise organization of receptors and signaling proteins is likely to be essential for function, but the full nature and role of the different arrangements are yet to be established.

An intriguing feature of array function is the means by which the balance between kinase-stimulating and inhibiting states is established. Receptor covalent modification stabilizes the kinase-stimulating state and increases the attractant concentration needed to achieve inhibition (17–22). In addition, attractant-mediated inhibition of receptor-associated CheA displays significant cooperativity (21, 22). Recent observations suggest that changes in receptor density are involved: attractant binding induces an apparent increase in the separation of receptors in E. coli and Bacillus subtilis cells (23, 24). Neither the physical mechanisms nor the relationships to signaling is understood.

To address this issue, we used liposome-mediated assembly to make complexes of membrane-associated aspartate receptor c-domain, W, and A at defined surface concentrations. Previously, we demonstrated that the c-domain in these complexes regained the functions it lacked in solution: assembled c-domains stimulated kinase activity and were effective substrates for methylation (25–28). Here, we measured the methylation and kinase stimulation as a function of protein surface concentration. Both methylation and kinase activity changed substantially as a function of receptor density, but in opposite directions. Kinase activity increased with receptor density in a cooperative manner. FRET measurements conducted with labeled c-domains demonstrated that the activity increase coincided with an increase in FRET efficiency, signifying a cooperative change in the arrangement of the membrane-associated proteins. In addition, the density dependence of kinase activity was influenced strongly by receptor modification: high modification levels stabilized the kinase-stimulating state over the entire range of surface concentrations. By contrast, the rate of receptor methylation decreased as surface concentration increased. We attributed the rate decrease to a density-induced change in the arrangement of c-domain that lowered its effectiveness as a methyltransferase substrate.

These results are incorporated in a model for signal regulation, which takes into account the opposing influences of receptor density on the kinase and methylation activities, and provides an explanation for the compensating effects of chemotactic stimuli and receptor modification. Finally, we suggest that the underlying mechanism is more general than its specific application to the E. coli chemotaxis system; it may occur in other membrane signaling processes, especially when larger assemblies of membrane-associated proteins are involved.

Results

Assembly at Defined Surface Concentrations.

Fig. 1A depicts an E. coli methyl-accepting chemoreceptor dimer (green). The organization of the ligand-binding domain that the major E. coli chemoreceptors typify is the most prevalent organization in a diverse group of sensing domains found among the microbial chemoreceptors (7, 29). The c-domain exhibits greater conservation; it is often comprised of a single membrane-proximal HAMP domain (30) connected to the methyl-accepting (MA) signal domain. The MA domain defines the chemoreceptor superfamily (16, 31): the dimer of a coiled-coil hairpin forms an extended four-helix bundle that contains the methylation sites and Che-protein binding sites (9, 13). Fig. 1A also shows A and W bound near the end of the dimer—the region of greatest sequence identity, and the location of the trimer-of-dimer contact sites observed in the crystal structure of the E. coli serine receptor c-domain (9). The c-domain fragment (CF) used in this study is the same CF used previously (25–27); it possesses an N-terminal hexa-histidine tag fused to the E. coli aspartate receptor MA/signaling domain and ends at the natural C terminus of the intact receptor (residues 257–553). The CF contains the methylation sites and the methyltransferase (CheR) tethering segment—an unstructured region that is required for efficient methylation in E. coli (32, 33).

Fig. 1.

The chemoreceptor dimer and membrane-assembled signaling complexes. (A) The receptor dimer (green) depicted with thickened line segments for α-helices and domains labeled (Ligand Binding, HAMP, Cytoplasmic). W (magenta) and the A dimer (A₂, red) are bound to the c-domain, and the four major methylation sites are shown partly modified (QEQE) as filled (sites 1 and 3 modified) and unfilled (unmodified) circles. (B) QEQE c-domain dimers (CF, green) are assembled with A₂ and W on a vesicle outer surface (blue). (C) A membrane section viewed from above, in which CF dimers (green circles) are shown (in cross-section) bound end-on to the lipids (blue capsules), corresponding to a situation with 30 μM CF and 580 μM total lipid. The CF cross-sectional area (≈4 nm²) is scaled to depict the surface area occupied by CF at a lipid/CF ratio of 20 (≈14 nm² per CF dimer).

