BopE is a type III secreted protein from Burkholderia pseudomallei, the aetiological agent of melioidosis, a severe emerging infection. BopE is a GEF (guanine-nucleotide-exchange factor) for the Rho GTPases Cdc42 (cell division cycle 42) and Rac1. We have determined the structure of BopE catalytic domain (amino acids 78–261) by NMR spectroscopy and it shows that BopE78–261 comprises two three-helix bundles (α1α4α5 and α2α3α6). This fold is similar to that adopted by the BopE homologues SopE and SopE2, which are GEFs from Salmonella. Whereas the two three-helix bundles of SopE78–240 and SopE269–240 form the arms of a ‘Λ’ shape, BopE78–261 adopts a more closed conformation with substantial interactions between the two three-helix bundles. We propose that arginine and proline residues are important in the conformational differences between BopE and SopE/E2. Analysis of the molecular interface in the SopE78–240–Cdc42 complex crystal structure indicates that, in a BopE–Cdc42 interaction, the closed conformation of BopE78–261 would engender steric clashes with the Cdc42 switch regions. This implies that BopE78–261 must undergo a closed-to-open conformational change in order to catalyse guanine nucleotide exchange. In an NMR titration to investigate the BopE78–261–Cdc42 interaction, the appearance of additional peaks per NH for residues in hinge regions of BopE78–261 indicates that BopE78–261 does undergo a closed-to-open conformational change in the presence of Cdc42. The conformational change hypothesis is further supported by substantial improvement of BopE78–261 catalytic efficiency through mutations that favour an open conformation. Requirement for closed-to-open conformational change explains the 10–40-fold lower kcat of BopE compared with SopE and SopE2.
- bacterial pathogen
- guanine-nucleotide-exchange factor
- protein–protein interaction
- protein structure
- Rho GTPase
- type III secretion
Burkholderia pseudomallei is a Gram-negative bacterium that is the aetiological agent of melioidosis, a severe emerging infection of humans and animals that is endemic in South-East Asia and tropical Australia and that has the potential to spread worldwide [1–3]. Melioidosis has a range of clinical manifestations, including rapidly fatal septicaemia, pneumonia, skin and soft tissue abscesses, and osteomyelitis or septic arthritis. Infection is usually via contaminated soil, dust or water [4–6]. Asymptomatic infection is common in areas where the infection is endemic and progression to disease depends on the condition of the host . Between the fatal and asymptomatic extremes, the infection may be chronic or may run a relapsing course. Latency and relapse are common even in patients treated with appropriate antibiotics . B. pseudomallei is closely related to Burkholderia mallei, the pathogen that causes glanders, a disease of horses and other solipeds. B. mallei can also affect humans and is often fatal if left untreated . Due to the severity of the infection, aerosol infectivity and worldwide availability, both B. pseudomallei and B. mallei are considered to be potential bio-weapons . There is currently no vaccine against B. pseudomallei .
The molecular mechanisms of B. pseudomallei pathogenesis are not completely understood . B. pseudomallei has a 7.3 Mb genome, unusually large for a prokaryote, comprising two chromosomes with 16 genomic islands possibly acquired through very recent lateral transfer . The B. pseudomallei genome contains at least three loci encoding putative TTS systems (type III secretion systems) . One of these, Bsa, is homologous with the inv/spa/prg TTS system of Salmonella serotype Typhimurium [13–15]. TTS systems resemble molecular syringes for the injection of multiple bacterial effector proteins into the host cell cytoplasm that modify host cell physiology to the benefit of the pathogen [16,17]. TTS systems are central to the virulence of many Gram-negative pathogens, including Salmonella, Shigella, Yersinia, enteropathogenic Escherichia coli and the four major genera of plant pathogenic bacteria [18,19].
BopE, encoded within the Bsa locus, is secreted via the Bsa TTS system and influences invasion of HeLa cells probably via its function as a GEF (guanine-nucleotide-exchange factor) for Rho GTPases that regulate the actin network . BopE shares sequence homology with the Salmonella translocated effector proteins SopE [21,22] and SopE2 [23,24] (Supplementary Figure S1 at http://www.BiochemJ.org/bj/411/bj4110485add.htm), which play an important role in Salmonella invasion of non-phagocytic intestinal epithelial cells. SopE is a potent GEF for the mammalian Rho GTPases Cdc42 (cell division cycle 42) and Rac1 in vitro and in vivo, whereas SopE2 efficiently activates Cdc42 but not Rac1 . The structures of SopE  and SopE2  are entirely different from those of the best characterized eukaryotic GEFs, which comprise a catalytic DH (Dbl homology) domain and an adjacent PH (pleckstrin homology) domain [28–30], although there are similarities in the catalytic mechanisms .
