• We are sorry, but NCBI web applications do not support your browser and may not function properly. More information
Logo of pnasPNASInfo for AuthorsSubscriptionsAboutThis Article
Proc Natl Acad Sci U S A. Feb 20, 2007; 104(8): 2885–2890.
Published online Feb 13, 2007. doi:  10.1073/pnas.0609359104
PMCID: PMC1797150
From the Cover
Microbiology

Evolutionary genomics reveals conserved structural determinants of signaling and adaptation in microbial chemoreceptors

Abstract

As an important model for transmembrane signaling, methyl-accepting chemotaxis proteins (MCPs) have been extensively studied by using genetic, biochemical, and structural techniques. However, details of the molecular mechanism of signaling are still not well understood. The availability of genomic information for hundreds of species enables the identification of features in protein sequences that are conserved over long evolutionary distances and thus are critically important for function. We carried out a large-scale comparative genomic analysis of the MCP signaling and adaptation domain family and identified features that appear to be critical for receptor structure and function. Based on domain length and sequence conservation, we identified seven major MCP classes and three distinct structural regions within the cytoplasmic domain: signaling, methylation, and flexible bundle subdomains. The flexible bundle subdomain, not previously recognized in MCPs, is a conserved element that appears to be important for signal transduction. Remarkably, the N- and C-terminal helical arms of the cytoplasmic domain maintain symmetry in length and register despite dramatic variation, from 24 to 64 7-aa heptads in overall domain length. Loss of symmetry is observed in some MCPs, where it is concomitant with specific changes in the sensory module. Each major MCP class has a distinct pattern of predicted methylation sites that is well supported by experimental data. Our findings indicate that signaling and adaptation functions within the MCP cytoplasmic domain are tightly coupled, and that their coevolution has contributed to the significant diversity in chemotaxis mechanisms among different organisms.

Keywords: chemotaxis, methyl-accepting chemotaxis protein, signal transduction

The signal transduction system controlling chemotaxis in Escherichia coli is an important model for the investigation of molecular information processing and is one of the best-studied signaling systems in biology. The key features of this system are high sensitivity, wide dynamic range, signal integration, memory, and adaptation (14). Principal components of the system include chemoreceptors [methyl-accepting chemotaxis proteins (MCPs)]; a histidine kinase, CheA; a receptor-coupling protein, CheW; and receptor-modification enzymes, CheR and CheB, all arranged into a protein complex (5, 6). MCPs detect various environmental and intracellular signals and control the activity of CheA, which in turn communicates the signal to the flagellar motor by phosphorylating its cognate response regulator, CheY (6).

MCPs are typically embedded in the cytoplasmic membrane (Fig. 1) and consist of an extracellular ligand-binding domain and a cytoplasmic signaling and adaptation domain (7, 8) connected by a HAMP linker domain (9, 10). The cytoplasmic domain includes two regions that contain sites of covalent modification (methylation helices 1 and 2) and the signaling subdomain implicated in CheA and CheW binding. The x-ray crystal structures of the cytoplasmic domain of the E. coli serine MCP Tsr (11) and the Thermotoga maritima MCP TM1143 (12), as well as scanning mutagenesis (13), revealed a dimeric four-helix bundle composed of two symmetric antiparallel coiled coils. No distinct structural subdomains that would correspond to known functional regions are apparent in either structure. The dimers of the Tsr cytoplasmic domain are organized into trimers in the crystal lattice (11, 14), an architecture also supported by genetic and biochemical studies (15, 16) and proposed as a major element of the signaling complex. Unlike Tsr, crystal packing of the cytoplasmic domain of the TM1143 chemoreceptor reveals a hedgerow of dimers (12), opening the possibility that different MCPs can be arranged into different types of assemblies. MCPs, CheW, and CheA likely form a 2D lattice maintained by a complex series of interactions (1114, 17). The exact nature of these interactions and the molecular mechanism of receptor signaling are not well understood, although experimental studies agree on the importance of interactions between MCPs for signal amplification (15, 16, 18, 19). Furthermore, there is significant diversity in the mechanism of signaling and adaptation in chemotaxis among different microbial species. For example, in E. coli, positive stimuli such as attractant increases inhibit CheA activity, whereas in Bacillus subtilis, the opposite is true. In E. coli, MCP demethylation increases in response only to negative stimuli, whereas in B. subtilis, it occurs in response to both positive and negative stimuli (20).

