Biochemical Analysis of SopE from Salmonella typhimurium, a Highly Efficient Guanosine Nucleotide Exchange Factor for RhoGTPases*

RhoGTPases are key regulators of eukaryotic cell physiology. The bacterial enteropathogen Salmonella typhimurium modulates host cell physiology by translocating specific toxins into the cytoplasm of host cells that induce responses such as apoptotic cell death in macrophages, the production of proinflammatory cytokines, the rearrangement of the host cell actin cytoskeleton (membrane ruffling), and bacterial entry into host cells. One of the translocated toxins is SopE, which has been shown to bind to RhoGTPases of the host cell and to activate RhoGTPase signaling. SopE is sufficient to induce profuse membrane ruffling in Cos cells and to facilitate efficient bacterial internalization. We show here that SopE belongs to a novel class of bacterial toxins that modulate RhoGTPase function by transient interaction. Surface plasmon resonance measurements revealed that the kinetics of formation and dissociation of the SopE·CDC42 complex are in the same order of magnitude as those described for complex formation of GTPases of the Ras superfamily with their cognate guanine nucleotide exchange factors (GEFs). In the presence of excess GDP, dissociation of the SopE·CDC42 complex was accelerated more than 1000-fold. SopE-mediated guanine nucleotide exchange was very efficient (e.g.exchange rates almost 105-fold above the level of the uncatalyzed reaction; substrate affinity), and the kinetic constants were similar to those described for guanine nucleotide exchange mediated by CDC25 or RCC1. Far-UV CD spectroscopy revealed that SopE has a high content of α-helical structure, a feature also found in Dbl homology domains, Sec7-like domains, and the Ras-GEF domain of Sos. Despite the lack of any obvious sequence similarity, our data suggest that SopE may closely mimic eukaryotic GEFs.

GTPases of the Rho subfamily act as molecular switches and are key regulators of cellular functions such as formation of stress fibers and focal adhesions, motility, cell shape change, membrane ruffling, cytokinesis, cell aggregation, and cell-to-cell adhesion, as well as smooth muscle contraction (1). They cycle between a biologically inactive GDP-bound and an active GTP-bound conformation (2). As spontaneous guanine nucleotide dissociation and GTP hydrolysis are fairly slow, activation and inactivation are dependent on the activity of guanine nucleotide exchange factors (GEFs) 1 and GTPase-activating proteins (3). GEFs enhance nucleotide exchange rates by binding to and stabilizing the nucleotide-free GTPase, whereas GTPase-activating proteins accelerate the intrinsic rate of GTP hydrolysis through stabilization of the transition state via a critical arginine residue (4). Mammalian GEFs are divided into several subfamilies that share sequence or structural similarity. GEFs of one family are generally specific for a certain subfamily of GTPases (5). So far, all GEFs known to act on RhoGTPases have a common sequence motif, the Dbl homology domain. This is the catalytically active part of the molecule responsible for binding to RhoGTPases and catalysis of guanine nucleotide exchange (6). Only in the active, GTP-bound state can the Rho-GTPases interact with their "effectors," which are the downstream elements of the signal transduction cascades that mediate the cellular responses.
A large number of bacterial virulence factors exert their toxic effects by interfering with RhoGTPase signaling of host cells (7). Some inhibit downstream signaling of RhoGTPases, whereas others mediate their permanent activation by inhibition of the GTP hydrolyzing activity. So far, all of these toxins have been found to modulate RhoGTPase function by introducing covalent chemical modifications. These modifications include ADP-ribosylation, glucosylation, the transfer of N-acetylglucosamine groups, and the deamidation of glutamine residues (7). Modulation of RhoGTPase function by bacterial toxins via a noncovalent interaction has not yet been demonstrated.
The bacterial enteropathogen Salmonella typhimurium is one of the leading causes of diarrhea in developed countries. S. typhimurium employs a large array of mechanisms to colonize, replicate, and survive within hosts (8). The specialized type III protein secretion system encoded in the Salmonella pathogenicity island 1 is important during the gut associated stages of the infection (9,10). It facilitates the secretion and translocation of several bacterial toxins (effector proteins) into the cytosol of host cells (11). The translocated effector proteins induce responses that range from the induction of apoptosis in macrophages to increased chloride secretion, the production of IL-8, and the induction of membrane ruffling, which facilitates bacterial entry into nonphagocytic host cells (12). The activation of RhoGTPases is thought to be a key step in the signal transduction cascades triggered by S. typhimurium (12,13).
SopE is one of the effector proteins translocated into host cells by the type III protein secretion system of S. typhimurium. In the host cell cytosol, it activates the RhoGTPases Cdc42, and Rac1 and induces nuclear responses and membrane ruffling (14,15). Thereby, SopE can promote efficient entry of the bacterium into host cells (16,17). Purified SopE enhances guanine nucleotide exchange in several RhoGTPases, such as Cdc42 and Rac1, but not in the more distantly related GTPase Ras, as shown in in vitro nucleotide exchange assays (14). However, the exact mechanism by which SopE enhances the rate of guanine nucleotide exchange in RhoGTPases has not yet been investigated, and it has remained unclear whether SopE may activate RhoGTPase signaling by covalently modifying these central regulators of host cell function.
