|Home | About | Journals | Submit | Contact Us | Français|
Site-directed sulfhydryl chemistry and spectroscopy can be used to probe protein structure, mechanism and dynamics in situ. The aspartate receptor of bacterial chemotaxis is representative of a large family of prokaryotic and eukaryotic receptors that regulate histidine kinases in two-component signaling pathways, and has become one of the best characterized transmembrane receptors. We report here the use of cysteine and disulfide scanning to probe the helix-packing architecture of the cytoplasmic domain of the aspartate receptor.
A series of designed cysteine pairs have been used to detect proximities between cytoplasmic helices in the full-length, membrane-bound receptor by measurement of disulfide-bond formation rates. Upon mild oxidation, 25 disulfide bonds form rapidly between three specific pairs of helices, whereas other helix pairs yield no detectable disulfide-bond formation. Further constraints on helix packing are provided by 14 disulfide bonds that retain receptor function in an in vitro kinase regulation assay. Of these functional disulfides, seven lock the receptor in the conformation that constitutively stimulates kinase activity (‘lock on’), whereas the remaining seven retain normal kinase regulation. Finally, disulfide-trapping experiments in the absence of bound kinase reveal large-amplitude relative motions of adjacent helices, including helix translations and rotations of up to 19 Å and 180°, respectively.
The 25 rapidly formed and 14 functional disulfide bonds identify helix–helix contacts and theirregister in the full-length, membrane-bound receptor–kinase complex. The results reveal an extended, rather than compact, domain architecture in which the observed helix–helix interactions are best described by a four-helix bundle arrangement. A cluster of six lock-on disulfide bonds pinpoints a region of the subunit interface critical for kinase activation, whereas the signal-retaining disulfides indicate that signal-induced rearrangements of the four-helix bundle are small enough to be accommodated by disulfide-bond flexibility (≤ 1.2 Å). In the absence of bound kinase, helix packing within the cytoplasmic domain is highly dynamic.
Site-directed sulfhydryl chemistry and spectroscopy can provide diverse information ranging from the backbone architecture of a protein of unknown structure to the molecular mechanism and backbone dynamics of a structurally defined protein [1–33]. A unique feature of this approach is that it can be applied to functional soluble proteins, membrane proteins and multiprotein complexes in their native environment, enabling molecular studies of structure, mechanism and dynamics in situ. In this technique, directed mutagenesis is used to introduce cysteine residues and thus target the sulfhydryl group and its diverse chemistry to specific regions of a protein. Cysteine substitutions are typically relatively nonperturbing, in part owing to the flexibility and small size of the cysteine sidechain. In addition, the polarity of the sulfhydryl group can adjust to its environment: in nonpolar surroundings the sulfhydryl is fully protonated and relatively hydrophobic, whereas in a polar environment a significant degree of sulfanion character can develop. Intrinsic cysteine residues can often be replaced by alanine or serine without significant perturbation ; alternatively, if the intrinsic cysteines are far from the engineered cysteine positions they can be retained. The introduced sulfhydryl group in the mutant protein provides a chemical labeling site that can be selectively modified by alkylating agents under mild conditions, enabling easy coupling to chemical or spectroscopic probes [1,3,7,14,17,22,23]. Alternatively, pairs of cysteines can be cross-linked via disulfide formation or bridging bifunctional reagents [1–3,5,6,13].
A number of methods exploiting engineered cysteines have been developed in several laboratories and successfully applied to a range of protein systems. These include sulfhydryl reactivity as a measure of solvent exposure or membrane topology [1,3,7]; disulfide cross-linking as a measure of proximity [3,18,31]; disulfide trapping as a detector of long-range backbone motions [2,4]; site-directed fluorescence spectroscopy [1,3,7]; site-directed spin label electron paramagnetic resonance (EPR) spectroscopy [22–26]; cysteine-scanning mutagenesis to detect essential sidechains ; disulfide scanning to map out intramolecular contacts or to identify lock-on disulfides [5–9]; cross-linking of protein fragments by disulfide bonds or chemical reagents to detect intramolecular contacts [13,14,27]; formation of metal-binding sites to characterize solvent exposure or intramolecular proximity [29,30]; substituted cysteine accessibility method (SCAM) [15,17]; and protein interactions by cysteine modification (PICM) . These methods have illuminated the physical and functional features of soluble proteins, extrinsic membrane-binding proteins, transmembrane receptors, transporters and channels, large protein complexes and even proteins in living cells [4,6,9,14,17,20,21,32]. The information provided is not at atomic resolution, but can map out the secondary structure and the overall fold of a tertiary structure and can define intermolecular docking surfaces [5,9,31]. If a structural model is available, experiments can be devised to resolve different mechanistic models or to characterize conformational changes and thermal backbone dynamics [4,33]. In signaling proteins, an appropriately positioned disulfide bond can reversibly lock the molecule in its ‘on’ or ‘off’ state [6,7].