Fig. 1.
Schematic representation of the most common MCP topology and domain organization. Two transmembrane regions (TM1 and TM2) anchor the receptor in the membrane (shaded in gray). The sensory (ligand-binding) domain is extracellular, whereas the linker HAMP ...

The recent availability of genomic information for hundreds of microbial species enables the identification of regions and positions in protein sequences that are conserved over long evolutionary distances and therefore are critically important for protein function. MCPs were identified in many genomes of motile bacteria and archaea even before the explosion of genomic data (8). In this study, we used a large-scale genomic approach combined with analysis of structural data to identify major classes of the MCP cytoplasmic domain, resolve subdomains that are distinctive in both function and structure, and identify structurally and functionally important amino acid residues. The flexible bundle subdomain, which has not been previously recognized in MCPs, appears to be a conserved element critical for signal transduction. Our findings indicate that signaling and adaptation functions within the MCP cytoplasmic domain are tightly coupled, and that their coevolution has contributed to the significant diversity in chemotaxis mechanisms among different organisms.

Results

Phyletic Distribution and Domain Classes.

Applying well defined and strict domain boundaries and alignment procedures, as described in Materials and Methods, we detected 2,125 MCP cytoplasmic domain (MCP_CD) sequences in 152 genomes of bacteria and archaea and constructed their multiple alignment [supporting information (SI) Table 1 and SI Dataset 1]. MCP_CD is present in most major bacterial phyla and is noticeably absent from Crenarchaeota and eukaryotes (except for the genome of Anopheles gambia, where its presence is likely due to contamination with a prokaryotic DNA).

The most striking characteristic of MCP_CD is symmetric pairs of insertions or deletions (indels) consisting of multiples of heptads (Fig. 2A), regions of seven consecutive amino acid residues (labeled a-b-c-d-e-f-g), and corresponding exactly to two α-helical turns. Based on (i) sequence conservation and (ii) the presence of symmetric indels, we now define seven major MCP classes (Fig. 2A). Three of the seven classes have been recognized (21); however, placement of some of the indels in them was incorrect because of a lack of structural information and the limited number of sequences available for analysis. In all seven MCP classes, the first and fourth (a and d) residues of each heptad, called knobs, are highly conserved (Fig. 2B; SI Fig. 6) and tend to be hydrophobic; this heptad repeat pattern is characteristic of coiled-coil proteins (22) and is supported by available crystal structures of this domain. We assigned 1,915 MCP sequences (≈90% of all detected MCPs) to these seven major MCP classes (SI Table 1). Of these, a small fraction (80 sequences) have genome-specific symmetric indels forming five minor length subclasses (SI Table 1 and SI Fig. 7). The remaining 210 sequences did not confidently match domain models of any major class either because of truncation (mostly the result of genome assembly problems) or poor conservation and the presence of asymmetric indels. Remarkably, the latter was often concomitant with apparent genome-specific domain fusion and loss events. For example, many unaligned sequences represent a fusion of MCP_CD to an N-terminal multitransmembrane region module (41 sequences) or a stand-alone MCP_CD (93 sequences).

Fig. 2.
Schematic representation of MCP_CD features revealed by the multiple sequence alignment. The complete alignment is shown in SI Dataset 1. (A) Subdomain structure of major domain classes. The three subdomains, methylation helices, flexible bundle, and ...

Identification of Subdomains.

Sequence conservation features that are common to all major classes allowed us to characterize and define the boundaries of three distinct subdomains within MCP_CD. The central region of the alignment corresponding to the hairpin turn is highly conserved (Fig. 2B) and has been recognized as a highly conserved domain (23). We now define its exact boundaries to be heptads 01–04 based on sequence conservation and gap placement in the multiple alignment (Fig. 2). Two regions of medium conservation correspond to two methylation units that were recognized but not previously well defined (21, 24), where sites of covalent modification have been experimentally determined in several proteins from two major classes (Fig. 2A). We now specifically define the boundaries of the methylation subdomain as heptads 13–22 (Fig. 2). The signaling and methylation subdomains are separated by two poorly conserved regions (Fig. 2B), consisting of heptads 05–12, where most of the gaps in the alignment are located (Fig. 2A). These regions have not been explicitly recognized in MCPs previously. We have termed these regions the flexible bundle subdomain based on a further detailed examination (see below).

Signaling Subdomain.