In the present study, we have analyzed the catalytic activity of SopE from S. typhimurium in more detail. Analysis of the kinetic properties of Rac1 that had been preincubated with SopE revealed that SopE does not covalently modify RhoGTPases. This was confirmed by mass spectrometry analysis of Rac1 that had been affinity purified from a complex with SopE. Analysis of the kinetics of formation and dissociation of a Cdc42⅐SopE complex and the kinetic analysis of the SopEmediated nucleotide exchange revealed striking parallels to eukaryotic GEFs. In addition, far-UV CD analysis suggested that the catalytic domain of SopE has a high content of ␣-helical structure, a feature also found in GEFs for ARF, Ras and RhoGTPases (18 -24). Despite the lack of any detectable sequence similarity, our biochemical and biophysical data demonstrate that SopE modulates RhoGTPase function by a noncovalent mechanism and suggest that this bacterial virulence factor may closely mimic GEFs of the host.

Preparation of Recombinant Proteins
Preparation of recombinant proteins was performed essentially as described (14,25). Briefly, all proteins used in this study were overexpressed from derivatives of the isopropyl-1-thio-␤-D-galactopyranosideinducible expression vector pGEX (Amersham Pharmacia Biotech) as GST fusion proteins in Escherichia coli, recovered from bacterial extracts by binding to glutathione-Sepharose 4B (Amersham Pharmacia Biotech) and either eluted with 10 mM glutathione (GST⅐SopE 78 -240 ) or cleaved off the column by digestion with thrombin protease (Cdc42 proteins, SopE 78 -240 ) or with factor Xa (Rac1). All steps of the purification were performed in buffers containing 20 mM Tris-HCl, pH 7.5, 5 mM MgCl 2 , Ն100 mM NaCl, 2 mM dithiothreitol. The quality of the preparations was assayed by SDS-polyacrylamide gel electrophoresis and staining with Coomassie Blue (Fig. 1). Protein concentrations were estimated from Coomassie-stained gels and by the method of Bradford (26). Activity of the RhoGTPases was determined by [ 3 H]GDP binding assays (27). Proteins were snap frozen and stored at Ϫ80°C. Due to the design of the expression vectors, the proteins carry the following additional amino acids: GST⅐SopE 78 -240 carries the 15 additional amino acids PGISGGGGGILEFEM between the thrombin cleavage site and Leu-78 of SopE. SopE 78 -240 carries 17 additional N-terminal amino acids: GSPGISGGGGGILEFEM. Rac1 carries 7 additional N-terminal amino acids (GIDPGAT), and mass spectrometry analysis of proteolytic fragments of Rac1 revealed that a fraction of the protein was C-terminally truncated after Lys-186, Arg-187, and Lys-188. Wild type Cdc42 carries 14 additional N-terminal amino acids (GSRRASVGSKIISA). Cdc42 ⌬C (residues 1-178) carries two additional N-terminal residues (GS), and the 13 C-terminal amino acids (PPEPKKSRRCVLL) are absent. The Cdc42 G12V carries three additional N-terminal residues (GSP). Preparation of the mant-GDP (mGDP)⅐Cdc42 G12V complex was performed as recently described (25).

Filter Binding Assays
Rac1 and Cdc42 proteins were loaded at 30°C for 15 min with [ 3 H]GDP in Buffer A (50 mM Tris-HCl, pH 7.6, 50 mM NaCl, 5 mM MgCl 2 , 5 mM dithiothreitol) supplemented with 10 mM EDTA. Loading was stopped by addition of 10 mM MgCl 2 and 3 volumes of cold Buffer A. Exchange reactions were started by adding [ 3 H]GDP-loaded GTPase to Buffer A containing the indicated concentrations of GDP and GST, SopE 78 -240 , or EDTA. Alternatively, exchange reactions were started by adding GDP to Buffer A containing the [ 3 H]GDP-loaded GTPase and either GST, SopE 78 -240 , or EDTA. Aliquots were withdrawn and analyzed in a nitrocellulose filter binding assay as described (27). Filters were washed three times with 50 mM Tris-HCl, pH 7.6, 50 mM NaCl, 5 mM MgCl 2 , dried, and radioactivity bound to the filters was analyzed by scintillation counting in a Wallac 1450 Micro Beta Trilux apparatus.