Fig. 1B depicts signaling complexes organized on the outside of a liposome in a manner inferred from transmission electron microscopic observations (26). Such complexes form spontaneously from mixtures of CF, A, and W with liposomes that contain a nickel-chelating lipid. Fig. 1 B and C illustrates the membrane-associated CF as a dimer—the most likely form present because of its requirement for function. The 2D concentration of proteins on the unilamellar liposome surface was varied by adjusting the CF binding site density, present as the mole percent of nickel-chelating lipid in a binary mixture with DOPC. The property of near-complete binding of the CFs to liposomes [≈95%, determined by the sedimentation of template-assembled signaling complexes with the liposomes (data not shown)] was used to estimate the average surface concentration of CFs from the total CF concentration and lipid surface area. Fig. 1C depicts the fractional coverage for a specific membrane composition (50% nickel-chelating lipid), in a view looking down on the membrane surface to which CF dimers are bound end-on. Based on structure-derived estimates for the relative cross-sectional areas of phospholipid and CF, this composition has been calculated to correspond to ≈30% surface coverage (26). The percent surface coverage was also estimated from binding isotherms of the CFs to vesicles. Samples comprised of 50% nickel-chelating lipid were found to have 45% of the surface concentration at saturation [supporting information (SI) Fig. S1 and Table S1; also provided are Fig. S2, Tables S2–S4, and SI Materials and Methods]. In the experiments described below, the c-domain surface concentration was varied 10-fold (≈5–50% of saturation) by using different nickel lipid/DOPC compositions.

Motivated by observations of density effects in signaling (23, 24), we set out to test the relationships between receptor density, covalent modification, and signaling activity. Previously, surface-assembled complexes prepared at a high surface density with CFs in a low, intermediate, or high level of covalent modification stimulated kinase activity to similar extents (26, 27). The activity stimulated by unmodified CF (CF_4E) was significant and essentially equal to activities produced by CFs in intermediate (CF_QEQE) and high (CF_4Q) levels of glutamine substitution at the four major methylation sites; these substitutions are known to mimic methylation and allow preparation of CF in defined modification states (17, 18). The high level of kinase activity generated with all three forms of the CF was unexpected, because similar measurements conducted by using intact receptors had found an increase in activity with covalent modification, where 4E (unmodified receptors) stimulated CheA much less than 4Q (17, 22, 34).

A possible relationship between receptor density to signal activity and the differences exhibited by the template-assembled and intact receptor samples prompted us to compare kinase activities of complexes assembled on vesicles with low, intermediate, and high binding site densities (5, 10, and 50 mol % nickel lipid, respectively). The kinase activities displayed in Fig. 2 A and B demonstrated that surface density, covalent modification, and, to some extent, liposome size all influenced activity. The density dependence of CF_4E-containing complexes was the largest of the three modifications. On small and large unilamellar vesicles (SUVs and LUVs), the activity of CF_4E-containing samples increased substantially between 5% and 60% nickel-chelating lipid: kinase activity increased from 0.5 to 13 s⁻¹ on SUVs and from 5 to 20 s⁻¹ on LUVs. Increasing the covalent modification to an intermediate level (CF_QEQE) dampened this trend, and it was absent at the highest level (CF_4Q).

Fig. 2.

Kinase activity versus CF surface concentration. (A) Kinase activity of CheA assembled on SUVs with CheW and CF_4E, CF_QEQE, or CF_4Q (open, gray, and black, respectively) containing nickel lipid at the stated mole percentages. (B) Activity of CheA assembled on LUVs (as in A). (C) Cooperative increase in the activity of CheA assembled, with W and CF_4E, on membrane vesicles containing the stated mole percentages of nickel lipid: SUVs (open circles) and LUVs prepared by extrusion through filters with pore diameters of 50 nm (filled circles) or 100 nm (open squares). The lines are fits of the data to the Hill equation, as described in SI Materials and Methods. (D) f_B(CheA), the fraction of CheA that co-sedimented with CF-decorated vesicles (symbols as in C). Protein and lipid concentrations: 30 μM CF, 5.0 μM W, 1.2 μM CheA, 290 μM nickel-chelating lipid, and DOPC (to achieve the stated mole percentage of nickel lipid). Activities and uncertainties are means and standard deviations of triplicates, respectively.