High sequence conservation of the signaling subdomain is found in all MCP classes (Fig. 3A). Residues in the subdomain fall into two categories, intra- and interdimer interaction sites, corresponding to residues in heptad registers adeg and bcf, respectively (Fig. 4B). Intradimer interaction sites contribute to dimer stability by self-interaction between the same residues on different monomers. At interdimer interaction sites, cognate residues from each monomer in a dimer face outward in opposite directions (Fig. 3B). Interdimer interaction sites may contribute to both interactions between MCPs (15) and with CheA and CheW (23, 25). Intradimer interaction sites are significantly more conserved within the signaling subdomain than outside it (Fig. 2B), probably because of the stringent constraints generated there by conserved interdimer and CheA/CheW interactions. The three glycines in heptads N01 and C01 are strictly conserved because they maintain the tight structure of the hairpin turn. Thus, overall dimer organization and its interaction with the signaling complex appear to be conserved in all MCP classes. Residues conserved within but not among classes are of particular interest. The most conspicuous such position is the interdimer interaction site at N03b (Fig. 3A). In 36H-class receptors like Tsr, the Phe residue there stabilizes the trimer of dimers (Fig. 3B) posited to be important for receptor cooperativity (14, 15). In other receptor classes, this position is usually occupied by a charged residue, which would tend to destabilize cooperative interactions among receptors whether they are organized in trimers (11) or not (12).

Fig. 3.
Family- and class-specific conservation in the signaling subdomain. (A) Sequence conservation in the seven major MCP classes. Two representative interdimer (N03b) and intradimer (C02d) interaction sites are indicated by arrows. (B) Visualization of selected ...
Fig. 4.
Flexible bundle subdomain. (A) Representative knob (black) into hole (gray) packing in the E. coli Tsr protein. Residue numbers and heptad registers are shown. (B) Schematic representation of the arrangement of coiled-coil heptads in a four-helical bundle ...

Flexible Bundle Subdomain.

The flexible bundle subdomain consists of heptads 05–12 (Fig. 2). Knobs are the only conserved residues in this subdomain, which prompted us to examine their conservation pattern more closely. In a coiled coil, each knob fits into a pocket formed by four hole residues on an adjacent helix (Fig. 4A). The short side chains of alanine knobs often favor one side of the pocket, forming a smaller triangular pocket of three hole residues (26). Fig. 4B shows the arrangement of heptad registers in the MCP_CD four-helix bundle. Knob residues in the core of the bundle are arranged in square a-d layers. When large hydrophobic (ILMV) and small (AGST) residues are paired in a knob layer, its shape tends to skew from square to diamond shaped because the hydrophobic interaction between the large residue pair is stronger (Fig. 4C). When the direction of skew in adjacent knob layers alternates, four-stranded coiled coils are stable (27). We propose that long stretches of knob layers skewed in the same direction are unstable (Fig. 4D). We identified two such stretches in the flexible bundle subdomain (SI Fig. 8). In the 28H MCP class, this subdomain has been almost entirely deleted. In the other six classes, heptads 10–12 have a small knob on the N-terminal helix and a large knob on the C-terminal helix. Conversely, heptads 05–09 tend to have a large knob on the N-terminal helix and a small knob on the C-terminal helix (SI Fig. 8). It therefore appears that all MCP dimers have two tracts of unstable knob layers, skewed in one direction at the top of the flexible bundle subdomain and in the opposite direction at its base near the signaling subdomain.

These sequence-based results are further confirmed by structural analysis. Fig. 4E shows the regions in the E. coli Tsr (class 36H) and T. maritima TM1143 (class 44H) structures where coiled coils were predicted by using the SOCKET algorithm. In both structures, rigid coiled coils are predicted in the methylation and signaling subdomains; however, no strong coiled-coil packing was found in the flexible bundle subdomain, in part because the algorithm does not account for the triangular pockets favored by alanine knobs. Calculation of cross-diagonal distances in knob layers in these two structures also suggested that the flexible bundle subdomain has stacks of layers skewed in the same direction, whereas in the methylation and signaling subdomains, there are stacks of alternating skewed layers (SI Table 2). In knob layers of the TM1143 structure, one diagonal is on average 3.6 Å (34%) shorter than the other. We call the helices with a stretch of large knobs “bones” and of small knobs “tendons.” In both the Tsr and TM1143 crystal structures, experimental temperature factors show that the flexible bundle subdomain is more flexible than the other subdomains, and its “tendon” helices are more flexible than its “bone” helices (Fig. 4F and SI Table 3).

