Crystal structure of human RhoA in a dominantly active form complexed with a GTP analogue.

The 2.4-A resolution crystal structure of a dominantly active form of the small guanosine triphosphatase (GTPase) RhoA, RhoAV14, complexed with the nonhydrolyzable GTP analogue, guanosine 5'-3-O-(thio)triphosphate (GTPgammaS), reveals a fold similar to RhoA-GDP, which has been recently reported (Wei, Y., Zhang, Y., Derewenda, U., Liu, X., Minor, W., Nakamoto, R. K., Somlyo, A. V., Somlyo, A. P., and Derewenda, Z. S. (1997) Nat. Struct. Biol. 4, 699-703), but shows large conformational differences localized in switch I and switch II. These changes produce hydrophobic patches on the molecular surface of switch I, which has been suggested to be involved in its effector binding. Compared with H-Ras and other GTPases bound to GTP or GTP analogues, the significant conformational differences are located in regions involving switches I and II and part of the antiparallel beta-sheet between switches I and II. Key residues that produce these conformational differences were identified. In addition to these differences, RhoA contains four insertion or deletion sites with an extra helical subdomain that seems to be characteristic of members of the Rho family, including Rac1, but with several variations in details. These sites also display large displacements from those of H-Ras. The ADP-ribosylation residue, Asn41, by C3-like exoenzymes stacks on the indole ring of Trp58 with a hydrogen bond to the main chain of Glu40. The recognition of the guanosine moiety of GTPgammaS by the GTPase contains water-mediated hydrogen bonds, which seem to be common in the Rho family. These structural differences provide an insight into specific interaction sites with the effectors, as well as with modulators such as guanine nucleotide exchange factor (GEF) and guanine nucleotide dissociation inhibitor (GDI).

Rho is a small GTPase 1 that was first purified from mam-malian tissue membrane (1) and cytosol (2) fractions and was identified as the gene product of the ras homologue gene, rho (3). Rho has three mammalian isoforms, RhoA, RhoB, and RhoC, that exhibit high sequence homology with 83% identities (4). Rho cycles between GTP-bound and GDP-bound forms in a similar manner as Ras and other small GTPases. The level of the active GTP-bound form is regulated by its own GDI, GEF, and GAP. The interconversion between the GTP-bound and GDP-bound forms allows Rho to act as a molecular switch that regulates intercellular signaling pathways. Rho is implicated in the cytoskeletal responses to extracellular signals including lysophosphatidic acid and certain growth factors, which result in the formation of stress fibers and focal adhesion (5)(6)(7). Recent isolation and characterization of putative target proteins for Rho from the bovine brain (8,9) have led to a possible mechanism by which Rho regulates cytokinesis, cell motility, or smooth muscle contraction (10,11). These proteins contain the MBS made up of myosin phosphatase and a novel serine/threonine kinase, Rho-kinase, that has been shown to phosphorylate MBS to inactivate myosin phosphatase and also to phosphorylate the MLC. Accumulation of phosphorylated MLC induces a conformational change in myosin II that increases its interaction with actin and enables the formation of myosin filaments (12). Rho-kinase is identical to ROK from the rat brain (13) and p160 ROCK from human megakaryocytic leukemia cells (14), which are members of a growing family of serine/ threonine protein kinases that include myotonic dystrophy kinase. Other target proteins for Rho contain PKN (15,16), Rhophilin (16), Rhotekin (17), and Citron (18). Further signaling pathways for actin polymerization have appeared to involve PtdIns 4-phosphate 5-kinase (19) and p140mDia (20), as downstream effectors.
Rho has two related small GTPases, Rac and Cdc42, that are also involved in regulating the organization of the actin cytoskeleton, whereas the cell morphological effects induced by these GTPases are clearly different in appearance. Rac regulates lamellipodium formation and membrane ruffling, and Cdc42 regulates filopodium formation. Rac is also known to be involved in the activation of NADPH oxidase in phagocytes. Rac has two mammalian isoforms, Rac1 and Rac2, that exhibit a high sequence homology with 90% identities. Rac and Cdc42 also share a significant homology with 68% identities and, actually, bind to some common target proteins for activation. Rho, however, exhibits a relatively low similarity to those GT-Pases, an approximately 45% identity with both Rac and Cdc42. These differences in similarity are thought to be essential for the activation of several downstream target proteins of each small GTPase, although we do not yet understand the molecular basis of the specificities. Based on these differences, RhoA, RhoB, and RhoC are hereafter referred to as the RhoA subfamily, and Rac1, Rac2, and Cdc42 as the Rac1 subfamily. The Rho-binding domains of the target proteins consist of less than 100 residues and have been classified into at least two motifs (9,21). The class 1 of the Rho-binding motif is characterized as a polybasic region followed by a leucine-zipper-like motif and is found in PKN, Rhophilin, Rhotekin, and MBS. Rho-kinase and Citron make up another class of the Rhobinding motif, the class 2, that has a putative coiled-coil motif located at the C terminus of the segment that is similar to myosin rod. It is of considerable interest that these sequences of the Rho-binding domains have no similarity to the binding domain of an activated Cdc42Hs-associated kinase (ACK) (22), a p21(Cdc42/Rac1)-activated protein kinase (PAK) (23), or the Ras-binding domain of Raf-1 (24).
