SH3 domain regulation of RhoGAP activity: Crosstalk between p120RasGAP and DLC1 RhoGAP

RhoGAP proteins are key regulators of Rho family GTPases and influence a variety of cellular processes, including cell migration, adhesion, and cytokinesis. These GTPase activating proteins (GAPs) downregulate Rho signaling by binding and enhancing the intrinsic GTPase activity of Rho proteins. Deleted in liver cancer 1 (DLC1) is a tumor suppressor and ubiquitously expressed RhoGAP protein; its activity is regulated in part by binding p120RasGAP, a GAP protein for the Ras GTPases. In this study, we report the co-crystal structure of the p120RasGAP SH3 domain bound directly to DLC1 RhoGAP, at a site partially overlapping the RhoA binding site and impinging on the catalytic arginine finger. We demonstrate biochemically that mutation of this interface relieves inhibition of RhoGAP activity by the SH3 domain. These results reveal the mechanism for inhibition of DLC1 RhoGAP activity by p120RasGAP and demonstrate the molecular basis for direct SH3 domain modulation of GAP activity.

as a gene deleted in multiple primary hepatocellular carcinoma (HCC) tumors and liver cancer cell lines 15 and upon cloning was observed to be homologous to the rat p122RhoGAP 16 . Deletion of DLC1 is embryonic lethal, and thus its critical role in cytoskeletal organization was established 17 . DLC1 has been shown to act on Rho family GTPases RhoA, RhoB, and RhoC, and to a lesser extent on Cdc42, Rac1 and TC10 7,9 . In addition to the RhoGAP domain located in their central region, each of the DLC proteins contain a sterile alpha motif (SAM) domain in their N-terminus, and a steroidogenic acute regulatoryrelated lipid transfer (START) domain near their C-terminus ( Fig. 1a 11,15 ). A serine-rich linker region (LR) between the SAM and RhoGAP domains is also present and includes a focal adhesion targeting (FAT) region. Localization of DLC proteins and regulation of the RhoGAP domain are thought to be mediated by the accessory domains 9,18 , and by partner proteins, including tensin proteins, focal adhesion kinase and talin [19][20][21][22][23][24] . However, the molecular mechanisms of regulation and targeting remain poorly understood.
One major binder and regulator of DLC1 is the RasGAP protein, p120RasGAP (RASA1; RasGAP, p120GAP). This was the first RasGAP identified and is a key regulator of Ras signaling [25][26][27][28] . p120RasGAP is a multidomain protein containing two Src homology-2 (SH2), one Srchomology-3 domain (SH3), a Pleckstrin homology (PH) domain, presumptive calcium-binding C2 domain, and a RasGAP domain (Fig. 1a). It is recruited to phosphotyrosine containing proteins by its SH2 domains, and the PH and C2 domains are thought to mediate membrane localization 29 ; however, the role of the SH3 domain is noncanonical, as it has not been demonstrated to bind to poly-proline (PxxP) motifs 30,31 . Instead, the SH3 domain was first observed by yeast two hybrid to bind the DLC1 RhoGAP domain and suppress its RhoGAP activity 32 , but the molecular mechanism was not determined 33 .
The roles and divergence of Src homology domains away from their canonical features continue to be identified [34][35][36] , including the recent discovery of a unique mode of phosphotyrosine binding by one of the SH2 domains of p120RasGAP 37 . We, therefore, hypothesized that the interaction of the p120RasGAP SH3 domain with the DLC1 RhoGAP domain may represent an atypical mode of SH3 binding partner recognition. Furthermore, direct regulation of a GTPase activating protein by an SH3 domain, one of the most frequent domain 'modules' in the human genome with over 330 sequences 38 , may represent a previously under-appreciated mode of control for GTPase signaling cascades. We, therefore, determined the co-crystal structure of DLC1 RhoGAP domain in complex with p120RasGAP SH3 domain. This structure of a RhoGAP-SH3 interaction reveals that the RhoGAP domain cradles the SH3 domain in a manner competitive with its binding to RhoA. Nonetheless, the binding sites encompass different regions of the RhoGAP domain, allowing site directed mutagenesis in the SH3 binding site to selectively disrupt DLC1 suppression by p120RasGAP without impinging on RhoGAP activity. Overall, this study reveals the molecular basis of a noncanonical role for SH3 domains in regulation and signaling of small GTPase signaling cascades.