Protein Sequence Analysis
Rac1 purified by GST⅐SopE 78 -240 affinity chromatography (see "Results") was digested with proteinase Glu-C (Roche Molecular Biochemicals) in 25 mM NH 4 HCO 3 at a protease:substrate ratio of 1:40 at 25°C for 16 h. The resulting peptides were analyzed by mass spectrometry using a Bruker Reflex mass spectrometer (matrix-assisted laser desorption mass spectrometry with ␣-cyano-4-hydroxy-cinnamic acid as matrix). Spectra were recorded either directly from the digest or from HPLC-separated peptides; HPLC separation was performed using a 2.1 ϫ 250 mm 5 m/300 Å Vydac C 18 column (The Separation Group) on a Waters 600 high pressure liquid chromatograph (Millipore) using 0.1% (v/v) trifluoroacetic acid in water as aqueous phase and 0.085% trifluoroacetic acid in acetonitrile as organic phase. After an initial lag phase of 5 min at 2% organic phase, a discontinuous gradient of 2-30% organic phase over 70 min (0.4%/min), followed by 30 -90% organic phase over 30 min (2%/min), was developed. The flow rate was 0.2 ml/min, and UV absorption was monitored at 215 nm.
Average masses were compared with the masses of proteolytic fragments predicted using the PeptideMass peptide characterization software (28).
For some peptides, the assignment was confirmed by analysis of fragment ions generated by postsource decay (29), using the FAST TM method (Bruker).
The N-terminal sequence (40 residues) was obtained by direct Edman degradation of the affinity-purified protein using a model 473A protein sequencer (Applied Biosystems).

Spectroscopic Techniques
Fluorescence-All fluorescence measurements were performed at 25°C in Buffer Y (40 mM HEPES/NaOH, pH 7.4, 100 mM NaCl, 5 mM MgCl 2 ) on a Fluoromax TM spectrofluorometer (Spex Industries). Emission spectra were recorded after excitation at 366 nm (1 nm bandwidth) with a time constant of 1 s and 2 nm bandwidth. The time courses of mGDP release from Cdc42 G12V in the presence of 1 mM nonfluorescent GDP were monitored at 435 nm (8 nm bandwidth).
For measurement of the single turnover kinetics of SopE 78 -240 -catalyzed nucleotide release, increasing concentrations of SopE 78 -240 (final concentration, 0.5-32 M) were mixed with mGDP⅐Cdc42 G12V (final concentration, 0.1 M) at 25°C in 40 mM HEPES/NaOH, pH 7.4, 100 mM NaCl, 5 mM MgCl 2 in a stopped flow apparatus (Applied Photophysics) with the monochromators set to 1.5 nm and a cut-off filter of 408 nm. Single exponentials were fitted to all kinetic data using the program Grafit (Erithacus Software).
Surface Plasmon Resonance-Association and dissociation reactions involving Cdc42 ⌬C and GST⅐SopE 78 -240 were studied by surface plasmon resonance using a BIAcore TM system (BIAcore AB, Uppsala, Sweden). Sensor chips were covalently coated with ␣-GST antibodies according to the protocol of the manufacturer (30). GST proteins were applied at a concentration of 0.5 M in binding Buffer X (10 mM HEPES/ NaOH, pH 7.4, 150 mM NaCl, 5 mM MgCl 2 , 0.005% Igepal CA-630 (Sigma)), and experiments were performed at 20°C in the same buffer. The flow rate was 5 l/min in all experiments. At the end of the assay, all noncovalently bound material was removed from the matrix by washing with 10 mM glycine/HCl, pH 2.0, 0.05% SDS, leaving the immobilized immunoglobulin at full activity. To eliminate the contribution of nonspecific binding, equivalent control experiments using immobilized GST instead of GST⅐SopE 78 -240 were used to calculate specific signal changes. Netto binding and dissociation curves were obtained by subtracting the GST controls from the GST⅐SopE curves to take into account unspecific binding of Cdc42 to GST. Parameters were fitted using a single exponential with the BIAevaluation software (version 2.1).
Far UV CD Spectroscopy-The far-UV CD spectrum of 5 M SopE 78 -240 was recorded at room temperature on a Jasco J710 spectropolarimeter in 10 mM potassium phosphate, pH 7.0, with a sensitivity of 20 mdeg, a time constant of 1 s, a scan speed of 20 nm/min and a bandwidth of 1 nm in a 1-mm quartz cuvette. The spectrum was accumulated 10-fold, corrected for buffer contributions, and converted to mean residue ellipticities according to Schmid (31). Data were analyzed for secondary structure content using the manufacturer's Jasco software.