Rho and the related small GTPases are the most common targets for bacterial toxins and are of major importance for the entry of bacteria into mammalian host cells. It is well known that various bacterial toxins can modify Rho by ADP-ribosylation, -glucosylation, and -deamidation. These toxins are classified into three families, C3-like exoenzymes such as Clostridium botulinum C3 ADP-ribosyltransferase, large clostridial cytotoxins such as Clostridium difficile toxins A, and Rhoactivating toxins such as Escherichia coli CNFs (25). The C3like exoenzymes act on members of the RhoA subfamily, but most of the large clostridial cytotoxins inactivate all members of the Rho family. The Rho-activating toxins activate members of the RhoA subfamily and Cdc42. No activity for Ras, Rap, and Ran has been reported for the bacterial toxins of these three families, but Clostridium sordelli HT, one of the large clostridial cytotoxins, is known to inactivate Ras and Rap. There is no interpretation for these emerging differences in the specificity of the small GTPases. Hence, it becomes essential to examine the three-dimensional structures of Rho to understand how their interactions with the target proteins control the various signaling processes and how the modifications by bacterial toxins change the activities of their target GTPases for bacterial invasion. We report here the crystal structure of recombinant human RhoA, which is dominantly activated with substitution of Gly 14 by valine (RhoA V14 ), complexed with GTP analogue, GTP␥S, and we compare it with the structures of H-Ras and other related GTPases.

EXPERIMENTAL PROCEDURES
Preparation and Crystallization of RhoA V14 -The cloning, expression, and purification of the dominantly active form of recombinant human RhoA V14 complexed with GTP␥S and Mg 2ϩ were carried out according to the methods described previously (8,9,11,15). Detail procedures will be described elsewhere. The resulting active sample, used in this study, is verified with MALDI-TOF MS (JMS-ELITE, PerSeptive Inc.) and N-terminal analysis (M492, Applied Biosystems). The protein is truncated at Ala 181 and has one additional serine residue at the N terminus. Crystals were obtained at 4°C by the hanging-drop vapor diffusion method from solutions containing 10 mg/ml GTP␥S-RhoA V14 , 50 mM Tris-HCl buffer, pH 8.5, 10% PEG8000, 7.5% 1,4dioxane equilibrated against 100 mM concentration of the same buffer containing 20% PEG8000 and 15% 1,4-dioxane. Plate-like crystals (Form A) grew within a few days and were found to diffract up to 2.4 Å resolution. The crystals belong to space group P2 1 2 1 2 (a ϭ 62.02 Å, b ϭ 74.78 Å, c ϭ 50.52 Å), with one molecule in the asymmetric unit. Hexagonal crystals (Form B) also were obtained from solutions contain-ing 10 mg/ml GTP␥S-RhoA V14 , 50 mM sodium acetate buffer, pH 4.6, 10% 2-propanol equilibrated against 100 mM of the same buffer containing 20% 2-propanol. Crystals had hexagonal or trigonal lattice parameters with a rather long c axis (a ϭ b ϭ 60.80 Å, c ϭ 214.56 Å) and diffracted at 3.0 Å.
Data Collection and Structure Determination-The structural analysis was performed using Form A. Intensity data were collected at 10°C using an R-AXIS IIc imaging plate detector with CuK ␣ x-rays generated by a rotating anode RU-300H (RIGAKU, Japan). The diffraction data were processed with PROCESS (RIGAKU). A summary of the data processing statistics are given in Table I. The initial phases were calculated by molecular replacement with the program AMoRe (26) using a search model based on the structure of human H-Ras (Protein Data Bank code 5P21, Brookhaven National Laboratory), with which RhoA shares a 27.5% identity. Several searches with a polyalanine model using different ranges of intensity data and integration radii resulted in a unique solution. Rigid body refinements of the searched model were performed with X-PLOR (27). The model obtained was divided into the secondary structure elements, and again, rigid body refinements were performed, followed by solvent flattening/histogram matching with the program DM (28). Four regions of insertions and deletions were inspected on the resulting 2F o -F c map that was generated with the program O (29). The structure was built and refined through alternating cycles using the programs O and X-PLOR, respectively.
Structure Refinement-Three regions were poorly defined in the resulting map. The first is at the loop and ␤-strand residues, Asp 28 -Gly 50 , which contain the switch I region connected to strand B2, and the second is at the residues of the switch II region. All residues in these regions were rebuilt on their omit maps. Structures of these parts were found to have large displacements from those of H-Ras (see text). The last is at the inserted residues, Glu 125 -Glu1 37 , which is specific for members of the Rho-family. After several cycles of refinements incorporating solvent water molecules located at regions other than the inserted residues, we defined the residues forming a short 3 10 -helix connected to an ␣-helix. The GTP␥S molecule and the Mg 2ϩ ion were identified unequivocally by their appearance in 2F o -F c maps. The ␥-sulfur atom of GTP␥S was also identified by its appearance in F o -F c maps and its standard sulfur-phosphorus bond distance (1.9 Å), which is longer than the nonbridging oxygen-phosphorus bond (1.5 Å). Three N-terminal residues have uninterpretable densities implying complex disorder. The structure consists of one RhoA V14 molecule of 178 residues, one GTP␥S, one Mg 2ϩ ion, and 38 water molecules. The side chains of Arg 5 , Glu 54 , Arg 68 , and His 126 are poorly defined in the current structure. There are no residues in disallowed regions as defined in PROCHECK (30). A summary of the refinement statistics is given in Table I. The structure was inspected using the program QUANTA (Molecular Simulators Inc).