We have previously shown that BopE is monomeric in aqueous solution, adopts a single conformation that is predominantly α-helical, is stable over a wide range of pH values and is able to refold independently . Now, as part of our examination of the structural and mechanistic relationships between BopE and its counterparts SopE and SopE2 from Salmonella, we report here the three-dimensional structure in solution of the catalytic domain of BopE (BopE residues 78–261, where 261 is the C-terminal residue of the full-length protein) and NMR and kinetic analyses of the interaction of BopE78–261 with the Rho GTPase Cdc42.
Biophysical and biological characterization, NMR sample generation and NMR spectroscopy of recombinant BopE78–261
The methods used to obtain BopE78–261 NMR samples and to derive backbone and side-chain resonance assignment, plus biophysical characteristics of BopE78–261, have been described previously [32,33]. The 1H, 13C and 15N chemical shifts of BopE78–261 are in the BioMagResBank database (http://www.bmrb.wisc.edu) under accession number BMRB-5974. The biological activity of exactly the same BopE78–261 construct as used here has been demonstrated previously: BopE78–261 was shown to have guanine nucleotide-exchange activity towards Cdc42 and Rac1 in vitro .
All NMR data were acquired at 25 °C on a Varian Unity INOVA spectrometer operating at a nominal proton frequency of 600 MHz, using a triple resonance 5 mm probe equipped with z-axis pulsed field gradients. NMR data were processed using the NMRPipe/NMRDraw software suite  and analysed using the SPARKY assignment program (http://www.cgl.ucsf.edu/home/sparky/). NOE (nuclear Overhauser effect) distance restraints were obtained by analysis of 1H-1H two-dimensional NOESY  (100 and 175 ms mixing times), 15N-NOESY HSQC (heteronuclear single-quantum coherence)  (50, 100 and 150 ms mixing times) and simultaneous three-dimensional 15N/13C-edited NOESY  (100 ms mixing time) spectra. Backbone 1DNH RDC (residual dipolar coupling) restraints were measured for BopE78–261 aligned with respect to the magnetic field by using a stretched polyacrylamide gel; gels were made using an apparatus based on that described previously . RDCs were measured using IPAP (in-phase anti-phase)–HSQC .
Each NOE was assigned to one of four restraint distances based on the peak intensity: 1.8–2.8, 1.8–3.3, 1.8–5.0 and 1.8–6.0 Å (1 Å=0.1 nm), corresponding to strong, medium, weak and very weak NOEs. Distances involving methyl groups, aromatic ring protons and non-stereospecifically assigned methylene protons were represented as a (Σr−6)−1/6 sum . For strong and medium NOE restraints involving amide protons, 0.2 Å was added. Backbone dihedral angles ϕ and ψ were predicted from 13Cα, 13Cβ, 13C', 1Hα and backbone 15N chemical shifts using TALOS . The ϕ dihedral angles were restrained to TALOS-predicted values ±30o for α-helices and ±40o for β-strands and ψ dihedral angles were restrained to TALOS-predicted values ±50o. Hydrogen bond restraints were obtained from hydrogen–deuterium exchange experiments: uniformly 15N-labelled BopE78–261 in NMR buffer was freeze-dried and resuspended in 99.96% 2H2O. A series of 1H-15N HSQC spectra was then recorded to determine amide protons protected from exchange with the solvent. For hydrogen bond distance constraints, the NH–O distance was assigned lower and upper distance bounds of 1.5 and 2.5 Å, and the N–O distance was assigned lower and upper distance bounds of 2.5 and 3.5 Å.
Structures were calculated using the Python interface of Xplor-NIH 2.16.0 [43,44], using simulated annealing starting from random extended structures. Default values were used for all force constants and molecular parameters. The ensemble of NMR structures was analysed for violated restraints using the VMD-Xplor visualization package . The structure determination was carried out iteratively whereby consistently violated restraints were reassigned, wherever possible, using existing structures or removed until a consistent set of constraints was obtained with few violations in the ensemble. The ensemble of structures was further refined with Xplor-NIH standard refinement protocols by using the final set of restraints. The quality of the structures was assessed by using PROCHECK-NMR .