The aspartate receptor is one of six closely related cell-surface receptors in Escherichia coli and Salmonella typhimurium that regulate cellular chemotaxis [34–38]. Homologous taxis receptors are widely distributed in prokaryotes, and belong to a large superfamily of receptors that regulate ubiquitous two-component signaling pathways in prokaryotes and lower eukaryotes [39–43]. Crystallographic studies have defined the structure of the periplasmic ligand-binding domain of the aspartate receptor, revealing a homodimeric pair of four-helix bundles (Figure 1a) [44–46]. In the initial structural determination, an engineered disulfide bond facilitated crystallization of the ligand-binding domain by stabilizing its subunit interface . In the full-length homodimeric receptor, the ligand-binding domain is continuous with the transmembrane region, wherein disulfide-mapping studies have shown that two helices from each subunit extend across the membrane to yield a four-helix bundle with well defined packing faces [5,6,18–20,31].
The transmembrane signal has been identified as a piston displacement of one of the four transmembrane helices, α4/TM2, toward the cytoplasm upon aspartate binding . This displacement is observed in crystal structures of the isolated periplasmic domain, and has been confirmed in situ by lock-on and lock-off disulfide bonds that trap the signaling helix at the extremes of the piston movement, thereby locking the associated kinase activity on and off, respectively [6,33]. EPR spectra that monitor the interaction between pairs of spin labels coupled to engineered cysteines on adjacent helices have identified the piston displacement as the largest component of the ligandinduced rearrangement . Independent studies of a related chemoreceptor have detected the same piston displacement by the changes in patterns of interhelix disulfide bond formation rates that are triggered by ligand binding . Disulfide bonds that prevent the displacement of the signaling helix strongly inhibit regulation of the kinase, as predicted by the piston model ; in contrast, disulfide bonds that covalently stabilize the periplasmic or transmembrane regions of the dimer interface have little or no effect on kinase regulation, indicating that a global rearrangement of the periplasmic and transmembrane subunit interface is not essential for signal transduction . In short, site-directed cysteine chemistry and spectroscopy have played a major part in defining the structure and mechanism of both the periplasmic ligand-binding and transmembrane signaling domains.
The C-terminal end of the transmembrane signaling helix emerges into the cytoplasm, where it couples directly to the cytoplasmic domain. Structural analysis of the isolated cytoplasmic domain has been slowed by its extensive dynamics, although hydrodynamic and circular dichroism studies suggest an extended, largely helical conformation [48–50]. Each subunit possesses four cytoplasmically exposed adaptation sites consisting of glutamate sidechains that are methyl esterified and demethylated by the methyl transferase (CheR) and esterase (CheB) components respectively of the receptor adaptation pathway [36,37]. The cytoplasmic domain also associates with the dimeric histidine kinase CheA, forming a ternary complex stabilized by the coupling protein CheW [51–54]. The transmembrane and adaptation signals regulate CheA autophosphorylation, thereby modulating the rate of phosphotransfer from phospho–CheA to the two response regulators CheB and CheY that control the adaptation response and swimming motor, respectively [36,37].
Early studies of cytoplasmic domain architecture focused on secondary structure predictions and patterns of hydrophobicity in the primary structure, which implied the presence of α helices participating in coiled coils or four-helix bundles [55,56]. In particular, sequence alignments of the cytoplasmic domains of 56 related taxis receptors revealed conserved heptad repeat patterns characteristic of coiled or bundled helices [39,40]. These sequence alignments suggest that each cytoplasmic domain subunit possesses up to five distinct α-helical regions termed α5 to α9. Figure 2 illustrates the locations of these putative α helices in the primary structure. Previous experimental studies of secondary structure have scanned a single-cysteine substitution through the indicated regions of the cytoplasmic domain (see Figure 2c). The measured chemical reactivities of these cysteines have revealed periodic patterns of solvent exposure and burial that are characteristic of α-helical elements with exposed and buried faces [7–9,57] (SE Winston and JJF, unpublished observations). These experimental analyses yielded essentially the same five α-helical regions deduced from the sequence alignments, and the agreement between these complementary approaches provides strong evidence that the cytoplasmic domain is dominated by amphiphilic α helices. It should be emphasized that the ends of the helical regions are not precisely determined by either the sequence analysis or the chemical reactivity scan. For example, it is possible that the regions termed α5, α6 and α7 are sections of one continuous helix, whereas regions α8 and α9 could represent a second long helix. For brevity, we will treat these regions as distinct helices.
