Molecular Microbiology (2014) 91(5), 853–857 ■
doi:10.1111/mmi.12516 First published online 27 January 2014
MicroCommentary The HAMP signal-conversion domain: static two-state or dynamic three-state? Valley Stewart* Department of Microbiology & Molecular Genetics, University of California, Davis, CA 95616-8665, USA.
Summary The 50-residue HAMP domain converts input signal into output response in a variety of transmembrane signal transduction proteins, including methylaccepting chemotaxis proteins and histidine kinases. HAMP domains are present in many other contexts as well. Despite focused study over the past decade, the question remains: How does this small domain play such a large role for so many different proteins? Analysis of structural models for the Afl1503 and Aer2 HAMP domains has generated hypotheses in which the HAMP domain assumes either of two discrete forms that generate opposing signal output. In contrast, genetic analysis of the HAMP domain from the Tsr methyl-accepting chemotaxis protein resulted in a distinct hypothesis, the biphasic dynamic bundle. In this hypothesis, signalling involves differential packing stabilities of the HAMP domain four-helix bundle, marked by at least three distinct states. Here I summarize and compare these hypotheses in the context of a deletion analysis that further explores the biphasic dynamic bundle hypothesis. Describing locked signal output mutants of the Escherichia coli transmembrane receptor Tsr (taxis to serine and repellants), Peter Ames and John (Sandy) Parkinson suggested that the ‘linker’ transmits conformational changes from the periplasmic input domain to the cytoplasmic output domain (Ames and Parkinson, 1988). This linker earned an acronym, HAMP (Aravind and Ponting, 1999), after homologous elements were recognized in sequences of disparate homodimeric signalling proteins, including
Accepted 9 January, 2014. *For correspondence. E-mail vjstewart@ ucdavis.edu; Tel. (+1) 530 754 7994; Fax (+1) 530 752 9014. Conflict of interest: None.
© 2014 John Wiley & Sons Ltd
histidine kinases (Collins et al., 1992), adenylyl cyclases, other methyl-accepting chemotaxis proteins (MCPs), and certain phosphatases. Indeed, HAMP domains are present, sometimes as tandem arrays, in a multiplicity of both transmembrane and cytoplasmic signalling proteins (Dunin-Horkawicz and Lupas, 2010). Accordingly, elucidating HAMP domain function is essential for comprehending signal transduction. Now, more than 25 years after its identification as a critical signalling element, distinct hypotheses for HAMP action are stimulating high-quality experimentation and thoughtful analysis. Ames and Parkinson, together with Qin Zhou, present further genetic tests of their idea that HAMP signalling involves a dynamic bundle populating a range of packing stabilities (Fig. 1) (Ames et al., 2014). For context, let us first consider hypotheses inspired by structural analyses.
Structure-based hypotheses for HAMP action Studies from several groups revealed that the HAMP domain monomer contains three elements of approximately 15 residues each: two amphipathic α-helices, termed AS1 and AS2, and a non-helical connector (Williams and Stewart, 1999). Specific interactions between these elements were revealed in a structural model for the dimeric HAMP domain from the Afl1503 protein of Archaeoglobus fulgidus, a hyperthermophilic archaeon (Hulko et al., 2006). In this NMR structural model, the AS1 and AS2 helices form a four-helix parallel coiled-coil, with the connector looped around the perimeter. The overall veracity of the four-helical structure is well-supported by mutational and cross-linking analyses of HAMP domains in native signal-transducing proteins (Parkinson, 2010). The Afl1503 structural model displays two distinctive features. First, the AS1 and AS2 helices are offset by one helical turn, with the AS1 amino-terminal and the AS2 carboxyl-terminal helical turns extending out from the four-helix bundle (Fig. 1). Second, the interhelical association, involving hydrophobic residues at the coiled-coil heptad a and d positions, exhibits an unusual x-da
854 V. Stewart ■
Fig. 1. The biphasic dynamic bundle hypothesis for Tsr and other MCPs. The four-helix coiled-coil HAMP domain helices AS1 and AS2 are coloured blue and gold respectively. The adaptation (methylation) region of the four-helix coiled-coil kinase control (output) domain is coloured green. In the presence of attractant (ATT), tight packing of the HAMP bundle enforces loose packing of the output bundle to generate the kinase-off CCW(A) conformation. In the presence of repellant (REP), intermediate packing of the HAMP bundle enforces intermediate packing of the output bundle to generate the kinase-on CW conformation. These two conformations, which delimit the physiological signalling range, are substrates for the adaptation methylation and demethylation enzymes (methylation sites are indicated as open or closed circles). Finally, certain missense and deletion mutants cause loose packing of the HAMP bundle. This drives tight packing of the output bundle to generate the severely kinase-off CCW(B) conformation, which lies outside of the physiological signalling range. This conformation requires packing by the three carboxyl-terminal residues in AS2, indicated by dotted lines. Figure courtesy of J.S. Parkinson.
packing geometry (Hulko et al., 2006) instead of the orthodox a-d pattern (Crick, 1953). This suggests a helix rotation (‘gearbox’) hypothesis, in which the HAMP domain interconverts between distinct x-da and a-d packing conformations (Hulko et al., 2006; Ferris et al., 2011; 2012). Congruent with this hypothesis, the HAMP domain displays more orthodox packing geometry in the X-ray structural model for the Streptococcus mutans VicK histidine kinase (Wang et al., 2013). The Aer2 (aerotaxis transducer) from Pseudomonas aeruginosa is a soluble protein containing three tandem HAMP domains at the amino-terminus. In the X-ray structural model, the HAMP1 and HAMP3 domain structures resemble that of the Afl1503 HAMP domain, with the notable exception that their packing geometries represent a mixture of x-da, a-d and x-x interactions (Airola et al., 2010). By contrast, the HAMP2 domain adopts a different conformation in which the AS2 helices form a two-helix rather than four-helix coiled-coil. This generated the distinct hypothesis, supported by analysis of select missense substitutions and chimeric MCPs (Airola et al., 2013), that the HAMP1 and HAMP2 structures represent opposing signalling states (Airola et al., 2010).
