|Home | About | Journals | Submit | Contact Us | Français|
Many bacteria are motile and use a conserved class of transmembrane sensory receptor to regulate cellular taxis toward an optimal living environment. These conserved receptors are typically stimulated by extracellular signals, but also undergo adaptation via covalent modification at specific sites on their cytoplasmic domains. The function of the cytoplasmic domain is to integrate the extracellular and adaptive signals, and to use this integrated information to regulate an associated histidine kinase. The kinase, in turn, triggers a cytoplasmic phosphorylation pathway of the two-component class. The high-resolution structure of a receptor cytoplasmic domain has recently been determined by crystallographic methods and is largely consistent with a structural model independently generated by chemical studies of the domain in the full-length, membrane-bound receptor. These results represent an important step toward a mechanistic understanding of receptor-to-kinase information transfer.
Both prokaryotic and eukaryotic cells possess cell-surface sensory receptors that detect specific extracellular cues, generate a transmembrane signal and regulate intracellular pathways. In many cases, the receptor is activated by the binding of a ligand molecule and is also modulated by covalent modification at specific adaptation sites on its cytoplasmic surface. The cytoplasmic domain, in turn, transmits information to one or more cytoplasmic signaling proteins. Thus, the cytoplasmic domain plays a central role as an information integrator and relay system, and is essential to both signal transduction and adaptation to constant background signals.
This review summarizes recent structural advances that have shed light on the conserved cytoplasmic domains of the sensory receptors that regulate bacterial taxis pathways. Two independent structural studies using crystallographic and chemical approaches have characterized the cytoplasmic domains of the closely related Escherichia coli serine receptor and Salmonella typhimurium aspartate receptor, respectively [1••,2••]. In both proteins, the cytoplasmic domain is a dimer of two identical helical hairpins, yielding a highly extended four-helix bundle. In general, the new structures are in excellent agreement and, together, demonstrate that chemical and crystallographic approaches provide complementary information about structure and mechanism. The integration of the new structures with genetic and biochemical data yields novel insights into the mechanisms of receptor adaptation and kinase regulation.
In prokaryotes, the most common type of signaling pathway is the ubiquitous two-component pathway found in all prokaryotes to date, as well as in lower eukaryotes. Two-component pathways begin with a sensor protein, often a transmembrane receptor, which regulates a cytoplasmic histidine kinase, which, in turn, modulates an aspartate kinase. Together, the histidine and aspartate kinases comprise a conserved phosphorelay system [3–6].
A large superfamily of two-component sensor proteins is composed of receptors that regulate bacterial motility pathways [4,7,8]. These receptors, termed the methyl-accepting taxis proteins (MTPs), control cellular taxis and possess conserved cytoplasmic domains that are covalently methylated and demethylated by adaptation enzymes. MTP receptors direct a diversity of bacterial behavior, including chemotaxis, phototaxis, osmotaxis, thermotaxis, pH taxis and aerotaxis. Currently, 146 members of the MTP superfamily have been identified by sequence analysis .
The MTP superfamily can be divided into at least four subfamilies on the basis of the disposition of hydrophobic regions that are predicted to form transmembrane helices by hydropathy analysis . Family A is the largest and is characterized by two putative transmembrane helices separated by a substantial periplasmic domain. This family includes the enterobacterial chemotaxis receptors for aspartate, serine, ribose and galactose, peptides, citrate and osmolarity [7,8]. Family B members also possess two distinct putative transmembrane helices but, in this case, the hydrophobic regions are connected by only a short hydrophilic loop, so that a substantial periplasmic domain does not exist. Such an arrangement is illustrated by the receptors that associate with sensory rhodopsin I or II . Family C includes proteins that contain just one segment with a high transmembrane helix potential, such as the aerotaxis receptor [12,13]. Family D members appear to lack transmembrane helices altogether and are predicted to be peripheral membrane or soluble proteins .