Disulfide-scanning studies have begun to define the packing interactions of the five cytoplasmic helices. Certain engineered cysteine positions on the buried faces of helices α5 and α6 rapidly form symmetric, α5–α5′ or α6–α6′ disulfide bonds that span the subunit interface [8,58]. Moreover, a subset of these disulfides yields constitutive kinase activation in vitro in the receptor-mediated kinase regulation assay . These results indicate that helices α5 and α6 are located at the subunit interface, where they pack against their symmetric counterparts α5′ and α6′ from the other subunit to form parallel helix–helix interactions. Other disulfide-scanning studies have indicated that helices α7 and α8 are separated by a sharp turn that enables them to form a helical hairpin [9,57]. The existence of this hairpin is supported by matching indels (insertions or deletions of identical lengths) in the α7 and α8 helices, which are detected by multiple sequence alignment [39,40]. Matching indels also suggest a direct interaction between helices α6 and α9, which together possess all of the four adaptation sites. The putative α6–α9 interaction has, however, not yet been tested by disulfide scanning.
Most of the available evidence is consistent with two different helix-packing models for the cytoplasmic domain. The extended model (Figure 1a) places the α7–α8 helical hairpin at the dimer interface, where it associates with the symmetric α7′–α8′ hairpin of the other subunit to form a four-helix bundle [36,39]. This model is consistent with the observation that an isolated fragment corresponding to the putative hairpin region, termed the signaling domain, is a stable independent structure that can couple to the CheA kinase and regulate its activity [9,56,59]. In the compact model (Figure 1b), by contrast, the α7–α8 hairpin is packed against the methylation helices α6 and α9, such that the hairpin is distant from the subunit interface . This model is supported by preliminary estimates of the length of the cytoplasmic domain, which favor the shorter length of the compact model [60,61].
The present study is designed to resolve the extended and compact models, to further investigate the mechanism of signaling and kinase regulation by the cytoplasmic domain, and to probe helix dynamics in the hairpin region. The two folding models yield contrasting predictions regarding the proximity of specific helix pairs, which can be tested by interhelix disulfide-bond formation between engineered cysteine pairs. Our disulfide-scanning analysis reveals helix–helix interactions between α7 and α7′, α8 and α8′, and α7 and α8, but not between α6 and α7, nor α6 and α8. These findings disfavor the compact model, but are consistent with the extended model of cytoplasmic domain structure. The effects of interhelix disulfide bonds on receptor function are also examined in order to determine the location of helix-packing surfaces that are tightly coupled to kinase activation. Interestingly, in the absence of bound kinase, the helices of the cytoplasmic domain exhibit extensive thermal movements, as detected by disulfide trapping.
In a dynamic protein, two nearby cysteines can collide and, under oxidizing conditions, form a disulfide bond. Collision frequency depends on their distance apart and thus the disulfide-bond formation rate can provide limited information on the cysteine–cysteine separation. Studies of disulfide-bond formation rates in a protein of known structure have shown that the rate generally decreases as the average distance between the cysteines increases; however, local sterics, electrostatics and dynamics also modulate the reaction . We have tested the predictions of extended and compact models for the cytoplasmic domain by engineering pairs of cysteines into the following helix combinations: α7–α7′, α8–α8′, α7–α8, α6–α7 and α6–α8 (where the prime distinguishes the symmetric helix from a different subunit). Subsequently, E. coli membranes containing the full-length receptors were isolated and subjected to mild oxidation conditions for 1 min at 25°C (catalyst 2.0 mM Cu(II)(1,10-phenanthroline)3 buffered with 5 mM EDTA), followed by quenching and analysis of the extent of disulfide-bond formation. This mild oxidation reaction provides comparatively slow disulfide-bond formation rates and optimizes the specificity of disulfide-bond formation for proximal pairs of cysteines [8,58]. Stronger oxidation conditions are used in applications requiring complete disulfide formation ([5,6] and below). Figure 3a illustrates the use of sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS–PAGE) to resolve different reaction products: formation of an intrasubunit disulfide bond yields a small increase in the mobility of the receptor monomer because of the introduction of an intramolecular loop, while an intersubunit disulfide bond greatly slows the mobility due to the formation of a covalent dimer.