NMR titration of Cdc42Δ7 against BopE78–261
Binding of unlabelled human Cdc42Δ7 to 15N-labelled BopE78–261 was monitored by recording 1H-15N HSQC spectra as a function of the BopE78–261/Cdc42Δ7 ratio. Cdc42Δ7 is Cdc42 lacking seven C-terminal amino acids; it was shown previously that C-terminal truncation of Cdc42 does not interfere with SopE GEF activity . Cdc42Δ7 was purified from E. coli BL21(DE3) as previously described . The NMR titration was performed as previously described [27,48]. Briefly, two initial NMR samples were prepared in 0.5 ml of NMR buffer (20 mM sodium phosphate, pH 5.5, and 50 mM NaCl) with 10% 2H2O. Sample A contained 0.5 mM 15N-labelled BopE78–261 (1.0:0.0 molar ratio of BopE78–261/Cdc42Δ7) and sample B contained 0.5 mM 15N-labelled BopE78–261 and 1.34 mM Cdc42Δ7 (1.0:2.7 molar ratio of BopE78–261/Cdc42Δ7). The buffer composition of both samples was identical as both samples were extensively exchanged into the same batch of sample buffer. Throughout the titration, the concentration of BopE78–261 was maintained at a constant concentration of 0.5 mM and the Cdc42Δ7 concentration was varied to give a series of BopE78–261/Cdc42Δ7 molar ratios from 1.0:0.0 to 1.0:2.7. A 1H-15N HSQC spectrum was acquired at each titration point with 512 complex 1H points and 192 complex 15N points with 32 scans per increment and spectral widths of 8000 Hz in 1H and 2000 Hz in 15N. The initial NMR samples represented the end points of the titration. Intermediate values of BopE78–261/Cdc42Δ7 were obtained by simultaneously taking equal aliquots from both sample A and sample B and then transferring the aliquots to the other NMR tube (i.e. from tube A to tube B and vice versa). This procedure was repeated until a series of 12 1H-15N HSQC experiments at BopE78–261/Cdc42Δ7 molar ratios between 1.0:0.0 and 1.0:2.7 was completed.
Generation and characterization of BopE mutants
BopE78–261 double mutants N224P/R230Q (mutant 1), N216P/L226P (mutant 2) and R207E/N216P (mutant 3) were made using the following pairs of primers (shown as 5′–3′; ‘for’ is forward; ‘rev’ is reverse): TCGCCCACGCTCGTCGAGTTCCAGCAGACGGT (N224PR230Q for) and CTGCTGGAACTCGACGAGCGTGGGCGAACGCTC (N224PR230Q rev); CGCCCGCGTTGCCGGCCGAGCGTTCGAACACGCCCGTCGAGT (N216PL226P for) and ACGGGCGTGTTCGAACGCTCGGCCGGCAACGCGGGCGCGACGA (N216PL226P rev); TGCGGAGCAGCAGGCGATCGATCTCGTCGCGCCCGCGTTGCC (R207EN216P for) and CGCGGGCGCGACGAGATCGATCGCCTGCTGCTCCGCATAC (R207EN216P rev). The mutants were constructed by overlapping PCR. The two overlapping primers (for and rev) were used in PCR with upstream and downstream primers to amplify the two parts of the gene (upstream-rev and for-downstream respectively). The resulting DNA fragments were purified, mixed and used as a template for a third PCR with upstream and downstream primers to amplify the mutated gene. The resulting DNA fragment in each case was digested with EcoRI and BamHI and cloned into pGEX4T1 (GE Healthcare). The cloned DNA was then sequenced. The mutant proteins were expressed and purified in the same way as wild-type BopE78–261 .
Filter binding assays
Cdc42Δ7 was loaded at 25 °C for 10 min with [3H]GDP in a reaction buffer containing 30 mM Hepes, 100 mM KCl, 0.1 mM EDTA (pH 7.5), 1 μg of creatine phosphokinase (Sigma) and 0.5 mM DTT (dithiothreitol). MgCl2 was added to a final concentration of 2.8 mM and the mixture was incubated for another 2 min. Exchange reactions were started by adding the respective GEF and unlabelled GDP to the reaction mixture containing Cdc42Δ7 and [3H]GDP. BSA (Sigma) was used as a negative control and SopE269–240 was used as a positive control. Aliquots were withdrawn and the reaction was stopped by quenching in ice-cold wash buffer, containing 30 mM Hepes, 100 mM KCl, 0.1 mM EDTA and 5 mM MgCl2 (pH 7.5), followed by analysis with the nitrocellulose filter binding assay . Filters were washed twice with wash buffer, containing 30 mM Hepes, 100 mM KCl, 0.1 mM EDTA and 5 mM MgCl2 (pH 7.5) and dried, and the radioactivity bound to the filters was analysed by scintillation counting in a Tri-Carb liquid-scintillation counter 1600 TR (Packard, Meriden, CT, U.S.A.).