RESULTS
Overall Structure-The major features of the fold, consisting of a six-stranded ␤-sheet surrounded by helices connected with loops, are basically conserved as found in H-Ras (31,32) and other related small GTPases (33-36) (Fig. 1). The ␤-sheet is formed by the anti-parallel association of two extended ␤-strands (B2 and B3) and the parallel association of five extended ␤-strands (B3, B1, B4 -B6). RhoA V14 contains five ␣-helices (A1, A3, A3Ј, A4, and A5) and three 3 10 -helices (H1-H3). There are three insertion and one deletion sites, which are common in the members of the Rho family, as can be seen from the sequence and secondary structure element alignment of RhoA and H-Ras (Fig. 2). The 13-residue insertion (Asp 124 -Gln 136 ) is located at the loop between strand B5 and helix A4. Excluding the deletion and insertion residues, the C ␣ -carbon atoms of RhoA V14 and the corresponding dominantly activated H-Ras V12 , which is complexed with GTP (37), superimpose with a root mean square (r.m.s.) deviation of 1.68 Å for 163 common C ␣ -carbon atoms. This superposition yields an r.m.s. deviation of 0.87 Å for the atoms of the guanine nucleotide. Segments that involve major differences are located at the switch I and II regions and part of the antiparallel ␤-sheet, consisting of the C-terminal half of strand B2 and the N-terminal half of strand B3, in addition to the insertion and deletion sites described above (Fig. 3A). Recently, Wei et al. have reported the crystal structure of RhoA bound to GDP (38). Compared with this RhoA-GDP structure, the significant conformational changes were found to be localized in the switch I and II regions (Fig.  3B), as described for H-Ras (31,39). Excluding these regions, the C ␣ -carbon atoms of RhoA V14 -GTP␥S and RhoA-GDP superimpose with a r.m.s. deviation of 0.48 Å. The nonhydrolyzable nucleotide GTP␥S binds to the protein with a Mg 2ϩ ion that has a typical octahedral coordination sphere (Fig. 4).
Insertion Regions-The N-terminal segment (Glu 125 -Lys 133 ) of the 13-residue insertion forms an ␣-helix designated as A3Ј, which is followed by an extended loop. A short 3 10 -helix, designated as H3 (Fig. 1), is induced at the segment flanking the N terminus of this insertion, with large displacements of Arg 122 and Asn 123 from those of H-Ras V12 (3.0 Å and 6.1 Å, respectively). Compared with RhoA-GDP, however, no significant conformational change exists in the 13-residue insertion and its N-terminal flanking regions. This folding seems to be basically similar to that of Rac1 complexed with GMP-PNP (36) but shows many differences in details. It is notable that the sequences of this region of members in the RhoA subfamily is rather different from those of the Rac1 subfamily. Among the key residues in stabilization of helices H3 and A3Ј (Fig. 5A), Arg 122 and Asp 124 are conserved in the Rho family but Arg 128 , Glu 137 , and Lys 140 are variant in the Rac1 subfamily. No water molecule is found to be involved in the structural stabilization of the RhoA insertion region, though Rac1 forms a watermediated hydrogen bond between the main chains. The conserved residues Leu 131 and Pro 138 of helix A3Ј form a hydrophobic patch with Thr 127 and Pro 89 , which are also conserved or conservatively replaced in the Rho family. Rac1 adds Ile 126 (Arg 128 of RhoA) to the hydrophobic patch.
The outer surface of helix A3Ј is covered with charged residues whose side chains form hydrogen bonds and/or ion pairs, Glu 125 -Arg 129 and Glu 130 -Lys 133 pairs. These residues are conserved or conservatively substituted in the RhoA subfamily, but are replaced by other amino acid residues in the Rac1 subfamily. It is interesting that, in the Rac1 subfamily, Glu 125 and Arg 129 are substituted by lysine and glutamic acid, respectively, and Glu 130 and Lys 133 are substituted by lysine/arginine and glutamic acid, respectively. Therefore, these pairs of acidic and basic residues of Rac1 could form hydrogen bonds or ion pairs as observed in the current structure, though most of the and F c (h) are observed and calculated reflections. d R free is R cryst which was calculated using 10% of the data, chosen randomly and omitted from the subsequent molecular replacement and structure refinement. e ⌬ is the deviation of the peptide torsion angle from 180°.
FIG. 1. Structure of RhoA V14 -GTP␥S. Shown is a ribbon representation of RhoA V14 complexed with GTP␥S (yellow) and Mg 2ϩ (a gray ball) with ␤-strands (red), ␣-helices (green), and 3 10helices (blue). Three water molecules (pink balls) are also illustrated. One water molecule (Wat1) participates in the guanine-base recognition of GTP␥S, the second (Wat2) participates in the binding of the ribose, and the last (Wat3) is a putative nucleolytic water molecule. The secondary structure elements, Mg 2ϩ ion, water, and GTP molecules are labeled as well as the N and C termini. exposed side chains of the residues of Rac1 corresponding to the residues 125-135 of RhoA are highly mobile.