SopE Does Not Covalently Modify
RhoGTPases-Tissue culture cell experiments, as well as in vitro nucleotide exchange assays, have established that SopE is capable of stimulating guanine nucleotide exchange in a number of RhoGTPases, including Cdc42 and Rac1 (14). However, it has remained unclear whether SopE activates guanine nucleotide exchange by stabilizing the nucleotide free state of RhoGTPases. Alternatively, SopE might activate nucleotide exchange by introducing a so far unknown type of chemical modification that may increase the intrinsic rate of nucleotide exchange of the RhoGTPase. Most of the types of chemical modifications (namely ADP ribosylation, glucosylation, and the transfer of N-acetylglucosamine), which are introduced by other bacterial toxins (7), can be ruled out because SopE-mediated nucleotide exchange activity can be observed in vitro in the absence of cofactors such as UDP-glucose, ATP, or UDP-N-acetylglucosamine. In contrast, mechanisms involving the deamidation of glutamine or asparagine residues or other types of chemical modification of RhoGTPases that do not depend on the presence of special cofactors have not been ruled out. To address this hypothesis, we used a filter binding assay and analyzed the interaction of SopE with Rac1 in more detail. In these and all subsequent experiments, SopE 78 -240 was used, a 162-amino acid fragment of SopE that makes up the catalytic domain of the protein (14).
To detect putative SopE-mediated covalent modifications via their effect on the intrinsic nucleotide exchange rates of Rac1, we started out by analyzing the effect of SopE on the rates of [ 3 H]GDP release from [ 3 H]GDP-loaded Rac1 in more detail. As long as no additional unlabeled nucleotide was added to the assay, the presence of 1 M SopE 78 -240 had no effect on the binding of [ 3 H]GDP to Rac1 (Fig. 2a). However, when [ 3 H]GDPloaded Rac1 was incubated with 1 M SopE 78 -240 in the presence of a large excess of unlabeled GDP, release of [ 3 H]GDP from Rac1 was completed within less than 10 s. Therefore, if SopE were mediating nucleotide release by introducing a chemical modification, 10 s would be sufficient for complete modification of Rac1 and for nucleotide release from the "modified" Rac1. In other words, the presence of a putative modification should be detectable by virtue of a greatly increased rate of intrinsic nucleotide exchange of Rac1.
In the first approach to detect putative chemical modifications, we have measured the rate of . Under these conditions, nucleotide exchange was completed to about 90% within 14 min (Fig. 2b, ࡗ, control experiment). Therefore, under these condi-tions 14 min must be sufficient to allow SopE to interact with (and possibly modify) about 90% of the [ 3 H]GDP-loaded Rac1 in the assay. No [ 3 H]GDP release was observed during the first 14 min when GDP was omitted (Fig. 2b, E). If modification of Rac1 would have taken place during this time, intrinsic guanine nucleotide exchange rates of Rac1 should be strongly increased and nucleotide exchange should be completed to 90% in less than 10 s after addition of excess unlabeled GDP. However, when unlabeled GDP (final concentration, 1 mM) was added at min 14, nucleotide exchange was initiated and proceeded with the same slow kinetics (k obs ϭ 0.24 Ϯ 0.03 min Ϫ1 ; Fig. 2b, E) as observed in the control experiment (k obs ϭ 0.28 Ϯ 0.04 min Ϫ1 ; Fig. 2b, ࡗ). This indicates that SopE-mediated guanine nucleotide exchange is not due to covalent modification of Rac1.
In a second approach to detect putative chemical modifications we have analyzed the properties of Rac1 that had been affinity purified on a GST⅐SopE 78 -240 column. For purification of Rac1, we have taken advantage of the high affinity binding of SopE to Rac1 in the absence of guanine nucleotides. In the presence of excess guanine nucleotide binding is much weaker (14) (see below), a property also reported for small GTPase⅐GEF complexes (32,33). GST⅐SopE 78 -240 was immobilized on a glutathione-Sepharose 4B column, Rac1 was loaded onto this column in a buffer containing no free guanine nucleotide, and bound Rac1 was eluted with the same buffer supplemented with 1 mM GDP (Fig. 3a). Unbound GDP was removed by dialysis, and the intrinsic rate of [ 3 H]GDP release of the affinity purified Rac1 (Rac1 GDP-el. ) was analyzed in a filter binding assay (Fig. 3b). For this purpose Rac1 GDP-el. was loaded with [ 3 H]GDP and subsequently assayed in the presence (Fig. 3b, q) or absence (ࡗ) of 1 mM unlabeled GDP. In both cases, virtually no [ 3 H]GDP release was observed during the entire assay. The 20% reduction in the amount of radioactivity bound to Rac1 GDP-el. that was observed upon addition of 1 mM unlabeled GDP was also observed with the "original" Rac1 (Fig. 2a, q) and might be attributable to some [ 3 H]GDP binding nonspecifically to Rac1 GDP-el. or to proteins misfolded during the loading procedure that can be easily displaced by unlabeled GDP. In the presence of 1 mM unlabeled GDP and 1 M SopE 78 -240 , however, all [ 3 H]GDP was displaced in less than 10 s, demonstrating that the Rac1 GDP-el. preparation was still active and that it was capable of interacting with SopE for a second time (Fig. 3b).