RESULTS AND DISCUSSION
Structure determination of BopE78–261
A semi-automated procedure for iterative NOE assignment was used to generate the structure of BopE78–261. The final structures were generated using 2452 NOE-derived distance restraints (comprising 784 intraresidue, 1151 sequential and medium-range and 517 long-range NOEs, where ‘long range’ means they are five or more amino acids apart in the sequence), 192 hydrogen bond restraints, 255 ϕ and ψ dihedral angle restraints (132 ϕ and 123 ψ) and 98 backbone 1DNH RDC restraints (Table 1). The ensemble of 20 final simulated annealing structures, selected from 40 calculations on the basis of the lowest energy, and the average structure are shown in Figure 1. Over the regular secondary-structure elements, the ensemble of structures has a backbone RMSD (root mean square deviation) from the mean of 0.65 Å and an RMSD of 1.13 Å for all non-hydrogen atoms. A Ramachandran plot of the structures with PROCHECK-NMR  indicates that 96.7% of the residues (excluding glycine and proline residues) lie in the most favoured or additionally allowed regions. The few non-glycine residues to fall into the generously allowed regions and disallowed regions correspond to residues located at the termini or loop regions where the NMR restraint density is low.
Three-dimensional structure of BopE78–261 and comparison with Salmonella SopE78–240 and SopE269–240
BopE has been identified  as a homologue of the Salmonella effector proteins SopE and SopE2 (Supplementary Figure S1). Overall, BopE has approx. 16 and 17% sequence identity with SopE and SopE2. Within the catalytic domain (comparing residues 78–240 of SopE and SopE2 with residues 78–240 of BopE), the sequence identity/similarity with SopE and SopE2 is approx. 25%/40% and 24%/39% respectively.
BopE78–261 consists of six major α-helices termed α1 to α6 arranged in two three-helix bundles, α1α4α5 and α2α3α6. The three-helix bundles are connected by a loop between α1- and α2-helices, a β-hairpin (residues 162–168), followed by a loop that contains the putative (by comparison with SopE, which has a G166AGA169 catalytic motif) G171AGT174 catalytic motif between α3- and α4-helices, and a loop between α5- and α6-helices (Figure 1).
The BopE78–261 fold is similar to that of its Salmonella counterparts SopE78–240 and SopE269–240, but is more closed and compact with substantial interaction between the two three-helix bundles (Figures 2 and 3). As an illustration of the more extensive association between the bundles in BopE78–261, the buried surface areas between the three-helix bundles are 1693 Å2 in SopE78–240, 1849 Å2 in SopE269–240 and 2148 Å2 in BopE78–261. Also, we have assigned 56 interbundle NOEs in BopE78–261 compared with 20 such NOEs in our previous structure determination of SopE269–240 . The greater conservation of bundle structure relative to bundle–bundle orientation is quantitatively illustrated by RMSD values for superimposed Cα traces and by comparison of interhelical angles. When the catalytic domains are superimposed, the RMSD values are 2.5 Å (SopE versus SopE2), 3.9 Å (SopE2 versus BopE) and 5.0 Å (SopE versus BopE). [Note that the buried surface area and RMSD values plus visual inspection (Figure 2) show that SopE269–240 is somewhat intermediate as it has a slightly more closed conformation than SopE78–240; it must be emphasized, however, that the only available SopE78–240 structure is from the complex with Cdc42, so it is possible that unbound SopE78–240 also has a more closed SopE269–240-like conformation.] When individual three-helix bundles are superimposed, the corresponding values are 2.3, 2.9 and 2.3 Å for the α1α4α5 bundle and 1.6, 2.8 and 2.8 Å for the α2α3α6 bundle. Calculation of the interhelical angles shows that the angles between helices in different bundles tend to differ considerably between BopE78–261 and the two Salmonella GEFs (Table 2).
The interactions between the two three-helix bundles of BopE78–261 constitute an intricate network of charge and hydrophobic interactions. Among the residues involved are five arginine residues at sequence positions 100, 182, 200, 207 and 230 that are almost unique to BopE: SopE and SopE2 do not possess arginine residues in any of the corresponding positions (Supplementary Figure S1), but the putative bacterial GEF family member CopE from Chromobacterium violaceum (accession AAQ57975) has arginine residues corresponding to BopE Arg200 and Arg207. Three of the BopE arginine residues, Arg200, Arg207 and Arg230, form part of the association between α5- and α6-helices, while Arg182 and Glu125 are suitably located to link α4- and α2-helices at the putative Cdc42-binding face (based on the SopE78–240–Cdc42 complex structure ) of BopE through a potential salt bridge. Arg100 (in α1-helix) occupies a hydrophobic pocket between α2- and α5-helices.
BopE residue Pro204 (corresponding to Ala199 in SopE and SopE2) promotes these interbundle interactions by disrupting α5-helix into two parts termed α5′ and α5″. As a consequence, α5′ is positioned to bridge the α1α4α5 and α2α3α6 bundles and its residues are able to interact with residues in α2 and α6 of the α2α3α6 bundle (Figure 3).
