Using Structural Information to Change the Phosphotransfer Specificity of a Two-Component Chemotaxis Signalling Complex

Analysis of the crystal structure of a phosphotransfer complex from the Rhodobacter sphaeroides chemotaxis pathway allowed reengineering of molecular recognition in a two-component signalling system.


Introduction
Bacteria, Archaea, and some eukaryotes use two-component signalling pathways to detect environmental conditions and bring about appropriate changes in cellular behaviour [1,2]. Twocomponent pathways comprise sensor histidine kinases (HPK) and response regulators (RRs). Environmental stimuli control the rate at which the HPK autophosphorylates on a conserved histidine residue. Once phosphorylated, the HPK transfers the phosphoryl group to an aspartate residue within the receiver domain of the cognate RR. The phosphorylated RR (RR-P), often a transcriptional regulator, then effects a response appropriate to the original stimulus. Some bacteria have over 150 different HPK and RR pairs, and the specificity of the phosphorylation reactions between them needs to be tightly controlled to prevent HPKs from inappropriately phosphorylating and activating noncognate RRs. A number of mechanisms contribute to this specificity [3,4], although the primary one is molecular recognition, in which a HPK shows a strong kinetic preference for its cognate RRs [5,6]. Understanding the mecha-nisms involved in molecular recognition will not only allow prediction of interacting pairs, but potentially allow the rewiring of bacterial sensory pathways for use in synthetic biology.
Many two-component systems utilize an additional element, the Hpt (histidine-containing phosphotransfer) domain in phosphotransfer; these include multistep phosphorelays and chemotaxis signalling pathways. In multistep phosphorelays, the Hpt domain serves as an intermediate in the transfer of phosphoryl groups from the receiver domain phosphorylated by the HPK and the output RR. In contrast, the histidine within the Hpt domain is the initial site of phosphorylation in the chemotaxis HPK, CheA, and is phosphorylated using ATP as the phosphodonor [7]. Subsequently, the phosphoryl group is transferred from the histidine residue in the Hpt (P1) domain of CheA to an aspartate residue on either of two RRs, CheY or CheB [8]. In this study, we present the structure of the Hpt (P1) domain of CheA 3 from R. sphaeroides in complex with its cognate RR CheY 6 , which, to our knowledge, is the first structure of a Hpt domain of a CheA protein in complex with its RR.
In groundbreaking work, the Laub group used a mutual information bioinformatics approach to identify coevolving residues in HPKs and their cognate RRs [28]. They reasoned that many of the coevolving residues would be the specificity determinants for the phosphotransfer reaction and went on to show that it is possible to switch the RR substrate specificity of the HPK EnvZ by mutating these residues [28]. In this study, we use a structure-based approach to identify specificity determinants for the phosphotransfer reaction between CheA 3 P1-P and CheY 6 . By introducing these residues into the noncognate RRs, we have been able to change their kinase specificity, allowing them to be phosphorylated by CheA 3 P1-P.