Vesicle size, which probes the influence of membrane curvature, was found previously to influence kinase activity, by comparing complexes formed at high densities on SUVs and LUVs (26). The trend held up when this comparison was extended to include activity measurements as a function of surface concentration: the kinase activities of complexes on LUVs were generally greater than complexes on SUVs under comparable conditions (Fig. 2 A versus B). At low surface concentrations especially, the larger curvature of SUVs dampened the formation of kinase-stimulating complexes. Compared to LUVs, SUVs accentuated the differences among the CFs, which was reflected in a more pronounced activity increase with covalent modification, but complexes formed with CF_4Q stimulated kinase activity over the entire range of surface concentrations on SUVs and LUVs. Altogether, these observations are compatible with the known effects of glutamine substitution in promoting both CF self-association and the kinase-stimulating state (15, 18).

Kinase Activity Increases Cooperatively with Surface Concentration.

The increase in kinase activity seen with CF_4E-containing assemblies led to a more extensive set of CheA activity and binding measurements as a function of CF surface concentration (Fig. 2 C and D). A cooperative transition between low and high activity states was observed, although it was less cooperative on SUVs compared to LUVs (Fig. 2C). The lower cooperativity is likely related to the comparatively small percent of CheA bound to CF_4E-decorated SUVs at low surface concentrations: 26% CheA bound to SUVs with 5% nickel lipid (Fig. 2D); we regard this observation as more evidence that highly curved membranes are less effective templates for assembly and stimulation. By contrast, the majority of CheA present bound to CF_4E-decorated LUVs at all surface concentrations tested. Membrane curvature effects were judged not to be significant with LUV templates, because no significant difference in binding or stimulation was observed with samples prepared by using LUVs of two different mean diameters (Fig. 2 C and D, squares and filled circles). CheA binding was specific; in a previous study, CheA binding to CF-decorated vesicles was reduced substantially by amino acid substitutions in the CF near the A interaction site (25). When binding and activity data are considered together, the change in CheA activity is plausibly interpreted as a process that involves the clustering and/or density-dependent rearrangement of surface-associated CF–W–A complexes.

To test that the activity increase resulted from a change in physical proximity of CFs, FRET measurements were conducted with vesicle-assembled signaling complexes made with a mixture of donor- and acceptor-labeled CFs. Fig. 3A plots the kinase activity (open symbols) and FRET efficiency (filled symbols) as a function of surface density for complexes containing either CF_4E (circles) or CF_4Q (triangles). The remarkable correspondence between the increase in FRET and activity observed in samples containing CF_4E, and the unchanging activity and FRET in samples containing CF_4Q, led us to conclude that the increase in surface density of CF_4E (and A and W) induced the formation of the kinase-stimulating state.

Fig. 3.

FRET efficiency, kinase activity, and methylation rate versus CF surface concentration. (A) Kinase activity (open symbols) and FRET efficiency (filled symbols) of membrane-assembled A, W, and donor- and acceptor-labeled mixtures of CF_4E (circles) or CF_4Q (squares and triangles). The kinase activities shown were measured with donor-labeled/unlabeled CF mixtures. Acceptor-labeled/donor-labeled/unlabeled CF mixtures produced similar results, although the activities measured with labeled CFs were generally smaller than activities observed with unlabeled CFs. Activities and uncertainties are means and standard deviations of triplicates, respectively. LUVs were made with 50-nm pore-diameter filters. Concentrations at assembly: 30 μM CF, 1.2 μM A, 5 μM W, 290 μM nickel-chelating lipid, and DOPC (to achieve the stated mole percentage of nickel lipid). (B) Methylation (pmol of CH₃/min) of template-assembled CF_4E versus surface concentration, without A and W (open bars) and with 5.0 μM W and 1.2 μM A (filled bars).

Methylation Rates Decrease with Surface Concentration.

A parallel set of experiments was conducted to determine the dependence of receptor methylation on the surface concentration of the CF. In E. coli, receptor methylation occurs by a process in which CheR (R) binds to the extreme C terminus of the receptor (32). As a result, R is tethered near the methylation sites on the receptor dimer to which it is bound and gains access to sites on nearby receptor dimers (33, 35). We have shown previously that liposome-mediated assembly increases the rate of methylation of CF_4E substantially (27).