Compared with H-Ras, helix A3 has a one-residue insertion at the center and two Pro residues (Pro 96 and Pro 101 ), which cause a disruption of the normal hydrogen-bonding pattern of an ␣-helix, whereas H-Ras has no Pro residue on this helix. These differences induce a relatively large discrepancy (2.05 Å at Pro 96 ) of helix A3 from that of H-Ras V12 (Fig. 3A). Both Pro 96 and Pro 101 face the solvent region so as to induce pronounced kinks that serve to maximize the contacts with strands B1 and B4.
Phosphate-binding Loop-The G14V mutation of RhoA and the G12V mutation of H-Ras exhibit less than one-tenth the GTPase activity of the wild-type GTPases. Crystal structures of H-Ras V12 complexed with GDP (31,39) and GTP (37) show that the mutation causes no significant conformational change at the phosphate-binding region, though there are large differences in mobility and conformation predominantly localized in the switch II region (see below). Similar results were obtained in RhoA. The r.m.s. deviation of the 12 GXG(V)XXGKT/S 19 motifs between RhoA V14 -GTP␥S and RhoA-GDP is small (0.31 Å) and that between RhoA V14 -GTP␥S and H-Ras V12 -GTP is relatively small (0.84 Å). However, the bulky side chain of Val 14 contacts with Gly 62 and Gln 63 , and these contacts cause a displacement (0.71 Å) of the C ␣ -carbon atom of Val 14 from the corresponding atom of the wild-type RhoA (Fig. 5B).
Switch I-The switch I region in RhoA V14 -GTP␥S is well ordered in the crystal structure as well as that of RhoA-GDP. In contrast, in Rac1-GMP-PNP, residues 32-36 of switch I (34 -38 in RhoA) are disordered. Dramatic changes in the conformations of switch I and its C-terminal flanking regions, as compared with RhoA-GDP, occurs with the largest displacements (5.4 Å and 6.4 Å) at Pro 36 and Phe 39 , respectively (Fig.  5B). Similar conformational changes whether bound to GTP or GDP were reported for H-Ras (31,40), the G ␣ subunits of the trimeric GTPases such as transducin-␣ (41), thereby playing a key role as a molecular switch in signal transduction. Tyr 34 and Pro 36 of RhoA V14 flip their side chains toward the nucleotide so as to shield the triphosphate group from the solvent region. The phenolic ring of Tyr 34 stacks on the Pro 36 and also contacts with Ala 15 and Val 14 to close the entrance of the phosphatebinding pocket. These conformational changes are accompanied by the flipping out of hydrophobic residues, Val 35 , Val 38 , and Phe 39 , toward the solvent region. It should be noted that Val 38 and Phe 39 form a hydrophobic patch on the molecular surface together with Tyr 66 and Leu 69 of switch II. These contacts play a pivotal role to induce the stable conformation of switch II (see below). Switch I of RhoA V14 displays a signifi- cantly different conformation from that of H-Ras V12 -GTP. Large displacements of the residues of switch I begin from Asp 28 and end at Pro 36 with the largest displacement (4.1 Å) at Glu 32 (Figs. 5C). These displacements, which result in differences in recognition of the ribose of the guanine nucleotide, seem to be caused by Pro 31 , which restricts the main chain torsion angles. Since Pro 31 is well conserved in the Rho family but is replaced by other residues in Ras (Val), Rab (Val), and Ran (Asp), the displacements could be a common structural feature of members of the Rho family. While large displacements were observed in switch I as described, no significant difference in the position and orientation of Thr 37 , which coordinates to the Mg 2ϩ ion, is seen between RhoA V14 and H-Ras V12 .
Strands B2 and B3-The switch I loop is connected to the anti-parallel ␤-sheet of strands B2 and B3, which is followed by switch II. This two-stranded sheet is located at the edge of the six-stranded ␤-sheet and is sitting on helices A1 and A5 to form a hydrophobic core. Compared with H-Ras V12 -GTP, a large displacement of these strands expands between a stretch from Ala 44 to Val 53 , which moves toward helices A1 and A5, with the largest shift being 2.9 Å at Asp 45 . It is notable that the sequence of this region is highly conserved in the RhoA subfamily. This displacement seems to be caused mainly by differences in hydrophobic interactions between these strands and helices A1 and A5. Strands B2 and B3 are connected by a reverse turn of type II formed by 48 VDGK 51 . H-Ras V12 and H-Ras also have a type II reverse turn at this position with the corresponding segment, 46 IDGE 49 . Since the third residue of this type of reverse turn should have a Gly residue to avoid the steric clash with the main-chain carbonyl group of the second residue, mutations of this residue to any other residue destroy the reverse turn, which results in large conformational changes of the strands or displacement of the ␤-sheet. This may possibly explain why mutations of Gly 48 in H-Ras inhibits effector function (42) even though this residue is distal from the effectorbinding site encompassing switch I and the N-terminal half of strand B2, as observed in the crystal of Rap1A-Raf1 complexes (35).