In contrast with BopE Pro204, three SopE/E2 proline residues appear to impede interbundle interaction and therefore contribute to the more open conformation adopted by SopE269–240 in solution relative to BopE78–261. Near the apex of the Λ formed by the two three-helix bundles, the loop connecting α5 and α6 in SopE269–240 and SopE78–240 bulges (Figure 3), presumably due to the presence of Pro211, Pro219 and Pro221. Due to the lack of proline residues at positions corresponding to 219 (Asn224 in BopE) and 221 (Leu226 in BopE), BopE78–261 α6-helix begins earlier in the amino acid sequence than SopE/E2 α6 and the BopE78–261 α5–α6 connecting element is a three-residue turn rather than the seven-residue loop observed in SopE269–240 and SopE78–240 (Figure 3 and Supplementary Figure S1). We reason that this protrusion of the polypeptide chain in the α5–α6 loop at the apex of the Λ, not observed in BopE78–261 due to the key amino acid differences described here, counteracts extensive interbundle interaction in SopE78–240 and SopE269–240.
NMR investigation of the interaction between BopE78–261 and Cdc42
In order to probe BopE78–261 binding to Cdc42 in solution, 12 two-dimensional 1H-15N HQSC experiments on mixtures of varying ratios of uniformly 15N-labelled BopE78–261 and unlabelled human Cdc42Δ7 were performed. Two main types of behaviour were observed for peaks in BopE78–261 HSQC spectra upon increasing the ratio of Cdc42Δ7 to BopE78–261: general broadening of peaks characterized by intensity loss throughout the spectrum; and for more than one-third of residues, the appearance of one or more additional peaks per backbone amide NH, indicating that BopE samples have more than one conformation upon interaction with Cdc42 with slow exchange between the conformations.
BopE78–261 cross-peak broadening with increasing Cdc42Δ7 concentration
Almost all of the backbone NH peaks in 1H-15N HQSC spectra of BopE78–261 broadened as a function of increasing Cdc42Δ7 concentration (Figure 4) until, at the highest Cdc42Δ7/BopE78–261 ratio of 2.7:1, there was a subset of 15 peaks that remained relatively intense (14 of which can be assigned as Thr78, Gly79, Asp80, Glu109, Phe110, Gly160, Glu251, Lys252, Ala254, Thr255, Asn256, Ala257, Gly260 and Ala261 and hence comprise amino acids in presumably relatively flexible parts of the protein near the N- and C-termini plus the α1–α2 and pre-β-hairpin loops) plus a subset of readily detectable peaks [some of which can be assigned as Ala81, Lys82, Gln83, Ala84 (all near the N-terminus), Asp162, Gly165, Val166 (β-hairpin), Gly190 (α4–α5 loop), Glu221 (α6) and Ser248 (unstructured C-terminal region)] and about 40 further peaks that were still detectable just above the noise level. The remaining backbone NH peaks (in excess of 100) were broadened into the noise. Most asparagine and glutamine side-chain NH2 cross-peaks were still present at the highest Cdc42Δ7/BopE78–261 ratio of 2.7:1.0.
The rate of backbone NH peak broadening was reasonably uniform across the sequence, suggesting that the major contributors to broadening are the following: molecular mass increase upon complexation (a 1:1 BopE78–261–Cdc42Δ7 complex is just over double the molecular mass of BopE78–261), shape change upon complexation with potential for nonlinear increase in effective rotational correlation time, and exchange between free and bound BopE78–261. Due to peak overlap, the degree and rate of broadening could not be quantified for a quarter of the approx. 175 backbone NH peaks. At a Cdc42Δ7/BopE78–261 ratio of 1.0:1.0, many peaks were broadened to below 20% of their original height with the greatest concentrations of less rapidly broadened peaks found at the terminal regions, particularly the C-terminal region (Figure 4). The highest concentration of particularly rapidly broadened peaks (to noise level at a Cdc42Δ7/BopE78–261 ratio of 1.0:1.0) occurred in α2-helix; the equivalent SopE helix is involved in the interface between SopE78–240 and Cdc42 in the SopE78–240-Cdc42 crystal structure .