Meanwhile, biophysical analyses of the HtrII (halobacterial transducer of rhodopsin) protein from the archaeon Natronomonas pharaonis suggest two conformations, ‘compact’ (resembling the Afl1503 structural model) and ‘dynamic’ (resembling the Aer2 HAMP2 structural model) (Klare et al., 2011). Additionally, results from a disulphide mapping analysis of the E. coli Aer protein suggest different arrangements of the AS2 helices in different signalling states (Watts et al., 2011). Similarly, molecular dynamics simulations for the Afl1503 and Tsr proteins suggest that alternate signalling states display differential packing of the AS2 helices (Park et al., 2011; Hall et al., 2012). Thus, structure-based analyses have inspired two general hypotheses for HAMP function: gearbox, in which the four-helix coiled-coil domain rotates but remains intact (Hulko et al., 2006), and a helix rearrangement hypothesis, in which the domain alternates between the compact four-helix state and a less-compact state, in which the AS2 helices disengage from AS1 to form a two-helix coiled-coil (Airola et al., 2010). In these and other hypotheses (Swain and Falke, 2007), the HAMP domain is viewed as alternating between two discrete, rigid conformations with opposing signalling properties. © 2014 John Wiley & Sons Ltd, Molecular Microbiology, 91, 853–857
HAMP domain action 855
Biphasic dynamic bundle hypothesis for HAMP action The biphasic dynamic bundle hypothesis is based on a large collection of Tsr− (null) missense substitutions obtained from mutational saturation of the 53 residue HAMP coding region (Ames et al., 2008; Zhou et al., 2009; 2011; Kitanovic et al., 2011), and has been summarized in detail (Manson, 2008; 2009; 2011; Parkinson, 2010). Here, the HAMP domain operates as a dynamic three-state biphasic element (Fig. 1) (Zhou et al., 2011), rather than the static two-state device envisioned by the structure-based hypotheses described above. The biphasic dynamic bundle hypothesis comprises two main postulates. The first is that the HAMP domain traverses a dynamic range of packing stabilities. This derives partly from the observation that different Tsr− missense substitutions generate distinct locked-output phenotypes. One phenotype, kinase-off, causes counterclockwise (CCW) flagellar rotation, which in wild-type cells results from the presence of attractant (serine). Most of the locked kinase-off Tsr− substitutions are predicted to severely destabilize the HAMP bundle (e.g. charged replacements for critical hydrophobic packing residues). The reciprocal phenotype, kinase-on, produces clockwise (CW) flagellar rotation, which in wild-type cells results from the absence of attractant. This is the default state, because a strong CW-biased phenotype results from deletion of the entire HAMP domain (Ames et al., 2014). Most of the locked kinase-on Tsr− substitutions are less severe (e.g. polar uncharged replacements for critical hydrophobic residues) (Zhou et al., 2011). Thus, two different hypothetical kinase-off states bookend the kinase-on state (Fig. 1). One, termed CCW(A), results from stable HAMP bundle packing (i.e. wild-type cells cultured in the presence of attractant). The other, termed CCW(B), results from relatively severe destabilization of the HAMP bundle. The physiological operating range is hypothesized to span the high-stability kinase-off form through the intermediate-stability kinase-on form. Further destabilization (e.g. from mutational alteration) leads to the distinct CCW(B) kinase-off form (Zhou et al., 2011). The ‘biphasic’ component of the hypothesis derives from the notion that HAMP conformers at either end of the bundle stability range generate kinaseoff phenotypes, whereas conformers of intermediate stability result in kinase-on phenotypes (Fig. 1). The second postulate is that packing stabilities of the coiled-coil HAMP domain are inversely coupled to those of the distal coiled-coil output domain, thereby controlling activity. This results from a helical phase clash (embodied as a conserved stutter in the coiled-coil heptad repeat pattern) between the HAMP and output domains (Zhou et al., 2009). Indeed, for the E. coli NarX histidine kinase, deletion of the corresponding stutter element © 2014 John Wiley & Sons Ltd, Molecular Microbiology, 91, 853–857
results in locked signalling output (Stewart and Chen, 2010). In Tsr and other MCPs, the output (kinase control) domain is a long four-helix coiled-coil with three functionally distinct subregions. The adaptation (methylation) region is just distal to the HAMP domain. Evidence suggests that the coiled-coil structure in this region is more loosely packed in the presence of attractant and more tightly packed in its absence (Hazelbauer et al., 2008). Thus, according to the biphasic dynamic bundle hypothesis, the kinase-off state represents a tightly packed HAMP bundle that, acting through the helical phase clash, enforces a more loosely packed adaptation bundle. In the kinase-on state, these patterns are reversed. Control of kinase activity hypothetically results from propagation of these reciprocal conformations along the length of the output domain (Swain et al., 2009). During the attractant response, the adaptation region is methylated at specific Glu residues to reset the output response (Hazelbauer et al., 2008). Mutant Tsr receptors locked in the severely destabilized CCW(B) kinase-off state are refractory to the methyl-dependent adaptation system (Zhou et al., 2011), as described below. Thus, methylation state provides a convenient operational definition for classifying receptors in either the CCW(A) or CCW(B) states (Fig. 1).
Deletion analysis of the Tsr HAMP domain The biphasic dynamic bundle hypothesis is challenging. It depends largely on indirect inference, in sharp contrast to the seemingly concrete structure-based atomic representations of hypothetical signalling states. The underlying interpretations often are nuanced, and involve the full palette of genetic tests (complementation, suppression, epistasis). Perhaps most critically, the biphasic dynamic bundle hypothesis requires acceptance of missense substitution phenotypes as representing more- or less-stable packing conformations. The current paper addresses this latter point by presenting results from deletion analysis of the Tsr HAMP domain (Ames et al., 2014). Given the well-defined HAMP subdomain boundaries (Hulko et al., 2006; Zhou et al., 2009), it is possible to design deletions that unambiguously remove specific elements. Accordingly, 17 different deletions that remove part or all of the Tsr HAMP domain were evaluated with the same battery of tests used previously for the missense substitutions. All these deletions blocked sensory adaptation, whether by methylation or by mutational mimicry (i.e. changing methyl-accepting Glu residues to the methylated mimic, Gln). Thus, the HAMP domain is required for MCP sensory adaptation, presumably by controlling the adaptation region’s conformation.