Structure of the CheA 3 P1.CheY 6 Complex
In order to elucidate the molecular details of the interaction between CheA 3 and CheY 6 , the complex structure of CheY 6 and the unphosphorylated Hpt domain of CheA 3 (residues 1-135) has been solved to 1.40 Å using Seleno single-wavelength anomalous dispersion (SAD) for phasing. Data collection and refinement statistics can be found in Table 1.
CheY 6 has the typical (a/b) 5 topology seen for E. coli CheY [29] and other structurally characterized RRs (Figure 1). Comparison with E. coli CheY reveals a high degree of structural conservation with a root mean square deviation (rmsd) of 1.4 Å over 115 Ca atoms (26% sequence identity). Although crystallized in the presence of Mn 2+ , there was no additional density in the CheY 6 divalent cation binding site. However, the conformation of the metal coordinating residues Asp56, Asp9, and Asp10 is most similar to that found in the structure of Mg 2+ bound CheY [30]. Only the backbone carbonyl of Glu58 is facing away rather than pointing towards the potential metal binding site ( Figure S1). CheY 6 has an elongated loop region connecting b5 and a5 ( Figure 1). This loop comprises 13 residues (residues 107-119) in CheY 6 compared to only three residues (109-111) in E. coli CheY. In the crystal, this loop is only partially ordered, suggesting that it is highly flexible. Residues 113-118 could not be traced and were omitted from the final model. The N-terminal region of this b5-a5 loop in combination with a1 of CheY 6 form the vast majority of contacts to CheA 3 P1 (Figure 1). CheA 3 P1 forms a four-helix bundle (aA-aD) with an additional C-terminal helix (aE) (Figure 1). Despite low sequence identity, it is structurally very similar to previously determined CheA P1 structures [31][32][33]. Comparison with CheA P1 from Salmonella enterica serovar Typhimurium [32] and Thermotoga maritima [33] gives an rmsd of 1.2 Å over 116 Ca atoms (21% sequence identity) and 1.5 Å over 101 Ca (18% sequence identity), respectively. The site of phosphorylation, His51 in CheA 3 P1, is located on aB, in close proximity to the active site of the RR (Figure 1) with the phosphoacceptor Asp56 on CheY 6 being 7.5Å from His51 on CheA 3 P1. Helices aA and aB face the RR and together with the aB-aC loop form the interface with CheY 6 ( Figure 1). Structure of Phosphorylated CheA 3 P1 in Complex with CheY 6 (D56N, S83A) To show that the conclusions drawn from the unphosphorylated structure are also valid for the physiologically relevant, phosphorylated complex, we solved the structure of phosphorylated CheA 3 P1 in complex with CheY 6 . Since the rapid rate of phosphotransfer between CheA 3 P1-P and CheY 6 does not allow crystallization of the wild-type, phosphorylated complex, we formed a stable, complex by introducing two substitutions, D56N and S83A, in the active site of CheY 6 (see Materials and Methods for details). These substitutions have previously been shown to abolish phosphotransfer from CheA 3 P1-P [19]. The structure was solved to 2.8 Å by molecular replacement using the unphosphorylated structure as model. Data collection and refinement statistics can be found in Table 1. The structure shows clear additional density adjacent to Ne2 of His51 of CheA 3 P1 in agreement with phosphorylation of this residue ( Figure S2). Due to the moderate resolution of the analysis, residues 60-65, 85-97, and 111-121 on CheY 6 could not be traced and were not included in the final model.

Author Summary
The ability to respond to environmental stimuli is a universal feature of living cells. Evolution has created a vast array of signalling mechanisms that enable cells to react in many ways to extracellular changes. In bacteria, twocomponent signalling mechanisms, comprising a sensor protein kinase paired with its a cognate response regulator, are used widely to sense and respond to environmental changes. Some species of bacteria have over 150 different two-component pairs in a single cell, so the specificity between these pairs has to be tightly controlled to prevent ''crossed wires'' between signalling pathways. In this study, we have identified the determinants of this specificity in a two-component complex that controls the movement of Rhodobacter sphaeroides along a chemical gradient. By solving and analysing the crystal structure of this complex, we were able to pinpoint the amino acid residues that are crucially involved in formation of the complex. Knowledge of these crucial residues allowed us to convert noncognate response regulators into cognate response regulators simply by changing two amino acids. This reengineering of two-component signalling pathways paves the way for producing custom-designed circuits for applications in synthetic biology.
Compared to the unphosphorylated complex, CheA 3 P1-P undergoes a rigid body translation of 2.1 Å relative to the RR ( Figure 2A). This realignment of CheA 3 P1-P positions the phosphorylated His51 closer to the phosphoacceptor on CheY 6 , Asp56 ( Figure 2B). The phosphorylated His51 is facing away from the active site of CheY 6 ; however, a 180u flip of this side chain would put it in near linear geometry and place the phosphoryl group within 4.5 Å of Asp56 of CheY 6 ( Figure 2B). The orientation of the phosphorylated His51 seen in the crystal structure is likely caused by the lack of a divalent cation in the metal binding site of the RR leading to electrostatic repulsion from the CheY 6 active site.
The difference in binding affinity for CheY 6 between CheA 3 and CheA 3 -P is not known. In E. coli, the difference in K d values for CheY between CheA and CheA-P is relatively small, with both values being in the low micromolar range [34]. Consistent with this, we find very little change in the total buried surface area between the phosphorylated (605 Å 2 ) and unphosphorylated (530 Å 2 ) structures of CheA 3 P1 complexed with CheY 6 . The only difference is a hydrogen bond formed between Glu58 of CheY 6 and His51 of CheA 3 P1 in the unphosphorylated complex that is released in the phosphorylated complex. As this is not contributing to the interface of the physiologically important complex, it is therefore not  discussed further. Because the interfaces in the phosphorylated and unphosphorylated complex are highly similar, the following discussion of the interactions between CheA 3 P1 and CheY 6 are based on the high-resolution, unphosphorylated structure.

