Research Article

Mechanism of transmembrane signaling by sensor histidine kinases

See allHide authors and affiliations

Science  09 Jun 2017:
Vol. 356, Issue 6342, eaah6345
DOI: 10.1126/science.aah6345

Bacterial sensing mechanism revealed

Escherichia coli use a transmembrane sensor protein to sense nitrate in their external environment and initiate a biochemical response. Gushchin et al. compared crystal structures of portions of the NarQ receptor that included the transmembrane helices in ligand-bound or unbound states. The structures suggest a signaling mechanism by which piston- and lever-like movements are transmitted to response regulator proteins within the cell. Such two-component systems are very common in bacteria and, if better understood, might provide targets for antimicrobial therapies.

Science, this issue p. eaah6345

Structured Abstract

INTRODUCTION

Microorganisms obtain most of the information about their environments through membrane-associated signaling systems. One of the most abundant classes of membrane receptors, present in all domains of life, is sensor histidine kinases, members of two-component signaling systems (TCSs). Tens of thousands of TCSs are known. Many of these systems are essential for cell growth, survival, or pathogenicity and consequently can be targeted to reduce virulence. Several large families of transmembrane (TM) TCS receptors are known: (i) sensor kinases, which generally possess a periplasmic, membrane, or intracellular sensor module; a transmembrane domain; often one or more intracellular signal transduction domains such as HAMP, PAS, or GAF; and an intracellular autokinase module (DHp and CA domains), which phosphorylates the response regulator protein; (ii) chemoreceptors, which also possess the sensor module and the TM domain but lack the kinase domain and control a separate kinase protein (CheA) via a kinase control module; and (iii) phototaxis systems, which are similar to chemotaxis systems except that the sensor module—a light receptor sensory rhodopsin—is a separate protein.

RATIONALE

Despite the wealth of biochemical data, the structural mechanisms of transmembrane signaling by TCS sensors are poorly understood at the atomic level. In particular, high-resolution structures of the TM segments connected to the adjacent domains are lacking. Deciphering of the signaling-associated conformational changes would shed light on the details of long-range transmembrane signal transduction and might help in the development of novel classes of antimicrobials targeting TCSs.

RESULTS

We used the in meso crystallization approach and single-wavelength anomalous dispersion to determine the crystal structures, at resolutions of up to 1.9 Å, of a fragment of Escherichia coli nitrate/nitrite sensor histidine kinase NarQ that contains the sensor, TM, and HAMP domains in a symmetric ligand-free apo state and in symmetric and asymmetric ligand-bound holo-S and holo-A states. In all of the structures, the TM domain is an antiparallel four-stranded coiled coil (CC) consisting of nine CC layers. The sensor domain is connected to the TM domain through continuous α-helical linkers that are partially disrupted in the holo state. The intracellular HAMP domain is connected to the TM helices via flexible proline junctions and robust hydrogen bonds conserved in all signaling states. The structures reveal the mechanism of transmembrane signal transduction in NarQ and show that binding of ligand induces displacement of the sensor domain helices by ~0.5 to 1 Å. This displacement translates into rearrangements and ~2.5 Å pistonlike shifts of transmembrane helices and is later converted, via leverlike motions of the HAMP domain protomers, into 7 Å shifts of the output helices and changes of the CC helical phase. The structures also demonstrate that the signaling-associated conformational changes in the TM domain do not need to be symmetric.

CONCLUSION

The determined structures of the transmembrane and membrane-proximal domains of the nitrate/nitrite receptor NarQ in ligand-free and ligand-bound forms present a template for studies of other TCS receptors, establish the importance of the pistonlike displacements of the TM helices for TM signal transduction, and highlight the role of the HAMP domain as an amplifier and converter of a piston-like displacement into helical rotation. Overall, the results show how a mechanistic signal is generated and amplified while being transduced through the protein over distances of 100 Å or more. Because membrane-associated TCSs are ubiquitous in microorganisms and are central for bacterial sensing, we believe that our results will help to elucidate a broad range of cellular processes such as basic metabolism, sporulation, quorum sensing, and virulence. They may also provide insights useful for the development of novel antimicrobial treatments targeting TCSs.

The structures of histidine kinase NarQ in ligand-free and ligand-bound forms.

The structures reveal rearrangement of transmembrane α helices during signal transduction and show that pistonlike shifts of the transmembrane helices result in leverlike motions of the HAMP domain protomers.