856 V. Stewart ■
Twelve deletions conferred locked kinase-on phenotypes, the default signalling output, and therefore destroy all HAMP domain function. These include deletions of the connector, the connector plus the AS1 or AS2 helices, the AS2 helix only, and the entire HAMP domain. The remaining five deletions, which remove only the AS1 helix, bestow an opposite phenotype, locked kinaseoff CCW(B). Therefore, the connector plus AS2 helix in these mutant proteins must retain at least partial assembly and activity, in order to override the default kinase-on state. Indeed, the CCW(B) phenotype generated by these deletions requires three carboxyl-terminal hydrophobic residues involved in inter-helical packing, suggesting that the AS2 two-helix coiled-coil is required. Thus, the current paper by Ames et al. (2014) presents two substantive extensions of their prior work. First, congruent phenotypes are conferred by the deletions described here and by the corresponding missense substitutions described earlier. Second, the HAMP domain is essential for sensory adaptation per se, not just for substrate recognition by the methylation and demethylation enzymes.
A grand unified theory for HAMP domain action? The widely cited gearbox hypothesis, invoking axial rotation of all four HAMP domain helices, is appealing (Hulko et al., 2006). Conceptually simple, it is based on a substantial accumulation of in vitro and in vivo data (Hulko et al., 2006; Ferris et al., 2011; 2012). However, this hypothesis is difficult to reconcile with the observation that Tsr AS1 helix deletions retain substantial HAMP domain activity (albeit in the locked-off conformation) (Ames et al., 2014). Instead, the gearbox hypothesis requires AS1 and AS2 for both of its postulated signalling states. An alternative two-state model, derived from analysis of the Aer2 HAMP tandem array, proposes that ‘a key property of any CCW state may be the formation of a tight two-helix bundle at the [carboxyl]-terminal end of AS2’ (Airola et al., 2013). This seemingly is compatible with the phenotype conferred by the Tsr AS1 deletions (Ames et al., 2014). However, note that the two CCW states defined for Tsr have different properties: receptors in the kinase-off CCW(A) state exhibit adaptation, whereas mutant receptors driven into either the extreme kinase-on CW or the kinase-off CCW(B) states fail to adapt and therefore are more similar to each other (Ames et al., 2014). The dissimilarity between the CCW(A) and CCW(B) kinase-off states is more compatible with the view that these are at opposite ends of a dynamic continuum (Fig. 1). The biphasic dynamic bundle hypothesis therefore demands accepting this notion, that the extreme, nonphysiological CCW(B) state reflects an authentic extension of the helical stability range. However, Tsr proteins in this
state are severely perturbed. Not only unable to stimulate output kinase activity, they also are refractory to adaptation and cannot assemble into ternary signalling complexes (Zhou et al., 2011; Ames et al., 2014). Thus, one might posit that these HAMP lesions result simply in large-scale Tsr structural perturbations, and consequently are not relevant to considerations of physiological signalling. (Nevertheless, these mutant proteins do accumulate to near wild-type levels.) On the other hand, the CCW(B) state does require core hydrophobic residues essential for AS2 helix interactions, implying at least some relevance to normal function. It is suggested that other classes of HAMP-containing proteins might function normally over a different range of bundle stabilities (Ames et al., 2014). If so, analysis of such proteins might help to resolve this point. Considering the broad distribution, diverse arrangements and multiple sequence subfamilies of HAMP domains (Dunin-Horkawicz and Lupas, 2010), it is possible that different HAMP domains operate by substantially different mechanisms. Indeed, signal input for the E. coli Aer HAMP domain is hypothesized to involve lateral contact with the amino-terminal PAS domain (Campbell et al., 2010). Output from the Afl1503 protein, with its carboxylterminal HAMP domain, is unknown (Hulko et al., 2006), and the function of the Aer2 protein amino-terminal polyHAMP array, beyond serving as a dimerization determinant, likewise is unknown (Airola et al., 2010). By contrast, the Tsr HAMP domain represents the majority of known homologues, connecting an extracytoplasmic ligandbinding domain to a cytoplasmic output domain. In a different context, it was noted that ‘The search for unity, a powerful impetus in biology, can impede a proper appreciation of life’s diversity’ (Thaler and Stahl, 1988). It will not be surprising to learn that evolution has modified HAMP domains in different contexts for different functions. The Afl1503 and Aer2 proteins are potential examples. The biphasic dynamic bundle hypothesis exemplifies genetic analysis conducted at a very high level. The heroic effort to saturate the coding region with missense substitutions was just the beginning. This mutant collection has been exploited in numerous ways, making full use also of the many other well-characterized mutants and assays available for studying chemoreceptors. Accordingly, the hypothesis is on firm footing, and presents a worthy challenge to our understanding of two-state allosteric proteins in general. We can anticipate continued thoughtful analysis from efforts to reconcile disparate notions of how this tiny HAMP domain is so central in so many contexts.
