Structural Insights into the Interaction of ROCKI with the Switch Regions of RhoA*

The Rho-ROCK pathway modulates the phosphoryla-tion level of a variety of important signaling proteins and is thereby involved in miscellaneous cellular processes including cell migration, neurite outgrowth, and smooth muscle contraction. The observation of the involvement of the Rho-ROCK pathway in tumor invasion and in diseases such as hypertension and bronchial asthma makes it an interesting target for drug development. We herein present the crystal structure of the complex between active RhoA and the Rho-binding domain of ROCKI. The Rho-binding domain structure forms a parallel (cid:1) -helical coiled-coil dimer and, in contrast to the published Rho-protein kinase N structure, binds exclusively to the switch I and II regions of the guanosine 5 (cid:1) -( (cid:2) , (cid:3) -imido)-triphosphate-bound RhoA. The switch regions of two different RhoA molecules form a predominantly hydrophobic patch, which is complementarily bound by two identical short helices of 13 residues (amino acids 998– 1010). The identified ROCK-binding site of RhoA strik-ingly supports the assumption of a common consensus-binding site for effector recognition. A hallmark small is their ability to undergo structural changes in response to the of GDP or GTP. The inactive GDP-bound and the active GTP-bound states are recognized by different partner proteins, thereby allowing the small GTPases to function as molecular switches (1, 2). Nucleotide by PCR and cloned into the pGEX-4T1 vector (Amersham Biosciences) via BamHI/XhoI restriction sites. Recombinant glutathi-one S -transferase fusion proteins were purified by a GSH-Sepharose column (Pharmacia) using 50 m M Tris/HCl, pH 7.5, 100 m M NaCl, 5 m M MgCl 2 , and 5 m M dithioerythritol. After cleavage with thrombin, pro- teins were purified using a benzamidine column (Pharmacia) to remove the protease and subsequent gel filtration (Superdex G75, Pharmacia). Purified protein quality and the nucleotide binding capacity were ana-lyzed by SDS-PAGE and reversed-phase high pressure liquid chroma- tography using a C-18 column (ODS-Hypersil, 5 (cid:4) m, Bischoff, Leon-berg, Germany), respectively. The complex of RhoA (cid:2) Gpp(NH)p and ROCKI-RBD was isolated by gel filtration (Superdex 75, 16/60, Pharmacia). Crystallization and Data Collection— Crystals of truncated RhoA (residues 1–181) in complex with the non-hydrolyzable GTP analogue Gpp(NH)p and ROCKI-RBD (residues 947–1015) were grown by vapor diffusion in hanging drops at 20 °C by adding equal volumes of reservoir solution (100 m M Tris/HCl, pH 7.5, 12% polyethylene glycol 3350, 2% isopropyl alcohol) and protein solution (10 mg/ml in 30 m M Tris/HCl, pH 7.5, 4 m M MgCl 2 , 2 m M dithioerythritol). Crystals appeared after 2 days and grew as plates attached to spherulites to a final dimension of 0.1 (cid:2) 0.5 (cid:2) 0.05 mm in (cid:3) 10 days. The crystals belong to space group P2 1 with unit cell dimensions of a (cid:4) 34.4 Å, b (cid:4)

A hallmark of small GTPases is their ability to undergo structural changes in response to the binding of GDP or GTP. The inactive GDP-bound and the active GTP-bound states are recognized by different partner proteins, thereby allowing the small GTPases to function as molecular switches (1,2). Nucleotide binding, hydrolysis, and localization are regulated by three families of regulatory proteins: guanine nucleotide exchange factors (3)(4)(5), GTPase-activating proteins (6), and guanine nucleotide dissociation inhibitors (7). The GTP-bound state provides a platform for the selective interaction with downstream targets, the so-called effector proteins.
The Rho subfamily of small GTPases regulates diverse cellular processes via their specific effector proteins, which are either serine/threonine protein kinases such as ROCK, PKN 1 / PRK1, and citron kinase or scaffold proteins such as rhophilin, rhotekin, citron, and diaphanous (reviewed in Refs. 8 -10). Among the best-characterized Rho effectors are ROCK proteins. Two ROCK isoforms, ROCKI/ROK␤/p160ROCK and ROCKII/ROK␣/Rho kinase, that share 65% overall identity and 95% homology have so far been identified (11)(12)(13)(14). They were first found as mediators of stress fibers and focal adhesion formation (15), but further studies revealed that ROCK is involved in many other cellular processes including smooth muscle contraction, cell migration, and neurite outgrowth (16,17). Interest in the Rho-ROCK pathway emerges from the fact that abnormal activation of this pathway plays a role not only in tumor invasion and metastasis (18 -20) but also in diseases such as hypertension and bronchial asthma (21,22).