Table 1 summarizes the 44 cysteine pairings tested, and the extent of inter- or intrasubunit disulfide formation observed for each pairing. All positions chosen for cysteine substitution were located on the buried face of each helix [7–9,57], with the exception of the α7–α8 helix pairings where cysteine substitutions were introduced on all faces of helix α7 to test for helix rotations (see below). For the α6–α8, α6–α7 and α7–α8 pairings, the cysteine pairs were designed to cover a wide range of possible helix–helix registries, as the exact registries are not known and remain model dependent. Rapid disulfide-bond formation was observed for 25 cysteine pairs in the α7–α7′, α8–α8′ and α7–α8 helix pairings (Table 1). In contrast, no disulfide-bond formation was detected for any of the nine cysteine pairs tested in the α6–α7 and α6–α8 pairings. These negative results are especially striking given the extensive thermal dynamics observed in the cytoplasmic domain, which allow long-range collisions between cysteine pairs on nearby helices (see below). Overall, the results are consistent with the helix–helix contacts proposed by the extended model, but are difficult to reconcile with the predictions of the compact model.
Helix–helix interactions can also be detected through the effects of interhelix disulfide bonds on receptor-mediated kinase regulation. This approach avoids the potential collisional promiscuity of disulfide-bond formation rates, and instead defines precise contacts between adjacent residues in the assembled receptor–kinase signaling complex. Thus, disulfide scanning was used to test all the available cytoplasmic cysteine locations for formation of a functional, intersubunit disulfide linkage. The library of 188 single-cysteine substitutions spans the regions delineated in Figure 2, including the entire linker and α5 regions, and large segments of regions α6 to α9 [7–9,57] (SE Winston and JFF, unpublished observations). Because of the homodimeric nature of the receptor, a given single-cysteine substitution introduces a pair of symmetric cysteines into the dimer [2,3,44]. Each single-cysteine mutant receptor was isolated in E. coli membranes and subjected to strong oxidation conditions for 20–30 min at 37°C (catalyst 0.2 mM Cu(II)(1,10-phenanthroline)3), thereby driving intersubunit disulfide-bond formation towards completion. Typically, this oxidation yielded 80–100% covalent dimer formation as measured by SDS–PAGE. Subsequently, the purified soluble components CheA, CheW and CheY were added to reconstitute the receptor-regulated phosphorylation pathway [5,52–54]. Using conditions that ensure that the autophosphorylation of CheA is rate limiting, the ability of the disulfide-containing receptor to activate CheA in the absence of attractant, and to inhibit CheA upon addition of aspartate, was measured by monitoring the formation rate of product, [32P]-phospho–CheY.
Table 2 summarizes the 14 functional disulfide bonds found so far in the cytoplasmic domain. Altogether, seven signal-retaining disulfide bonds were detected, defined as those that retain at least 20% of the wild-type kinase activation capacity of the apo receptor, as well as full kinase downregulation upon addition of aspartate: T253C–T253C′, S260C–S260C′, G271C–G271C′, S272C–S272C′, A304C–A304C′, A387C–A387C′ and G467C–G467′C. The remaining seven lock-on disulfides were found to lock the receptor constitutively in its ‘on’ state, yielding at least 20% of maximal wild-type kinase activation even when the ligand-binding site is saturated with aspartate. These were located at: G278C–G278C′, G285C–G285C′, L300C–L300C′, F394C–F394C′, T482C–T482C′, N485C–N485C′ and S492C–S492C′. All 14 functional disulfide bonds were formed between symmetric buried positions on adjacent helices at the subunit interface. Four were previously reported: three lock-on disulfides (G278C–G278C′, G285C–G285C′ and L300Cys–L300C′) and one signal-retaining disulfide (A387C–A387C′) [7,9]. Together, the 14 functional disulfides demonstrate that the kinase-activating state of the receptor possesses helix–helix contacts between α5 and α5′ (T253C–T253C′), between α6 and α6′ (S260C–S260C′, G271C–G271C′, S272C–S272C’, G278C–G278C′, G285C–G285C′, L300C–L300C′ and A304C–A304C′), between loops α7–α8 and α7′–α8′ (A387C–A387C′ and F394C–F394C′), and between α9 and α9′ (G467C–G467C′, T482C–T482C′, N485C–N485C′ and S492C–S492C′). The seven signal-retaining disulfides, indicate that signal-induced rearrangements of the interfaces between α5–α5′, α6–α6′ and the α7–α8–α7′–α8′ loops are small enough to be accommodated by the limited flexibility of the cysteine disulfide linkage.