Methylation Subdomain.

The methylation subdomain (heptads 13–22) has been identified in all but one of the major MCP classes. Remarkably, the entire methylation subdomain appears to be missing from the 24H class (Fig. 2A and SI Dataset 1). A striking feature of the MCP_CD alignment is the conservation of Glx pairs that look like the consensus methylation sequence -[EQ]-[EQ]-x(2)-A-[ST]- found in E. coli MCPs (28) (SI Fig. 6). Structurally, the Glx pair and small residues in the motif lie in adjacent turns on the solvent-exposed side of a methylation helix. One residue of the Glx pair is the target for methylation by CheR and for deamidation and demethylation by CheB and CheD, a receptor-modifying deamidase lacking from E. coli but typical of many chemotaxis systems (29), whereas the flanking small residues are thought to be important for correct docking at the helix by the adaptation enzymes (30, 31). We collected sequences from each signaling class containing the methylation subdomain, where a Glx pair was identified in the bc heptad registers predicted to be on the solvent-exposed surface of the four-helix bundle and determined the consensus methylation sequence for each class. Merging class-specific information together, we generated a consensus methylation sequence for the MCP_CD family: -[ASTG]-[ASTG]-x(2)-[EQ]-[EQ]-x(2)-[ASTG]-[ASTG]-. This conservation pattern strongly supports the importance of small residues in the helical turns both upstream and downstream of the Glx pair. Sites matching this motif in all six classes are visualized in a dot plot of the MCP_CD alignment and in structural models (Fig. 5). It is evident that each MCP class has a different pattern of methylation sites, and these sites map onto different locations in the 3D structure of the domain.

Fig. 5.
Methylation sites are conserved and located at class-specific positions. (A) Dot plot of the MCP_CD alignment showing conserved predicted methylation sites. A total of 1,656 sequences are shown. Heptads (N22–C22) are indicated by alternating gray ...

Pentapeptide Tether.

All MCP sequences were scanned for the presence of the C-terminal pentapeptide tether (see SI Text) to which CheB and CheR bind in E. coli. The “assistance neighborhoods” of MCPs linked by this tethering mechanism have been found to enable precise adaptation in E. coli (4). Most MCPs contain no C-terminal extension; we found 217 MCPs in 67 of 152 genomes that do contain the consensus pentapeptide motif -x-[HFWY]-x (2)-[HFWY]-. All of these MCPs are of classes 36H and 34H, and all but two of the organisms where they are found are proteobacterial (SI Table 1), implying that the pentapeptide tether is a recently evolved mode of interaction between MCPs and adaptation enzymes.

Discussion

Flexible Bundle and Signaling.

Two functional regions have been recognized previously in MCP_CD: the methylation helices and the signaling subdomain, both with poorly defined boundaries (7, 11, 12, 24). Based on conservation of individual amino acid positions, heptad pattern, and indel locations, we defined boundaries of both regions and identified a third previously unrecognized one, which we termed the flexible bundle subdomain. We propose that the characteristic pattern of knob residue conservation in this subdomain is an important feature of the MCP signaling mechanism. The rigid “bones” are likely to stabilize the dimer structure, whereas the flexible “tendons” are likely to transmit sensory information to the signaling subdomain. Scanning mutagenesis of glycine residues in the Tar receptor of E. coli revealed the importance of the transition point between heptads 09 and 10 (32). Mutation of the d knob residue in heptad N10 (G338A and G338C) created a lock-on phenotype with constitutive kinase activation. Our analysis shows that this residue forms a conserved N10d/C10a knob layer, which is flanked above and below by “bone” and “tendon” helices in all major MCP classes except 28H, where most of this subdomain has been deleted (Fig. 5B and SI Fig. 8). Coleman et al. (32) named this region the glycine hinge and argued that its function is to allow receptor dimers to bend, either to promote the initial assembly of the trimer of dimers or as part of the signaling mechanism. Conservation of a flexible bundle region with a central glycine hinge strongly supports the hypothesis that bending is central to the signaling mechanism.