Abstract

One of the major and essential classes of transmembrane (TM) receptors, present in all domains of life, is sensor histidine kinases, parts of two-component signaling systems (TCSs). The structural mechanisms of TM signaling by these sensors are poorly understood. We present crystal structures of the periplasmic sensor domain, the TM domain, and the cytoplasmic HAMP domain of the Escherichia coli nitrate/nitrite sensor histidine kinase NarQ in the ligand-bound and mutated ligand-free states. The structures reveal that the ligand binding induces rearrangements and pistonlike shifts of TM helices. The HAMP domain protomers undergo leverlike motions and convert these pistonlike motions into helical rotations. Our findings provide the structural framework for complete understanding of TM TCS signaling and for development of antimicrobial treatments targeting TCSs.

Microorganisms obtain most of the information about their environments through membrane-associated signaling systems. One of the most abundant classes of membrane receptors, present in all domains of life, is sensor histidine kinases, members of two-component signaling systems (TCSs) (13). Tens of thousands of TCSs are known (4). Many of these systems are essential for cell growth, survival, or pathogenicity and consequently can be targeted to reduce virulence (57).

Membrane-associated sensor kinases generally function as homodimers and possess a periplasmic, membrane, or intracellular sensor module; a transmembrane (TM) domain; often one or more intracellular signal transduction domains (such as HAMP, PAS, or GAF); and an intracellular autokinase module (DHp and CA domains), which phosphorylates the response regulator protein (1, 8, 9) (fig. S1). Other TCS TM sensors, chemo- and photoreceptors, have domain organization similar to that of sensor histidine kinases; they possess the sensor module, TM, and HAMP domains but lack the autokinase module and control a separate kinase protein (CheA) via the kinase control module (fig. S1). Also, the core functional unit of chemoreceptors and phototaxis systems is not a dimer but a trimer of dimers (1012), which may then pack into higher-order oligomeric assemblies (13).

The molecular mechanisms of TM signal transduction in different TCS classes are expected to be somewhat similar, because a vast majority of the sensors feature a dimeric four-helix TM core in the TM domain, often followed by the HAMP domain downstream. At present, only two reports on atomic-resolution structures of the TM domains are available, both describing isolated TM fragments and thus providing little insight into the signal transduction in and out of the TM domain of a typical TCS sensor. One report (14) provided a crystallographic structure of the complex between the photoreceptor sensory rhodopsin and the TM domain of its transducer protein. The other report (15) described nuclear magnetic resonance (NMR) structures of nonphysiological monomeric TM domains of histidine kinases from three different classes. Additionally, several low-resolution models of the TM helices in the context of full-length proteins such as chemoreceptor Tar or histidine kinase PhoQ have been constructed on the basis of disulfide cross-linking and modeling (16, 17). Because of the limitations of all these studies, atomic details of TM signal transduction are still not resolved (8, 9). As for the HAMP domain, multiple experimental structures of the domains from different proteins are available, and there are several competing hypotheses of HAMP domain signaling (8, 9, 18).

To complement the above reports, we present our studies of the E. coli nitrate/nitrite sensor kinase NarQ. Similarly to NarX, another E. coli nitrate/nitrite sensor, NarQ consists of seven major structural elements: the periplasmic sensor domain, the TM domain, the HAMP domain, the signaling helix, the GAF-like domain, the DHp domain, and the kinase CA domain (1820) (fig. S1). Whereas no structural information is available for NarQ, crystal structures of NarX sensor domain in apo and holo states have been determined (21), as well as multiple structures of the HAMP, GAF, DHp, and kinase domains of other proteins (8, 9).

Here, we used the in meso (2224) and single-wavelength anomalous dispersion (SAD) (25) approaches to crystallize and determine the crystal structures, at resolutions of up to 1.9 Å, of a fragment of the nitrate sensor NarQ that contains the sensor, TM, and HAMP domains. We determined the symmetric apo (ligand-free) state structure of a conservative mutant and symmetric and highly asymmetric holo-S and holo-A (ligand-bound) state structures of the wild-type protein (Fig. 1 and fig. S2). Comparison of the apo and holo structures reveals extensive structural rearrangements and provides a direct demonstration of signal transduction through the TM and HAMP domains, including a pistonlike motion of TM helices and a leverlike rotation of HAMP domain protomers. Comparison of holo-S and holo-A structures shows that the signal transduction in NarQ can be achieved via different sets of conformational changes. Overall, our findings clarify the mechanism of TM signaling in TCS receptors.

Fig. 1 Overall structure of the crystallized NarQ fragment in apo and holo states.