It is important to ascertain whether the observed functional disulfide bonds are intramolecular, that is between the two subunits within a dimer, or are intermolecular, forming during dimer–dimer collisions. To identify those residues that form intra- or interdimer disulfide bonds, each of the 14 single-cysteine substitutions was combined with N36C to yield a double-cysteine mutant. The periplasmic N36C–N36C′ disulfide bond is known to form exclusively between two subunits of the same dimer [2,3,44]; therefore, when the double mutant is oxidized and visualized by SDS–PAGE the products will be dominated by either a dimer containing two intersubunit disulfide bonds, if the cytoplasmic cysteine forms an intradimer cross-link, or by tetramers and higher-order oligomers if the cytoplasmic cysteine forms an interdimer cross-link. All 14 functional disulfide bonds were found to form within the dimer, as illustrated for the A387C–A387C′ disulfide bond in Figure 3b. For the N36C/A387G double mutant, the time course of oxidation shows that the intradimer N36C–N36C′ disulfide bond is generally formed first, followed by the A387C–A387C′ disulfide to yield a doubly cross-linked dimer. The latter doubly cross-linked dimer is easily resolved because its mobility in SDS–PAGE is intermediate to the mobilities of the two singly cross-linked dimers. Thus, the 14 functional disulfides provide strong constraints on helix–helix contacts at the subunit interface of the fully active, membrane-bound receptor–kinase complex. Moreover, these disulfides have important implications for the mechanism of cytoplasmic domain signaling.
Finally, in order to probe the relative motions of the adjacent α7 and α8 helices, 16 dicysteine receptors were constructed, each possessing the A411C cysteine on helix α8 and a second cysteine on helix α7 in the same subunit. Table 1 shows that nine of the 16 interhelix cysteine pairs were able to collide and form detectable levels of disulfide bond during mild oxidation for 1 min at 25°C. It follows that A411C on helix α8 is able to collide with a 19 Å length of helix α7 from V365C through A377C. Furthermore, a subset of these collisions requires helix α7 to rotate at least 180° about its long axis. There are limits to the observed motions, however, as cysteine positions lying outside the V365C to A377C region of helix α7 failed to form detectable levels of disulfides with the A411C probe position on the adjacent helix. Given the significant amplitudes of the detected helix motions, the observation of a rapidly forming interhelix disulfide bond places the two helices in the same vicinity, but does not define the precise register of their packing.
Our results resolve the compact and extended models of the cytoplasmic domain illustrated in Figure 1b. Rapid disulfide-bond formation is observed in the membrane-bound receptor between cysteine pairs located on the α7–α7′ and α8–α8′ helices (Table 1), indicating that the α7–α8 helical hairpin is packed at the subunit interface against its symmetric partner as predicted by the extended model, rather than against the methylation helices as proposed by the compact model . In contrast, no rapid disulfide-bond formation is detected between cysteine pairs localized to the α6–α7 or α6–α8 helices (Table 1), despite their predicted close proximity in the compact model. Finally, the signal-retaining and lock-on disulfide bonds formed between the loops of the two symmetric α7–α8 hairpins (A387C–A387C′ and F394C–F394C′ respectively) show that the loop lies near the subunit interface in the receptor conformation that activates the kinase (Table 2), as predicted by the extended model.
The resulting extended model is summarized in Figure 4. This model possesses two colinear four-helix bundles containing the α6, α9, α6′ and α9′ methylation helices and the α7, α8, α7′ and α8′ signaling-domain helices, respectively . Because the ends of the cytoplasmic helices are not well defined, it is possible that regions α6 and α7 combine to form one long continuous helix, as could α8 and α9. In this case the domain would contain a single four-helix bundle up to 170 Å long. Various pieces of evidence support the existence of a four-helix bundle in the methylation region: rapid disulfide-bond formation between buried positions at the α6–α6′ interface [8,58]; similar lengths of the α6 and α9 helices as judged by the number of consecutive heptad repeats in the sequence alignment [39,40]; matching indels in the aligned sequences of the α6 and α9 helices that suggest direct packing interactions between them [39,40]; and functional disulfide bonds that place both α6 and α9 near the subunit interface (Table 2). Moreover, both the α6 and α9 helices have one or more sites of adaptive methyation [36,37], which may well need to be close together in space so that they can be modified by the CheR methyltransferase which is bound to the receptor by a tether of limited length . It should be emphasized, however, that the proposed α6–α9 interaction has still to be confirmed experimentally.