Appearance of multiple cross-peaks per BopE78–261 backbone NH
The second major observation upon increasing the Cdc42Δ7/BopE78–261 ratio was the appearance of a peak or peaks in addition to the original backbone NH peak for approx. 70 of the 175 backbone NH peaks; single extra peaks accounted for approx. 75% of these 70 cases. In 56 instances, these additional peaks could be assigned to a particular amino acid by proximity to the corresponding original backbone NH peak. At least two of the 16 asparagine and glutamine side-chain NH2 groups also displayed a second pair of peaks in the presence of Cdc42Δ7. In the vast majority of cases with one or more extra peaks, upon increasing the Cdc42Δ7/BopE78–261 ratio the Cdc42Δ7-induced extra peaks increased in height or sometimes reached a plateau as the original backbone NH peaks decreased in height. The chemical shift difference between the original backbone NH peak and Cdc42Δ7-induced additional peak(s) at a Cdc42Δ7/BopE78–261 ratio of 1.0:1.0 was calculated according to the formula Δδave=[(ΔδHN2+(ΔδN2/25))/2]1/2, where ΔδHN and ΔδN correspond to the chemical shift difference in the amide 1H and 15N chemical shifts between the original NH peak and the Cdc42Δ7-induced extra peak(s); the Δδave values are shown in Figure 5(A). In the cases where more than one Cdc42Δ7-induced extra peak could be assigned to a specific amino acid, the value plotted is the average of the Δδave values. For 67 residues, only one backbone NH peak was observed throughout the titration; the approximate sequence positions of these residues are highlighted in Figure 5(A). For the remaining 40 or so backbone NH peaks, overlap hindered the observation of peak behaviour during the titration.
The presence of the Cdc42Δ7-induced additional peaks for residues in several parts of BopE78–261 indicates that BopE78–261 samples have more than one conformation in the presence of Cdc42Δ7 with the Cdc42Δ7-induced conformations in slow exchange with the initial Cdc42Δ7-free conformation. The fact that in approx. 75% of cases with more than one NH peak the additional peak was a single peak indicates that one Cdc42Δ7-induced conformation was predominant. Clusters of residues exhibiting multiple backbone NH peaks are located in the α1–α2 loop and adjacent parts of α1 and α2, the β-hairpin and loops adjacent to the β-hairpin including the putative 171GAGT174 catalytic motif, and around the α5–α6 loop (Figure 5). There is also a sequence of such residues in α6.
Comparison of BopE78–261-Cdc42 and SopE269–240-Cdc42 titration results
Very similar NMR titrations, both using Cdc42Δ7 and the same protocol, have now been carried out to study the BopE78–261–Cdc42 (the present study) and SopE269–240–Cdc42  interactions. BopE78–261 and SopE269–240 both experienced widespread backbone NH peak broadening upon increasing the ratio of Cdc42Δ7 to BopE78–261/SopE269–240. The broadening was, if anything, more rapid in the SopE269–240–Cdc42Δ7 titration. The SopE269–240 NH peaks that underwent Cdc42Δ7-induced chemical shift changes fall into two groups, one of which showed very good agreement with the SopE78–240 residues involved in important intermolecular interactions in the SopE78–240–Cdc42 crystal structure : this group included SopE269–240 residues Gln109 (α2), Asp124 (α2), Gly165 (adjacent to catalytic motif), Gly166, Gly168, Ala169 (all catalytic motif), Val174 (α4), Gln194 (α5) and Lys198 (α5). The second group of perturbed SopE269–240 residues comprised several scattered internal residues and isolated residues on the opposite side of the molecule to the binding interface. In contrast with SopE269–240, slow exchange between unbound and Cdc42Δ7-bound conformations of BopE78–261 was observed during the BopE78–261–Cdc42Δ7 titration. The chemical shift differences between these states of BopE78–261 were, in general, 4–5 or more times the magnitude of the Cdc42Δ7-induced chemical shift changes observed in the SopE269–240–Cdc42Δ7 titration. The BopE equivalents (BopE residues Asp128, Gly171, Gly173, Thr174 and Thr179) of five of the Cdc42-perturbed SopE269–240 residues (SopE269–240 residues Asp124, Gly166, Gly168, Ala169 and Val174) listed above were involved in the Cdc42-induced slow conformational exchange, whereas Ser170, Tyr199 and Gln203, the BopE equivalents of SopE residues Gly165, Gln194 and Lys198, were not. The behaviour of Gln113 (BopE equivalent of SopE Gln109) during the titration could not be monitored due to peak overlap. Of the BopE equivalents of a further two SopE78–240 residues that interact with Cdc42 in the SopE78–240–Cdc42 crystal structure but that were not significantly perturbed in the SopE269–240–Cdc42Δ7 NMR titration , Ala135 (α2–α3 loop) was involved in the Cdc42-induced slow conformational exchange, but the behaviour of Asp107 could not be monitored due to peak overlap. The significance of the positions of slowly exchanging residues in BopE78–261 is discussed in the next section.