Interactions between CheA 3 P1 and CheY 6
The interface between CheA 3 P1 and CheY 6 is dominated by hydrophobic interactions with only one hydrogen bond and no salt bridges formed between the two proteins. The buried surface area of the interface is small (530 Å 2 ), indicative of a weak interaction, consistent with the transient nature of the complexes [35]. A continuous interface is formed by two sites of interaction complemented by a hydrogen bond formed between Ser83 on CheY 6 and Arg58 on CheA 3 P1. One of these sites lies between the elongated loop region between b5 and a5 in CheY 6 and aB and the following loop connecting aB and aC on CheA 3 P1 ( Figure 1A). The other site lies between the N-terminal end of a1 on CheY 6 and aA/aB on CheA 3 P1 ( Figure 1B). The latter is the major binding site including ,70% of the total buried surface area (358 Å 2 ).
In this region, helices aA and aB of CheA 3 P1 form a hydrophobic pocket comprising Ile11, Leu14, and Tyr15 on aA, and Asn56, Val59, and Leu60 on aB ( Figure 3A and 3B). This pocket is situated adjacent to a1 on CheY 6 and in ideal position to accommodate Met13 on the N-terminal region of this helix. This Met finger is protruding from a1, and its side chain fits snugly into the hydrophobic pocket, with 99.7% of its accessible surface area being buried ( Figure 3B). The hydrophobic interaction is complemented by Ala12, Leu16, and Tyr17 of a1 on CheY 6 , all facing the same side on CheA 3 P1 as Met13 ( Figure 3A). Ala12 of CheY 6 extends the hydrophobic interface by interaction with Leu14 of CheA 3 P1, whereas Leu16 of CheY 6 interacts with Ile11 and the aliphatic part of Glu10 on aA of CheA 3 P1. Tyr17 extends the binding surface towards the b5-a5 loop by interacting with Val59 at the C-terminus of aB. The elongated loop between b5 and a5 of CheY 6 forms van der Waals interactions with Ser109, Gly110, and Thr111 of CheY 6 contacting Ile59, Gly61, and Ser63 on aB and the following loop region on CheA 3 P1 ( Figure 3C). A main chain hydrogen bond is formed between Gly110 and Val59.
Of these interactions, the Met13 finger has the largest contribution towards the binding interface on CheY 6 , accounting for almost a third (156 Å 2 ) of the total buried surface area. Together with Ala12, Leu16, and Tyr17, it is accounting for over 60% (321 Å 2 ) of the interface. On CheA 3 P1, the residues involved in the interaction with CheY 6 are slightly less clustered and spatially farther apart. Yet the three residues with the biggest contribution towards the total buried surface area, Val59 (103 Å 2 , 20% of total buried surface area), Leu14 (64 Å 2 , 12%), and Ile11 (48 Å 2 , 9%), together account for almost half of the overall binding surface.
This analysis shows that only a small number of residues are necessary to form the majority of interactions within the complex. On CheY 6 , these are all located on a1. Within this helix, Met13 has the most crucial role in the hydrophobic interaction. A sequence alignment of all identified RRs in R. sphaeroides shows that only CheY 6 and CheB 2 have a Met residue at position 13, whereas all others either have a Ser, Thr, or Ala at this position ( Figure 4). As CheA 3 -P serves as a phosphodonor for both CheY 6 and CheB 2 this suggests that Met13 has an important role in determining the specificity for binding to CheA 3 P1.
Substitution of M13 in CheY 6 Reduces Interaction and the Rate of Phosphotransfer from CheA 3 P1-P Residues 11-17 of CheY 6 collectively account for ,67% of the buried surface area of CheY 6 in the CheA 3 P1-CheY 6 complex, with M13, L16, and Y17 each contributing ,30%, ,16%, and ,8%, respectively. To confirm that this surface of CheY 6 is involved in the interactions with CheA 3 P1-P that lead to phosphotransfer in solution, we substituted residues at this surface to mimic those found in the noncognate RRs, CheY 3 and CheY 4 . M13 was changed to S, and Y17 was changed to M as found in both CheY 3 and CheY 4 . CheY 3 and CheY 4 both have C at position 16; however, it has previously been shown that CheY 3 (C16S) and CheY 4 (C16S) behave indistinguishably from CheY 3 and CheY 4 in phosphotransfer assays (S. L. Porter, unpublished data), and therefore, to avoid any potential problems from disulphide bond formation during the purification, we changed L16 to S rather than C. Two mutant proteins were produced, CheY 6 (M13S) and CheY 6 (M13S,L16S,Y17M).
Surface plasmon resonance (SPR) assays showed that wild-type CheY 6 binds to CheA 3 P1 with an affinity of 218 mM ( Figure 5). This weak interaction is in agreement with the transient nature of the complex and the small buried surface area between the two proteins as observed in the crystal structure. Both the singlemutant protein, CheY 6 (M13S), and the triple-mutant protein CheY 6 (M13S,L16S,Y17M) showed a remarkable decrease in affinity, with a K d of more than 1 mM ( Figure 5 and Figure S3).
Consistent with the SPR binding assays, the rate of phosphotransfer from CheA 3 P1-P to each of the two mutant CheY 6 proteins was much slower than to wild-type CheY 6 ( Figure   6A-6C), with phosphotransfer to the triple-mutant protein CheY 6 (M13S,L16S,Y17M) being slowest. These results underline the conclusions drawn from the structural characterization of the complex and stress the importance of the Met finger at position 13 for binding of CheY 6 to CheA 3 P1. In addition, the hydrophobic interaction mediated by Leu16 and Tyr17 also adds significantly to recognition of CheY 6 by CheA 3 P1.