All of the MTP receptors for which structural information exists are family A chemoreceptors from the homologous chemotaxis pathways of E. coli and S. typhimurium [1••,2••,4,7,15–27]. These chemoreceptors are homodimers and possess a large periplasmic domain formed by the association of two identical, antiparallel four-helix bundles, one from each subunit [15–19]. The transmembrane region is a single antiparallel four-helix bundle formed by the pairing of two membrane-spanning α helices from each subunit [20–27]. Thus, the periplasmic and transmembrane architectures of selected family A chemoreceptors are well understood and are likely to be representative of many family A members. In general, however, the structural details of the periplasmic and transmembrane regions are variable, as these regions are specialized to recognize different stimuli. For example, the sequence conservation among distantly related periplasmic domains is low and often this domain is missing entirely (families B and C), whereas the total number of transmembrane helices per dimeric receptor is predicted to range from four (families A and B, perhaps C) to two (perhaps family C) and to zero (family D).
In contrast to the specialized periplasmic and transmembrane regions, the cytoplasmic domain of MTP receptors (MTPCD) exhibits large regions of homology, even between the most distantly related sequences, indicating a highly conserved domain architecture [8,10]. These conserved regions are presumed to play ubiquitous roles in receptor adaptation and histidine kinase regulation. Indeed, functional chimaeric receptors have been generated by swapping the cytoplasmic domains of closely or distantly related MTP receptors, indicating that the cytoplasmic domain is a fully interchangeable signaling module [28–31]. Early circular dichroism, hydrodynamic and NMR studies of isolated MTPCDs have suggested that the domain structure is predominantly α-helical and highly elongated [32–34]. Moreover, sequence alignments have revealed a pattern of heptad repeats throughout large regions of the domain, in which the first and fourth positions of the heptad are typically hydrophobic, as expected for bundles of α helices in a coiled-coil arrangement [8,10]. Within the putative α-helical regions lie two distinct groups of conserved adaptation sites; these are specific glutamate sidechains targeted for methyl esterification and demethylation by adaptation enzymes [4–8]. Furthermore, the two putative methylation helices of each subunit have been proposed to associate to form an antiparallel four-helix bundle in the homodimer [4,8,35]. The most highly conserved region of the MTPCD is that which has been shown to be required for histidine kinase coupling and regulation [8,10]. This region, termed the signaling domain, has been found to fold independently and to regulate kinase activity when generated as an isolated fragment [36–39]. Another conserved region is the linker, which contains the HAMP motif and bridges the C terminus of the transmembrane domain and the N terminus of the cytoplasmic domain [40,41••,42].
Cysteine scanning and solvent exposure studies have probed the secondary structure of the cytoplasmic domain and have identified six regions of α helix, some of which could represent different sections of a longer continuous helix ([43–45,46••]; also S Winston, JJ Falke, unpublished data). These chemical studies analyzed a representative family A chemoreceptor, the S. typhimurium aspartate receptor, in its full-length, membrane-bound state. The solvent exposures of consecutive engineered cysteines were estimated from their chemical reactivities toward an aqueous, fluorescent alkylating agent. In many (but not all) regions examined, when cysteine was scanned through the sequence, the relative solvent exposures of consecutive positions displayed a helical periodicity, indicating the presence of α-helical structures with distinct exposed and buried faces.