The protein is observed as a symmetric dimer in both the apo and holo-S states in the space groups I212121 and F222, respectively, and as an asymmetric dimer in holo-A state in the space group P2. Termini of the second protomer are denoted with primes. The crystallized fragment comprises the transmembrane domain (helices TM1 and TM2), the periplasmic sensor domain (helices H1 to H4) with the nitrate ion binding at the dimerization interface, and the intracellular HAMP domain (helices AS1 and AS2). The structures are aligned by sensor domains. Hydrophobic membrane boundaries (black) were calculated using the OPM server (69) with the symmetric ligand-bound structure as a template.

Structure of the sensor domain

In all of the presented structures, including the asymmetric ligand-bound form, the NarQ sensor domain is a symmetric dimer of two monomers comprising four α helices, H1 to H4, with helices H2 and H4 broken into subhelices (fig. S3). The loop between helices H2 and H3 is anchored into the membrane by residues Trp89 and Tyr90. Nitrate binds at the symmetry axis between helices H1 and H1′ of the two protomers, where it is coordinated by the side chains of the residues Arg50 and Arg50′ (fig. S3E). The ligand was observed in all crystals of the wild-type protein, including those obtained without supplementation of the protein purification and crystallization solutions with nitrate. This might be a consequence of several factors: (i) the high ligand affinity of the crystallized fragment for nitrate, (ii) the presence of nitrate impurities in the aforementioned solutions, and (iii) the very small amount of nitrate needed to saturate the binding sites in the ~0.5 mM receptor solution used for crystallization. Consequently, we generated a conservative Arg50 → Lys (R50K) mutant to obtain the crystal structure of NarQ in its apo state. This mutation is known to lock the receptor in the off state (26) but was not expected to alter the local structure of the sensor domain, because Arg and Lys side chains are similar in size and charge. Indeed, the structures of the NarQ sensor domain in the apo R50K and holo-S and holo-A wild-type forms reported here closely resemble those of the apo and holo wild-type NarX sensor domain (21), respectively. In particular, the positions of the membrane-proximal ends of helices H1 and H4, responsible for downstream signaling, are identical in the corresponding structures (fig. S3, C and F), and we thus conclude that introduction of the mutation results in correct reproduction of the structural features of the wild-type ligand-free protein.

Structure of the TM domain

In all of the structures presented here, the NarQ TM domain is an antiparallel four-stranded coiled coil (CC) consisting of nine CC layers (Figs. 1 and 2 and fig. S4). Several ordered lipid fragments are observed in the holo-S structure, one fragment is observed in the holo-A structure, and no fragments are observed in the apo structure. However, reliable identification of the corresponding lipidic moieties is not possible because of the small size of the fragments.

Fig. 2 Signaling-associated conformational changes in the NarQ TM domain.

Superposition of the apo state structure (gray) and symmetric holo state structure (colored) is shown. Left: Changes in conformation of helix TM1. Center: Changes in conformation of helix TM2. Right: Changes in the arrangement of the TM helices. Cα atom positions of the residues from CC layer 9 are marked with the spheres to highlight the displacements in the direction perpendicular to the membrane. The structures are aligned by the sensor domains.

There are notable variations in the packing of the TM helices. The apo state structure is best described as a dimeric CC comprising helices TM1 and TM1′, with TM2 and TM2′ flanking this. The symmetric holo-S state structure is a traditional four-helix CC, switching from a 7/2-period left-handed CC at the periplasmic side to an 11/3-period structure with close to zero supercoil twist on the cytoplasmic side. Finally, in the asymmetric holo-A state, the TM domain is a highly distorted CC, where the relative arrangement of the helices in layers 1 to 4 and 9 resembles that observed in the symmetric holo-S state and the arrangement in layers 7 and 8 resembles that seen in the apo state (fig. S4).

In general, the cytoplasm-proximal layers are more hydrophobic and are packed tightly, whereas those proximal to the periplasm contain hydrated cavities and multiple polar (Ser22, Ser25, Thr26, Thr32, Ser154, and Thr163) and somewhat disordered glycine residues (Gly27, Gly157, Gly158, and Gly160; fig. S2E). Although the role of water molecules and glycine residues in the NarQ TM domain is not clear, similar α helix–destabilizing features are observed in the TM domains of other histidine kinases (fig. S5), and they probably impart to the TM domain the flexibility needed to accommodate signaling-associated transitions, similarly to the glycine hinge in the cytoplasmic part of chemoreceptors (27).