Implications of BopE78–261 tertiary structure and BopE78–261–Cdc42 NMR titration for BopE interaction with Rho GTPases
The question arises as to whether the conformational difference between the catalytic domain of BopE and those of SopE and SopE2 has implications for interaction with Rho GTPases. Analysis of the interface between SopE78–240 and Cdc42 in the SopE78–240–Cdc42 complex crystal structure  reveals that the interaction can be broken down into two major components: a groove on SopE78–240 accommodates a ridge on Cdc42 formed by residues 35–41 (switch region I) and the gap between the two three-helix bundles of SopE78–240 accommodates Cdc42 residues Val36 and Asp38 (Supplementary Figure S2 at http://www.BiochemJ.org/bj/411/bj4110485add.htm). The latter interaction, in particular, indicates that, in its closed conformation, BopE78–261 would experience steric clashes with Cdc42. The resulting implication is that BopE catalytic domain must undergo a change from its closed conformation to a more open conformation like those of SopE and SopE2 catalytic domains in order to carry out its guanine nucleotide-exchange function. A requirement for such a large-scale conformational change is consistent with, and may at least partially explain, the observed differences in catalytic-centre activity for guanine nucleotide exchange between BopE78–261 and its Salmonella counterparts: a kcat of 0.48 s−1 was measured for BopE78–261-induced guanine nucleotide exchange in Rac1 (a similar rate was measured for Cdc42) , whereas the kcat values for guanine nucleotide exchange in Cdc42 are 5±1 and 19±3 s−1 for SopE78–240 and SopE269–240 respectively .
It might then be asked whether BopE catalytic domain exists in equilibrium in solution between closed and open forms or whether it undergoes a conformational change upon interaction with the target protein. These two possibilities are not necessarily mutually exclusive – there may be equilibrium in solution for unbound BopE catalytic domain but one that lies strongly towards the closed conformation. The results of the BopE78–261–Cdc42Δ7 titration are consistent with a significant conformational change in BopE78–261 upon binding to Cdc42: when superimposed on the structure of BopE78–261 (Figure 5B), it is apparent that many of the amino acids that sampled one or more Cdc42Δ7-induced conformations during the BopE78–261–Cdc42Δ7 titration are located in potential hinge areas for a closed-to-open conformational change involving relative reorientation of the two three-helix bundles of BopE78–261. These hinge areas include the α1–α2 loop and adjacent residues in α1 and α2, residues in the region between α3 and α4 that includes the β-hairpin and G171AGT174 putative catalytic motif, and residues in and around the α5–α6 turn. Residues in the central part of α2 also show slow exchange between initial and Cdc42Δ7-induced conformations, consistent with a change in conformation and/or position of the α3–α4 loop C-terminal to the β-hairpin that associates with this part of α2 in Cdc42-free BopE78–261 (Figure 1). It is also striking that a few of the amino acids with multiple NH peaks are located in areas that would be involved in any intrabundle conformational changes, suggesting that the three-helix bundles themselves remain largely unchanged. The considerably greater magnitude of the Cdc42Δ7-induced chemical shift differences between free and Cdc42Δ7-bound states of BopE78–261 compared with the magnitude of the chemical shift changes observed in the SopE269–240–Cdc42Δ7 titration underpins the conclusion that BopE78–261 undergoes greater structural change than SopE269–240 upon binding of the Rho GTPase.
Guanine nucleotide-exchange activity of BopE78–261 and BopE78–261 mutants
In order to investigate further the requirement for a conformational change in BopE for catalysis of nucleotide exchange in Rho GTPases, three BopE78–261 double mutants were made. These were N224P/R230Q (mutant 1), N216P/L226P (mutant 2) and R207E/N216P (mutant 3). The mutations were selected according to their potential for changing BopE78–261 from its relatively closed conformation to a more open conformation closer to those observed for SopE78–240 in its complex with Cdc42  and unbound SopE269–240 , as follows: N224P to induce a SopE/E2-like bulge in the α5–α6 loop; R230Q to further disrupt the α5–α6 interaction; N216P and L226P to induce a SopE/E2-like bulge in the α5–α6 loop; R207E to disrupt the α5–α6 interaction and N216P to induce a SopE/E2-like bulge in the α5–α6 loop.