The recent structural elucidation of the MTPCD illustrates the complementary information provided by chemical and crystallographic approaches. The primary advantages of the chemical approach, which employs site-directed sulfhydryl chemistry and spectroscopy to map out secondary structure and tertiary contacts, include the ability to probe structure in the full-length, membrane-bound protein in vitro or even in vivo [20–27,47–50]. This approach begins with a cysteine scanning and solvent exposure analysis of the targeted region, which, as described above, has identified extensive helical regions within the cytoplasmic domain. Subsequently, tertiary contacts are identified by measuring relative disulfide bond formation rates between pairs of cysteines, although considerable caution must be used in this analysis as the disulfide reaction is sensitive to local structure, electrostatics and dynamics, as well as to the cysteine–cysteine separation. Finally, the effects of the engineered cysteines and disulfides on protein activity are quantitated. This last step provides a wealth of functional and mechanistic information, as essential sidechains are identified by the activity loss that occurs upon cysteine substitution, whereas the effects of disulfide bonds on activity provide additional data regarding tertiary contacts. Activity-retaining disulfide bonds directly identify contacts between structural elements that are proximal in the working structure and, in a signaling protein, certain disulfide bonds can covalently lock the protein in the on or off state. The latter lock-on and lock-off disulfide bonds provide unique mechanistic insights and place strong constraints on models for on-off switching. The chief disadvantage of the chemical approach, however, is its relatively low resolution. By contrast, the spatial resolution of the crystallographic approach is unmatched, yielding essential structural details that are not accessible to other methods. Although it remains difficult to crystallize many transmembrane proteins, the approach can be applied to smaller independent folding units, including cytoplasmic domains, which can often be generated as soluble fragments.
The chemical and crystallographic methods have independently elucidated the structures of two MTPCDs from the closely related chemotaxis pathways of E. coli and S. typhimurium [1••,2••]. Each of these transmembrane receptors forms a stable ternary complex with the chemotaxis coupling protein CheW and the chemotaxis histidine kinase CheA, yielding a cytoplasmic receptor–kinase signaling complex that is regulated by periplasmic or cytoplasmic signals [4,5]. The chemical approach has elucidated the architecture of the cytoplasmic domain in the full-length, membrane-bound S. typhimurium aspartate receptor, whereas the crystallographic approach has solved the high-resolution structure of a major 227-residue fragment from the cytoplasmic domain of the E. coli serine receptor. Overall, the cytoplasmic domains of these receptors are 341 and 337 residues in length, respectively, and exhibit 71% sequence identity, including 86% identity in the signaling subdomain [8,10]. Position numbers are offset by two residues between these two cytoplasmic domains, such that homologous position numbers in the serine receptor are two larger than in the aspartate receptor (except for the C-terminal 10 residues, where a deletion has occurred in the serine receptor). Both new structures reveal that the cytoplasmic domain is a homodimeric, four-helix bundle formed by the intimate association of two long α-helical hairpins, one from each subunit. Schematic Figure 1 displays the cytoplasmic domain in the context of the full-length receptor and identifies the functionally important regions.
The chemical analysis of the cytoplasmic domain architecture, carried out in the intact, membrane-bound aspartate receptor [2••], began with the prior knowledge of large helical regions defined by the aforementioned cysteine scanning and solvent exposure studies ([43–45,46••]; also S Winston, JJ Falke, unpublished data). The packing of these helical regions was elucidated by measuring disulfide bond formation rates between a set of 44 cysteine pairs engineered at buried positions to test for specific helix–helix contacts within and between different subunits [2••]. Altogether, 21 of the 44 cysteine pairs rapidly formed disulfide bonds, whereas 23 formed slowly or not at all. This strategy tested only a small subset of the combinatorial possibilities, but the observed 21 rapidly forming cross-links provided sufficient information to define the relative helix positions. Moreover, the 44 cysteine pairs selected for the study were carefully designed to resolve the two conflicting published models of the cytoplasmic domain structure [4,7]. Indeed, the observed pattern of disulfide bond formation excludes a compact architecture formed by the side-to-side association of two short helical bundles; instead, the results establish the architecture of the domain as the extended four-helix bundle displayed in the schematic Figure 2.
Further confirmation of the extended four-helix bundle structure was provided by an analysis of the effects of 188 intersubunit disulfide bonds on receptor activity in the reconstituted receptor–CheW–CheA signaling complex [2••]. Figure 2 displays the seven disulfide bonds found to retain receptor-regulated kinase activity, as well as the seven disulfides that lock the receptor in the kinase-activating state, in both the absence and the presence of saturating aspartate ligand. The signal-retaining and lock-on disulfide bonds bridging the two subunits at the hairpin turn (A387C–A387C′ and F394C–F394C′, respectively) further exclude the compact tertiary model, while supporting the extended structure. Moreover, the presence of the cluster of intersubunit lock-on disulfides in the vicinity of the adaptation sites highlights the importance of this interfacial region to kinase regulation, as discussed further below.