The α-helical composition of the TM region of TCS receptors has long been predicted (8, 9, 1417). However, the structure of the NarQ TM domain is notably different from that of the sensory rhodopsin transducer TM domain (fig. S5). In NarQ, the TM region is either a simple four-helix CC or has a dimeric CC core consisting of the TM1 helices, whereas NpHtrII has a dimeric CC core consisting of the TM2 helices (14, 28). Also, in NarQ, there is either zero or left-handed supercoil twist, whereas in NpHtrII this changes from close to zero at the periplasmic side to right-handed at the cytoplasmic side (fig. S5). As for the comparison with the TM fragments of ArcB, QseC, and KdpD, the NMR structures unfortunately do not provide information about the dimerization interfaces and CC packing in the physiological dimers of any of these kinases (15).

Junctions between TM and adjacent domains

Our crystal structures of NarQ reveal details of the interactions of the TM domain with the membrane-proximal domains (fig. S6). On the periplasmic side of the protein, in the apo state, there is a continuous transition of the TM1 and TM2 helices into the H1 and H4 helices, respectively, whereas in both holo-A and holo-S states there is a break in the α-helical structure between TM1 and H1 (see fig. S6, A and D, for comparison of apo and holo-S states).

At the cytoplasmic side, in all states the cytoplasmic side of TM1 and the protein’s N terminus make extensive contacts with the HAMP domain. The highly conserved residue Glu207 of the membrane-proximal region of HAMP domain helix AS2 (29) forms two hydrogen bonds with the backbone nitrogens of helix TM1 residues (fig. S6, B and E), similarly to what has been observed in the structure of the soluble mutant of protein Af1503 (30) (fig. S7A). As for the TM2-AS1 junction, there is a proline-induced kink (fig. S6, C and F), again as also seen for Af1503 (30) (fig. S7B).

The HAMP domain

The NarQ HAMP domain displays a typical parallel four-stranded CC fold (18), with each protomer consisting of two α helices, AS1 and AS2, separated by an unstructured linker (fig. S8). In all of the structures, the HAMP domain is less ordered than the sensor and TM domains, and in the apo state structure, the positions of the side chains of Leu225 and Tyr226 are not resolved. Overall, the apo state structure is similar to the structure of the Af1503 HAMP domain (fig. S9). The structure observed for the symmetric holo-S state HAMP domain is influenced by crystal contacts: The AS2 helices are interspersed with the same helices of an adjacent dimer (figs. S2 and S8), and consequently there is no contact between the Leu225 side chains facing toward the CC core. Although such conformation would be consistent with an antiparallel four-helix CC downstream domain, such as in bacterial chemoreceptors (31), it is unusual in the structural context of histidine kinases. The crystal contacts mentioned above are absent in the asymmetric holo-A state structure (space group P2). There, the hydrophobic Leu225 side chains are less exposed to the solvent and are in contact with each other. Thus, we conclude that the holo-A structure represents a better model of the activated NarQ HAMP domain.

The signaling helix region

To obtain insights into signal transduction through and downstream of the HAMP domain, we constructed atomic models of the adjacent signaling helix region (residues 224 to 245; see NarQ domain architecture in fig. S1). The signaling helix is a common sequence motif in sensor kinases (32) and sensory rhodopsin transducers (33), and the positions of hydrophobic and hydrophilic residues indicate that two CC heptad repeat pattern assignments are possible for NarQ (fig. S10). Neither of the models is compatible with the holo state HAMP domain structures, but both—especially model 2—are compatible with the structure of the apo state HAMP domain (see layer 6 in fig. S8 and layer 1 in fig. S10). Model 1 comprises a phase stutter that is also observed in other histidine kinases and chemoreceptors (18, 19, 34) and may be required for destabilization of the signaling helix for it to be able to adopt different signaling states. However, model 2 appears to be more probable, judging from the protein sequence (19).

Discussion

Signal transduction in NarQ

Comparison of the apo and holo state crystal structures of NarQ reveals that binding of nitrate causes extensive rearrangements in the TM region that can be represented as a combination of changes in the lateral arrangement of the TM helices and pistonlike shifts of the helices in the direction perpendicular to the membrane plane. At the same time, comparison of the holo-S and holo-A states reveals profound differences in the middle of the TM segment but identical conformations at its periplasmic and intracellular sides (see movie S1 for comparison of apo and holo-S state structures, and movie S2 for comparison of holo-S and holo-A structures). We make several observations from these comparisons.