Receptor methylation has been thoroughly investigated in vitro but usually with receptor samples that do not contain A and W. This omission can be plausibly expected to produce significant differences, because A and W promote receptor cluster formation and probably alter the conformation of receptor dimers and the arrangements among dimers (3, 36). The results of methylation experiments plotted in Fig. 3B provide evidence that A and W, as well as the CF surface concentration, influence methylation rates substantially (Fig. 3B, open versus filled bars). Without A and W, methylation increased between 10% and 30% nickel lipid (corresponding to a decrease in the area per CF from 36 to 12 nm²). The increase in rate is an expected consequence of receptor transmethylation (32, 33), yet increasing the nickel lipid concentration still further (to 60%, ≈6 nm² per CF) decreased the rate. We postulate that the rate decrease results from crowding, which reduces the access of R to the methyl-accepting glutamate residues.

A markedly different dependence of rate on surface concentration was found when A and W were present: the rate was largest at the smallest CF_4E surface concentration (10% nickel lipid), and it was 4-fold greater than the same sample but without A and W (Fig. 3B). With A and W the methylation rate changed more substantially as a function of the CF_4E surface concentration: it was 25-fold smaller in samples with vesicles containing 60% nickel lipid compared to vesicles with 10%. A few experiments were conducted to determine the methylation rate as a function of the W concentration. At a single CF_4E surface concentration (10% nickel lipid), the largest rate was observed with 10 μM W, whether or not CheA was also present (Fig. S2). Coincidentally, this concentration corresponds to a W:CF ratio similar to the ratio in the cell (37). The increase in the methylation rate produced by A and W in the template-assembled system is qualitatively similar to the rate increase observed with the (full-length) serine receptor in membranes (36). This suggests that the variation in methylation rate as a function of the CF_4E surface concentration with W (and A) present—larger rates at smaller surface concentrations—might also be observed with intact receptors in membranes. The markedly different trend in the methylation rate when A and W are present, versus their absence, demonstrates that several factors influence receptor methylation. When A and W are present, we hypothesize that the greatest influence is exerted by density-dependent changes in the arrangement of surface-associated CF, A, and W, such that the rate of methylation is lower in the high-density state.

Discussion

Transmembrane receptors and membrane-associated signaling proteins occupy environments in the cell membrane where the protein concentration in two dimensions exerts a large influence on clustering and signaling (38). The transmembrane signaling process in the bacterial chemotaxis system represents a situation that has received less attention: one in which protein arrays comprised of receptors and signaling proteins are thought to communicate primarily via allosteric and cooperative interactions (1, 2). It is also a system in which the 2D protein concentration, e.g., the receptor density, changes in response to ligand binding (23, 24), but it is not understood how receptor clustering and density-dependent changes in the arrangement of clustered receptor are linked to signaling.

A Model for Regulation Based on Surface Concentration.

A model for signal regulation that explains the role of receptor density is depicted in Fig. 4; it takes into consideration the data presented here with well known and recently reported effects of ligand binding on kinase activity, methylation, and the arrangement of receptor clusters (17, 22–24, 36). The model is based in the observation that, as the surface concentrations of receptors and signaling proteins increase, the kinase activity also increases and the rate of receptor methylation decreases. Fig. 4A illustrates this: the kinase activity and receptor methylation data are plotted together as a function of surface concentration (CFs per nm²). From these results, it is evident that large surface concentrations favor kinase stimulation and small concentrations support larger methylation rates. These results are consistent with and extend previous observations to generate a model for signal regulation (Fig. 4B).

Fig. 4.

Effects of surface concentration on kinase activity and methylation combine to regulate signaling. (A) Kinase activities (circles) and methylation rates (bars) versus surface concentration (CF per nm²) of CF_4E–A–W complexes (from Figs. 2C and 3B, respectively). (B) A model based on low- and high-density states and the factors that influence their formation (curved arrows). Arrays receptor dimer (green), W (magenta), and A (red) reside on the membrane (blue lines) in low- or high-density states (left and right, respectively). At high density, the methylation level, represented by filled circles, is larger, and CheA is stimulated to phosphorylate CheY (Y, yellow) and the methylesterase CheB (B, brown). Formation of the low-density state is promoted by attractant binding (filled circles) and receptor demethylation by phospho-CheB (CheB-P). In the low-density state, receptors do not stimulate CheA, and attractant-bound receptors are methylated more rapidly by CheR (R, blue), which favors formation of the high-density state.