References Airola, M.V., Watts, K.J., Bilwes, A.M., and Crane, B.R. (2010) Structure of concatenated HAMP domains provides © 2014 John Wiley & Sons Ltd, Molecular Microbiology, 91, 853–857
HAMP domain action 857
a mechanism for signal transduction. Structure 18: 436–448. Airola, M.V., Sukomon, N., Samanta, D., Borbat, P.P., Freed, J.H., Watts, K.J., and Crane, B.R. (2013) HAMP domain conformers that propagate opposite signals in bacterial chemoreceptors. PLoS Biol 11: e1001479. Ames, P., and Parkinson, J.S. (1988) Transmembrane signaling by bacterial chemoreceptors: E. coli transducers with locked signal output. Cell 55: 817–826. Ames, P., Zhou, Q., and Parkinson, J.S. (2008) Mutational analysis of the connector segment in the HAMP domain of Tsr, the Escherichia coli serine chemoreceptor. J Bacteriol 190: 6676–6685. Ames, P., Zhou, Q., and Parkinson, J.S. (2014) HAMP domain structural determinants for signalling and sensory adaptation in Tsr, the Escherichia coli serine chemoreceptor. Mol Microbiol 91: 875–886. Aravind, L., and Ponting, C.P. (1999) The cytoplasmic helical linker domain of receptor histidine kinase and methylaccepting proteins is common to many prokaryotic signalling proteins. FEMS Microbiol Lett 176: 111–116. Campbell, A.J., Watts, K.J., Johnson, M.S., and Taylor, B.L. (2010) Gain-of-function mutations cluster in distinct regions associated with the signalling pathway in the PAS domain of the aerotaxis receptor, Aer. Mol Microbiol 77: 575–586. Collins, L.A., Egan, S.M., and Stewart, V. (1992) Mutational analysis reveals functional similarity between NARX, a nitrate sensor in Escherichia coli K-12, and the methylaccepting chemotaxis proteins. J Bacteriol 174: 3667– 3675. Crick, F.H.C. (1953) The packing of α-helices: simple coiledcoils. Acta Crystallogr 6: 689–697. Dunin-Horkawicz, S., and Lupas, A.N. (2010) Comprehensive analysis of HAMP domains: implications for transmembrane signal transduction. J Mol Biol 397: 1156–1174. Ferris, H.U., Dunin-Horkawicz, S., Mondéjar, L.G., Hulko, M., Hantke, K., Martin, J., et al. (2011) The mechanisms of HAMP-mediated signaling in transmembrane receptors. Structure 19: 378–385. Ferris, H.U., Dunin-Horkawicz, S., Hornig, N., Hulko, M., Martin, J., Schultz, J.E., et al. (2012) Mechanism of regulation of receptor histidine kinases. Structure 20: 56–66. Hall, B.A., Armitage, J.P., and Sansom, M.S. (2012) Mechanism of bacterial signal transduction revealed by molecular dynamics of Tsr dimers and trimers of dimers in lipid vesicles. PLoS Comput Biol 8: e1002685. Hazelbauer, G.L., Falke, J.J., and Parkinson, J.S. (2008) Bacterial chemoreceptors: high-performance signaling in networked arrays. Trends Biochem Sci 33: 9–19. Hulko, M., Berndt, F., Gruber, M., Linder, J.U., Truffault, V., Schultz, A., et al. (2006) The HAMP domain structure implies helix rotation in transmembrane signaling. Cell 126: 929–940. Kitanovic, S., Ames, P., and Parkinson, J.S. (2011) Mutational analysis of the control cable that mediates transmembrane signaling in the Escherichia coli serine chemoreceptor. J Bacteriol 193: 5062–5072.
© 2014 John Wiley & Sons Ltd, Molecular Microbiology, 91, 853–857
Klare, J.P., Bordignon, E., Engelhard, M., and Steinhoff, H.J. (2011) Transmembrane signal transduction in archaeal phototaxis: the sensory rhodopsin II-transducer complex studied by electron paramagnetic resonance spectroscopy. Eur J Cell Biol 90: 731–739. Manson, M.D. (2008) The tie that binds the dynamic duo: the connector between AS1 and AS2 in the HAMP domain of the Escherichia coli Tsr chemoreceptor. J Bacteriol 190: 6544–6547. Manson, M.D. (2009) A mutational wrench in the HAMP gearbox. Mol Microbiol 73: 742–746. Manson, M.D. (2011) Not too loose, not too tight–just right. Biphasic control of the Tsr HAMP domain. Mol Microbiol 80: 573–576. Park, H., Im, W., and Seok, C. (2011) Transmembrane signaling of chemotaxis receptor Tar: insights from molecular dynamics simulation studies. Biophys J 100: 2955– 2963. Parkinson, J.S. (2010) Signaling mechanisms of HAMP domains in chemoreceptors and sensor kinases. Annu Rev Microbiol 64: 101–122. Stewart, V., and Chen, L.-L. (2010) The S helix mediates signal transmission as a HAMP domain coiled-coil extension in the NarX nitrate sensor from Escherichia coli K-12. J Bacteriol 192: 734–745. Swain, K.E., and Falke, J.J. (2007) Structure of the conserved HAMP domain in an intact, membrane-bound chemoreceptor: a disulfide mapping study. Biochemistry 46: 13684–13695. Swain, K.E., Gonzalez, M.A., and Falke, J.J. (2009) Engineered socket study of signaling through a four-helix bundle: evidence for a yin-yang mechanism in the kinase control module of the aspartate receptor. Biochemistry 48: 9266–9277. Thaler, D.S., and Stahl, F.W. (1988) DNA double-chain breaks in recombination of phage lambda and of yeast. Annu Rev Genet 22: 169–197. Wang, C., Sang, J., Wang, J., Su, M., Downey, J.S., Wu, Q., et al. (2013) Mechanistic insights revealed by the crystal structure of a histidine kinase with signal transducer and sensor domains. PLoS Biol 11: e1001493. Watts, K.J., Johnson, M.S., and Taylor, B.L. (2011) Different conformations of the kinase-on and kinase-off signaling states in the Aer HAMP domain. J Bacteriol 193: 4095– 4103. Williams, S.B., and Stewart, V. (1999) Functional similarities among two-component sensors and methyl-accepting chemotaxis proteins suggest a role for linker region amphipathic helices in transmembrane signal transduction. Mol Microbiol 33: 1093–1102. Zhou, Q., Ames, P., and Parkinson, J.S. (2009) Mutational analyses of HAMP helices suggest a dynamic bundle model of input-output signalling in chemoreceptors. Mol Microbiol 73: 801–814. Zhou, Q., Ames, P., and Parkinson, J.S. (2011) Biphasic control logic of HAMP domain signalling in the Escherichia coli serine chemoreceptor. Mol Microbiol 80: 596–611.
doi:10.1111/mmi.12516 First published online 27 January 2014
MicroCommentary The HAMP signal-conversion domain: static two-state or dynamic three-state? Valley Stewart* Department of Microbiology & Molecular Genetics, University of California, Davis, CA 95616-8665, USA.