First, binding of nitrate in the vicinity of Gly47 causes disruption in the α helix H1 and appearance of a 310-helix element: Gly47 carbonyl switches from Met51 amide hydrogen to Arg50 amide hydrogen (from i → i + 4 helical hydrogen bonding to i → i + 3; figs. S3 and S11). 310 helices have a higher helical rise per residue than α helices (35), and as a result the membrane-proximal part of H1 moves slightly toward the membrane and rotates. This rotation of the membrane-proximal part of H1 opens a “hole” near residues Ile43 and Ala46, which is then occupied by a Val136 “knob” (36). Consequently, H4 rotates as well and moves away from the membrane, which results in a pistonlike displacement of H4 relative to H1 for ~0.5 to 1.0 Å. Also, the H4 and H4ʹ helices are brought much closer to the H1 and H1ʹ helices. This interpretation is supported by mutagenesis data: Mutation of Ala46 to Ser is inconsequential, mutation to Thr results in impaired signaling, and mutations to the much bulkier Asn and Ile lock NarQ in the “off” state (37). The observed conformational changes in NarQ are identical to the changes observed in isolated NarX sensor domain (fig. S3) (21).

The rotation of the membrane-proximal parts of H1 and H1′ and the approach of H4 and H4ʹ upon binding of nitrate result in disruption of the TM1/H1-TM1′/H1′ CC interface, appearance of discontinuity between α helices TM1 and H1, and distancing of the TM1 and TM1′ periplasmic sides (Fig. 3, fig. S11, and movie S1). The introduced discontinuity between TM1 and H1 further increases the relative pistonlike displacement from ~1 Å for H1 relative to H4 [as in NarX (21)] to ~2.5 Å for TM1 relative to TM2 (fig. S11). The importance of the TM1-H1 helical break for kinase activation is also highlighted by a mutagenesis study in which the introduction of a proline into the TM1-H1 junction (mutation E41P) locks the receptor in the “on” state, whereas other nonconservative mutations of Glu41 to Arg, His, and Leu (E41R, E41H, and E41L) do not disrupt the function of the receptor (37).

Fig. 3 Details of the signal transduction from the TM domain to and through the HAMP domain.

(A) Inactive apo state. (B) Active holo-S state. (C) Superposition of the apo (gray) and holo-S (colored) states. A pistonlike displacement of the cytoplasmic end of TM1 relative to TM2 and the TM2-AS1 hinge is transmitted to the membrane-proximal end of AS2 and results in leverlike rotations of the HAMP domain protomers around the hinges. Because the HAMP domain protomers move in opposite directions, the positions of membrane-distal ends of the AS2 helices also change relative to each other. Positions of the Leu225 Cα atom are marked by spheres. The gray bar shows the position of TM1 ends in the apo state structure. The domains are aligned by residues 175 to 177.

Conformational changes in the sensor domain and at the sensor-TM junction are moderate and identical in holo-A and holo-S structures; however, the rearrangements in the TM domain are much more pronounced and are different between holo-A and holo-S. The TM1 helices, which form a tight CC interface in the apo state, are not fully separated in the holo-A structure (fig. S4). In the holo-S state, the four TM helices form a traditional four-helix CC (Fig. 2 and fig. S4). But there is a similarity between the holo-A and holo-S structures: In both, the ~2.5 Å relative pistonlike displacement of the TM helices is fully transduced toward the HAMP domain (Figs. 2 and 3 and movies S1 and S2). Conformational changes at the TM-HAMP interface are also identical in the holo-S and holo-A structures as compared to the apo state structure. Thus, comparison of the asymmetric holo-A and symmetric holo-S state structures offers insight into the receptor flexibility and robustness of signal transduction by the TM domain despite the ~9 Å difference in the position of the HAMP domain (Figs. 1 and 3 and figs. S4 and S8).

Finally, in the cytoplasmic part of NarQ, the pistonlike displacements of the TM helices, caused by binding of the ligand, are amplified through the leverlike rotations of the HAMP domain protomers around the Pro179 hinge into roughly 7 Å displacements of the output ends of the AS2 helices (Fig. 3). During these conformational rearrangements, each HAMP domain protomer behaves essentially as a rigid body. The RMSD values of the backbone atom positions of helices AS1 (residues 179 to 192) and AS2 (residues 206 to 225) are ~0.6 Å between the apo and both holo-S and holo-A structures, and ~0.4 Å between the holo-S and holo-A structures; these values are comparable to crystal-to-crystal variation and typical crystallographic error of atomic positions. However, as a result of the relative motions of the HAMP domain protomers, the whole dimeric HAMP domain itself changes profoundly: The RMSDs of the backbone atom positions of helices AS1 and AS2 are ~2.9 Å between the apo and holo-S structures and ~3.0 Å between the apo and holo-A structures, as opposed to ~1.2 Å between the holo-S and holo-A structures.