ROCK proteins consist of an N-terminal kinase domain followed by a central putative coiled-coil region, a pleckstrin homology domain, and a cysteine-rich domain at the C terminus ( Fig. 1A) (11,12,23). The Rho-binding domain (RBD) within the predicted amphipathic ␣-helical coiled-coil (Fig. 1A) is responsible for the recognition and binding of the active Rho proteins (13). Coiled-coil structures have also been predicted for the RBDs of most Rho effectors (rhotekin, rhophilin, Citron kinase, and Kinectin) (9). The most recent crystal structure of bovine ROCKII showed that two long helical strands of the RBD formed a parallel coiled-coil dimer (24). A different scenario has been described for the PKN where two adjacent domains at the N terminus of PKN, called homology region (HR)1a and HR1b, have been identified to bind RhoA (25,26). Both domains form anti-parallel coiled-coil (ACC) structures called ACC finger (27,28). HR1a of PKN binds RhoA at two different contact sites (I and II) from which only contact site II overlaps with the switch regions (27). Mutagenesis studies have shown that HR1b is also capable of binding Rac1, another GTPase of the Rho family, at a region corresponding to the contact site I of RhoA (28).
To examine the interaction of ROCK with RhoA, we determined the crystal structure of the complex between the human ROCKI-RBD (residues 947-1015) and the truncated form of human RhoA (residues 1-181) bound to the non-hydrolyzable GTP analogue Gpp(NH)p. Whereas the overall structure of the ROCKI-RBD parallel coiled-coil is distinct from the structure of unbound ROCKII-RBD, the C-terminal part where the interaction with RhoA takes place shares high structural similarity. A short stretch of 13 residues at the C terminus creates a minimal Rho-interacting motif of ROCK by forming a parallel coiled-coil dimer. This motif represents a novel type of interaction with a small GTPase by employing both helices interacting with the switch regions of RhoA in a complementary manner.
Crystallization and Data Collection-Crystals of truncated RhoA (residues 1-181) in complex with the non-hydrolyzable GTP analogue Gpp(NH)p and ROCKI-RBD (residues 947-1015) were grown by vapor diffusion in hanging drops at 20°C by adding equal volumes of reservoir solution (100 mM Tris/HCl, pH 7.5, 12% polyethylene glycol 3350, 2% isopropyl alcohol) and protein solution (10 mg/ml in 30 mM Tris/HCl, pH 7.5, 4 mM MgCl 2 , 2 mM dithioerythritol). Crystals appeared after 2 days and grew as plates attached to spherulites to a final dimension of 0.1 ϫ 0.5 ϫ 0.05 mm in ϳ10 days. The crystals belong to space group P2 1 with unit cell dimensions of a ϭ 34.4 Å, b ϭ 89.5 Å, c ϭ 98.1 Å, and ␤ ϭ 91.9°. For data collection at 100 K, the solution containing the crystals was adjusted to 100 mM Tris/HCl, pH 7.5, 12% polyethylene glycol 3350, 15% glycerol, and 3% isopropyl alcohol. Cryoprotected crystals were then suspended in a rayon loop (Hampton Research) and flash-frozen in liquid nitrogen. X-ray diffraction data were collected on an ADSC Q4 CCD detector at the ID29 beamline at the European Synchrotron Radiation Facility and processed using the programs DENZO and SCALE-PACK (29). Because of the technical reasons, the data set was collected only at a 2.6-Å resolution.
Structure Determination and Refinement-Initial phases were calculated after molecular replacement using the program AMoRe (30) with a search model of RhoA based on the RhoA-PKN structure (27). The R cryst value was 36% after rigid body refinement performed with the program CNS (31), and the resultant initial map showed clear electron density for most of RhoA but only for a part of the RBD molecule. The model was manually built and refined by using the programs O (32) and CNS (31), respectively. Non-crystallographic restraints were applied on two RhoA molecules and on the C-terminal parts of the ROCKI molecules in the asymmetric units. The final model refined to a R cryst ϭ 21.9% and R free ϭ 25.7% contains 69 and 70 residues of the two RBD molecules spanning residues 947-1013 and 947-1014 (with two additional residues at the N termini due to the thrombin cleavage), 179 residues of RhoA (residues 3-181), 2 Gpp(NH)p molecules, 2 magnesium ions, and 55 water molecules (Table I).