Summary The 50-residue HAMP domain converts input signal into output response in a variety of transmembrane signal transduction proteins, including methylaccepting chemotaxis proteins and histidine kinases. HAMP domains are present in many other contexts as well. Despite focused study over the past decade, the question remains: How does this small domain play such a large role for so many different proteins? Analysis of structural models for the Afl1503 and Aer2 HAMP domains has generated hypotheses in which the HAMP domain assumes either of two discrete forms that generate opposing signal output. In contrast, genetic analysis of the HAMP domain from the Tsr methyl-accepting chemotaxis protein resulted in a distinct hypothesis, the biphasic dynamic bundle. In this hypothesis, signalling involves differential packing stabilities of the HAMP domain four-helix bundle, marked by at least three distinct states. Here I summarize and compare these hypotheses in the context of a deletion analysis that further explores the biphasic dynamic bundle hypothesis. Describing locked signal output mutants of the Escherichia coli transmembrane receptor Tsr (taxis to serine and repellants), Peter Ames and John (Sandy) Parkinson suggested that the ‘linker’ transmits conformational changes from the periplasmic input domain to the cytoplasmic output domain (Ames and Parkinson, 1988). This linker earned an acronym, HAMP (Aravind and Ponting, 1999), after homologous elements were recognized in sequences of disparate homodimeric signalling proteins, including
Accepted 9 January, 2014. *For correspondence. E-mail vjstewart@ ucdavis.edu; Tel. (+1) 530 754 7994; Fax (+1) 530 752 9014. Conflict of interest: None.
© 2014 John Wiley & Sons Ltd
histidine kinases (Collins et al., 1992), adenylyl cyclases, other methyl-accepting chemotaxis proteins (MCPs), and certain phosphatases. Indeed, HAMP domains are present, sometimes as tandem arrays, in a multiplicity of both transmembrane and cytoplasmic signalling proteins (Dunin-Horkawicz and Lupas, 2010). Accordingly, elucidating HAMP domain function is essential for comprehending signal transduction. Now, more than 25 years after its identification as a critical signalling element, distinct hypotheses for HAMP action are stimulating high-quality experimentation and thoughtful analysis. Ames and Parkinson, together with Qin Zhou, present further genetic tests of their idea that HAMP signalling involves a dynamic bundle populating a range of packing stabilities (Fig. 1) (Ames et al., 2014). For context, let us first consider hypotheses inspired by structural analyses.
Structure-based hypotheses for HAMP action Studies from several groups revealed that the HAMP domain monomer contains three elements of approximately 15 residues each: two amphipathic α-helices, termed AS1 and AS2, and a non-helical connector (Williams and Stewart, 1999). Specific interactions between these elements were revealed in a structural model for the dimeric HAMP domain from the Afl1503 protein of Archaeoglobus fulgidus, a hyperthermophilic archaeon (Hulko et al., 2006). In this NMR structural model, the AS1 and AS2 helices form a four-helix parallel coiled-coil, with the connector looped around the perimeter. The overall veracity of the four-helical structure is well-supported by mutational and cross-linking analyses of HAMP domains in native signal-transducing proteins (Parkinson, 2010). The Afl1503 structural model displays two distinctive features. First, the AS1 and AS2 helices are offset by one helical turn, with the AS1 amino-terminal and the AS2 carboxyl-terminal helical turns extending out from the four-helix bundle (Fig. 1). Second, the interhelical association, involving hydrophobic residues at the coiled-coil heptad a and d positions, exhibits an unusual x-da
854 V. Stewart ■
Fig. 1. The biphasic dynamic bundle hypothesis for Tsr and other MCPs. The four-helix coiled-coil HAMP domain helices AS1 and AS2 are coloured blue and gold respectively. The adaptation (methylation) region of the four-helix coiled-coil kinase control (output) domain is coloured green. In the presence of attractant (ATT), tight packing of the HAMP bundle enforces loose packing of the output bundle to generate the kinase-off CCW(A) conformation. In the presence of repellant (REP), intermediate packing of the HAMP bundle enforces intermediate packing of the output bundle to generate the kinase-on CW conformation. These two conformations, which delimit the physiological signalling range, are substrates for the adaptation methylation and demethylation enzymes (methylation sites are indicated as open or closed circles). Finally, certain missense and deletion mutants cause loose packing of the HAMP bundle. This drives tight packing of the output bundle to generate the severely kinase-off CCW(B) conformation, which lies outside of the physiological signalling range. This conformation requires packing by the three carboxyl-terminal residues in AS2, indicated by dotted lines. Figure courtesy of J.S. Parkinson.
packing geometry (Hulko et al., 2006) instead of the orthodox a-d pattern (Crick, 1953). This suggests a helix rotation (‘gearbox’) hypothesis, in which the HAMP domain interconverts between distinct x-da and a-d packing conformations (Hulko et al., 2006; Ferris et al., 2011; 2012). Congruent with this hypothesis, the HAMP domain displays more orthodox packing geometry in the X-ray structural model for the Streptococcus mutans VicK histidine kinase (Wang et al., 2013). The Aer2 (aerotaxis transducer) from Pseudomonas aeruginosa is a soluble protein containing three tandem HAMP domains at the amino-terminus. In the X-ray structural model, the HAMP1 and HAMP3 domain structures resemble that of the Afl1503 HAMP domain, with the notable exception that their packing geometries represent a mixture of x-da, a-d and x-x interactions (Airola et al., 2010). By contrast, the HAMP2 domain adopts a different conformation in which the AS2 helices form a two-helix rather than four-helix coiled-coil. This generated the distinct hypothesis, supported by analysis of select missense substitutions and chimeric MCPs (Airola et al., 2013), that the HAMP1 and HAMP2 structures represent opposing signalling states (Airola et al., 2010).