Reengineering CheY 4 into a Cognate RR of CheA 3 P1-P
The sequence of CheY 4 is 36% identical to CheY 6 ; however, unlike CheY 6 , CheY 4 is not a cognate RR of CheA 3 P1-P ( Figure 6D). Consistent with this, SPR assays failed to detect a significant interaction between CheA 3 P1 and CheY 4 ( Figure 5).   In an attempt to alter the phosphotransfer specificity of CheY 4 to allow phosphotransfer from CheA 3 P1-P, we substituted the CheY 4 residues corresponding to the positions shown to interact in the CheA 3 P1-CheY 6 structure, so that they matched CheY 6 . Two mutant proteins were produced: CheY 4 (P12A,S13M) and CheY 4 (P12A,S13M,C16L,M17Y). Ala12 was included in both mutant proteins since CheY 4 has a Pro at position 12 that might influence the orientation of a1 and thus interfere with the proper positioning of Met13 for interaction with the hydrophobic pocket. SPR assays showed that both mutant proteins bind CheA 3 P1 ( Figure 5 and Figure  S3), and in phosphotransfer assays, both mutant proteins were phosphorylated by CheA 3 P1-P ( Figure 6E and 6F). Phosphotransfer was fastest from CheA 3 P1-P to CheY 4 (P12A,S13M), with most of the initial CheA 3 P1-P dephosphorylated within 10 s ( Figure 6E).
Interestingly, although the rate of phosphotransfer from CheA 3 P1-P to CheY 4 (P12A,S13M) was faster than to CheY 4 (-P12A,S13M,C16L,M17Y), the SPR results show that CheA 3 P1 interacts slightly more strongly with CheY 4 (P12A,S13M, C16L,M17Y) than with CheY 4 (P12A,S13M) ( Figure 5 and Figure  S3). This apparent discrepancy could be explained by the alignment of the phosphorylated histidine and the phosphorylatable aspartate in the CheY 4 (P12A,S13M,C16L,M17Y).CheA 3 P1-P complex, which might be slightly less optimal for catalysis than in the CheY 4 (P12A,S13M).CheA 3 P1-P complex. Nevertheless, both methods show that mutating CheY 4 so that it resembles CheY 6 at the contact sites with CheA 3 P1 enhances both binding affinity and phosphotransfer rate. These results demonstrate that substitution of just two residues is sufficient to change the phosphotransfer specificity of CheY 4 .