Results
Co-crystal structure of p120RasGAP SH3 domain with DLC1 RhoGAP domain To reveal the molecular basis of SH3 binding and suppression of GAP activity, we sought a co-crystal structure of p120RasGAP SH3 domain and DLC1 RhoGAP domain. We recombinantly expressed the isolated DLC1 RhoGAP and p120RasGAP SH3 domains, and copurified the complex. We determined the co-crystal structure to 3.2 Å resolution ( Fig. 1 and Table 1), and observe clear electron density throughout, including the key residues that mediate the protein-protein interaction ( Supplementary Fig. 1). There are 4 copies of the DLC1-p120RasGAP complex found in the asymmetric unit, and they are experimentally identical (RMSD between 0. 3  ( Supplementary Fig. 2a). This structure thus provides an experimentally determined molecular-level structure describing an interaction between a RhoGAP and SH3 domain.
The DLC1 RhoGAP domain (residues 1075-1240) adopts a conserved overall structure that is highly similar to the GAP domain of other RhoGAP proteins, including the previously determined structure of the isolated DLC1 RhoGAP domain (PDB ID: 3KUQ 39 ). Its closest structural neighbor is the GAP domain of p50RhoGAP (PDB ID: 1TX4 40 , RMSD 2.0 over 188 Cα atoms, 27% identity, Dali server, Supplementary  Fig. 2b). The DLC1 RhoGAP domain consists of nine helices, termed αA0, αA, αA1, αB, αC, αD, αE, αF, and αG according to the nomenclature of Barrett et al. 41 . At the core of the domain is a four-helix bundle of αA-αB and αE-αF. αA-αB is flanked by αA1 and capped by αA0. αC-αD form a hairpin that protrudes from the four-helix bundle and, together with αE-αF, forms a V-shaped groove for αG. The long FG loop (residues 1234-1259), which is of variable length, sequence and structure in the RhoGAP family 7 , is partially disordered in this structure (Fig. 1b) and is located adjacent to the solvent channel in all four copies. In GAP proteins, the critical catalytic residue is a highly conserved arginine in the αA-αA1 loop (or the "finger loop"), termed the arginine finger 5,6 . In DLC1 RhoGAP this residue is Arg-1114. Comparison with the unpublished structure of apo DLC1 RhoGAP (PDB ID: 3KUQ 39 ) shows only minor conformational changes (RMSD value of 0.7 Å over 186 Cα atoms), potentially due to crystal packing and distal from the SH3 binding site (residues 1130-1134 in αA1-αB located approximately 20-30 Å from the SH3 binding site, Supplementary Fig. 2c).
The p120RasGAP SH3 domain (residues 281-341) adopts a beta sandwich fold comprised of five strands (β1-β5) that form a β-barrel. The strands are connected by loops termed the RT loop (β1-β2), n-Src loop (β2-β3), distal loop (β3-β4), and a 3/10 helix (β4-β5) (based on Src nomenclature 42 ). Several previous structure determinations of the apo form of p120RasGAP SH3 domain have been reported: PDB IDs: 2J05 and 2J06 43 , 4FSS (crystal structure, unpublished), 2GQI (NMR, unpublished), and 2M51 (NMR, unpublished). The structure here superposes well with these previous structures with RMSD values ranging from 0.4 to 0.8 Å over 58 equivalent Cα positions ( Supplementary Fig. 2d). We observe no material conformational change in either the SH3 domain or RhoGAP domain upon complex formation when compared to the previous apo structures. Therefore, the binding mode can be defined as 'lock-and-key' type rather than as an 'induced fit' interaction.