For simplicity, receptor complexes are depicted by two states: low and high density. The properties of these states and the established relationships of ligand binding, receptor methylation, and kinase activity in the E. coli signaling pathway provide for the establishment of regulated signaling via negative feedback. For example, a system prepared in the low-density, low-methylation-level state tends toward a state with higher methylation levels. The increase in methylation promotes the formation of the high-density kinase-stimulating state (left to right in Fig. 4B). This process becomes self-limiting via negative feedback, because the stimulation of CheA leads to the phosphorylation and activation of the methylesterase (CheB). A similar process qualitatively explains how a system prepared in the high-density state will tend toward lower methylation due to a low methylation rate and the presence of activated esterase (CheB∼P). As the receptor methylation level drops, a steady state is achieved through a greater contribution from the low-density, kinase-inactive, methylation-active state.

This model can also explain the restoration of steady-state kinase and methylation activities following a chemotactic stimulus. The attractant binding process is depicted as an arrow that shifts the equilibrium toward the low-density state (Fig. 4B), because attractant binding to receptors has been observed to increase the physical separation among receptors (23, 24). Attractant binding also produces well known biochemical effects: a reduction in kinase activity, a rise in the methylation rate, and a lowering of the demethylation rate (22, 35, 39). These effects set in motion the return toward the prestimulus kinase activity (adaptation). For instance, the shift away from equilibrium initiated by the attractant stimulus serves to increase the methylation level, which promotes the shift back toward the high-density (kinase-active/methylation-inactive) state. Stimuli involving either a decrease in an attractant concentration or changes in repellent concentrations should also result in adaptation through the counterbalancing surface concentration mechanism.

Biochemical Activity and Ligand Binding May Be Coupled by Surface Concentration.

A detailed picture is still emerging for the structure of the receptor array in the cell. Nonetheless, we can cite two ways that template-assembled proteins are similar to cellular arrays. (i) CheA associates to the same degree in kinase-stimulating and nonstimulating states (Fig. 2D) (40). (ii) The average surface concentrations of CF in the template-assembled complexes are compatible with the dimensions of receptor patches observed in cells and the numbers of receptors that these patches are expected to contain (4). The average surface concentration at which the kinase and methylation activities change rapidly is ≈0.07 nm⁻² (Fig. 4A). At this concentration, a representative circular patch of receptors (≈5,000) will have a 150-nm radius, which is similar to the dimensions of signaling arrays visualized in frozen E. coli cells by electron tomography (4).

The cooperative increase in CheA activity in template-assembled signaling complexes as a function of the increasing surface concentration agrees with several lines of evidence that support a role for clustered receptors in the regulation of kinase activity (3, 4, 6, 15, 21, 22, 41). By contrast, the change in the rate of receptor methylation, which decreased as the surface concentration increased, was not expected. Yet this feature, with the observation that attractant binding increases the physical separation of receptors (23, 24), forms a consistent picture in the density model (Fig. 4B), where attractant binding drives an increase in separation among receptors and receptor methylation subsequently restores the prestimulus separation. In the experiments involving template-assembled signaling proteins, the average surface concentration is fixed by the protein-to-lipid ratio. The samples have distinguishably different properties at low and high concentrations, which is evident from the cooperative changes in the kinase activity and FRET. We postulate that the transition from the high- to the low-density state is accompanied by a conformational change in the CF that is similar to the conformation change induced by attractant binding in the full-length receptor.

Although we cannot rule the contributions of crowding to the reduction in the methylation rate at larger surface concentrations, crowding probably does not contribute significantly at the concentrations where the methylation and kinase activities are both changing rapidly (10–30 mol %; Fig. 2C and 3B). When A and W were not present, the methylation rate increased, suggesting that crowding was not a significant factor. By contrast, the rate decreased over this same range when A and W were present. Because A and W are associated with receptors in the cell (40), we think that the more relevant dependence is a decreasing methylation rate with increasing surface concentration. In addition, a new model for the architecture of the chemoreceptor array in Caulobacter crescentus, produced by cryo-electron tomography (42), places the receptor concentration at 0.09 nm⁻²—in the range of surface concentrations where the biochemical activities of template-assembled signaling complexes change rapidly (Fig. 4A). The agreement might be coincidental, but such models are consistent with the idea that CheR has ample access to the sites of methylation (42).

The molecular weights of CheB (B) and R (38 and 33 kDa, respectively) imply that both proteins have similar (steric) access to the methylation sites, yet B is expected to display greater activity toward receptors in the high-density state. This greater activity may arise in part from differences in the receptor conformation in stimulating versus inhibiting states, but it will also arise from CheB phosphorylation, which increases esterase activity ≈100-fold (39).