region of RhoA V14 is well defined and has a single conformation at the present resolution. RhoA V14 has two 3 10 -helices, H1 ( 64 EDY 66 ) and H2 ( 70 RPL 72 ), which are separated by a short loop of three residues ( 67 DRL 69 ). The sequence of this region is well conserved in the Rho family but is different from those regions in the Ras family (Fig. 2). A similar conformation is also seen in the crystal structure of Rac1-GMP-PNP. In contrast, H-Ras V12 has a 3 10 -helix at the position corresponding to the short loop of RhoA V14 with an ␣-helix of five residues corresponding to residues 71-75 of RhoA V14 . These differences induce a large displacement of Glu 64 (3.8 Å), together with reorientations of the side chains of Gln 63 , Glu 64 and Asp 65 from the corresponding residues of H-Ras V12 (Fig. 3A). Lys 98 and Glu 102 , both of which are located at helix A3, play crucial roles in the conformation of the segment by forming multiple hydrogen bonds to switch II (Fig. 5D). These two residues are conserved in members of the Rho family but are replaced in H-Ras. Moreover, the segment from helices H1 to H2 makes hydrophobic contacts strands B2 and B3. In this hydrophobic core, H-Ras V12 has an additional residue Tyr 71 , which is replaced by a small residue (Ser 73 ) in RhoA. This difference causes a movement of the helix H2 toward strand B2. These differences seem to be one of the main reasons why the conformations of the segment are so different between RhoA V14 and H-Ras V12 .
Magnesium Ion Binding-The strong GTP/GDP-binding and the GTPase activity of small GTPases have been shown to be absolutely dependent on the presence of divalent ions. The Mg 2ϩ ion of the present structure is located at a position similar to those in H-Ras V12 -GTP and in H-Ras-GMP-PNP, as well as that in RhoA-GDP. The displacement of the ion from the corresponding position in the GDP-bound form is 1.04 Å. The Mg 2ϩ ion plays a key role in bringing together the functional regions of the phosphate-binding, switches I and II, as observed in H-Ras. Actually, the stereochemistry of Mg 2ϩ coordination is identical to that in H-Ras-GMP-PNP. In contrast, the stereo-chemistry of Mg 2ϩ coordination in RhoA-GDP is different from the current form but also is different from that in H-Ras-GDP.
Guanosine Nucleotide Binding-The glycosyl conformation of GTP␥S is anti with the C2Ј-endo sugar pucker. The guanine base is trapped in a hydrophobic pocket, in a manner similar to H-Ras V12 -GTP, to be recognized by several interactions with the conserved residues of the 116 GXKXDL 121 and 160 SAK 162 motifs (Figs. 5C). A major difference in base recognition is the water-mediated hydrogen bonds to the N7 and O6 atoms of the guanine base. The water molecule (Wat-1) is completely buried inside the hydrophobic binding pocket with a hydrogen bond to Gly 17 . The space for the accommodation of this water molecule is mainly produced by a rearrangement of the side-chain packing of the pocket, involving Leu 21 , Asn 117 , and Cys 159 (Fig. 5E). In H-Ras, Cys 159 of RhoA is replaced by a Thr residue. In addition to the base recognition, the 2Ј-hydroxyl group of the ribose also has a water-mediated hydrogen bond to switch I, although in H-Ras V12 -GTP, the hydroxyl group of the ribose forms direct hydrogen bonds with the main chains corresponding to Pro 31 and Glu 32 of RhoA. As mentioned above, these differences in the recognition of the ribose are caused by the large displacements of switch I. Similar water-mediated hydrogen bonds in base and sugar recognition have also been found in RhoA-GDP and in Rac1-GMP-PNP.
Triphosphate Binding-GTP and GDP bind to small GT-Pases with dissociation constants on the order of nanomolar. This strong binding affinity is well demonstrated in the current structure. The triphosphate moiety of GTP␥S has 21 direct and 9 water-mediated hydrogen bonds to the protein, together with 2 magnesium coordinations. These involve six residues of the phosphate-binding loop, four residues of switch I, three residues of switch II, and four residues of base-recognition motifs. It is notable that most of these residues of the phosphatebinding loop interact with the triphosphate through their main-chains, especially the amino groups. This is the reason FIG. 4. GTP␥S bound to RhoA V14 . A cartoon of GTP␥S binding to RhoA V14 with Mg 2ϩ and water molecules. All dashed lines correspond to hydrogen bonding interactions (distance less than 3.5 Å), and the corresponding distances (Å) are indicated. The residues whose main chains participate in the hydrogen bonding are represented by rectangles, and the residues whose side chains participate in the hydrogen bonding are represented by ovals. The coordination bonds to the Mg 2ϩ ion are indicated by arrows. The possible hydrogen bond between Gln 63 and Wat-3 has a longer distance (3.8 Å). The hydrogen bonds observed in the current structure but not in H-Ras are highlighted in red.
why the amino acid sequence of the 12 GXGXXGKT/S 19 motif contains many variant residues. The residues whose side chains participate in the interactions with the triphosphate are invariant Lys 18 , Tyr 34 , and Asp 59 . The conformation of the triphosphate exhibits similarity to that of GDP bound to RhoA, with relatively small displacements of the ␣and ␤-phosphates from those of GDP, 0.75 Å and 0.67 Å, respectively. Two oxygen atoms of the ␥-phosphate make contact with the protein by several hydrogen bonds, together with the coordination to the Mg 2ϩ ion buried inside the pocket formed by switches I and II and the phosphate-bonding loops. These heavy interactions allow the ␥-phosphate to orient the ␥-sulfur atom toward Val 14 , Tyr 34 , and Pro 36 . Similar configurations of the ␥-thiophosphate were also observed in the crystal structures of transducin-␣ (44) and G i␣1 (45) complexed with GTP␥S. In all these crystals complexed with GTP␥S, the ␥-sulfur atom is the closest atom to the side-chain amide group of Gln 63 (Gln 200 of transducin-␣ and Gln 204 of G i␣1 ) among the ␥-thiophosphate atoms.