Reengineering Other CheYs
Having successfully reengineered the phosphotransfer specificity of CheY 4 by introducing A12 and M13, we used the same approach to change the specificity of CheY 1 , CheY 3 , CheY 5 , and E. coli CheY (Figure 7). These proteins share between 30% and 33% sequence identity with CheY 6 . In all cases, the mutant proteins containing the alanine and methionine substitutions were phosphorylated more rapidly by CheA 3 P1-P than were their wildtype counterparts. Although CheY 1 and E. coli CheY both lack a methionine residue at the position corresponding to M13 of CheY 6 , they are both phosphorylated by CheA 3 P1-P ( Figure 7A and 7C); however, phosphotransfer to their corresponding alanine and methionine substitution mutant proteins proceeded even faster, with almost complete dephosphorylation of CheA 3 P1-P within 10 s ( Figure 7B and 7D). Similar to CheY 4 , wild-type CheY 3 and CheY 5 were not phosphorylated by CheA 3 P1-P ( Figure 7E and 7G), and the effect of the alanine and methionine substitutions was to allow rapid phosphotransfer from CheA 3 P1-P ( Figure 7F and 7H). These results demonstrate that the RR residues at the positions equivalent to A12 and M13 of CheY 6 play a major role in determining phosphotransfer specificity.

Discussion
Two-component signalling systems depend on a high level of selectivity between histidine kinases and their cognate RRs to prevent cross-talk between different pathways. Here, we present structural and functional data that elucidate how this specificity is mediated on a molecular level. We identified the key residues for molecular recognition and thereby were able to reengineer phosphotransfer specificity. Considering the high level of structural homology within the receiver and Hpt domain families, the presented structure together with complementary information, e.g., analysis of amino acid covariation [28,36,37], is likely to facilitate the rational reengineering of CheA-like histidine kinase-RR pairs for use, for example, in synthetic signalling circuits.

Conserved Features of the CheA 3 P1.CheY 6 Complex
The structure of CheA 3 P1 in complex with CheY 6 is, to our knowledge, the first structure of a Hpt domain of a CheA protein in complex with a RR. To date, there are only two other structures of a Hpt domain in complex with a RR, namely Spo0B/Spo0F [38,39] from Bacillus subtilis and YPD1/SLN1 [40,41] from Saccharomyces cerevisiae. Although the overall structure of the RR is well conserved in all three complexes, there are remarkable differences in the Hpt domains. Spo0B exists as a dimer in which the two a-helices of the helical hairpin domain of each protomer associate to form a four-helix bundle. In addition, Spo0B has a Cterminal domain with an a/b fold, which is involved in binding of the RR. YPD1 is a monomeric Hpt protein consisting of a fourhelix bundle with an additional short helix at the N-terminus. This N-terminal helix is involved in RR binding and complements the interface formed by helices aB and aC. When the three complexes were superimposed with respect to the receiver domains, CheA 3 P1 aligned well with YPD1 but only poorly with Spo0B ( Figure S4).

Phosphotransfer Specificity Determinants
CheA 3 P1.CheY 6 shows the smallest binding interface amongst the three complexes solved so far, with 530 Å 2 compared to 953 Å 2 in YPD1/SLN1 and 1,200 Å 2 in Spo0F/Spo0B. The interface is smaller since CheA 3 P1 has neither the C-terminal a/b fold domain seen in Spo0B nor the additional N-terminal helix seen in YPD1. Despite this small interface, CheA 3 P1 shows binding and efficient phosphotransfer to CheY 6 . It was previously proposed that the P2 domain of CheA-like proteins might be necessary for binding of the RR to CheA by increasing the binding interface [40]. However, CheA 3 does not have a P2 domain, and the experimental data presented here and elsewhere [26] show that the P1 domain alone is sufficient for binding of and specific phosphotransfer to the cognate RR. Moreover, the structural and mutational analysis suggests that their interaction is mediated by very distinct residue clusters on the RR and the Hpt domain, the former being located on the N-terminal region of a1 and the loop region between b5 and a5, and the latter being located on the Nand C-terminal region of aA and aB, respectively, and the loop region connecting aB and aC. This is in good agreement with a recently published computational analysis of amino acid coevolution of cognate histidine kinase-RR pairs [28] and the binding site found in the YPD1/SLN1 [40,41] or Spo0B/Spo0F [38,39] complex.