Meanwhile, biophysical analyses of the HtrII (halobacterial transducer of rhodopsin) protein from the archaeon Natronomonas pharaonis suggest two conformations, ‘compact’ (resembling the Afl1503 structural model) and ‘dynamic’ (resembling the Aer2 HAMP2 structural model) (Klare et al., 2011). Additionally, results from a disulphide mapping analysis of the E. coli Aer protein suggest different arrangements of the AS2 helices in different signalling states (Watts et al., 2011). Similarly, molecular dynamics simulations for the Afl1503 and Tsr proteins suggest that alternate signalling states display differential packing of the AS2 helices (Park et al., 2011; Hall et al., 2012). Thus, structure-based analyses have inspired two general hypotheses for HAMP function: gearbox, in which the four-helix coiled-coil domain rotates but remains intact (Hulko et al., 2006), and a helix rearrangement hypothesis, in which the domain alternates between the compact four-helix state and a less-compact state, in which the AS2 helices disengage from AS1 to form a two-helix coiled-coil (Airola et al., 2010). In these and other hypotheses (Swain and Falke, 2007), the HAMP domain is viewed as alternating between two discrete, rigid conformations with opposing signalling properties. © 2014 John Wiley & Sons Ltd, Molecular Microbiology, 91, 853–857
HAMP domain action 855
Biphasic dynamic bundle hypothesis for HAMP action The biphasic dynamic bundle hypothesis is based on a large collection of Tsr− (null) missense substitutions obtained from mutational saturation of the 53 residue HAMP coding region (Ames et al., 2008; Zhou et al., 2009; 2011; Kitanovic et al., 2011), and has been summarized in detail (Manson, 2008; 2009; 2011; Parkinson, 2010). Here, the HAMP domain operates as a dynamic three-state biphasic element (Fig. 1) (Zhou et al., 2011), rather than the static two-state device envisioned by the structure-based hypotheses described above. The biphasic dynamic bundle hypothesis comprises two main postulates. The first is that the HAMP domain traverses a dynamic range of packing stabilities. This derives partly from the observation that different Tsr− missense substitutions generate distinct locked-output phenotypes. One phenotype, kinase-off, causes counterclockwise (CCW) flagellar rotation, which in wild-type cells results from the presence of attractant (serine). Most of the locked kinase-off Tsr− substitutions are predicted to severely destabilize the HAMP bundle (e.g. charged replacements for critical hydrophobic packing residues). The reciprocal phenotype, kinase-on, produces clockwise (CW) flagellar rotation, which in wild-type cells results from the absence of attractant. This is the default state, because a strong CW-biased phenotype results from deletion of the entire HAMP domain (Ames et al., 2014). Most of the locked kinase-on Tsr− substitutions are less severe (e.g. polar uncharged replacements for critical hydrophobic residues) (Zhou et al., 2011). Thus, two different hypothetical kinase-off states bookend the kinase-on state (Fig. 1). One, termed CCW(A), results from stable HAMP bundle packing (i.e. wild-type cells cultured in the presence of attractant). The other, termed CCW(B), results from relatively severe destabilization of the HAMP bundle. The physiological operating range is hypothesized to span the high-stability kinase-off form through the intermediate-stability kinase-on form. Further destabilization (e.g. from mutational alteration) leads to the distinct CCW(B) kinase-off form (Zhou et al., 2011). The ‘biphasic’ component of the hypothesis derives from the notion that HAMP conformers at either end of the bundle stability range generate kinaseoff phenotypes, whereas conformers of intermediate stability result in kinase-on phenotypes (Fig. 1). The second postulate is that packing stabilities of the coiled-coil HAMP domain are inversely coupled to those of the distal coiled-coil output domain, thereby controlling activity. This results from a helical phase clash (embodied as a conserved stutter in the coiled-coil heptad repeat pattern) between the HAMP and output domains (Zhou et al., 2009). Indeed, for the E. coli NarX histidine kinase, deletion of the corresponding stutter element © 2014 John Wiley & Sons Ltd, Molecular Microbiology, 91, 853–857
results in locked signalling output (Stewart and Chen, 2010). In Tsr and other MCPs, the output (kinase control) domain is a long four-helix coiled-coil with three functionally distinct subregions. The adaptation (methylation) region is just distal to the HAMP domain. Evidence suggests that the coiled-coil structure in this region is more loosely packed in the presence of attractant and more tightly packed in its absence (Hazelbauer et al., 2008). Thus, according to the biphasic dynamic bundle hypothesis, the kinase-off state represents a tightly packed HAMP bundle that, acting through the helical phase clash, enforces a more loosely packed adaptation bundle. In the kinase-on state, these patterns are reversed. Control of kinase activity hypothetically results from propagation of these reciprocal conformations along the length of the output domain (Swain et al., 2009). During the attractant response, the adaptation region is methylated at specific Glu residues to reset the output response (Hazelbauer et al., 2008). Mutant Tsr receptors locked in the severely destabilized CCW(B) kinase-off state are refractory to the methyl-dependent adaptation system (Zhou et al., 2011), as described below. Thus, methylation state provides a convenient operational definition for classifying receptors in either the CCW(A) or CCW(B) states (Fig. 1).
Deletion analysis of the Tsr HAMP domain The biphasic dynamic bundle hypothesis is challenging. It depends largely on indirect inference, in sharp contrast to the seemingly concrete structure-based atomic representations of hypothetical signalling states. The underlying interpretations often are nuanced, and involve the full palette of genetic tests (complementation, suppression, epistasis). Perhaps most critically, the biphasic dynamic bundle hypothesis requires acceptance of missense substitution phenotypes as representing more- or less-stable packing conformations. The current paper addresses this latter point by presenting results from deletion analysis of the Tsr HAMP domain (Ames et al., 2014). Given the well-defined HAMP subdomain boundaries (Hulko et al., 2006; Zhou et al., 2009), it is possible to design deletions that unambiguously remove specific elements. Accordingly, 17 different deletions that remove part or all of the Tsr HAMP domain were evaluated with the same battery of tests used previously for the missense substitutions. All these deletions blocked sensory adaptation, whether by methylation or by mutational mimicry (i.e. changing methyl-accepting Glu residues to the methylated mimic, Gln). Thus, the HAMP domain is required for MCP sensory adaptation, presumably by controlling the adaptation region’s conformation.