Putative Nucleophilic Water Molecule-We identified one water molecule (Wat-3) that is close enough to the ␥-phosphate to perform an in-line nucleophilic attack. The water molecule is located at a position 10°off from this line at a distance of 3.6 Å from the phosphorus atom and forms a hydrogen bond (3.3 Å) to the ␥-sulfur atom, although the distance to the ␥-oxygen atoms of the phosphate group is too long to form a hydrogen bond (3.6 Å for both). Similar water molecules have been located in analogous positions close to the ␥-phosphate in the crystal structures of transducin-␣ and G i␣1 complexed with GTP␥S, as FIG. 5. Various parts of RhoA V14 -GTP␥S. A, the 13-residue insertion subdomain of RhoA V14 . The carbon, nitrogen, oxygen, and sulfur atoms are in white, blue, red, and yellow, respectively. The C ␣ -carbon atom tracing is in brown. The hydrogen bonds involving the side chains are indicated by thin white lines. B, interactions in switches I and II and P-loop regions of RhoA V14 (white) bound to GTP␥S (magenta). The C ␣ -carbon atom tracings of the corresponding regions of RhoA-GDP (green), together with the side chains of switch I, are superimposed. C, part of RhoA V14 switch I that displays large displacements from the corresponding part of H-Ras V12 . The corresponding part of switch I and the guanosine moiety of H-Ras V12 -GTP complex are shown with green lines. The guanosine (the carbons in brown) of GTP␥S bound to RhoA V14 is also shown with residues in the guanine-binding site. The two water molecules are indicated by red balls. D, superposition of C ␣ -carbon atom tracings of switch II of RhoA V14 (white) and H-Ras V12 (green). The side chains that stabilize the unique conformation of the RhoA V14 switch II are added as well as the functionally important residues. For clarity, the hydrogen bond between the side chain of Arg 70 and the main chain of Ala 61 (see text) is not shown in this figure. E, rearrangement of the side-chain packing of the guanine recognition site for accommodation of a water molecule (Wat-1) of the RhoA V14 -GTP␥S complex. The corresponding part of H-Ras V12 -GTP is superimposed with the carbon in green. F, modification sites by bacterial toxins located at the C terminus of switch I and part of ␤-sheet B2 (white) -B3 (brown). The stacking interaction between the side-chain carbonyl group of Asn 41 and the indole ring of Trp 58 is indicated by a broken line. The target residue, Thr 37 , for glucosylation by large closteidal cytotoxins is located at the N terminus of switch I. The Mg 2ϩ ion and the triphosphate group of GTP␥S is shown with a ball-and-stick model. well as of H-Ras, Rac1, and EF-Tu (46) complexed with GMP-PNP, although no water molecule corresponding to Wat-3 has been found in RhoA-GDP. Gln 63 positions the side-chain oxygen atom at a distance of 3.8 Å from the hydrolytic water and the side-chain nitrogen atom at a distance of 3.8 Å from the side-chain carboxyl group of Asp 65 .

Modification Sites by Bacterial Toxins-C. botulinum C3
ADP-ribosyltransferase transfers an ADP-ribose moiety of NAD to Asn 41 of Rho (47). The side-chain of Asn 41 , which is located at strand B2, forms a hydrogen bond (3.1 Å) to the main-chain carbonyl group of Glu 40 . This hydrogen bond allows Asn 41 to interact with the indole ring of Trp 58 of strand B3 (Fig.  5F). The distances between the nearest atoms of the indole ring and the carbonyl oxygen atom of the side chain of Asn 41 range from 3.3 to 3.5 Å, which indicates the existence of a stacking interaction between them. Because the indole ring is a strong electron-donor, this interaction may help to enhance the nucleophilic properties of the side-chain nitrogen atom of Asn 41 . It should be noted that the hydrophobic side chains of Val 38 , Phe 39 , and Val 43 are exposed to the solvent region around Asn 41 , together with Trp 58 . This unusual feature of the molecular surface may be related to the interaction with C3-like exoenzymes. Asn 41 orients the side chain away from the switch I loop. This is consistent with the fact that the ADP-ribosylation on Rho affected neither the GTP␥S binding nor its intrinsic GTPase activity. Furthermore, the ADP-ribosylation on Rho did not affect its interaction with rhoGAP (48). Recent data using Swiss 3T3 cells indicates that the ADP-ribosylation of Rho enhances its binding to PtdIns 4-phosphate 5-kinase and acts as a dominantly negative inhibitor (19,49). This also suggests that the ADP-ribosylation does not impair the intrinsic properties of the switch I conformation though PtdIns 4-phosphate 5-kinase could bind to the GDP-bound form, and therefore, the binding may be different from those of other effectors that do not bind to the GDP-bound form. Rac and Cdc42 are not subjected to ADP-ribosylation (50). This may be related to the global conformation of the anti-parallel ␤-sheet formed by strands B2 and B3 since the sequence of this region of RhoA subfamily is conserved but is different from that in the Rac1 subfamily. On the molecular surface around Asn 41 , Val 43 is replaced by Ser/Ala and Glu 40 is replaced by Asp in Rac and Cdc42.