RESULTS AND DISCUSSION
Overall Structure-The asymmetric unit of the crystal contains a ␣-helical coiled-coil ROCKI-RBD dimer and two RhoA molecules (Fig. 1B). A 2-fold non-crystallographic symmetry axis runs through the center of the coiled-coil dimer at the interacting interface with RhoA (Fig. 1C). The structure of RhoA⅐Gpp(NH)p is very similar to the GTP␥S-bound RhoA structure alone (33) (with a root mean square deviation (r.m.s.d.) of 0.42 Å for 181 C␣ atoms) and in the complex with PKN-RBD (r.m.s.d. of 0.33 Å for 181 C␣ atoms) (27). The binding of Gpp(NH)p and the coordination of the magnesium ion are conserved.
The ROCKI-RBD forms a 95-Å long ␣-helix that assembles into a parallel coiled-coil ( Fig. 1D) with a total buried solventaccessible surface of 3200 Å 2 . Coiled-coil arrangement of ROCKI-RBD is consistent with the published data on bovine ROCKII-RBD (24,34), which showed that ROCK-RBDs exist as dimers. A periodicity of seven residues over two helical turns called heptad repeat is the characteristic feature of a canonical left-handed coiled-coil. These seven residues are commonly designated as abcdefg (35). The interaction between the helices in a canonical coiled-coil is mediated by the hydrophobic residues at the positions a and d and oppositely charged residues at the positions e and g (35,36). Seven heptad repeats were found along the coiled-coil of ROCKI-RBD between the residues Ile-952 and Arg-1012 ( Fig. 1D) together with three so-called "stut-ters" at the residue positions Met-966, Lys-977, and Ile-981 (marked by red arrows). Stutters are insertions of four residues into the heptad repeat pattern that cause a non-canonical right-handed twist of a coiled-coil (35). The twist of ROCKI-RBD is therefore left-handed at the N-and C-terminal part but right-handed in the central region (residues 965-985) (Fig. 1D). The coiled-coil structure of bovine ROCKII alone (24) reveals a persistent heptad repeat motif without stutters forming a canonical left-handed coiled-coil. Although we cannot exclude that the observed differences in the overall spatial arrangement compared with the human ROCKI are caused by crystal packing, it is very likely that they emerge because of deviations in the N-terminal and central regions of the primary structure of these isoforms. In contrast to these differences, the C-terminal parts of both structures where ROCKI contacts RhoA (see next section), are very similar (Fig. 1, D-F), implicating that this region is structurally conserved. The repetition of hydrophobic residues at the a and d positions within the heptad repeat is not strictly conserved in the middle of the ROCKI-RBD, and as a consequence the structure is less ordered. It is worth noting that three of four amino acids at positions e and g in the C-terminal part are hydrophobic and increase the exposed hydrophobic surface, facilitating the interaction with RhoA (see below).
The RhoA-ROCK-RBD Interface-The interface between the helical ROCKI dimer and one of the RhoA molecules buries a solvent-accessible surface of 1328 Å 2 comparable to the interfaces of Ras-PI3K␥ (37) and Rac-p67 phox (38). The structure of the RhoA⅐ROCKI complex defines a 13-residue left-handed coiled-coil at the C-terminal part of the ROCK-RBD (residues 998 -1010) as the minimal Rho-interacting motif (Fig. 1A). This Rho binding motif is invariant in all of the ROCK proteins from different organisms. Both switch regions of RhoA are involved in the interaction with ROCKI. One of the coiled-coil helices (Fig. 2, B and C, cyan) faces switch I, whereas the second coiled-coil helix (blue) faces switch II. The interface between the RhoA molecules and the RBD dimer is formed by a combination of hydrophobic and electrostatic interactions (Fig. 2, B and C). The interaction between the hydrophobic patches of RhoA (Pro-36, Val-38, Phe-39, Tyr-66, Leu-69, and Leu-72) and ROCK (Ala-1002, Val-1003, Leu-1006, Ala-1007, and Met-1010) 2 is stabilized by a number of electrostatic interactions at their edge (Glu-40/Lys-1005, Asp-65/Lys-999, and Arg-68/Asn-1004). Two polar residues (Tyr-66 and Lys-1005) also partially contribute to the hydrophobic interaction, whereas their hydrophilic parts interact with the main chain carbonyl groups of Leu-998 and Phe-39, respectively. Remarkably, four amino acids of RhoA (Val-38, Phe-39, Asp-65, and Tyr-66) interact with both RBD helices and stabilize their dimeric structure (Fig. 2, B and C). Two key residues (Leu-998 and Lys-1005) of the right helix (cyan) of the ROCK-RBD create a hydrophobic cluster where Leu-998 tightly binds to a pocket formed by Pro-36, Gln-63, Asp-65, and Tyr-66 of RhoA, whereas Lys-1005 contacts Val-38 and Phe-39 (Fig. 2, B and C). In addition, Lys-1005, stabilized by an intramolecular salt bridge with Glu-1008 ( Fig. 2A), mediates two electrostatic interactions with the carboxyl group of Glu-40. Furthermore, Lys-999 and Asn-1004 from the left helix (blue) of the RBD form hydrogen bonds with Asp-65 and Arg-68 of RhoA, respectively.