The crystal structure of the serine receptor fragment, corresponding to residues 294–520, provides a very similar picture of the cytoplasmic domain [1••]. The crystal structure, determined to 2.6 Å resolution and illustrated in Figure 3, also reveals an extended four-helix bundle in which the hairpin turn of each subunit lies at the same location identified chemically in the full-length receptor. Moreover, the four helices of the bundle show the same exposed and buried faces detected by the chemical approach. The only significant differences between the two structures are observed near the ends of the fragment, where, during crystal formation, the helical ends of one dimer form coiled coils with two other dimers, causing the four-helix bundle to splay apart in the vicinity of the adaptation sites. This separation of the bundle is not observed in the full-length receptor, wherein multiple signal-retaining and lock-on disulfide bonds are formed between the four helices in the adaptation site region [2••]. Future crystallographic studies of longer cytoplasmic fragments or of type D soluble cytoplasmic domains will ultimately yield a high-resolution view of the adaptation sites.
The structural detail provided by the crystallographic approach reveals extensive supercoiling of the helices in the four-helix bundle [1••], as seen in Figure 3. This supercoiling explains the heptad repeating pattern of exposed and buried positions observed in the aspartate receptor using cysteine scanning and solvent exposure measurements, and is a conserved feature of MTPCDs, as indicated by sequence alignment [8,10,43–45,46••]. Furthermore, the crystal structure reveals a second heptad repeat superimposed on the first to yield a ‘double-heptad repeat’, arising from the fact that each helix is supercoiled with two other helices [1••]. Overall, assuming that the supercoiled helices extend at least from the adaptation sites to the hairpin turn, the length of the cytoplasmic domain exceeds 130 Å. Thus, as the periplasmic and transmembrane regions are approximately 80 and 40 Å in length, respectively, both the serine and aspartate receptors are highly extended, rod-like molecules with lengths exceeding 250 Å. In fact, if the linker region is also helical, these receptors could be 380 Å in length, as in the model shown in Figure 4. Such extended structures may allow a small ligand-induced conformational or dynamic change in the periplasmic domain to be amplified into a larger change at the regulatory receptor–kinase interface in the cytoplasm [23,26,51].
One interesting feature of both the chemical and the crystallographic structures is a group of hydrophobic residues near the hairpin turn of each subunit [1••,2••]. In the crystal, these hydrophobic sidechains, as well as hydrophilic sidechains from the same dimer, interact with residues from two other dimers to form a trimer of dimers, as shown in Figure 5. This observation may help explain the clustering of the E. coli chemoreceptors that is known to occur under certain conditions in vitro and in vivo [52–54,55••]. Moreover, the exposed hydrophobic surface could represent a docking site for CheA or CheW, as previously proposed [46••], such that the trimer would separate into individual dimers upon formation of the receptor–CheA–CheW ternary complex. Additional studies are needed to ascertain the physiological roles of the hydrophobic surface and the trimer.