Recent biochemical data have shown that Thr 37 is glucosylated by the major virulence factors of C. difficile, toxin A and B (51). The glucosylated RhoA induces the disaggregation of actin filaments. It also appears that GDP-bound RhoA is a superior substrate for Toxin B to GTP-bound RhoA. This is Structure of RhoA V14 -GTP␥S complex consistent with the crystal structures: in RhoA-GDP, the side chain of Thr 37 does not participate in either Mg 2ϩ ion or phosphate binding, whereas it participates in both in the current structure. Since Thr 37 orients the side chain inside the loop, its glucosylation must accompany a structural deformation of the loop. This structural change could extend to strand B2. Actually, it has been shown that the glucosylation of Thr 37 inhibits ADP-ribosylation by C3-like exoenzymes (52).
CNFs from E. coli and dermonecrotic toxins (DNTs) from Bordetella species induce the massive reorganization of the actin cytoskeleton and inhibit cell division, leading multinucleated cells. Recently, CNF1 has been shown to cause the deamidation of Gln 63 of RhoA, resulting in a dominantly active form, RhoA E63 (53)(54). CNF1 acts preferentially with RhoA but also inhibits the GAP-stimulated GTPase activity of Cdc42 and of Rac at high concentrations. These actions of CNF1 may be related with the unique conformation and/or conformational properties of switch II. The differences in the CNF1 activity on RhoA, Cdc42 and Rac probably indicate that this toxin may interact with these small GTPases through segments other than switch I, though it remains unclear.
GTP/GDP Switching and Effector Binding-The fundamental mechanism of the molecular switch, which involves the significant conformational changes in switch I and II regions, in signal transduction seems to be common in small GTPases and G ␣ subunits of trimeric GTPases, as described for H-Ras (31), transducin-␣ (41), and G i␣1 (55) and Rap2A (56). However, the present RhoA V14 structure reveals large conformational deviations from H-Ras V12 in the regions containing switches I and II. The structures of two other small GTPases, ADP-ribosylation factor (Arf) (33) and Ran (34), also showed significant conformational variations in the switch regions, whose structures are also different from those of the present RhoA V14 : switch II of Arf forms a long ␤-strand, and Ran has a completely different orientation of switch II that contains a short ␤-strand. All these results indicate that small GTPases from different families may have a similar fold but with significant variations in the switch regions.
There are several biochemical data indicating that the switch I region of RhoA is involved in its effector binding. It has been suggested from analyses of the chimeric proteins of Rho and Ras that the switch I region (residues 32-42) is essential for the induction of actin stress fiber formation (57). Either Cdc42 or Rac shows no significant binding to the target proteins of RhoA, such as Rho-kinase and the others described above. At switch I and its franking regions of RhoA V14 , residues that are exposed to the solvent region are well conserved in the RhoA subfamily but are replaced in the Rac1 subfamily. It is of interest that most of the side chains of these residues protrude onto the same molecular surface (Fig. 6). The double mutant Rap1A (58), which mimics Ras, binds to the Ras-binding domain of c-Raf1 through several residues that are located at the same side of the corresponding molecular surface of RhoA V14 . Among them, residues whose side chains form the specific hydrogen bonds to the Ras-binding domain are located at the N-terminal half of strand B2. It is notable that most of these residues are replaced by non-conservative, mainly hydrophobic, residues (Val 33 , Val 35 , Phe 39 -Asn 41 , and Val 43 ) in RhoA V14 -GTP␥S to form hydrophobic patches on the molecular surface, as described.
In addition to switch I, the second effector site is suggested in the C-terminal two-thirds of the molecule (57). However, little is currently known about the possible second effector site of RhoA. Recent mutagenesis experiments have indicated that the 13-residue insertion region of Rac1 participates in the interaction with p67phox but not in the interaction with PAK, and a combinational use of the multiple effector-binding sites has therefore been proposed (59). Since there are several structural differences in the 13-residue insertion regions between RhoA V14 -GTP␥S and Rac1-GMP-PNP, it may be possible that RhoA also utilizes the insertion region in the specific binding with its own effector proteins, but this remains to be seen in future experiments. It should be noted that the 13-residue insertion region has no significant displacement from that in RhoA-GDP and, therefore, has no switching function between GTP-bound and GDP-bound forms. It is well known that the G ␣ subunits of trimeric GTPases contain four insertion regions if compared with small GTPases. The 13-residue insertion region Molecular surface of RhoA V14 . Residues whose mutations abolish the interaction with GEF are in yellow. Asn 41 is also highlighted in green. Switches I and II are shown in red and blue, respectively. This surface also contains most of the residues corresponding to the effector-binding residues as seen in the complex between the Ras-binding domain of Raf1 and a double mutant Rap1A (E30D/K31E), which mimics Ras. of the members of the Rho family corresponds to the third insertion region that forms an additional helix at the N-terminal portion of the segment (44), although no homology has been detected between RhoA and each of G ␣ subunits and no possible function has been assigned to this insertion region.