Strong conservation of the intradimer interaction sites suggests that all MCPs from different classes and organisms form dimers; however, there is no evidence for a preferential pattern of higher-order organization. Interestingly, the most conspicuous class-specific residue in the signaling subdomain is Phe in the N03b position (Fig. 3B) of the 36H class exemplified by the E. coli Tsr receptor. This is a strong aromatic contact site of the trimer of dimers (11, 14). On the other hand, this position in other MCP classes holds a strongly conserved charged residue and is consistent with the hedgerow organization of dimers of the 44H class receptor TM1143 (12). Only an MCP dimer and not a trimer of dimers can be accommodated in the model of CheA/CheW interactions recently proposed for T. maritima (12). In cryoelectron microscopy studies in E. coli, three MCP dimers cluster together near CheA and CheW but do not form a trimer (5), suggesting that the model from T. maritima may apply more widely. Our analysis is consistent with both models, suggesting that MCP dimers of different classes might form different patterns of higher-order organization. However, we favor the hypothesis that these structural data represent static snapshots of a conserved, dynamic receptor array. Although the exact nature of its higher-order organization remains to be determined, we suggest that a critical element of this dynamic signaling mechanism might be the oscillation of each MCP dimer between straight and bent conformations, one of which might favor higher-order clustering more than the other. This “forest of dimers” model provides a structural basis for theoretical Ising models that explore how receptor cooperativity can enable high gain and wide dynamic range in chemotactic signaling (17). MCP classes of different lengths and with different interdimer contact residues may have evolved as a way to tune the “infectivity” of ligand binding, a key parameter of such models that quantifies how many neighboring dimers are affected by ligand binding at one MCP dimer. We admit that current experimental evidence favors an alternative view of receptor bending in E. coli, specifically that it allows dimers to remain associated (e.g., as trimers of dimers) in both signaling states despite significant conformational changes. The analysis presented here may help to design experiments that would directly test alternative hypotheses.

Coevolution of Signaling and Adaptation.

We have shown that symmetric pairs of indels are a key pathway of evolutionary change in the MCP cytoplasmic domain but have not established a mechanism that can generate such unusual changes in sequence. We propose that cytoplasmic MCPs may provide a reservoir of decreased selective pressure where the maintenance of symmetry at the ends of the domain away from the hairpin turn is less important for structural stability than in their membrane-bound counterparts. Cytoplasmic MCPs in archaeal species, for example, fit the 44H class domain model less well than their membrane-bound counterparts because of gaps in this region (data not shown) and may represent an evolutionary bridge between class 44H and 40H receptors. This hypothesis could also explain the significant number of unaligned MCP_CD sequences from cytoplasmic MCPs with apparently asymmetric indel locations.

Sensory inputs from receptor to kinase are oppositely wired in E. coli and B. subtilis: in E. coli, positive stimuli decrease kinase activity, whereas in B. subtilis, positive stimuli increase kinase activity (20). The adaptation mechanisms of the two organisms also differ. In the class 36H receptors of E. coli, positive stimuli increase methylation at all sites and negative stimuli decrease methylation at all sites. In contrast, the class 44H McpB receptor of B. subtilis has a differential methylation mechanism: one residue in MH2 (E630 at alignment position C18c) is methylated in response to positive stimuli and demethylated in response to negative stimuli, whereas another (E637 at C19c) is the opposite (33). The bending model of kinase activation is consistent with both of these mechanisms. The indel in MH2 that distinguishes class 44H from class 36H receptors includes position C18c (Fig. 2A and SI Fig. 6), so there is a correlation between the change in pattern of methylation sites and the change in adaptation mechanism. Based on our data linking methylation pattern and receptor length in all major MCP classes (Fig. 5), we propose that signaling and adaptation are closely linked not just in E. coli and B. subtilis, but in all organisms, and that signaling and adaptation mechanisms have coevolved throughout the natural history of chemotaxis.

Materials and Methods

For details, see SI Text.

Multiple Sequence Alignment.

MCP protein sequences were identified by scanning 312 prokaryotic genomes against the Pfam MCPsignal domain model (Pfam accession no. PF00015) by using the MiST database (34). A total of 2,133 matching sequences were found in 152 organisms. The Pfam model does not depict the full-length MCP cytoplasmic domain correctly; therefore, we defined sequence markers for start, center, and end positions in the domain and used them to determine correct domain boundaries (see SI Text). Ungapped alignments of each length class were generated by using ClustalW 1.83 (35) with default settings and manually edited; then HMM domain models were generated from these seed alignments by using HMMER 2.3.2 (36) with default parameters. A sequence was assigned to the top-scoring domain model if the model's Z-score was at least 50 units higher than the second-best-scoring model. This procedure resulted in the assignment of a length class to 1,846 of 2,125 MCPs; eight sequences were discarded because all of their Z-scores were negative and they had poor matches to the highly conserved domain HMM. Of the remaining 279 MCPs, 69 were classified manually by examining the HMMER output, leaving 210 unaligned MCPs. The 12 individually aligned classes were merged into a single multiple alignment guided by the information content and amino acid consensus determined for each class, by profile–profile alignments, and by structural information from the Tsr (11) and TM1134 (12) crystal structures.