Genetic and biochemical studies carried out in several laboratories have highlighted the functional importance of specific regions of the cytoplasmic domain. Firstly, studies of aspartate receptor heterodimers in which one cytoplasmic domain is truncated have shown that the linker region is essential for function, suggesting its importance to critical intersubunit or intrasubunit interactions [56,57]. In addition, cysteine scanning and other studies have revealed a high density of mutations in this region that block receptor accumulation or prevent function [42,44]. The poorly characterized structure of the linker region lies outside the region characterized by crystallographic analysis, but is partially defined as α-helical by cysteine scanning and solvent exposure measurements, and by the occurrence of heptad repeat patterns in its primary structure [8,10,40,44]. Secondly, the new structural evidence reviewed herein indicates that the adaptive methylation sites of the dimer lie in close proximity within a narrow band wrapping around the exposed face of the four-helix bundle [1••,2••]. Numerous studies have revealed evidence for the functional importance of the adaptation site region, especially the region containing the three adaptation sites on helix HP1. Mutagenic analysis of the region has identified it as a key sensor of not only adaptation signals in chemotaxis, but also temperature in thermotaxis . Moreover, site-directed mutations and engineered intersubunit disulfide bonds demonstrate that the region is tightly coupled in an unknown way to CheA kinase activation, so that modifications of buried positions within the region often lock the kinase in its activated state [43,59•]. These findings suggest that the helix–helix packing within the adaptation region of the four-helix bundle is critical for kinase regulation. Thirdly, the signaling subdomain retains at least partial kinase regulation as an isolated fragment and mutagenesis studies have implicated this region in the docking of CheA and CheW to form the ternary complex [36–39,45,46••,60]. The new structures are in excellent agreement within the signaling subdomain; both reveal a four-helix bundle composed of two helical hairpins [1••,2••]. The exposed hydrophobic residues lying on the surface of the dimer, near the hairpin turn, are probably involved in the intermolecular interactions that stabilize either the ternary complex or receptor clustering, as noted above [46••,53,60].
Although the recently determined structures of MTPCDs represent a first step toward a structural understanding of receptor adaptation and kinase regulation, a number of important pieces are still missing from the mechanistic puzzle. The conserved linker region connecting the second transmembrane helix to the cytoplasmic domain is widely observed in MTP receptors and certain other proteins as well, and is known to be essential for effective signal transduction by chemotaxis receptors. The structure of this linker remains unknown, although it is bordered by α helices and appears to include helical elements [40,41••,42–44]. Assuming that these helical regions extend throughout the linker, the simplest working model for the linker structure is an α-helical coiled coil formed by the association of two helices, one from each subunit of the dimer (Figure 4). As the linker appears to be essential for the transmission of information from the transmembrane domain to the cytoplasmic domain, its structural elucidation remains a high priority. Another priority is a high-resolution view of the cytoplasmic four-helix bundle in the vicinity of the adaptation sites, which is crucial to obtaining a molecular understanding of adaptation and the coupling of this region to kinase regulation [59•]. Moreover, although three-dimensional structures are now available for the cytoplasmic domain, CheA and CheW ([61••,62–64]; FW Dahlquist, personal communication), the structure of the assembled ternary complex remains unknown. Further structural studies will probe the architecture of the assembled complex using the cytoplasmic domain fragment or a soluble family D receptor as the framework for CheA and CheW docking. Together, high-resolution views of the adaptation sites and the ternary complex will shed light on the mechanism by which the transmembrane and adaptation signals are transmitted by the cytoplasmic domain to the histidine kinase. Finally, recent biochemical studies have revealed evidence of limited cooperative interactions between chemotaxis receptors during kinase regulation, and chemoreceptors are known to cluster at the poles of E. coli in living cells [53,54,55••,65••,66••]. Such clustering is critical to the adaptation of certain receptors that lack the C-terminal docking site for adaptation enzymes that is present on other receptors, including the aspartate and serine receptors [67–69]. It is not yet clear, however, whether clustering plays an important role in transmembrane kinase regulation. The trimer of dimers observed in the crystal structure of the cytoplasmic domain may be involved in such higher order interactions, but elucidating the functional role and mechanism of receptor clustering requires further study.
We thank our collaborators at the University of Colorado, University of California and Lawrence Berkeley National Laboratories for their essential contributions, including Randal Bass, Scott Butler, Steve Chervitz, Matthew Coleman, Mark Danielson, Kyeong Kyu Kim, Matthew Trammel, Susanna Winston and Hisao Yokota. This work was supported by National Institutes of Health grants R01 GM-40731 (JJF) and R01 CA-78406 (S-HK), and by funds from the Director, Office of Science, Office of Biological and Environmental Research, US Department of Energy under contract number DE-AC03-76SF00098 (S-HK).
Papers of particular interest, published within the annual period of review, have been highlighted as:
• of special interest
•• of outstanding interest