Details of the signal transduction in NarQ downstream of the HAMP domain are less clear because of truncation of the crystallized construct. In the apo state, the CC register of the AS2 output ends is consistent with the CC register of the signaling helix. However, in the holo-A and holo-S states the CC helical phase is much less consistent with that of the signaling helix. Therefore, it appears that the signaling helix should be destabilized and/or dissociated upon binding of the ligand (Fig. 4). Signaling-related dissociation or destabilization of signaling helix was also observed in Af1503-EnvZ chimeras (34) and in the histidine kinases BvgS and DesK (38, 39) and may thus be the general mechanism for signal propagation toward the DHp domain in TCS sensor kinases.

Fig. 4 Mechanism of NarQ transmembrane signaling.

Binding of the ligand results in a pistonlike displacement toward the periplasm of the TM1 helices relative to TM2 and consequent leverlike conformational changes in the HAMP domain, which cause dissociation or destabilization of the signaling helix.

Mechanism of transmembrane signaling by TCS receptors

Several hypotheses concerning TM signaling mechanisms in TCS receptors have been proposed (8, 9). For sensor histidine kinases, the crystal structures of α-helical NarX and TorS sensor domains hinted at pistonlike conformational changes (21, 40). The crystal structures of the mixed α/β tandem PAS sensor domain of HAMP-less kinase LuxQ revealed rotation of signal output helices (41); this was also demonstrated in another HAMP-less kinase, AgrC, where the sensor domain is integral to the membrane (42). A recent cross-linking study revealed scissorlike diagonal displacement of the TM helices in the kinase PhoQ (17). For chemoreceptors with α-helical sensor domains, such as aspartate receptor Tar, the major model is pistonlike displacement of the TM helices in the direction normal to the membrane plane (13, 31). For chemoreceptor McpB with mixed α/β tandem PAS sensor domain, the signal was found to be transduced via rotation of TM helices (43). Finally, in the sensory rhodopsin–cognate transducer complex, where the sensor is integral to the membrane, a combination of a pistonlike motion and a rotation of the signal output helices was observed (44). Although the pistonlike displacements, helical rotations, CC phase changes, and scissoring motions are not mutually exclusive and are possibly coupled to each other, consensus on the TM signaling mechanism in TCS sensors is currently lacking (8, 9).

Our results show that binding of ligand to NarQ causes a pistonlike displacement of the TM helices, which is accompanied by extensive symmetric or asymmetric rearrangements and scissoring of the TM helices. The rearrangements are different in the two presented holo state structures, but the pistonlike displacement is perfectly conserved. Thus, the latter appears to be a more robust mechanism of TM signal transduction.

Also, our data show that symmetrical changes in the sensor domain may result in both symmetrical changes in the TM domain (as in the holo-S structure) and asymmetrical changes (as in the holo-A structure). Asymmetry only changes the position of the HAMP domain relative to the TM domain and not its signaling state (Fig. 1). Sensor domains of several histidine kinases are known to reside in asymmetric conformations in the inactive state and in symmetric conformations in the active state (40). The aspartate chemoreceptor can be activated by asymmetric changes in the sensor domain upon binding of a ligand (31, 45) and presumably also by symmetric conformational changes upon mutations of amino acids at the membrane-water interfaces (46, 47). Thus, it is still unclear whether asymmetric changes in the sensor domain can result in symmetric changes in the TM domain, and the exact relations between asymmetry and activation also remain obscure.

Signal transduction through the HAMP domain

Currently, there are several competing hypotheses concerning the mechanism of signal transduction through the HAMP domain (8, 9, 18): The gearbox model postulates rotation of the HAMP domain helices in opposite directions (48, 49); the dynamic bundle model proposes that signal transduction is associated with changes in HAMP domain stability (50, 51); and several reports based on cross-linking and crystallographic studies propose a variety of scissoring models (5254).

Our results indicate that in NarQ, the HAMP domain serves as an amplifier and converter of the pistonlike conformational changes in the TM domain. Pistonlike displacements of TM1 and TM1′ cytoplasmic ends for 2.5 Å relative to the proline hinge result in leverlike rotations of the protomers around their respective proline hinges and consequent 7 Å displacements in opposite directions of AS2 and AS2′ membrane-distal ends. This results in a ~90° change in the helical phase and the output CC register (Fig. 3 and fig. S8). Thus, the membrane-proximal HAMP domain acts as a converter of the pistonlike motions of TM domain helices into helical rotation at its output helices. These conformational changes are similar to the scissoring models (5254) and are possibly accompanied by changes in the stability of the HAMP or adjacent domains. However, our results do not support the gearbox hypothesis. We observe that the HAMP domain protomers move relative to each other in a rigid-body fashion, without rotation of the AS1 and AS2 helices within each protomer. Also, observations of slightly different HAMP domain conformations in holo-A and holo-S structures and increased B factors in the HAMP domain region support its dynamic nature. It is possible that the domain exists in an ensemble of similar conformations in vivo, whereas crystallographic structures provide only static snapshots. Finally, as a result of the absence of the adjacent signaling helix region in the crystallized construct, the observed conformational changes may under- or overestimate the actual transformations associated with signal transduction.