Our structural data of the binding interface between RhoA and ROCKI nicely match with the mutational analysis of previous studies. Two regions of RhoA (residues 23-40 and 75-92) have been implicated as binding determinants for ROCKI (39). Fig. 2, B and C, shows that the C-terminal part of the first region overlaps with the switch I region that interacts with ROCKI-RBD. Accordingly, the mutations of Phe-39 and Glu-40 have been shown to disrupt ROCKI binding (40,41), supporting the crucial role of these residues in the RhoA-ROCKI interaction described in this study. Unlike the previous study (39), our structure does not confirm a direct involvement of the 2 For better visibility, the ROCK residues are indicated in italics.

FIG. 1. Crystal structure of the RhoA⅐ROCKI⅐RBD complex.
A, schematic view of the ROCKI domain architecture. The RBD (residues 950 -1012) of ROCKI is an integral part of the amphipathic ␣-helix. CRD, cysteine-rich domain; PH, pleckstrin homology domain. The Rhointeracting motif at the C terminus of the RBD-(998 -1010) is depicted in red. B, the crystal-packing diagram shows 4 molecules/asymmetric unit: a ␣-helical coiled-coil of two ROCKI-RBDs (blue and cyan) and two RhoA molecules (gold and beige). C, zoomed top view of the complex between two RhoA molecules (gold and beige) and the Rho-interacting motif of the RBD molecules (blue and cyan) shown in ribbon representation. The bound Gpp(NH)p molecule (black) is shown as a ball-and-stick model. Switch region I and II of both RhoA molecules are highlighted in red and orange, respectively. The 2-fold symmetry axis is depicted as a dot. D, twist of the coiled-coil. The coloring of the ROCKI helices is the same as in B. The twist of the coiled-coil is specified above. Red arrows indicate the location of stutters. The box at the C terminus shows the minimal Rho binding motif. E, the structure of bovine ROCKII-RBD coiled-coil (24) of which only the C-terminal part was superimposed on the C-terminal part of the ROCKI-RBD. The dashed-lined box points at the Rho binding motif. F, detailed comparison of the RhoA binding motifs of ROCKI (blue and cyan) and ROCKII (light green and dark green). The depicted amino acids present the residues of RhoA that contribute to the binding of ROCKI. second region (residues 75-92) in the ROCK-RBD interaction. As the central part of this region is buried in the hydrophobic pocket of both RhoA and Rac1 proteins, it is very likely that substitutions such as F78I or I80M (Rac-Rho chimera) may cause overall structural changes creating or disrupting the GTPase-effector interaction. Mutations of ROCKII at residues Lys-999, Gln-1001, Asn-1004, and Lys-1005, which are all important components of the interface of both ROCK helices, resulted in the loss of RhoA binding (11). It has also been proposed by Shimizu et al. (24) that the C-terminal part of the coiled-coil structure of ROCKII-RBD binds the RhoA molecule. However, the described positive electrostatic potential of this RBD serves only as a long range driving force for initial recognition because the RBD-RhoA interface is predominantly mediated by hydrophobic interactions (Fig. 3B).