856 V. Stewart ■
Twelve deletions conferred locked kinase-on phenotypes, the default signalling output, and therefore destroy all HAMP domain function. These include deletions of the connector, the connector plus the AS1 or AS2 helices, the AS2 helix only, and the entire HAMP domain. The remaining five deletions, which remove only the AS1 helix, bestow an opposite phenotype, locked kinaseoff CCW(B). Therefore, the connector plus AS2 helix in these mutant proteins must retain at least partial assembly and activity, in order to override the default kinase-on state. Indeed, the CCW(B) phenotype generated by these deletions requires three carboxyl-terminal hydrophobic residues involved in inter-helical packing, suggesting that the AS2 two-helix coiled-coil is required. Thus, the current paper by Ames et al. (2014) presents two substantive extensions of their prior work. First, congruent phenotypes are conferred by the deletions described here and by the corresponding missense substitutions described earlier. Second, the HAMP domain is essential for sensory adaptation per se, not just for substrate recognition by the methylation and demethylation enzymes.
A grand unified theory for HAMP domain action? The widely cited gearbox hypothesis, invoking axial rotation of all four HAMP domain helices, is appealing (Hulko et al., 2006). Conceptually simple, it is based on a substantial accumulation of in vitro and in vivo data (Hulko et al., 2006; Ferris et al., 2011; 2012). However, this hypothesis is difficult to reconcile with the observation that Tsr AS1 helix deletions retain substantial HAMP domain activity (albeit in the locked-off conformation) (Ames et al., 2014). Instead, the gearbox hypothesis requires AS1 and AS2 for both of its postulated signalling states. An alternative two-state model, derived from analysis of the Aer2 HAMP tandem array, proposes that ‘a key property of any CCW state may be the formation of a tight two-helix bundle at the [carboxyl]-terminal end of AS2’ (Airola et al., 2013). This seemingly is compatible with the phenotype conferred by the Tsr AS1 deletions (Ames et al., 2014). However, note that the two CCW states defined for Tsr have different properties: receptors in the kinase-off CCW(A) state exhibit adaptation, whereas mutant receptors driven into either the extreme kinase-on CW or the kinase-off CCW(B) states fail to adapt and therefore are more similar to each other (Ames et al., 2014). The dissimilarity between the CCW(A) and CCW(B) kinase-off states is more compatible with the view that these are at opposite ends of a dynamic continuum (Fig. 1). The biphasic dynamic bundle hypothesis therefore demands accepting this notion, that the extreme, nonphysiological CCW(B) state reflects an authentic extension of the helical stability range. However, Tsr proteins in this
state are severely perturbed. Not only unable to stimulate output kinase activity, they also are refractory to adaptation and cannot assemble into ternary signalling complexes (Zhou et al., 2011; Ames et al., 2014). Thus, one might posit that these HAMP lesions result simply in large-scale Tsr structural perturbations, and consequently are not relevant to considerations of physiological signalling. (Nevertheless, these mutant proteins do accumulate to near wild-type levels.) On the other hand, the CCW(B) state does require core hydrophobic residues essential for AS2 helix interactions, implying at least some relevance to normal function. It is suggested that other classes of HAMP-containing proteins might function normally over a different range of bundle stabilities (Ames et al., 2014). If so, analysis of such proteins might help to resolve this point. Considering the broad distribution, diverse arrangements and multiple sequence subfamilies of HAMP domains (Dunin-Horkawicz and Lupas, 2010), it is possible that different HAMP domains operate by substantially different mechanisms. Indeed, signal input for the E. coli Aer HAMP domain is hypothesized to involve lateral contact with the amino-terminal PAS domain (Campbell et al., 2010). Output from the Afl1503 protein, with its carboxylterminal HAMP domain, is unknown (Hulko et al., 2006), and the function of the Aer2 protein amino-terminal polyHAMP array, beyond serving as a dimerization determinant, likewise is unknown (Airola et al., 2010). By contrast, the Tsr HAMP domain represents the majority of known homologues, connecting an extracytoplasmic ligandbinding domain to a cytoplasmic output domain. In a different context, it was noted that ‘The search for unity, a powerful impetus in biology, can impede a proper appreciation of life’s diversity’ (Thaler and Stahl, 1988). It will not be surprising to learn that evolution has modified HAMP domains in different contexts for different functions. The Afl1503 and Aer2 proteins are potential examples. The biphasic dynamic bundle hypothesis exemplifies genetic analysis conducted at a very high level. The heroic effort to saturate the coding region with missense substitutions was just the beginning. This mutant collection has been exploited in numerous ways, making full use also of the many other well-characterized mutants and assays available for studying chemoreceptors. Accordingly, the hypothesis is on firm footing, and presents a worthy challenge to our understanding of two-state allosteric proteins in general. We can anticipate continued thoughtful analysis from efforts to reconcile disparate notions of how this tiny HAMP domain is so central in so many contexts.