GEF and GDS Binding-GEFs for small GTPases of the Rho family have been identified as Dbl-containing proteins that contain a region with a sequence homology to the dbl oncogene product (60). While most of these Dbl-containing proteins can act on multiple members of the Rho family in vitro, some have a limited specificity for one type of the GTPases in vivo. Among them, Lbc shows selectivity for Rho (61) but Tiam-1 for Rac (62) and Cdc24 for Cdc42 (63). Analysis of RhoA/Cdc42Hs chimeric proteins has suggested that residues of switch I and switch II are involved in the specific interaction with Lbc (64). Based on mutation analyses, Lys 27 , Tyr 34 , Thr 37 , and Phe 39 in switch I and Asp 76 in switch II have been identified as Lbc-sensitive residues. These residues are located at nearly the same side of the molecule, which may form a surface of interaction for Lbc. It is of interest that this surface is almost the same as that for a tentative effector binding (Fig. 6). The side chains of all these residues are highly projected toward the solvent region but Thr 37 is buried inside the switch I loop. This might be one reason why the mutation of T37A has an affinity for Lbc comparable with the wild-type RhoA although most of the other mutants failed to associate with Lbc. In contrast, extensive mutations in switch II have been reported to have no significant change in their sensitivity to Lbc. Most of these residues are found on the other side of the molecule or inside the protein.
It is notable that the ADP-ribosylated residue by C3-like exoenzymes, Asn 41 , is also located at this surface and may inhibit the binding to Lbc. There is another protein having guanine nucleotide exchange activity, smgGDS, that has no homology to the Dbl-containing proteins but has some homology to Cdc25 of yeast (65,66). smgGDS shows wider specificity than Dbl homologues and acts on Ki-Ras, Rap in addition to the Rho family members. Rho seems to interact with smgGDS through its molecular surface containing Asn 41 since ADP-ribosylation of Rho by C3-like exoenzymes is reduced in the presence of smgGDS (67).
GDI Binding-In addition to inhibiting nucleotide dissociation, GDIs mediate partitioning their cognate small GTPases between the membrane and the cytosol (68). RhoGDI inhibits the guanine nucleotide exchange of all members of the Rho family. Recent structural studies have suggested that rhoGDI binds to the cognate GTPases via an immunoglobulin-like domain that has a hydrophobic pocket for binding to the Cterminal isoprenyl group (69,70). Although this immunoglobulin-like domain has little effect on the rate of nucleotide dissociation from the GTPases, it has been suggested that this binding directs the flexible N-terminal arm of rhoGDI to GTPases, resulting in the inhibition of nucleotide exchange. It is of interest to question how the N-terminal arm interacts with GTPases because the C terminus having the isoprenyl group is located on the molecular surface of the GTPases opposite to the nucleotide binding surface. It has been reported that GDI effectively prevents ADP-ribosylation by C3-like exoenzymes and the nucleotide-exchange activity of smgGDS (67). Furthermore, the nucleotide-exchange activity of Dbl also was remarkably reduced (66). Taken together, these results suggest that the N-terminal arm of rhoGDI may interact with GTPases on the molecular surface, which has residues interact with GEF and GDS as well as C3-like exoenzymes (Fig. 6). This hypothesis provides a framework for analyzing the interactions of rhoGDI, GEF, and GDS with RhoA.
Effects of the ␥-Sulfur Atom and the G14V Mutation on the GTPase Activity-The GTP␥S molecule exhibits its resistance to hydrolysis, which is conferred by the ␥-thiophosphorothioate. In the present structure, the ␥-thiophosphate turns the sulfur-phosphorus bond toward Gln 63 and positions the ␥-sulfur atom to come into contact with the putative nucleophilic water molecule. Therefore, the bulky sulfur atom, which has a van der Waals radius (1.8 Å) much larger than that of the oxygen atom (1.4 Å), sterically shields the phosphorus atom from the close approach of the nucleophilic water molecule and could interfere with the stabilization of the transition state by Gln 63 . The ␥-sulfur atom could also interfere with the stabilization of the transition state by the Arg residue from GAP (71). Similar mechanisms for the resistance of the GTP␥S molecule to hydrolysis, which is conferred by the ␥-thiophosphorothioate, are possible for transducin-␣ and G i␣1 . Based on the crystal structure of transducin-␣ complexed with GTP␥S, it has also been pointed out that Arg 174 , which is a key residue stabilizing the transition state, prevents the thiophosphate from reaching the transition state, due to a steric clash between the firmly anchored guanidino group and the sulfur atom (44). It has been suggested that the role of the key Gln residue (Gln 63 of RhoA) is to stabilize the transition state by direct hydrogen bonds doubly bonded to the ␥-phosphate and the putative nucleophilic water molecule. The transition state should induce conformational changes around the active site of the current structure since the present conformation of Gln 63 directs the side-chain carbonyl group toward the nucleolytic water molecules (3.8 Å) but also positions the side-chain amide group away from the phosphate. The ␥-sulfur atom of the phosphate is closest to the side-chain amide group of Gln 63 , as described above, but the distance between them is more than 5 Å. The contacts of the branched side chain of Val 14 with the N terminus of switch II seem to push Gln 63 away from the ␥-phosphate group and reduce the conformational flexibility of the side chain of Gln 63 . Actually, the C␥ carbon atoms of Val 14 and Gln 63 have a contact of 3.6 Å. Any rotation around the sidechain torsions of Gln 63 could not bring the side-chain amide group to a position close enough to interact with the ␥-sulfur atom because of the steric hindrance of the bulky side chain of Val 14 . We postulate that these steric effects are a possible means of inhibiting GTP hydrolysis by the dominantly active mutation V14 of RhoA. Thus, the mechanism of dominant activation by the G14V mutation of RhoA seems to be similar to that of G12V of H-Ras even though the conformations of switches I and II are quite different.