A conformational change coupled to receptor density is compatible with the long-standing notion that the receptor dimer is the functional unit of methylation, i.e., ligand binding increases the methylation rate via a conformational change propagated within dimers and a single dimer can function efficiently as a substrate for R (32, 33, 35, 41, 43). The density model extends this idea to couple the ligand-induced conformational change to interactions between dimers, thereby generating density dependences in both the kinase and methylation activities. The precise molecular arrangements between receptors, A, and W are important to this process, and our results provide an impetus to test these ideas further.

We suspect that other examples of density-dependent signal regulation will be found. In prokaryotes, the methyl-accepting chemotaxis proteins are widespread; the organization of the E. coli pathway is representative of many, but not all, aspects (7, 16). Attractant-induced unclustering of receptor signaling complexes—an important facet of the mechanism—is found in both Gram-positive and negative bacteria (23), but variations in signaling logic imply that density dependences will be implemented in different ways. More generally, the membrane bilayer dictates the 2D organization of proteins in all cells. Thus, counterbalancing effects in surface reactions are likely to be found in dynamic membrane processes that involve the orchestration of large numbers of membrane-associated proteins, as in the cell–cell interactions that are widespread in cellular immunology and neurobiology (44, 45).

Materials and Methods

Vesicles and Protein Assembly.

CF, A, W, CheY, and R were purified according to established protocols (26, 32). Vesicles of DOPC (1,2-dioleoyl-sn-glycero-3-phosphocholine) and the nickel-chelating lipid 1,2-dioleoyl-sn-glycero-3-[[N(5-amino-1-carboxy-pentyl)iminodiacetic acid]succinyl] (Avanti Polar Lipids) were prepared by sonication or extrusion; size distributions were determined by dynamic light scattering as described in ref. 26. Vesicle/CF/W/A assemblies were generated by incubating components together at 25°C, using the specified concentrations (assembly conditions) for 4 h.

Enzyme Assays.

Kinase activity was measured with an enzyme-coupled ATPase assay, as described previously, immediately after a 100-fold dilution from assembly conditions (26, 27). Methylation assays (100-μl volume) were conducted in ATPase buffer with 25 μM CF_4E, LUVs present at 290 μM nickel-chelating lipid (prepared by extrusion through 50-nm pore-diameter filters), 1.2 μM A, and 5.0 μM W, unless specified otherwise. The samples also contained 10 μM [³H]-s-adenosyl-l-methionine (NEN Life Sciences), and R (to give 3.6 μM) was added to initiate methylation. Fourteen-microliter aliquots were removed over time, quenched with 2× SDS/PAGE buffer, resolved by SDS/PAGE (12.5% acrylamide), stained for 10 min (GelCode Blue; Pierce), and rinsed with H₂O. CF bands were excised and methyl group incorporation was measured by scintillation after 24 h of vapor-phase transfer (35). Initial rates (pmol of CH₃/min) were determined from data collected over a 2-min period.

Fluorescence Resonance Energy Transfer.

FRET experiments were conducted with engineered CF_4E and CF_4Q, in which six residues in the R-tethering segment were replaced with a tetra-cysteine motif (⁵¹³FRLAASPLTNKPQTP⁵²⁷ → ⁵¹³FRLAACCPGCCPQTP⁵²⁷; numbers denote locations in the full-length E. coli aspartate receptor) that binds bis-arsenical derivatives of fluorescein (FlAsH) and resorufin (ReAsH) with high affinity (46). FlAsH- and ReAsH-labeled CFs were prepared in ATPase buffer (which also contained 10 mM TCEP) within 24 h of use by mixing tetraCys CF and FlAsH or ReAsH in a 1:1 ratio, followed by incubation in the dark for 1 h. Donor fluorescence was measured in the absence or presence of acceptor with matched pairs of samples. In donor-only samples, the unlabeled/FlAsH-labeled CF mixture was 0.85/0.15. In samples with donor- and acceptor-labeled CFs, the unlabeled/ReAsH-labeled/FlAsH-labeled mixture was 0.50/0.35/0.15. Vesicle–protein mixtures were assembled and assayed for kinase activity in the conventional manner. For FRET, emission spectra (510–590 nm) were collected with a 495-nm excitation wavelength and the FRET efficiency, E, was calculated as 1 − F′_D/F_D (47), where F′_D and F_D are donor fluorescence intensities at 530 nm with and without the acceptor, respectively.