Comparison with the RhoA-related Proteins-Whereas RhoA has been the most thoroughly investigated member of the Rho subfamily, the role of the highly related RhoB and RhoC that share over 85% amino acid identity with RhoA has not been completely elucidated (42). It has been shown that ROCKII binds to RhoA as well as to its isoforms RhoB and RhoC (11). Because the region of RhoA contacting ROCK-RBD is identical to the corresponding sequence of RhoB and RhoC, both of these isoforms are likely to interact with ROCKI in a same way as RhoA. On the other hand, Rnd proteins, a GTPase-deficient Rho-related protein subfamily, have been shown to antagonize some effects of the RhoA-induced cytoskeletal reorganization (43,44). Although the Rnd3 structure reveals remarkable similarities in the arrangement of the switch I region, it shows major amino acid and surface deviations at the switch II region compared with RhoA (45,46). The deviation at positions Gln-63, Glu-64, and Asp-65 of RhoA to SPY in Rnd3 (45) is apparently the reason why Rnd3 does not bind the ROCKI-RBD, 3 because both Gln-63 and Asp-65 of RhoA are essential for the interaction with ROCKI (Fig. 2, B and C). Thus, a competition between Rnd3 and RhoA for ROCK binding can be excluded as a possible explanation for the antagonistic effect of Rnd3 on Rho signaling. Most recently, it has been shown that Rnd proteins bind the N terminus of ROCKI close to the kinase domain (47) and thereby prevent Rho-ROCK interaction. Another model proposes a down-regulation of Rho by activation of Rho-specific GAPs via Rnds (48,49).
ROCK Specificity for Rho versus Rac or Cdc42-It has been previously shown that ROCKII binds exclusively RhoA but not Rac1 or Cdc42 (11). The residues involved in ROCK-RBD binding differ among these GTPases only in one position, namely in Glu-40 of RhoA, which corresponds to Asp-38 in Rac1 and Cdc42. Glu-40 interacts with the Lys-1005 of ROCKI and stabilizes its orientation, thus facilitating the hydrophobic contacts to Val-38 and Phe-39 (Fig. 2). Its change to aspartic acid may weaken the interaction with Lys-1005 because of the longer distance between the corresponding electrostatic counter ions. To analyze the corresponding residues on the structural level, we superimposed the GTP␥S-bound RhoA (33), the Gpp(NH)p bound Rac1 (50), and the Gpp(NH)p-bound Cdc42 in complex with RhoGAP (51) with respect to the ROCK-binding regions of RhoA in the presented structure (Fig. 3A). Strik-3 P. Chardin, personal communication. ingly, the peptide backbone and the relative orientation of most side chains are well conserved with the exception of residues Phe-39, Tyr-66, and Leu-69, which strongly contribute to the extended hydrophobic interaction with ROCK as described above. A comparison of the hydrophobic patches on the RhoA, Rac1, and Cdc42 surface shows similarities at positions Pro-36, Val-38, Tyr-66, and Leu-72 in RhoA (Pro-34, Val-36, Tyr-64, and Leu-70 in Rac and Cdc42) (Fig. 3B). However, differences in the orientation of Phe-39 and Leu-69 in RhoA (Phe-37 and Leu-67 in Rac and Cdc42) are rather significant and probably contribute to the specificity of Rho proteins for ROCK by affecting the overall charge, hydrophobic, and shape complementarity.
Comparison with Other GTPase Effector Structures-Although the proposed RhoA contact site I of PKN consists of the ␣ 1 helix (amino acids 25-28), ␤ 2 and ␤ 3 strands (amino acids [43][44][45][46][47][48][49][50][51][52][53][54], and the C-terminal ␣ 5 helix (164 -172) (Fig. 4), the contact site II described as a symmetry-related contact site for the ACC finger (27) overlaps remarkably well with the ROCKIbinding site described in this study (Figs. 2D and 4). With the exception of four additional residues (Asn-41, Trp-58, Ser-73, and Asp-76), PKN interacts with RhoA via the same amino acids as shown here for ROCKI-RBD (Figs. 2, B and D, and 4). Interestingly, the general type of interaction, i.e. the contact of complementary hydrophobic patches stabilized at the edge by electrostatic interactions, is highly conserved. The structure of ROCK-RBD is similar to the structure of PKN-RBD (27) in the respect that they form a coiled-coil, but the amino acids of PKN that contact RhoA at this site do not correspond to those found in ROCK-RBD (Figs. 2, B and D, and 4). This structural type of a GTPase-binding domain represents a new and clearly distinct category of target proteins in comparison with other known effector protein structures of the Rho family. The CRIB (Cdc42/ Rac-interactive binding) motif of Cdc42-associated tyrosine kinase (52), Wiskott-Aldrich syndrome protein (53), and p21activated kinase (54), for example, is a largely extended structure containing a short region of anti-parallel ␤-strands.