We also analyzed Rac1 GDP-el. for the presence of possible chemical modifications that may have been introduced by SopE. Rac1 GDP-el. was digested with protease Glu-C, a highly specific protease that cleaves proteins at glutamate residues. Therefore, deamidation of any of the glutamine residues in Rac1 GDP-el. should yield an additional Glu-C cleavage site. The proteolytic fragments of the Glu-C digest were purified by HPLC and analyzed by mass spectrometry. Within the limits of experimental error (Ϯ 1 Da), each of the proteolytic fragments detected had the same mass as predicted for unmodified Rac1 (data not shown). No proteolytic fragments attributable to additional glutamate residues were detected.
In conclusion, these results establish that SopE belongs to a novel class of bacterial toxins interfering with RhoGTPase function of host cells. In contrast to all other known bacterial toxins addressing these central regulators of host cell function, SopE does not introduce covalent chemical modifications. Instead, SopE enhances nucleotide exchange rates of RhoGTPases by transient interaction.
Kinetic and Thermodynamic Analysis of SopE-mediated Guanine Nucleotide Exchange-SopE does not share any sequence similarity with the active dbl domain or any other parts of eukaryotic GEFs. Therefore it was of interest to study the SopE-mediated nucleotide exchange reaction in more detail.
G12V-and C-terminal Deletion Mutants of Cdc42 Are Efficient Substrates for SopE-Several functionally important residues are highly conserved among small GTP-binding proteins. Ras and Cdc42 mutants carrying a valine instead of the universally conserved glycine 12 are constitutively active in tissue culture experiments (34 -36). This is presumably due to strongly reduced rates of GTP hydrolysis in the absence or presence of GTPase-activating protein, as inferred from biochemical analyses of Ras and Cdc42 (4, 37, 38). 2 X-ray analyses of GDP-bound Ras and Cdc42 have shown that the structure of the G12V mutant is virtually identical to that of the wild type protein (39,40), and rates of GDP dissociation are only marginally affected (37).
The C terminus of all RhoGTPases carries the highly conserved CaaX motif. As shown for Ras (41), this domain is thought to provide a site for prenylation that is necessary for anchoring the protein to a particular membrane compartment. Presumably, the CaaX motif does not play a significant role in guanine nucleotide binding and GTPase function.
We have analyzed whether a Cdc42 G12V mutant or the GTPbinding domain (a truncated version of Cdc42 lacking the Cterminal 13 amino acids, Cdc42 ⌬C ) are efficient substrates for SopE 78 -240 . Both mutant proteins, as well as wild type Cdc42, were purified as GST fusion proteins from E. coli. The Cdc42 proteins were loaded with [ 3 H]GDP (see under "Materials and Methods"), and the rates of guanine nucleotide release in the presence of 1 mM GDP were determined in a filter binding assay using 1 M SopE 78 -240 , 10 mM EDTA, or 1 M GST (Fig.  4). Similar to the results obtained with Rac1 (Fig. 2a), [ 3 H]GDP release from each of the Cdc42 variants was completed within 10 s in the presence of 1 M SopE 78 -240 (t 1/2 Ͻ Ͻ 10 s; Fig. 4, q). EDTA-mediated [ 3 H]GDP release was somewhat slower (t 1/2 ϭ 15 Ϯ 5 s; Fig. 4, ‚), whereas little [ 3 H]GDP release was observed in the GST control experiments (Fig. 4, ࡗ). This demonstrates that both Cdc42 G12V and Cdc42 ⌬C are efficient substrates for SopE and implies that neither the highly conserved Gly12 nor the C-terminal amino acids are directly involved in the interaction with this bacterial virulence factor.
Association/Dissociation of the GST⅐SopE 78 -240 Complex with Cdc42 ⌬C -The kinetics of formation and dissociation of the SopE⅐Cdc42 complex were analyzed using surface plasmon resonance. This technique allows one to study binding kinetics in real time by measuring the change in mass on the surface of a sensor chip. First, GST⅐SopE 78 -240 (or GST as a control) was bound to a sensor chip coated with an ␣-GST antibody. Then, the kinetics of binding of nucleotide free Cdc42 ⌬C were measured. Cdc42 ⌬C bound to GST⅐SopE 78 -240 , whereas no specific binding was observed in the control experiments using GST alone. The observed rates for formation of the GST⅐SopE 78 -240 complex with Cdc42 ⌬C were dependent on the concentration of Cdc42 ⌬C applied (Fig. 5). Assuming a simple one-step bimolecular association reaction, a linear fit to the observed associa-2 M. G. Rudolph, unpublished data. tion constants yields the association rate constant for the formation of the GST⅐SopE 78 -240 ⅐Cdc42 ⌬C complex as k ass ϭ 6 Ϯ 1 ϫ 10 5 M Ϫ1 s Ϫ1 (Fig. 5c).