The complex between the SH3 domain of p120RasGAP and the RhoGAP domain of DLC1 represents a noncanonical mechanism for SH3 domain recognition of binding partners, and for RhoGAP domains to be inhibited. The p120RasGAP SH3 domain latches onto the DLC1 RhoGAP domain using its RT and n-Src loops and additional contacts are made between these latches by residues in the β2, β3 and β4 strands (Fig. 1b). On the RhoGAP domain, the bulk of the interface comprises residues on the solvent-accessible faces of αF and αG, with several more contributions from the αA-αA1 loop, αB, and the FG loop (Fig. 1b). Mapping the interaction regions on the surface of the two proteins reveals a single continuous binding site that buries a total surface area of 1550 Å 2 comprising approximately 8% of the DLC1RhoGAP surface and 20% of the SH3 surface, with a shape complementarity score of 1.00 (Pisa server 44 , Fig. 2a). The interface is largely hydrophobic, yet also involves several negatively charged residues from SH3 and positively charged surface on RhoGAP.

Details of binding interface
The RT loop of the p120RasGAP SH3 domain appears critical to the interaction and makes contact along the length of the loop (Fig. 2b). The most notable feature of the RT loop binding is the direct interaction with the catalytic arginine finger residue Arg-1114 in DLC1 RhoGAP, located in the αA-αA1 loop. Specifically, Arg-1114 forms a salt bridge with the sidechain of Glu-298 and a hydrogen bond with the sidechain of Tyr-290, and additional van der Waals interactions with Val-293 in the RT loop and Trp-317 at the start of β3. At the tip of the RT loop, Asp-295 makes a salt bridge to Arg-1155 at the end of αB, the carbonyl of Thr-296 hydrogen bonds to Thr-1238 in the FG loop, and the sidechain of Asp-297 hydrogen bonds with the sidechain of Asn-1266 in αG (Fig. 2b, inset). Nearby, there are additional van der Waals interactions contributed by Val-1227, Pro-1231, and Leu-1236. In addition to its RT loop interactions, the SH3 domain n-Src loop residues Glu-314, Asp-315 and Trp-317 interact with DLC1 residues Thr-1221, Pro-1222 and Thr-1223, with a hydrogen bond forming between the sidechains of Asp-315 and Thr-1223 (Fig. 2c). Also, the β2-β3-β4 strands of p120RasGAP SH3 domain bind DLC1 along one face of αG (Fig. 2d). At the center of this interface are Leu-1263 and Leu-1267 of DLC1, which bind the hydrophobic region created by SH3 Trp-319, Leu-313 and Asn-311. Gln-1260 and Gln-1271 bracket this interface, with Gln-1260 forming a hydrogen bond to Gln-328. Taken together, there are extensive interactions between the p120RasGAP SH3 domain and DLC1 RhoGAP domain to create a large interface.

SH3 domain binding site overlaps RhoA binding site on DLC1
Previous work demonstrated that the p120RasGAP SH3 domain inhibits RhoGAP activity of DLC1 32,33 . We asked whether this inhibition is due to interference with the Rho GTPase binding site. There is currently no experimentally determined three-dimensional structure of the complex between the RhoGAP domain of any DLC protein and a GTPase. However, a consensus interface between RhoGAP and Rho GTPases has been postulated 7,45 based on highly conserved features observed in structures of Rho GTPase-RhoGAP complexes (e.g., RhoA-p50RhoGAP 40 and RhoA-p190RhoGAP-A 7 ). Because the RhoGAP domain of DLC1 is highly similar to other RhoGAPs (e.g., RMSD 1.6 Å over 179 equivalent Cα positions when compared to p190RhoGAP, PDB ID: 5IRC 7 ), we can compare the SH3 binding site with the predicted interface between DLC1 and RhoA: the RhoA binding site on DLC1 is predicted to be highly similar to that of p190RhoGAP (Fig. 3a). Comparison of the Rho binding site on p190RhoGAP with our experimentally determined SH3 binding site on DLC1 reveals that these sites partially overlap (Fig. 3b). This strongly suggests that DLC1 Rho-GAP cannot bind both RhoA and SH3 simultaneously, but instead supports that p120RasGAP SH3 can directly compete with RhoA for DLC1 binding. We, therefore, hypothesize that SH3 inhibits the DLC1 RhoGAP activity by direct competition with RhoA for interaction with DLC1. In more detail, the location of the SH3 interaction site suggests functional importance. It is well established that RhoGAP domains bind the GTP-loaded (active) conformation of Rho GTPases at a binding site that includes the switch I and switch II regions 4,40 . Our analysis suggests that the SH3 domain RT loop overlaps with the RhoA switch I and switch II regions, indicating that the most extensive region of SH3 domain recognition corresponds with the most functionally relevant region of RhoA recognition (Fig. 3c). Notably, the arginine finger of the DLC1 RhoGAP domain is critical to both RhoA and SH3 binding (Fig. 3c).