Turning to the signaling domain, several lines of evidence support a four-helix bundle in this region: rapid disulfide-bond formation between buried positions at the α7–α7′ and α8–α8′ interfaces (Table 1); matching indels in the α7 and α8 helices detected in the sequence alignment [39,40]; and the two functional disulfides that place the α7–α8 loop at the subunit interface (Table 2). No functional intersubunit disulfides have yet been found at the α7–α7′ and α8–α8′ interfaces, but this is the most highly conserved region in the entire sequence of the receptor [39,40]. Thus, the packing of this region may be too sensitive for the successful introduction of a functional disulfide bond. Alternatively, it is possible that helices α7 and α8 of the signaling domain undergo major structural rearrangement when the kinase docks, which would be blocked by an interhelix disulfide bond in this region. Our results cannot rule out such a rearrangement; however, the two functional disulfides in the interhelix loop do constrain this loop region to the subunit interface in the working signaling complex. Moreover, the kinase docking surface predicted by the PICM assay maps to a logical location in the proposed structure, as illustrated in Figure 4 [9,57].
The 14 functional disulfides that retain kinase activation provide the strongest constraints on helix packing in the active receptor–kinase complex. The range of allowed Cβ–Cβ distances in protein disulfides spans 1.2 Å [4,63]; thus, to a first approximation the translational flexibility of a disulfide linkage will be limited to this range. It follows that disulfide bonds that retain function are formed between pairs of positions that are adjacent, or even in van der Waals contact, in the active structure. Functional, intersubunit disulfide bonds are observed between helices α5-α5′, α6–α6′, α9–α9′, and between the loops of the two α7–α8 hairpins (Table 2), placing each of these structural components at the subunit interface in the on-state of the receptor–kinase complex. For the α6–α6′ and α9–α9′ interfaces, as well as the α7–α8 loop–loop′ interface, both lock-on and signal-retaining disulfides are detected, indicating that these contacts are tightly coupled to kinase activation and undergo a signal-induced rearrangement that is small enough to be accommodated by the flexibility of the disulfide linkage (≤ 2 Å). Such small rearrangements are consistent with the subtle 1.6 Å displacement of the α4/TM2 signaling helix detected in the periplasmic and membrane-spanning regions during transmembrane signaling ; thus signaling appears to induce small structural changes in both the external and internal receptor domains [64,65]. An important difference between the transmembrane and cytoplasmic signals, however, is that the transmembrane signal does not require rearrangements of the membrane-spanning region of the subunit interface [5,33], whereas small rearrangements in the cytoplasmic region of the subunit interface appear to be directly involved in kinase regulation [7,66].
Interestingly, all six of the lock-on disulfides in the α6 and α9 helices are clustered in a defined region of the subunit interface, bordered by signal-retaining disulfides on either side. Such structural constraints place strong limits on the mechanism of receptor-mediated kinase regulation. The results are consistent with a model in which these lock-on disulfides stabilize particular helix–helix contacts at a critical location within the four-helix bundle, thereby trapping the bundle in its kinase-activating conformation as previously proposed for the α6–α6′ contact in this region .
Finally, the disulfide trapping experiment described for the α7 and α8 helices provides a graphic view of the thermal backbone motions that occur in the membrane-bound receptor lacking the kinase, as illustrated in Figure 5. It has been shown previously that the isolated domain has a dynamic structure , and the range of motions observed here confirm that these dynamics are retained in the full-length, membrane-bound receptor. Disulfide bonds are observed to form between A411C on helix α8 and nine engineered cysteines ranging between V365C and A377C on helix α7. The span of these helix α7 positions is 19 Å along the helix axis, and as disulfide-bond formation requires sulfhydryl–sulfhydryl collisions it follows that the A411C sidechain must undergo collisions with a substantial length of the adjacent helix. Moreover, the positions tested on helix α7 are located on both the exposed and buried faces, indicating that the helix rotates at least 180° about its long axis. An alternative possibility is that one or both helices spontaneously melt at a substantial rate, but such a random process seems less likely given the well defined spatial limits of the observed collisions (Figure 5). Analogous long-range helix motions have been detected by disulfide trapping in the periplasmic galactose-binding protein . Together, these findings suggest that such motions could be common in protein frameworks. In the aspartate receptor, however, the observed motions are likely to be damped when the CheA and CheW proteins dock to the cytoplasmic domain. Figure 6 presents a schematic working model for the full receptor–kinase signaling complex.