After formation of the complex, the dissociation of the GST⅐SopE 78 -240 ⅐Cdc42 ⌬C complex was followed (Fig. 5a). Because the dissociation was very slow (Fig. 6, a and b), it was impossible to determine the exact value of the dissociation rate constant in these experiments, and the apparent k off of 1.4 Ϯ 0.1ϫ10 Ϫ4 s Ϫ1 (Fig. 6b) should only be viewed as a rough estimate. Nevertheless, the data indicate that the equilibrium dissociation constant as defined by K D ϭ k off /k ass for the GST⅐SopE 78 -240 ⅐Cdc42 ⌬C complex is in the low nanomolar range. In contrast, in the presence of 1 mM GDP, the dissociation of the GST⅐SopE 78 -240 ⅐Cdc42 ⌬C complex was accelerated more than 1000-fold to k off(GDP) ϭ 0.6 Ϯ 0.05 s Ϫ1 (Fig. 6c). This effect is well known for eukaryotic GEFs (see, e.g. Ref. 42).
Equilibrium Fluorescence Titration-Cdc42 G12V complexed with mGDP, a fluorescent GDP derivative, has been employed to analyze SopE-mediated nucleotide exchange in more detail. mGDP is frequently used in kinetic and equilibrium studies because the fluorescence of the mant moiety can change dramatically upon binding to GTPases. It has been shown that the fluorescence intensity of mGDP bound to Cdc42 is about 4-fold stronger than that of unbound mGDP (25,43). Generally, the presence of the fluorophore has only little impact on the kinet- ics of spontaneous nucleotide exchange, GEF-mediated nucleotide exchange, or GTPase function (42)(43)(44)(45).
As a control, the fluorescence of SopE (0.63 M) in the range between 380 and 500 nm was probed after excitation at 366 nm and found to be insignificant (Fig. 7, curve f). The fluorescence spectrum of the Cdc42 G12V ⅐mGDP complex (0.1 M) has a maximum at 435 nm, and fluorescence is gradually decreased and shifted to longer wavelengths upon addition of increasing concentrations of SopE 78 -240 (0.63, 1.3, and 1.9 M). At 1.9 M SopE 78 -240 the fluorescence intensity at 435 nm was reduced more than 2-fold and approached saturation. A further decrease in fluorescence intensity down to the magnitude of free mGDP (data not shown) was observed upon addition of 1 mM nonfluorescent GDP (Fig. 7). This indicates the possible existence of a ternary GST⅐SopE 78 -240 ⅐Cdc42 G12V ⅐mGDP complex. In addition, changes in the mGDP fluorescence can be used as a sensitive probe to analyze the kinetics of SopE-mediated nucleotide exchange.
Multiple Turnover Kinetics of SopE 78 -240 -mediated Nucleotide Exchange-The Michaelis-Menten parameters k cat and K m of the SopE-mediated nucleotide exchange were analyzed using the decrease in mant fluorescence upon the SopEcatalyzed dissociation of the Cdc42 G12V ⅐mGDP complex. For this, the dissociation rate constants (Fig. 8a) in the presence of 1 mM GDP and 25 nM SopE 78 -240 were determined. The observed rates of mGDP release increased with increasing Cdc42 G12V ⅐mGDP concentrations and reached a plateau at about 10 M (Fig. 8b). Fitting a hyperbola to the concentration dependence of the observed rates yielded the kinetic constants for multiple turnover catalysis: k cat ϭ 0.95 Ϯ 0.06 s Ϫ1 and K m ϭ 4.5 Ϯ 0.9 M. This corresponds to a catalytic efficiency of k cat /K m ϭ 2.1 ϫ 10 5 M Ϫ1 s Ϫ1 , which is not unusual for catalytic reactions involving large conformational changes.
FIG. 6. Surface plasmon resonance measurement of the dissociation of GST⅐SopE 78 -240 ⅐Cdc42 ⌬C complex. a, Cdc42 ⌬C (1 M; sec Ϫ400 -0) was bound to GST⅐SopE 78 -240 as described in Fig. 5a. Dissociation of the complex was followed for 3200 s. Then, 20 M GDP was added to the wash buffer, and dissociation of the complex was followed for an additional 500 s. Only the data showing the association of the complex and the subsequent dissociation reactions are shown. b, The graph shows the best fit (black line) of a single exponential to the data (gray line) for the dissociation of the GST⅐SopE 78 -240 ⅐Cdc42 ⌬C complex in the absence of GDP shown in a, yielding k off ϭ 1.4 Ϯ 0.1 ϫ10 Ϫ4 s Ϫ1 . c, the graph showing the best fit (black line) of a single exponential to the data (gray line) for dissociation of the GST⅐SopE 78 -240 ⅐Cdc42 ⌬C complex upon addition of 1 mM GDP shown in a, yielding k off ϭ 0.6 Ϯ 0.05 s Ϫ1 . Identical results were obtained when the experiment was performed a second time.