References Airola, M.V., Watts, K.J., Bilwes, A.M., and Crane, B.R. (2010) Structure of concatenated HAMP domains provides © 2014 John Wiley & Sons Ltd, Molecular Microbiology, 91, 853–857
HAMP domain action 857
a mechanism for signal transduction. Structure 18: 436–448. Airola, M.V., Sukomon, N., Samanta, D., Borbat, P.P., Freed, J.H., Watts, K.J., and Crane, B.R. (2013) HAMP domain conformers that propagate opposite signals in bacterial chemoreceptors. PLoS Biol 11: e1001479. Ames, P., and Parkinson, J.S. (1988) Transmembrane signaling by bacterial chemoreceptors: E. coli transducers with locked signal output. Cell 55: 817–826. Ames, P., Zhou, Q., and Parkinson, J.S. (2008) Mutational analysis of the connector segment in the HAMP domain of Tsr, the Escherichia coli serine chemoreceptor. J Bacteriol 190: 6676–6685. Ames, P., Zhou, Q., and Parkinson, J.S. (2014) HAMP domain structural determinants for signalling and sensory adaptation in Tsr, the Escherichia coli serine chemoreceptor. Mol Microbiol 91: 875–886. Aravind, L., and Ponting, C.P. (1999) The cytoplasmic helical linker domain of receptor histidine kinase and methylaccepting proteins is common to many prokaryotic signalling proteins. FEMS Microbiol Lett 176: 111–116. Campbell, A.J., Watts, K.J., Johnson, M.S., and Taylor, B.L. (2010) Gain-of-function mutations cluster in distinct regions associated with the signalling pathway in the PAS domain of the aerotaxis receptor, Aer. Mol Microbiol 77: 575–586. Collins, L.A., Egan, S.M., and Stewart, V. (1992) Mutational analysis reveals functional similarity between NARX, a nitrate sensor in Escherichia coli K-12, and the methylaccepting chemotaxis proteins. J Bacteriol 174: 3667– 3675. Crick, F.H.C. (1953) The packing of α-helices: simple coiledcoils. Acta Crystallogr 6: 689–697. Dunin-Horkawicz, S., and Lupas, A.N. (2010) Comprehensive analysis of HAMP domains: implications for transmembrane signal transduction. J Mol Biol 397: 1156–1174. Ferris, H.U., Dunin-Horkawicz, S., Mondéjar, L.G., Hulko, M., Hantke, K., Martin, J., et al. (2011) The mechanisms of HAMP-mediated signaling in transmembrane receptors. Structure 19: 378–385. Ferris, H.U., Dunin-Horkawicz, S., Hornig, N., Hulko, M., Martin, J., Schultz, J.E., et al. (2012) Mechanism of regulation of receptor histidine kinases. Structure 20: 56–66. Hall, B.A., Armitage, J.P., and Sansom, M.S. (2012) Mechanism of bacterial signal transduction revealed by molecular dynamics of Tsr dimers and trimers of dimers in lipid vesicles. PLoS Comput Biol 8: e1002685. Hazelbauer, G.L., Falke, J.J., and Parkinson, J.S. (2008) Bacterial chemoreceptors: high-performance signaling in networked arrays. Trends Biochem Sci 33: 9–19. Hulko, M., Berndt, F., Gruber, M., Linder, J.U., Truffault, V., Schultz, A., et al. (2006) The HAMP domain structure implies helix rotation in transmembrane signaling. Cell 126: 929–940. Kitanovic, S., Ames, P., and Parkinson, J.S. (2011) Mutational analysis of the control cable that mediates transmembrane signaling in the Escherichia coli serine chemoreceptor. J Bacteriol 193: 5062–5072.
© 2014 John Wiley & Sons Ltd, Molecular Microbiology, 91, 853–857
Klare, J.P., Bordignon, E., Engelhard, M., and Steinhoff, H.J. (2011) Transmembrane signal transduction in archaeal phototaxis: the sensory rhodopsin II-transducer complex studied by electron paramagnetic resonance spectroscopy. Eur J Cell Biol 90: 731–739. Manson, M.D. (2008) The tie that binds the dynamic duo: the connector between AS1 and AS2 in the HAMP domain of the Escherichia coli Tsr chemoreceptor. J Bacteriol 190: 6544–6547. Manson, M.D. (2009) A mutational wrench in the HAMP gearbox. Mol Microbiol 73: 742–746. Manson, M.D. (2011) Not too loose, not too tight–just right. Biphasic control of the Tsr HAMP domain. Mol Microbiol 80: 573–576. Park, H., Im, W., and Seok, C. (2011) Transmembrane signaling of chemotaxis receptor Tar: insights from molecular dynamics simulation studies. Biophys J 100: 2955– 2963. Parkinson, J.S. (2010) Signaling mechanisms of HAMP domains in chemoreceptors and sensor kinases. Annu Rev Microbiol 64: 101–122. Stewart, V., and Chen, L.-L. (2010) The S helix mediates signal transmission as a HAMP domain coiled-coil extension in the NarX nitrate sensor from Escherichia coli K-12. J Bacteriol 192: 734–745. Swain, K.E., and Falke, J.J. (2007) Structure of the conserved HAMP domain in an intact, membrane-bound chemoreceptor: a disulfide mapping study. Biochemistry 46: 13684–13695. Swain, K.E., Gonzalez, M.A., and Falke, J.J. (2009) Engineered socket study of signaling through a four-helix bundle: evidence for a yin-yang mechanism in the kinase control module of the aspartate receptor. Biochemistry 48: 9266–9277. Thaler, D.S., and Stahl, F.W. (1988) DNA double-chain breaks in recombination of phage lambda and of yeast. Annu Rev Genet 22: 169–197. Wang, C., Sang, J., Wang, J., Su, M., Downey, J.S., Wu, Q., et al. (2013) Mechanistic insights revealed by the crystal structure of a histidine kinase with signal transducer and sensor domains. PLoS Biol 11: e1001493. Watts, K.J., Johnson, M.S., and Taylor, B.L. (2011) Different conformations of the kinase-on and kinase-off signaling states in the Aer HAMP domain. J Bacteriol 193: 4095– 4103. Williams, S.B., and Stewart, V. (1999) Functional similarities among two-component sensors and methyl-accepting chemotaxis proteins suggest a role for linker region amphipathic helices in transmembrane signal transduction. Mol Microbiol 33: 1093–1102. Zhou, Q., Ames, P., and Parkinson, J.S. (2009) Mutational analyses of HAMP helices suggest a dynamic bundle model of input-output signalling in chemoreceptors. Mol Microbiol 73: 801–814. Zhou, Q., Ames, P., and Parkinson, J.S. (2011) Biphasic control logic of HAMP domain signalling in the Escherichia coli serine chemoreceptor. Mol Microbiol 80: 596–611.