Cysteine and disulfide scanning studies have led to a four-helix bundle model for the cytoplasmic domain of the bacterial aspartate receptor, and have provided constraints on the signaling mechanism and insights into backbone dynamics. As this cytoplasmic domain is highly homologous to the cytoplasmic domains of other members of the methyl-accepting taxis receptor family, the four-helix bundle model is likely to be directly relevant to these closely related homologs. More generally, the aspartate receptor belongs to a superfamily of architecturally, mechanistically and functionally homologous receptors that regulate histidine kinases in two-component signaling pathways. It follows that the present model will be broadly relevant to the prokaryotic and eukaryotic receptors that regulate histidine kinases. Site-directed sulfhydryl chemistry and spectroscopy will continue to be useful in a range of applications, especially those involving complex protein systems where it is advantageous to probe structure, dynamics and mechanism in situ.
E. coli strain RP3808 (Δ(cheA–cheZ)DE2209 tsr-1 leuB6 his-4 eda-50 rpsL136 [thi-1 Δ(gal-attl)DE99 ara-14 lacY1 mtl-1 xyl-5 tonA31 tsx-78]/mks/) was kindly provided by John S Parkinson, University of Utah . The receptor expression plasmid pSCF6 has been described previously . Expression strains and plasmids used to produce CheA (HB101/pMO4), CheW (HB101/pME5) and CheR (JM109/pME43), were generously provided by Jeff Stock, Princeton University. The strain and plasmid used to generate CheY (RBB455/pRBB40) were kindly provided by Bob Bourret, University of North Carolina. [γ-32P]-ATP was purchased from Amersham. Deoxyoligonucleotides were synthesized by Life Technologies Inc. Kunkel mutagenesis reagents (T7 DNA polymerase and deoxynucleoside triphosphates) were purchased from BioRad. Restriction enzymes MluI, MscI, PstI and StuI, as well as T4 DNA ligase and calf intestinal phosphatase (CIP), were purchased from New England Biolabs. Unless noted, all other reagents were obtained from Sigma and were reagent grade.
Engineered receptors containing single cysteines in the cytoplasmic domain were created by oligonucleotide-directed mutagenesis of the plasmid pSCF6 with subsequent expression and isolation in native E. coli membranes as described previously [5,9].
Double-cysteine receptors containing the N36C substitution in the periplasmic domain paired with a cysteine substitution in the cytoplasmic domain were created by combining restriction fragments of the appropriate mutant derivative of pSCF6 . A double digest with PstI and MluI produced an 0.9 kb fragment containing the N36C mutation or a 4.1 kb fragment containing the cytoplasmic domain mutation. The digest containing the 0.9 kb fragment was treated with CIP to prevent self-ligation in subsequent steps. All fragments were resolved on a 1% TAE gel, stained with ethidium bromide, illuminated with UV light and the appropriate band was excised. The fragments were then isolated with the Gene Clean kit (Bio 101), mixed and ligated overnight at 16°C with T4 DNA ligase. The resulting reconstructed pSCF6 derivative was transformed directly into the expression strain, RP3808, and receptor-containing membrane was prepared using the standard protocol [5,9]. Under appropriate oxidation conditions, a small percentage of the double-cysteine containing receptor could be forced to form trimers, tetramers and higher-order products, as observed by SDS–PAGE. Formation of these covalently linked receptor species confirmed that an engineered receptor possessed two cysteines per receptor subunit.
Double-cysteine receptors containing A411C on α8 and a second cysteine on α7 were created by combining plasmid fragments containing the α7 and α8 regions from previously characterized single-cysteine mutants. A double digest of mutant pSCF6 plasmids with PstI and MscI produced a 3.5 kb fragment containing the A411C mutation or a 1.5 kb fragment containing the α7 variable site. The digest containing the 3.5 kb fragment was treated with CIP to prevent self-ligation. Fragments were isolated on 1% TAE gels and the appropriated bands were visualized, excised and DNA isolated as for the N36C double-cysteine mutants above. Ligations to reconstruct the full-length double-cysteine mutant pSCF6 plasmid, transformation, expression and purification were carried out as above.