Reengineering Phosphotransfer Specificity
The specificity of protein-protein interactions is essential for most cellular processes. Despite the vast number of these interactions, our understanding of the molecular basis of their specificity is limited. The structure presented here has allowed us to identify specificity determinants for the CheA-CheY interaction. We successfully used this information to redesign noncognate RRs to allow them to be rapidly phosphorylated by CheA 3 P1-P (Figures 6 and 7). Whereas the Laub group have reengineered a HPK to specifically phosphorylate non-cognate RR substrates [28], we have now shown that it is possible to rationally reengineer a RR so that it can be phosphorylated by a noncognate HPK. The changing of phosphotransfer specificity described here represents, to our knowledge, the first example of the redesign of the intracellular part of the chemotaxis pathway and provides valuable insight into how cells mediate specificity in one of the most abundant signal transduction pathways, two-component signalling. The ability to reengineer phosphotransfer specificity coupled with recent work from the Bourret group [42], which has shown that RR autodephosphorylation rate can be manipulated, provide a platform for the future design of synthetic two-component circuits with customizable kinetics.
Sequence verified plasmids were transformed into M15pREP4 cells (Qiagen) and cultivated in Terrific Broth to an absorbance at 600 nm (A 600 ) of 0.8. Cultures were cooled to 20uC, induced with 0.25 mM IPTG, and then grown for ,15 h before harvesting. The bacterial pellets were resuspended in 25 mM sodium phosphate (pH 8.0), 500 mM NaCl, 0.5 mM b-mercaptoethanol, and EDTA-free protease inhibitor cocktail (Roche). Cells were lysed using a Basic Z model cell disruptor (Constant Systems) and fractionated by centrifugation (45,000g, 4uC, 60 min). CheA 3 P1, CheY 6 , and CheY 4 were purified from the supernatant by immobilized metal-affinity chromatography [19][20][21]. The samples were dialysed against standard buffer (30 mM Hepes [pH 7.5], 150 mM NaCl, 5 mM sodium acetate, 2 mM manganese chloride, 2 mM Tris(2-carboxyethyl)phosphine [TCEP]) and further purified by size-exclusion chromatography. Protein purity and concentration was measured as described [43]. Purified proteins were stored at 280uC. Seleno-methionine (SeMet)labelled CheY 6 was produced essentially as described [44]. The complex between CheY 6 and CheA 3 P1 was formed by mixing equimolar amounts of both proteins. The sample was incubated for 30 min at room temperature (20uC), and the complex was purified by size-exclusion chromatography. CheA 3 P1 was phosphorylated using ATP and CheA 4 , and purified as described previously [26]. The final preparation of CheA 3 P1-P was free of ATP and CheA 4 . Phosphorylation was verified by mass spectrometry. Equimolar amounts of phosphorylated CheA 3 P1 and CheY 6 (D56N,S83A) were mixed and set up for crystallization.

Site-Directed Mutagenesis
Mutations were introduced using the Quikchange Site-Directed Mutagenesis Kit (Stratagene).

Crystallization and Data Collection
Crystal trials were set up with the purified CheA 3 P1.CheY 6 complex at a concentration of 20 mg/ml. We set up nanoliter crystallization trials using a Cartesian Technologies robot (100 nl of protein solution plus 100 nl of reservoir solution) in 96-well Greiner plates [45], placed them in a TAP (The Automation Partnership) Homebase storage vault maintained at 295 K, and imaged them via a Veeco visualization system. Crystals of the phosphorylated, unphosphorylated and SeMet-labelled complex between CheY 6 and CheA 3 P1 could be obtained in 100 mM Bicine (pH 9), 1 M LiCl, 20% (w/v) PEG6000. Crystals were optimized using a dilution series of the initial condition. Crystals reached their final size of 40061006100 mm 3 after 12 h.
Diffraction data were collected at 100 K, crystals being flashcooled in a cryo N 2 gas stream. Before flash-freezing, crystals were cryo-protected with perfluoropolyether oil PFO-X125/03 (Lancaster Synthesis). The native dataset was collected at beamline I03 (l = 0.9757 Å ) and the SeMet dataset at beamline I04 (l = 0.979 Å ) at the Diamond Light Source. Data for the phosphorylated complex were collected on a MAR345 detector (Marresearch) mounted on a Miromax007 generator (Rigaku), equipped with Varimax HR Osmic mirrors (Rigaku). X-ray data were processed and scaled with the HKL suite [46]. Data collection statistics are shown in Table 1.