SH3 domain inhibits DLC1 RhoGAP activity in vitro
To validate the molecular basis of p120RasGAP inhibition of DLC1 we conducted biochemical GTPase activity assays. We established an in vitro RhoA GTPase activity assay that utilizes the malachite green colorimetric reagent to directly monitor production of inorganic phosphate 46 . Similar to assays presented for Ras 47 , this method has the advantage of directly monitoring the production of inorganic phosphate by hydrolysis without the need for a fluorescently tagged GTP analog. This assay also avoids any steric interference by a fluorescent moiety. DLC1 displays preference for RhoA, RhoB and RhoC over the other Rho GTPase family members 7 ; therefore, to assess GAP activity we assessed DLC1 activity on RhoA hydrolysis. Additionally, to allow for nucleotide exchange and RhoA to perform multiple hydrolysis cycles we performed these assays in the presence of both EDTA and magnesium.
Our initial assays demonstrated that the inefficient intrinsic GTP hydrolysis by purified RhoA is catalytically enhanced in the presence of increasing concentrations of DLC1 RhoGAP domain (Fig. 4a). In contrast, the Arginine-finger mutant R1114A of DLC1 RhoGAP does not stimulate hydrolysis (Fig. 4a). We next assessed the inhibition of DLC1 RhoGAP activity by the p120RasGAP SH3 domain. We performed a titration of SH3 domain from 0 to 20 μM in an assay containing 5 μM RhoA and 0.05 μM DLC1 RhoGAP. We observe a dose-dependent inhibition of DLC1 RhoGAP activity by the SH3 domain (Fig. 4b). Plotting the phosphate generation versus log p120RasGAP SH3 concentration allowed us to calculate an IC 50 of 0.6 μM (95% confidence interval 0.48-0.83 μM) ( Table 2). Although not a direct measurement of affinity, this calculated IC 50 value is in remarkably close agreement with the previously reported dissociation constant (K d ) of 0.6 μM measured by isothermal titration calorimetry and the K i of 0.6 μM determined by kinetics 33 .
The co-crystal structure of DLC1 RhoGAP domain in complex with p120RasGAP SH3 domain reveals a direct interaction of the Arginine finger Arg-1114 with SH3 residues Tyr-290 and Glu-298 (Fig. 2b). To test the importance of this interface, we generated single point alanine mutant SH3 domains Y290A and E298A and tested their ability to inhibit DLC1RhoGAP in a dose-dependent manner. Both mutant SH3 domains demonstrate reduced inhibition of DLC1 (Fig. 4b, c) with calculated IC 50 values 4-fold and over 30-fold lower for the Y290A and E298A mutants, respectively (Table 2). These inhibition assays therefore support the structural observation that the SH3 domain interacts directly with the DLC1 RhoGAP domain via these residues.