The signal converter role of the HAMP domain might reconcile the data on TM signal transduction in histidine kinases that contain and those that lack the HAMP domains. In sensors containing a membrane-proximal HAMP domain, binding of a ligand causes a pistonlike displacement of the TM helices (which may be accompanied by scissoring); this displacement is then converted into helical rotation by the HAMP domain. In sensors lacking the HAMP domain, binding of ligand directly causes rotation of the TM signal output helices (41, 42). In both cases, the signal that is passed downstream is the helical rotation. Whether pistonlike displacements are present during PhoQ activation remains to be tested.

The multiple conformational states observed for the NarQ TM domains present a template for studies of other TCS receptors, establish the importance of the pistonlike displacements of the TM helices, and highlight the role of the HAMP domain as an amplifier and converter of a pistonlike displacement into helical rotation. Overall, our results show how a mechanistic signal is generated and amplified while being transduced through the protein over distances of 100 Å or more. Because membrane-associated TCSs are ubiquitous in microorganisms and are central for bacterial sensing (14), we believe that the results reported here will help in understanding of a broad array of cellular processes, ranging from basic metabolism to sporulation, quorum sensing, and virulence. They may also provide insights for development of novel antimicrobial treatments targeting TCSs (57).

Materials and methods

Cloning, protein expression, and purification

The nucleotide sequence encoding residues 1 to 230 of NarQ was cloned from E. coli strain BL21 (DE3) and introduced into the pSCodon1.2 expression vector (StabyCodon T7, Eurogentec, Belgium) via Nde I and Xho I restriction sites. Consequently, the construct harbored a C-terminal 6×His tag. Mutation R50K was introduced by site-directed mutagenesis. Both the wild-type protein and the mutant version were expressed in E. coli strain SE1 (StabyCodon T7, Eurogentec). Cells were cultured in shaking baffled flasks in ZYP-5052 auto-inducing medium (55) containing ampicillin (100 mg/liter). After 2 hours of incubation at 37°C, the temperature was decreased to 30°C and incubation continued overnight. Harvested cells were disrupted in M-110P Lab Homogenizer (Microfluidics) at 25,000 psi in a PBS buffer with addition of DNase I (50 mg/liter; Sigma-Aldrich) and EDTA-free protease inhibitor cocktail Complete (Roche). The membrane fraction of cell lysate was isolated by ultracentrifugation at 90,000g for 1 hour at 4°C. The pellet was resuspended in a buffer containing 50 mM NaH2PO4/Na2HPO4, pH 8.0, 0.3 M NaCl, and 2% DDM (Anatrace, Affymetrix) and stirred overnight for solubilization. The insoluble fraction was then removed by ultracentrifugation at 90,000g for 1 hour at 4°C. The supernatant was loaded on a gravity flow column containing Ni-NTA resin (Qiagen) and the protein was eluted in a buffer containing 20 mM Tris, pH 8.0, 0.3 M NaCl, 0.2 M imidazole, and 0.1% DDM. Imidazole was removed by dialysis against 20 mM Tris, pH 8.0, 0.3 M NaCl, and 0.1% DDM for 3 hours. The eluate was subjected to size-exclusion chromatography on 125 ml Superdex 200 PG column (GE Healthcare Life Sciences) in a buffer containing 20 mM Tris, pH 8.0, 0.3 M NaCl, and 0.1% DDM. Protein-containing fractions were pooled and concentrated to 30 mg/ml for crystallization.

Crystallization

The crystals were grown using the in meso approach (2224), similarly to our previous work (14, 56, 57). The solubilized protein in the crystallization buffer was added to the monooleoyl-formed lipidic phase (Nu-Chek Prep). Crystallization trials were set up using the NT8 robotic system (Formulatrix). The crystals were grown at 22°C and reached the final size of 50 to 150 μm within 2 weeks. The best holo state crystals in space groups F222 and P2 were obtained using the precipitants 0.6 M KH2PO4/Na2HPO4, pH 4.6, 5 mM NaNO3, and 0.3 M KH2PO4/Na2HPO4, pH 7.6, 250 mM NaBr, respectively. The best apo state crystals in the space group I212121 were obtained using the precipitant 1.6 M (NH4)3PO4, pH 5.8. Before harvesting, the crystals were incubated for 5 min in the respective precipitant solutions supplemented with 20% glycerol. For iodide-SAD experiments, the soaking solution additionally contained 0.5 M NaI. All crystals were harvested using micromounts (MiTeGen) and were flash-cooled and stored in liquid nitrogen.