Coiled-Coil Analysis.

The pronounced conservation of knob residues (Fig. 2B) enabled the identification of heptad register throughout MCP_CD. Knobs and holes were also identified in the Tsr (11) and TM1143 (12) crystal structures using the SOCKET algorithm (22, 37) with a 7.8-Å cutoff.

Methylation Pattern Analysis.

Methylation sites were identified in each length class by locating adjacent sites in the b and c heptad registers where the information content of Glu or Gln residues in the multiple sequence alignment exceeded 0.5 bits at both sites. To visualize methylation sites in different MCP_CD length classes, a neighbor-joining tree was constructed from ungapped aligned sequences by using MEGA3 software (38) with the p distance model of amino acid substitution and complete deletion of gap residues. The tree was partitioned into maximal subtrees containing receptors from just one length class. Class 24H MCPs were analyzed separately because they lack the methylation subdomain. Methylation sites matching the consensus motif -[ASTG]-[ASTG]-x (2)-[EQ]-[EQ]-x (2)-[ASTG]-[ASTG]- were identified and visualized by using custom Perl scripts. Sequences in Fig. 5A were arranged in tree order but with subtrees rearranged to cluster all receptors of the same class.

Computer Graphics.

Structure images were generated in Pymol (www.pymol.org). Sequence logos were generated in WebLogo (39). All sequence-based figures were generated from sequences that matched their top-scoring domain model without gaps, except that the methylation motifs in Fig. 5A were generated from all categorized sequences.

Supplementary Material

Supporting Information:

Acknowledgments

We thank Kristin Wuichet and John S. Parkinson for discussions, careful reading of the manuscript, and helpful comments. This work was supported by National Institutes of Health Grant GM72285 and by the Science Alliance (University of Tennessee).

Abbreviations

MCP
methyl-accepting chemotaxis protein
MCP_CD
the MCP cytoplasmic domain.

Footnotes

The authors declare no conflict of interest.

This article is a PNAS direct submission.

See Commentary on page 2559.

This article contains supporting information online at www.pnas.org/cgi/content/full/0609359104/DC1.

