Cooperativity between bacterial chemotaxis receptors

Cellular sensory transduction pathways typically begin at the cell surface where membrane-spanning receptors detect an external stimulus, propogate a transmembrane signal, and regulate cytoplasmic signaling events. In many pathways these cell surface receptors have been thought to function independently, without significant interactions between receptor molecules. For example, in prokaryotic and eukaryotic cells the most prevalent types of cell surface receptors are histidine kinase-coupled receptors (HKCRs) and G protein-coupled receptors, respectively. Both of these receptor classes traditionally have been described as devoid of direct receptor–receptor interactions. By contrast, receptor–receptor interactions are known to play an important role in certain other receptor classes, including tyrosine kinase receptors (1).

In recent years, however, observations that HKCRs can be found in large clusters and that G protein-coupled receptors can oligomerize have necessitated a reanalysis of the idea that these receptors function as lone, independent signal transducers (2, 3). The bacterial chemosensory pathway that controls the swimming of Escherichia coli and Salmonella typhimurium in attractant concentration gradients is arguably the best-characterized signaling pathway in nature (4, 5), and the HKCRs of this pathway provide an ideal model system in which to probe the structural, mechanistic, and functional implications of receptor–receptor interactions. The paper by Ames, Studdert, Reiser, and Parkinson (6) provides important new information about the nature of interactions between HKCRs. These findings, together with recent results from other laboratories, are beginning to paint a picture in which HKCR receptor–receptor interactions within stable receptor oligomers play a number of important functional roles in a simple sensory pathway.

The trimer of dimers is the simplest oligomer that can explain the Hill coefficients.

The bacterial chemosensory pathway possesses a set of five cell surface receptors that are specific for different attractants and repellents (4, 5). Studies of receptor–receptor interactions have focused to date on the aspartate, serine, and ribose/galactose receptors. In addition, the pathway possesses six cytoplasmic components, all of which interact with the long, helical, highly conserved cytoplasmic domains of the transmembrane receptors to form a supramolecular signaling complex. These cytoplasmic components include: (i) the histidine kinase CheA whose autophosphorylation activity is regulated by the receptor, (ii) the coupling protein CheW that stabilizes the complex formed between CheA and the receptor, (iii) the response regulator CheY that is activated by phosphotransfer from CheA and diffuses to the flagellar motor where it regulates swimming activity, (iv) the dephosphorylation protein CheZ that stimulates the autohydrolysis of phospho-CheY, and (v) the adaptation enzymes CheB and CheR that covalently modify specific adaptation sites on the receptor cytoplasmic domains. The architecture shared by the transmembrane receptors has been extensively characterized, and the structures of all six soluble components are now known (4, 5), but much remains to be learned about the packing of components in the assembled signaling complex, including the receptor stoichiometry. Moreover, the signaling complexes form larger clusters containing hundreds or thousands of receptors that are typically localized to one or both poles of the cell (2, 7, 8). All evidence indicates that ligand binding does not alter the oligomeric state of the receptor within the signaling complex and the larger receptor clusters (5). Thus receptor–receptor interactions, if present, occur within stable chemoreceptor oligomers, in contrast to other receptor systems, wherein ligand binding changes the receptor oligomeric structure (1).

It is well established that the minimal unit of receptor structure is a symmetric dimer of two identical subunits (9–12). Typically, this dimer binds a single molecule of attractant (13, 14). The dimer remains tightly associated when the receptor is extracted from the membrane by detergents (10). Similarly, the isolated, water-soluble periplasmic domain is a homodimer that retains its ability to bind a single attractant molecule (11, 12, 15). By contrast, the crystal structure of the isolated, water-soluble cytoplasmic domain reveals a trimeric assembly of three approximately symmetric dimers (16) as schematically illustrated in Fig. 1. This trimer of dimers is stabilized by hydrophobic and electrostatic interactions near the tip of the cytoplasmic domain farthest from the membrane. The same region of the domain is also responsible for the docking of CheW and CheA (17–19), thus in principle the trimer could be an artifact of nonspecific associations between sticky docking surfaces during crystal packing.

Figure 1.

Shown is the trimer of receptor dimers (16) proposed to form the core of the bacterial chemotaxis signaling complex. Arrows indicate strong positive cooperativity. (A) In a mixed oligomer with two serine receptor dimers, a lone aspartate receptor signals independently (22). (B) In a homooligomer, the three aspartate receptor dimers exhibit strong positive cooperativity modulated by their adaptation sites (20, 21). (C) In a higher-order receptor cluster (2, 7, 8, 33), aspartate receptors exhibit strong positive cooperativity within oligomers, but little or no cooperativity between oligomers.

Growing evidence from several parallel lines of research is now revealing that an oligomer of multiple receptor dimers, probably the trimer of dimers, indeed forms the core of the active signaling complex. In vitro and in vivo studies of attractant regulation of receptor-coupled kinase activity have uncovered cooperative interactions between receptors. Using the reconstituted complex containing the aspartate receptor Tar, CheW, CheA, and CheY, a quantitative in vitro analysis of the effect of attractant concentration on kinase activity has revealed that the Hill coefficient for attractant regulation ranges between 1 and 3 (20, 21). Independent in vivo studies using a fluorescence resonance energy transfer assay to monitor the activity of the signaling complex have also found Hill coefficients ranging between 1 and 3 for attractant regulation of the aspartate receptor (ref. 22; V. Sourjik and H. C. Berg, personal communication). The trimer of dimers is the simplest oligomer that can explain these in vitro and in vivo Hill coefficients, because the observed maximum Hill coefficient of 3 is consistent with the theoretical maximum for a trimeric system of binding sites. A rough correlation is observed between the Hill coefficient and the modification state of the receptor adaptation sites, with more highly modified receptors yielding the maximal Hill coefficient (21, 23). Thus, the positive cooperativity between identical receptor dimers can be modulated by their adaptation sites.