Structure Determination and Refinement
The CheA 3 P1-CheY 6 complex structure was determined by SAD analysis. The positions of 12 selenium atoms were determined by using SHELXD [47]. This solution was put into AUTOSHARP [48] for phase calculation, improvement, and phase extension using the high-resolution native data to 1.4 Å resolution. The resulting map was of high quality and allowed tracing of the whole polypeptide chain. An initial model was built automatically using Arp/wARP [49] and manually adjusted using COOT [50]. The structure was refined using autoBUSTER [51], REFMAC [52], and Phenix [53]. The phosphorylated complex was solved by molecular replacement using Phaser [54] with the unphosphorylated complex as model.

Surface Plasmon Resonance Binding Studies
SPR experiments were performed using a Biacore T100 machine (GE Healthcare) at 25uC in standard buffer supplemented with 0.05% (v/v) Tween 20. Protein concentrations were determined from the absorbance at 280 nm using calculated molar extinction coefficients. Proteins for surface attachment were enzymatically biotinylated within an engineered C-terminal tag. These proteins were then attached to surfaces on which 3,000 response units (RU) of streptavidin were coupled via primary amines [59] yielding a density of 200-1,500 RU of biotinylated protein. All experiments were done in duplicates with independently purified proteins. The signal from experimental flow cells was corrected by subtraction of a blank and reference signal from a mock or irrelevant protein-coupled flow cell. In all experiments analyzed, the experimental trace returned to baseline after each injection and the data fitted to a simple 1:1 Langmuir model of binding. K d values were obtained by nonlinear curve fitting of the Langmuir binding isotherm (bound = Cmax/(K d + C), where C is analyte concentration and max is the maximum analyte binding) evaluated using the Biacore Evaluation software (GE Healthcare).

Detection of Phosphotransfer from CheA 3 P1-P to the RRs
Assays were performed at 20uC in TGMNKD buffer (50 mM Tris HCl, 10% [v/v] glycerol, 5 mM MgCl 2 , 150 mM NaCl, 50 mM KCl, 1 mM DTT [pH 8.0]). CheA 3 P1 was phosphorylated using [c-32 P] ATP and CheA 4 , and purified as described previously [26]. The final preparation of CheA 3 P1-32 P was free of ATP and CheA 4 . CheA 3 P1-32 P (1 mM) was mixed with 5 mM RR in a final reaction volume of 100 ml. Following the addition of RR, reaction aliquots of 10 ml were taken at the indicated time points and quenched immediately in 20 ml of 1. 56  Quenched samples were analyzed using SDS-PAGE and phosphorimaging as described previously [21]. Figure S1 Stereoview of a superposition of the metal binding site of CheY 6 and E. coli CheY. CheY 6 is shown in pale green, E. coli CheY with Mg 2+ bound (PDB-code: 1CHN) in yellow, and E.coli CheY without Mg 2+ (3CHY) in teal. Structures were aligned on their secondary structure elements using secondary structure matching (SSM) implemented in COOT [50]. Residues involved in the coordination of Mg 2+ in E. coli CheY are shown in stick representation for all three structures. Residues in E. coli structures are only labelled for the Mg 2+ bound form. Mg 2+ from the E. coli structure is shown as a grey sphere. CheY 6 resembles the Mg 2+ bound form of E. coli CheY despite not having a divalent cation bound to its metal binding site. Found at: doi:10.1371/journal.pbio.1000306.s001 (6.24 MB TIF) Figure S2 Electron density of phosphorylated His51 in CheA 3 P1 in the active complex structure. The orientation is similar to Figure 2B. The density represents a 2Fobs-Fcalc map contoured at 1.5 s and calculated after initial rigid body refinement in Phaser [54] and one round of positional refinement in autoBUSTER [51], both omitting the phosphate group from the model.  [50]. CheY 6 is shown in pale green, CheA3P1 in light blue, Spo0F in orange, Spo0B in teal, SLN1 in purple, and YPD1 in aquamarine. Only the four-helix bundle of the Spo0B dimer is shown; the Cterminal domain is omitted for clarity. The orientation is similar to Figure 1. CheA 3 P1 is structurally more similar to the monomeric YPD1 than to Spo0B. Found at: doi:10.1371/journal.pbio.1000306.s004 (2.64 MB TIF)