References

1. Segall JE, Block SM, Berg HC. Proc Natl Acad Sci USA. 1986;83:8987–8991. [PMC free article] [PubMed]
2. Sourjik V, Berg HC. Proc Natl Acad Sci USA. 2002;99:123–127. [PMC free article] [PubMed]
3. Alon U, Surette MG, Barkai N, Leibler S. Nature. 1999;397:168–171. [PubMed]
4. Endres RG, Wingreen NS. Proc Natl Acad Sci USA. 2006;103:13040–13044. [PMC free article] [PubMed]
5. Francis NR, Wolanin PM, Stock JB, DeRosier DJ, Thomas DR. Proc Natl Acad Sci USA. 2004;101:17480–17485. [PMC free article] [PubMed]
6. Wadhams GH, Armitage JP. Nat Rev Mol Cell Biol. 2004;5:1024–1037. [PubMed]
7. Falke JJ, Hazelbauer GL. Trends Biochem Sci. 2001;26:257–265. [PMC free article] [PubMed]
8. Zhulin IB. Adv Microb Physiol. 2001;45:157–198. [PubMed]
9. Aravind L, Ponting CP. FEMS Microbiol Lett. 1999;176:111–116. [PubMed]
10. Hulko M, Berndt F, Gruber M, Linder JU, Truffault V, Schultz A, Martin J, Schultz JE, Lupas AN, Coles M. Cell. 2006;126:929–940. [PubMed]
11. Kim KK, Yokota H, Kim SH. Nature. 1999;400:787–792. [PubMed]
12. Park SY, Borbat PP, Gonzalez-Bonet G, Bhatnagar J, Pollard AM, Freed JH, Bilwes AM, Crane BR. Nat Struct Mol Biol. 2006;13:400–407. [PubMed]
13. Falke JJ, Kim SH. Curr Opin Struct Biol. 2000;10:462–469. [PMC free article] [PubMed]
14. Kim SH, Wang W, Kim KK. Proc Natl Acad Sci USA. 2002;99:11611–11615. [PMC free article] [PubMed]
15. Ames P, Studdert CA, Reiser RH, Parkinson JS. Proc Natl Acad Sci USA. 2002;99:7060–7065. [PMC free article] [PubMed]
16. Studdert CA, Parkinson JS. Proc Natl Acad Sci USA. 2004;101:2117–2122. [PMC free article] [PubMed]
17. Bray D, Levin MD, Morton-Firth CJ. Nature. 1998;393:85–88. [PubMed]
18. Gestwicki JE, Kiessling LL. Nature. 2002;415:81–84. [PubMed]
19. Sourjik V, Berg HC. Nature. 2004;428:437–441. [PubMed]
20. Szurmant H, Ordal GW. Microbiol Mol Biol Rev. 2004;68:301–319. [PMC free article] [PubMed]
21. LeMoual H, Koshland DE. J Mol Biol. 1996;261:568–585. [PubMed]
22. Walshaw J, Woolfson DN. J Struct Biol. 2003;144:349–361. [PubMed]
23. Liu JD, Parkinson JS. J Bacteriol. 1991;173:4941–4951. [PMC free article] [PubMed]
24. Starrett DJ, Falke JJ. Biochemistry. 2005;44:1550–1560. [PMC free article] [PubMed]
25. Mehan RS, White NC, Falke JJ. Biochemistry. 2003;42:2952–2959. [PMC free article] [PubMed]
26. Gernert KM, Surles MC, Labean TH, Richardson JS, Richardson DC. Protein Sci. 1995;4:2252–2260. [PMC free article] [PubMed]
27. Munson M, Obrien R, Sturtevant JM, Regan L. Protein Sci. 1994;3:2015–2022. [PMC free article] [PubMed]
28. Terwilliger TC, Wang JY, Koshland DE. J Biol Chem. 1986;261:814–820.
29. Chao X, Muff TJ, Park SY, Zhang S, Pollard AM, Ordal GW, Bilwes AM, Crane BR. Cell. 2006;124:561–571. [PubMed]
30. Shapiro MJ, Panomitros D, Koshland DE. J Biol Chem. 1995;270:751–755. [PubMed]
31. Perez E, West AH, Stock AM, Djordjevic S. Biochemistry. 2004;43:953–961. [PMC free article] [PubMed]
32. Coleman MD, Bass RB, Mehan RS, Falke JJ. Biochemistry. 2005;44:7687–7695. [PMC free article] [PubMed]
33. Zimmer MA, Tiu J, Collins MA, Ordal GT. J Biol Chem. 2000;275:24264–24272. [PubMed]
34. Ulrich LE, Zhulin IB. Nucleic Acids Res. 2007;35:D386–D390. [PMC free article] [PubMed]
35. Chenna R, Sugawara H, Koike T, Lopez R, Gibson TJ, Higgins DG, Thompson JD. Nucleic Acids Res. 2003;31:3497–3500. [PMC free article] [PubMed]
36. Eddy SR. Bioinformatics. 1998;14:755–763. [PubMed]
37. Walshaw J, Woolfson DN. J Mol Biol. 2001;307:1427–1450. [PubMed]
38. Kumar S, Tamura K, Nei M. Brief Bioinform. 2004;5:150–163. [PubMed]
39. Crooks GE, Hon G, Chandonia JM, Brenner SE. Genome Res. 2004;14:1188–1190. [PMC free article] [PubMed]
40. Kristich CJ, Ordal GW. J Biol Chem. 2002;277:25356–25362. [PubMed]
41. Koch MK, Oesterhelt D. Mol Microbiol. 2005;55:1681–1694. [PubMed]
42. Perazzona B, Spudich JL. J Bacteriol. 1999;181:5676–5683. [PMC free article] [PubMed]
43. Perez E, Zheng H, Stock AM. J Bacteriol. 2006;188:4093–4100. [PMC free article] [PubMed]

Articles from Proceedings of the National Academy of Sciences of the United States of America are provided here courtesy of National Academy of Sciences
PubReader format: click here to try

Formats:

Related citations in PubMed

See reviews...See all...

Links

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...