The work of Ames et al. (6) provides additional evidence for receptor–receptor interactions, including the first evidence that different types of receptors can form mixed oligomers. This elegant genetic study constructed engineered serine receptors with single mutations at positions implicated by the crystal structure in trimer stabilization. Many of the mutant serine receptors could inhibit signaling by both the wild-type serine and aspartate receptors when expressed in the same cell (6). Inhibition of the wild-type serine receptor can be explained by the formation of nonfunctional mixed dimers containing both a normal and mutant subunit. However, such dimer spoiling cannot account for inhibition of the aspartate receptor, because the serine and aspartate receptors do not form mixed dimers (6, 24). The simplest explanation is that defective serine receptor dimers assemble with aspartate receptor dimers in a higher-order oligomer, most likely the trimer of dimers (ref. 6 and Fig. 1A). In a few cases the presence of wild-type aspartate receptor actually restored the attractant response of a defective serine receptor, which again is easiest to explain if the aspartate and serine receptors can form a mixed oligomer (6). Further support for the existence of mixed oligomers was provided by chemical crosslinking experiments, which detected direct contact between the serine and aspartate receptors and showed that these interactions could be disrupted by trimer contact site mutations (6).

Recent Hill coefficient measurements using the in vivo fluorescence resonance energy transfer assay strongly support the conclusion that aspartate and serine receptor dimers can form a mixed oligomer, but suggest these receptors have little or no direct cooperative interaction. In particular, the Hill coefficient measured for attractant regulation of the aspartate receptor decreases from 3 in the absence of the serine receptor to 1 in the presence of excess serine receptor (ref. 22; V. Sourjik and H. C. Berg, personal communication). Such findings suggest that asparate receptor dimers in a pure oligomer can exhibit strong positive cooperativity, whereas a lone aspartate receptor dimer in a mixed oligomer with serine receptors signals independently (Fig. 1 A and B). Thus, cooperative interactions appear to be strongest between like receptors and can be weak or nonexistent between different receptor types. Significantly, in the noncooperative limit where the Hill coefficient approaches 1 (because of low adaptation site modification or the presence of excess serine receptor (21, 22), a single aspartate receptor dimer can independently generate a transmembrane signal even without the active participation of other dimers. It follows that the mechanism of transmembrane signaling (5) does not require cooperation between dimers. Finally, the most recent evidence supporting the trimer of dimers is provided by the discovery of engineered cysteine pairs in the aspartate receptor that can be oxidized to generate a covalent hexamer of receptor subunits in high yield (I. Kawagishi, personal communication).

The most controversial aspect of receptor–receptor communication remains its role in the function of polar clusters. These clusters represent a higher-order level of receptor organization beyond the trimer of dimers, because a cluster contains hundreds or thousands of receptors (2, 7, 8). As reviewed by Ames et al. (6), the existence of large clusters has led to the development of a theory proposing long-range cooperative interactions between receptors within the clusters, such that the binding of a single attractant molecule cooperatively switches the regulatory states of many nearby receptor dimers (25). An attractive feature of the theory is that it could help account for the remarkably high sensitivity exhibited by the chemotaxis pathway for attractants that signal through the aspartate receptor (22, 26–28). This sensitivity amplifies small changes in aspartate receptor occupancy to yield large changes in cytoplasmic kinase activity. Evidence supporting long-range cooperativity between receptors has been provided by a recent study using a multivalent ligand to drive artificial receptor clustering, which was found to increase the apparent attractant affinity (29). However, this artificial clustering was observed to decrease the rate of receptor adaptation, presumably by squeezing the adaptation enzyme CheR out of its native binding site, which could shift the apparent attractant affinity in the observed direction (21, 23). Evidence that challenges the existence of long-range cooperativity includes the maximal Hill coefficient of 3 observed in vitro and in vivo for attractant regulation of the aspartate receptor (20–22), which is an order of magnitude smaller than the value predicted by the long-range cooperativity model (25). Interestingly, Hill coefficients significantly greater than 3 have been measured in vitro for the serine receptor, suggesting this receptor may exhibit longer-range cooperative interactions between trimers of dimers than the aspartate receptor (23). Because the serine receptor is the most abundant receptor in the wild-type cell, it would be ideally suited for transmitting cooperative interactions between large numbers of receptors.

Overall, the available data for the aspartate receptor are consistent with short-range cooperative interactions between aspartate receptor dimers in the same trimer of dimers (Fig. 1C). To explain pathway sensitivity to attractants that signal through the aspartate receptor, a mechanism other than long-range receptor cooperativity must be proposed to amplify small changes in receptor occupancy. One model proposes that attractant binding to a given receptor inhibits the kinase activity of that receptor directly and also inhibits other receptors by increasing the concentration of a receptor inhibitor (30). The most likely candidate is unphosphorylated CheB because this protein is essential for pathway sensitivity (22, 28, 30). Even if long-range receptor interactions do not play a significant role in sensitivity, large receptor clusters are essential for the proper function of the adaptation system, which requires proximity between receptors because only a subset of the receptors possess tethered binding sites for the adaptation enzymes (31). Moreover, clustering of the pathway components reduces the number of slow, long-range diffusion steps that occur during pathway function. Diffusion can significantly impede rapid signal transduction; in fact, the diffusion of phospho-CheY from the signaling complex to the motor is probably the rate-limiting step in this pathway (32). Finally, clustering could serve to increase the local concentrations of signaling components while limiting crosstalk with related pathways in the same cell.