Double-cysteine receptors pairing α6–α8 cysteine substitutions were created as described for α7–α8 double mutants, whereas constructions pairing α6–α7 mutations were made using a double digest with the enzymes PstI and StuI. This latter digestion produced fragments of approximately 1.3 kb and 3.7 kb containing the α6 and α7 cysteine mutations, respectively. Subsequent steps were as before.
To analyze both the single- and double-cysteine receptors for their ability to form disulfides, each receptor isolated in native E. coli membranes was diluted in a metal-buffering solution containing EDTA to slow the disulfide reaction to measurable rates (2 µM receptor monomer, 20 mM sodium phosphate and pH 7.0 with NaOH, 10% v/v glycerol and 5 mM EDTA) [8,58]. Oxidation was initiated by the addition of 2 mM Cu(II)(1,10-phenanthroline)3 and allowed to proceed at 25°C for 1 min. Reactions were then quenched with an equal volume of 2 × Laemmli sample buffer supplemented with 2 mM NaAsO4, 0.2 mM EDTA and 10 mM N-ethylmaleimide by heating to 95°C for 2 min . Samples were resolved by 10% Laemmli SDS-PAGE (acrylamide: bisacrylamide 40:0.2), stained with Coomassie blue, then visualized and quantitated with a digital camera (Alpha Innotec) .
Components of the in vitro receptor-coupled kinase assay (CheA, CheW and CheY) were prepared and the assay performed as described previously [9,52,53]. Assays were carried out with both reduced receptors and receptors exposed to oxidizing conditions to test the effect of disulfide-bond formation on receptor-mediated kinase activity. Briefly, membrane samples containing 12 µM receptor monomer were either used in their reduced state or incubated with 0.2 mM Cu(II)(1,10-phenanthroline)3 and ambient O2 (approximately 200 µM) for 20–30 min at 37°C. This oxidation reaction was inactivated by the addition of 0.1 mM sodium persulfate. With either reduced or oxidized membrane, formation of the supermolecular signaling complex was achieved by equilibrating wild-type or engineered receptors with soluble CheA and CheW and CheY for 30 min. The kinase reaction was initiated by the addition of [γ-32P]-ATP to the reaction mixture. After 10 s, aliquots were removed and quenched at 22°C with 2 × Laemmli sample buffer supplemented with 25 mM EDTA to chelate metal from the kinase CheA. The indicated conditions ensured that receptor-mediated CheA autophosphorylation was the rate-determining step, such that the rate of phosphotransfer to CheY was linearly proportional to the CheA autophosphorylation rate. [32P]-Phosphorylated CheY was resolved by 15% Laemmli SDS–PAGE (acrylamide to bisacrylamide ratio 40:1.25). Gels were dried and quantitated by phosphorimaging (Molecular Dynamics). In the case of the oxidized membranes the extent of disulfide formation was determined by analyzing on a 10% non-reducing Laemmli SDS–PAGE gel. Gels were subsequently Coomassie stained and monomer and dimer bands were quantitated with a digital camera and the percentage of dimer formation calculated.
To test whether functional disulfide bonds identified in the in vitro receptor-coupled kinase assay were formed between the two subunits of the same dimer, or rather by collisions between dimers, a set of double-cysteine mutants was created. Each mutant contained the periplasmic N36C substitution and a second cysteine shown to form a functional cytoplasmic disulfide bond [2,3,44]. E. coli membranes containing the double-cysteine receptor were first diluted to final concentration of approximately 2 µM receptor monomer. The oxidation reaction was initiated by the addition of 0.2 mM Cu(II)(1,10 phenanthroline)3. The reaction was allowed to proceed for the time points listed above at 37°C then quenched with 2 × Laemmli sample buffer supplemented with 2 mM NaAsO4, 0.2 M EDTA and 10 mM N-ethylmaleimide. Samples were resolved using 10% Laemmli SDS–PAGE and Coomassie stained.
The authors acknowledge stimulating discussions and information provided by former and present group members (S Chervitz, M Danielson, S Butler, S Winston, M Coleman, R Mehan) and colleagues in the field (S Parkinson, J Hazelbauer, S Mowbray, R Kaback, D Koshland, J Stock). We especially thank R Mehan for outstanding technical assistance in the production of Figure 3. This work was supported by NIH Grant GM R01-40731 (to JJF).