Like the wild-type recombinant BopE78–261, the mutants were cloned and expressed as GST (glutathione transferase) fusions. Mutant 1 was expressed relatively poorly in E. coli, but could be purified; mutant 2 was expressed at low levels, but disappeared during purification (perhaps this mutant is misfolded and therefore rapidly degraded); and mutant 3 was expressed well and could be purified. In filter binding assays  with BSA as the negative control, the order of nucleotide exchange catalytic efficiency was: BopE78–261 N224P/R230Q (mutant 1)>SopE269–240>wild-type BopE78–261>>BopE78–261 R207E/N216P (mutant 3); in fact, mutant 3 showed essentially no catalytic activity (Figure 6). The reason for the lack of nucleotide-exchange activity in mutant 3 is unclear, but the R207E/N216P double mutation obviously induces changes that disrupt rather than enhance BopE function. The N224P/R230Q double mutation in BopE78–261, on the other hand, produces a much more effective GEF than wild-type BopE78–261 and a better GEF than even SopE269–240 (Figure 6), itself a better GEF for Cdc42 than SopE78–240 . This result, showing that mutations designed to abrogate important interbundle interactions and thereby induce a more open conformation in BopE78–261 can substantially improve nucleotide-exchange catalytic efficiency, adds further strong experimental support to that from NMR titration for the hypothesis that BopE GEF domain undergoes Rho GTPase-induced change from a closed to an open conformation.
The molecular mechanisms of B. pseudomallei pathogenesis are not well understood. A number of putative type III secreted effector proteins have been identified by analysis of the B. pseudomallei genome sequence . One of these proteins, BopE, is a homologue of the potent GEFs SopE [21,50] and SopE2 [23,24] from Salmonella enterica (Supplementary Figure S1). SopE and SopE2 catalyse nucleotide exchange in mammalian Rho GTPases, contributing to disruption of the host cell membrane and invasion of the host cell [17,21,23,25,50,51]. BopE, likewise, acts as a GEF for the Rho GTPases Cdc42 and Rac1 in vitro and may play a role in the invasion of non-phagocytic epithelial cells . The present study shows that BopE and SopE/SopE2 catalytic domains adopt similar three-dimensional folds comprising two three-helix bundles but also shows that BopE has a more compact conformation, involving significant interbundle interactions, than its Salmonella homologues. The most open conformation of the three is for Cdc42-bound SopE78–240, with unbound SopE269–240 slightly more closed. It is worth noting, however, that SopE residues involved in contacting Cdc42 in the SopE78–240–Cdc42 complex crystal structure  are largely conserved or conservatively substituted in BopE (Supplementary Figure S1). SopE residues (Asp103, Gln109, Asp124 and Gly168) shown by mutation to be functionally important  are, moreover, conserved in BopE. It seems likely, therefore, that despite its more closed conformation, BopE ultimately utilizes the same mechanism as SopE and other Rho GEFs  in catalysing guanine nucleotide exchange in Rho GTPases. This would require that BopE change from closed to open conformations in the presence of Rho GTPase target proteins. Such a conformational change is evidenced here by the results of a BopE78–261–Cdc42 NMR titration and measurements of nucleotide-exchange catalytic efficiency comparing wild-type and mutant BopE GEF domain. Phosphorylation of BopE would not seem to be required for any conformational change as we and others have shown that BopE78–261 purified from E. coli exhibits GEF activity . Finally, given the sequence and conformational differences between BopE and SopE/E2 catalytic domains, it is possible that there are as yet unknown differences in specificity among the members of this family of bacterial GEFs, with the potential for modulation of the activities of small G-proteins in addition to Cdc42 and Rac1.
This work was supported at the University of Bath by The Wellcome Trust (grant no. 060998) and at IAH by the BBSRC (Biotechnology and Biological Sciences Research Council). The Wellcome Trust is acknowledged for purchase of the 600 MHz NMR spectrometer (grant no. 051902) used in this study. C. W. was supported by a Ph.D. studentship from the EPSRC (Engineering and Physical Sciences Research Council; U.K.). We thank Dr Julian Eaton (Department of Biology and Biochemistry, University of Bath, Bath, U.K.) for constructive criticism of this paper and help with RMSD calculations, Professor Lewis Kay (Department of Biochemistry, University of Toronto, Toronto, ON, Canada) for some of the pulse sequences used here, Dr Charles Schwieters (Center for Information Technology, National Institutes of Health Bethesda, MD, U.S.A.) for help with the Python interface of Xplor-NIH, Dr Kyoko Yap (Department of Structural and Chemical Biology, Mount Sinai School of Medicine, New York, NY, U.S.A.) for the program Interhlx, Graham Pavitt's group (Faculty of Life Sciences, University of Manchester, Manchester, U.K.) for instructions on filter binding assays and Mareike Posner (Department of Biology and Biochemistry, University of Bath, Bath, U.K.) for advice on enzyme assays.
Abbreviations: Cdc42, cell division cycle 42; for, forward; GEF, guanine-nucleotide-exchange factor; HSQC, heteronuclear single-quantum coherence; NOE, nuclear Overhauser effect; rev, reverse; RDC, residual dipolar coupling; RMSD, root mean square deviation; TTS system, type III secretion system
- © The Authors Journal compilation © 2008 Biochemical Society