Single Turnover Kinetics of SopE 78 -240 -mediated Nucleotide
Exchange-The rates of single turnover reactions are determined by the rate-limiting step during substrate binding and/or catalysis. In the presence of a large excess of enzyme, all subsequent steps of the catalytic cycle have no effect on the observed kinetics. A stopped flow apparatus was used to analyze the dissociation of the Cdc42 G12V ⅐mGDP complex under single turnover conditions catalyzed by SopE 78 -240 . Cdc42 G12V ⅐mGDP (final concentration, 100 nM) was rapidly mixed with GDP (final concentration, 1 mM) and a large excess of SopE 78 -240 (final concentrations, 0.5, 1, 2.9, 4, 5.9, 11.7 and 32 M), and the rates of mGDP release were determined via the decrease in mant fluorescence (Fig. 9a). The observed rate constants were plotted as function of the SopE 78 -240 concentration (Fig. 9b). As 32 M SopE 78 -240 was insufficient to reach saturation, it was impossible to determine the maximal acceleration rate for the single turnover guanine nucleotide release. However, it is clear that the single turnover rate constant k react is at least 2.5 s Ϫ1 , which is almost 10 5 -fold faster than the intrinsic rate of mGDP dissociation from Cdc42 G12V (k off intrinsic ϭ 3 ϫ 10 Ϫ5 s Ϫ1 ). 3 Analysis of the Secondary Structure of SopE 78 -240 by CD Spectroscopy-To gain insight into the structural composition of SopE, the far-UV CD spectrum of its catalytic domain (SopE 78 -240 ) (Fig. 10) was recorded. The spectrum shows minima at 222 and 208 nm and a strong maximum near 195 nm, which are typical for a predominantly ␣-helical protein. This observation is compounded by decomposition analysis of the spectrum (46), which suggests that SopE 78 -240 is highly organized and has a high ␣-helical content (57.4%), whereas unstructured regions (21.9%) and especially ␤-folded sheets (8%) or ␤-turns (12.6%) are far less prominent.

DISCUSSION
So far, all bacterial toxins known to interfere with host cell signaling by modulating the activity of RhoGTPases were found to be modifying enzymes. They modulate host cell signaling by deamidation of certain glutamine residues of RhoGTPases or by covalently attaching ADP-ribose, glucose, or N-acetylglucosamine groups to these key regulators (7). The data presented here demonstrate that SopE modulates RhoGTPase signaling by a different mechanism. In contrast to the other kown bacterial toxins specific for RhoGTPases, SopE is not a modifying enzyme. This has been shown by kinetic analysis of Rac1 preparations pretreated with SopE as well as by kinetic and protein sequence analysis of Rac1 that had been specifically eluted from the complex with SopE. Therefore, SopE belongs to a new category of bacterial virulence factors modulating RhoGTPase function via noncovalent interactions. This is reminiscent of GEFs, which regulate RhoGTPase activity in eucaryotic cells.
To explore the similarity between SopE and eukaryotic GEFs in more detail, we have analyzed the kinetic parameters of the SopE-mediated guanine nucleotide exchange. Neither the G12V amino acid exchange, nor deletion of the C-terminal 13 amino acids of Cdc42 had any effect on SopE-mediated nucleotide exchange. Formation and dissociation of the SopE⅐Cdc42 complex was analyzed by surface plasmon resonance using immobilized GST⅐SopE 78 -240 and Cdc42 ⌬C . In the absence of guanine nucleotide the association rate constant k ass was 6 Ϯ 1 ϫ 10 5 M Ϫ1 s Ϫ1 , and the dissociation of the GST⅐SopE 78 -240 ⅐Cdc42 ⌬C complex was very slow (apparent k off Ϸ 1.4 Ϯ 0.1 ϫ 10 Ϫ4 s Ϫ1 ). In the presence of free GDP, the rate of dissociation of the GST⅐SopE 78 -240 ⅐Cdc42 ⌬C complex increased more than 1000-fold (k off (GDP) ϭ 0.6 Ϯ 0.05 s Ϫ1 ). Binding of Cdc42 to GST⅐SopE in the presence of free GDP could not be analyzed by surface plasmon resonance, presumably because the formation of a trimeric complex of Cdc42, GDP, and SopE was not favored under the conditions tested (20°C, 1 M GDP-loaded Cdc42 ⌬C , 20 M GDP; data not shown). Interestingly, the kinetic constants determined for SopE 78 -240 were very similar to those described for the interaction of another small GTPase (Ras) with its cognate GEF Cdc25 (42) (k ass ϭ 3 ϫ 10 5 M Ϫ1 s Ϫ1 ; k off ϭ 1 ϫ 10 Ϫ3 s Ϫ1 as well as a much weakened binding in the presence of GDP: K D(nucleotide free) ϭ 3.3 nM; K D(GDP) ϭ 0.6 mM).