Implications for Rho-mediated Effector Activation-Effector  activation by small GTPases has been most extensively studied on PI3K, Raf kinase, and PAK. For the Ras-PI3K interaction, it has been implicated that the recognition is provided by the interaction of the Ras-binding domain of PI3K with the switch I region of Ras, whereas the activation is achieved by the binding of the kinase domain to the switch II region (37). A similar model has been suggested for the Raf kinase activation by Ras. The interaction between the switch I region of Ras and the Ras-binding domain of the Raf kinase may provide the initial binding process recruiting Raf to the plasma membrane (56,57). This process is probably followed by the binding of the cysteine-rich domain (CRD) of the Raf kinase to the switch II region, which may result in Raf kinase activation (reviewed in Ref. 58). Although the mechanism of Raf kinase activation by Ras is still controversial, the Cdc42-PAK1 interaction has been suggested to induce a conformational change, resulting in the dissociation of the PAK1 dimer and subsequently in autophosphorylation at several sites that prevent the kinase domain from reverting to the inactive conformation (59 -62). The structures of an autoinhibited ␣-PAK and the Cdc42⅐PAK-CRIB complex support this model (54,61).
The hydrophobic interaction of Cdc42 via Leu-67 and Leu-70 with PAK, ACK, and WASP, respectively, has been suggested to provide a common mechanism for activation of these effectors (54). A similar mechanism can be hypothesized for the activation of Rho effectors because the corresponding leucines in RhoA, Leu-69 and Leu-72, are strongly involved in hydrophobic interaction with ROCK (Leu-1006, Ala-1007, and Met-1010 in Fig. 2, B and C) and PKN (Leu-52, Ala-56, Leu-59, Val-73, and Leu-76 in Fig. 2D). The I1009A mutation has been previously defined as a dominant negative mutant of ROCKI that abolishes ROCKI-RBD interaction with RhoA (13) and inhibits Rho-induced formation of focal adhesions and stress fibers (63). Ile-1009 is not directly involved in the Rho-ROCK interaction but is adjacent to Leu-1006 and Met-1010 ( Fig. 2A). Its substitution may destabilize the Rho-interacting coiled-coil motif and consequently the complex formation with RhoA. The dominant negative effect of this mutant can be explained by the following model. The ROCK I1009A mutant when overexpressed in cells sequesters the endogenous ROCK (available at much lower concentration) by dimerization and the formed heterodimer is apparently no longer activated by RhoA. It is important to note that dimerization and autophosphorylation have been suggested to be major events for the regulation of these kinases (34). CONCLUSIONS The presented structure highlights fundamental principles of small GTPase interaction with effectors. The essential determinants of the GTP-dependent interaction with effectors are the two switch regions. Although nearly all of the effectors interact with switch I and switch II, they exhibit an amazing diversity in their structures. However, RhoA seems to interact preferably with coiled-coiled domains. The structure described here represents a novel interaction type of a small GTPase with a parallel ␣-helical coiled-coil dimer. We could show that the 160-kDa protein kinase ROCKI employs only 10 residues of a 13 amino acid stretch at the C-terminal part of the coiled-coil to bind a predominantly hydrophobic patch assembled by the switch regions of RhoA. Although the presented structure of ROCKI and the recently published structure of ROCKII alone do not form coiled-coils with the same twist, their C-terminal parts are structurally very similar. The specificity of ROCKI interaction with RhoA versus Rac1 and Cdc42 appears to be determined by the residues Phe-39, Tyr-66, and Leu-69 of RhoA. They exhibit a different orientation in RhoA concerning shape, charge, and hydrophobic complementarity with ROCKI.
Our structure also suggests that the contact site between PKN-HR1a and RhoA described as the symmetry-related site in the structure of RhoA⅐GTP␥S⅐PKN complex is most likely the contact site II. However, important questions regarding the structural rearrangement of full-length ROCKI upon interaction of the RBD with RhoA and the implication of multiple Rho binding sites in ROCK activation remain to be elucidated.