Acquisition and treatment of diffraction data

The diffraction data were collected at 100 K at ESRF beamline ID23-1 (58) equipped with a PILATUS 6M-F detector. The data collection statistics are reported in table S1. In all cases the diffraction was anisotropic as determined by decay of the CC1/2 values in 20° cones along the reciprocal cell directions a*, b*, and c* (59). In the space group I212121, the resolution limits along the directions a* and b* were 2.55 and 2.17 Å, and along the direction c* the CC1/2 value was ~0.9 at the resolution cutoff of 1.9 Å. In the native data set in the space group F222, the resolution limits along the directions a* and b* were 2.26 and 1.94 Å, and along the direction c* the CC1/2 value was ~0.8 at the resolution cutoff of 1.94 Å. In the space group P2, the resolution limit along the direction a* was 2.82 Å, and along the directions b* and c* the CC1/2 values were ~0.6 and 0.9 at the resolution cutoff of 2.42 Å. Diffraction images were processed using XDS (60). XSCALE (60) was used to merge different data sets and to scale the data for the phasing steps. POINTLESS (59) and AIMLESS (59) were used to merge, scale, assess the quality, convert intensities to structure factor amplitudes, and generate Free-R labels.

Structure determination and refinement

The holo-S state structure in the space group F222 was solved using experimental phasing (61), and the apo and holo-A state structures in the space groups I212121 and P2, respectively, were solved using molecular replacement with MOLREP (62) and the sensor domain from the structure in the space group F222 as a search model. For the F222 structure solution, eight isomorphous data sets from NaI-soaked crystals were collected and merged for subsequent SAD phasing. The iodide sites were determined using ShelxD (63) with the HKL2MAP (64) interface, with CCall and CCweak values of 40.0% and 15.4%. The resolution was extended to 1.94 Å using the native F222 data collected from a single crystal. 178 Ala residues were placed in 5×50 cycles of autotracing and density modification using ShelxE (63). The initial model was improved using automatic model building software ARP-wARP (65). The final F222 model was built using only the native data. For the I212121 and P2 structure solution, native data sets collected from single crystals were used in each case. All models were refined manually using Coot (66) and REFMAC5 (67). Intermediate conformations shown in movies S1 and S2 were calculated using the NOLB algorithm (68).

Supplementary Materials

www.sciencemag.org/content/356/6342/eaah6345/suppl/DC1

Figs. S1 to S11

Table S1

Movies S1 and S2

References and Notes

  1. Acknowledgments: Atomic coordinates and structure factors for the reported crystal structures have been deposited in the Protein Data Bank under accession codes 5IJI, 5JEF, and 5JEQ. The work was done in the framework of CEA(IBS)–HGF(FZJ) STC 5.1 specific agreement and supported by ERA.Net RUS PLUS and Ministry of Education and Science of the Russian Federation (project ID 323, RFMEFI58715X0011). The work used the platforms of the Grenoble Instruct Centre (ISBG; UMS 3518 CNRS-CEA-UJF-EMBL) with support from FRISBI (ANR-10-INSB-05-02) and GRAL (ANR-10-LABX-49-01) within the Grenoble Partnership for Structural Biology. I.M. is the recipient of a studentship funded by the Ph.D. program of the European Synchrotron Radiation Facility. We acknowledge the European Synchrotron Radiation Facility for provision of synchrotron radiation facilities. Author contributions: I.G., A.I., and V.G. designed the study; A.I. expressed and purified the wild-type protein; I.G., P.B., and T.B. helped with expression and purification of the wild-type protein; A.Yu. expressed and purified the R50K mutant; I.G., I.M., V.P., and A.I. crystallized the protein; E.R. helped with crystallization; A.P. solved the structures; G. Bourenkov and G.L. helped with structure solution; I.G. and I.M. refined the structures; I.G. analyzed the structures; S.G., V.B., D.W., and G. Büldt helped with structure analysis; I.G. and V.G. oversaw the study, analyzed the results, and prepared the manuscript with contributions from all other coauthors.
View Abstract

Navigate This Article