The kinetics of SopE-mediated nucleotide exchange were analyzed using the mant fluorescence of mGDP as spectroscopic readout. Under conditions of excess mGDP⅐Cdc42 G12V 3 M. G. Rudolph, unpublished data.  Fig. 8 was used. b, kinetic data as shown in a were analyzed by fitting single exponentials to them assuming pseudo-first-order conditions. The resulting rate constants (k obs ) were plotted against the SopE 78 -240 concentration used in the assay. and 1 mM free GDP, the Michaelis-Menten constant was determined to be K m ϭ 4.5 Ϯ 0.9 M. Interestingly, this corresponds closely to the dissociation constant K D(GDP) ϭ 10 M for binding of the exchange factor RCC1 to its cognate GTPase Ran (47), but it is 2 orders of magnitude lower than the K D(GDP) for CDC25 binding to Ras. It has been suggested that this difference in affinity may be related to the distinct cellular function of both systems: the high affinity of RCC1 for Ran may allow nucleotide exchange to proceed efficiently in solution, whereas in the case of CDC25/Ras, both proteins have to be tethered to a membrane for the exchange reaction to occur (42). Interestingly, at early times of infection, SopE has been found diffusely localized in the cytoplasm of the host cell within the region of the Salmonella-induced membrane ruffle. 4 Therefore, the comparatively high substrate affinity (K m ϭ 4.5 Ϯ 0.9 M) may ensure that all Cdc42 and Rac1, membrane bound as well as cytosolic fractions, in this region of the cell will become activated.
The rate constant for SopE-mediated mGDP release under conditions of excess Cdc42 G12V ⅐mGDP was determined to k cat ϭ 0.95 Ϯ 0.06 s Ϫ1 , which is almost 10 5 -fold faster than the uncatalyzed rate of intrinsic guanine nucleotide exchange. This translates into a catalytic efficiency of k cat /K m ϭ 2.1 ϫ10 5 M Ϫ1 s Ϫ1 , which is far slower than the rate predicted for a diffusion controlled reaction. However, this is a typical value for catalytic reactions involving large conformational changes.
Single turnover experiments using up to 32 M SopE 78 -240 confirmed that binding of the guanine nucleotide-Cdc42 G12V complex and catalysis of guanine nucleotide dissociation were very fast. At 32 M SopE 78 -240 , the rate of nucleotide exchange was k obs ϭ 2.5 s Ϫ1 . However, as saturation had not yet been reached, it remains unclear whether the single turnover rate constant may be even higher. The 2.5-fold difference between the single turnover rate k obs ϭ 2.5 and k cat ϭ 0.95 Ϯ 0.06 s Ϫ1 may be due to inaccuracies in the determination of the "active concentration" of the SopE 78 -240 preparation used. Overestimation of the actual protein concentration or the presence of a substantial fraction of catalytically inert SopE 78 -240 in the preparation would result in the underestimation of the rate constant k cat in the Michaelis-Menten experiments. The presence of such an inert fraction is suggested by our surface plasmon resonance analysis of Cdc42 ⌬C binding to GST⅐SopE 78 -240 . Alternatively, the differences between k obs ϭ 2.5 and k cat ϭ 0.95 Ϯ 0.06 s Ϫ1 might be attributable to a step after release of the nucleotide (e.g. a conformational change) that might be rate-limiting under the conditions of multiple turnover. This part of the catalytic cycle would not be relevant under single turnover conditions, resulting in higher rates of catalysis (k obs ). In conclusion, our data demonstrate that SopE is an efficient guanine nucleotide exchange factor. Despite the lack of any detectable sequence homology, the kinetic constants for SopE-mediated nucleotide exchange show some striking similarities to those determined for eukaryotic GEFs.
The far-UV CD spectrum of the catalytic domain SopE 78 -240 is indicative of a high ␣-helical content of about 58%. This is in good agreement with secondary structure predictions for the ␣-helical content of this polypeptide (ProtScale tool on the ExPASy molecular biology World Wide Web server of the Swiss Institute of Bioinformatics) (algorithms of Chou and Fasman (48), 50% ␣-helical content; Deleage and Roux (49), 57% ␣-helical content; and Levitt (50), 60% ␣-helical content). The Nterminal 17 amino acids of SopE 78 -240 have a very low probability for forming an ␣-helix. Considering that these residues were artificially added during the construction of the expression vector, this suggests that the actual helical content of the catalytic domain of SopE may be well above 60%. Interestingly, the structures of several GEF domains belonging to the Dbl family, which have been determined recently, have revealed that this catalytic domain is composed almost entirely of ␣-helices (20,22,23) and the RasGEF domain of Sos and GEFs for ARF also have a highly helical structure (18,19,21,24). It is intriguing to speculate whether the catalytic domain of SopE may display structural similarity with any eukaryotic GEF. In addition, the lack of detectable sequence similarity between SopE and GEFs of the Dbl family or any other GEF for small GTP-binding proteins suggests that additional GEFs belonging to neither of the classical GEF families may have gone unnoticed.