Since the p120RasGAP SH3 domain is closely flanked by SH2 domains (Fig. 1a), we next asked whether inhibition of DLC1 by the SH3 domain is affected by these flanking SH2 domains by titrating increasing amounts of p120RasGAP SH2-SH3-SH2 or SH3 domain protein and monitoring the phosphate release by the BIOMOL green assay (Enzo Life Sciences) which is complementary to the Malachite green assay used above. We plotted the phosphate release versus p120RasGAP concentration (Fig. 4d) and determine similar IC 50 values of the SH2-SH3-SH2 domain and SH3 domain of 2.9 and 1.7 μM, respectively ( Table 2). We also observe that mutation of the crystallographically-defined interface residue Glu-298 (Fig. 2b) in the context of the SH2-SH3-SH2 protein abrogates the inhibition (Fig. 4d and Table 2). Thus, the presence of the flanking SH2 domains does not strongly alter the inhibitory activity of the SH3 domain in p120RasGAP.

DLC1 mutants are not inhibited by p120RasGAP
Our structural analysis of the complex between DLC1 and p120RasGAP, and comparison with other RhoGAP co-crystal structures in complex with Rho family small GTPases (Fig. 3a, b) suggest different binding footprints for p120RasGAP SH3 domain when compared to small GTPases. On closer inspection, in αF of DLC1 the Thr-1223 sidechain forms a hydrogen bond to Asp-315 of the p120RasGAP SH3 n-Src loop and contacts Asp-314, Asp-315 and Trp-317 by van der Waals interactions (Fig. 2c). Similarly, in αG of DLC1 the Leu-1267 sidechain of is at the center of a hydrophobic patch and interacts with SH3 domain residues Asn-311, Leu-313 and Trp-319 (Fig. 2d). Comparison of the locations of these residues with the expected small GTPase binding footprint suggests that their point mutation may result in disruption of p120RasGAP binding but not of RhoA (Figs. 3a, b and 5a)   in a loss of inhibition by SH3 but leaving the RhoGAP activity intact. We, therefore, introduced aspartate mutations at these positions, T1223D and L1267D, and assessed GTP hydrolysis in the malachite green assay. We find that suppression of RhoGAP activity by p120RasGAP SH3 domain is lost for both T1223D and L1267D mutant DLC1 RhoGAP proteins (Fig. 5b), although both mutant DLC1 proteins maintain their GAP activity toward RhoA in the absence of SH3 domain ( Fig. 5b, 0 μM SH3 samples). These data reflect that these mutants interrupt the binding of p120RasGAP SH3 domain but do not impact activity towards RhoA. The co-crystal structure of DLC1 RhoGAP domain in complex with p120RasGAP domain, therefore, describes a binding site that overlaps with the Rho binding site, but that can be manipulated to selectively disrupt RhoGAP inhibition by p120RasGAP.
Lastly, we asked whether the p120RasGAP SH3 domain can inhibit its own RasGAP domain by measuring GTP hydrolysis by H-Ras in the presence of the isolated RasGAP and SH3 domains, and find that Ras-GAP, which stimulates the intrinsic hydrolysis of H-Ras, is not inhibited by excess amounts of SH3 domain (Fig. 5c). Thus, p120RasGAP does not employ an SH3/RasGAP interface as a mode of autoinhibition.

p120RasGAP has an atypical SH3 domain that binds RhoGAP domain
The SH3 domain of p120RasGAP is unusual compared to canonical SH3 domains, and its interaction with DLC1 RhoGAP domain may help explain its non-canonical features. Canonical SH3 domains exhibit canonical binding activity toward left-handed PxxP (x is any amino acid) polyproline II (PPII) helical peptides [48][49][50] . These SH3 domains contain three binding pockets on their surfaces between the RT and n-Src loops: two shallow xP dipeptide binding sites, and a third acidic specificity site that binds a positively charged peptide residue and helps determine the directionality of peptide binding 31  xP sites (xP1 and xP2) are comprised of aromatic residues in the RT loop, β4, and 3/10 loop and are conserved in most PxxP-binding SH3 domains (Fig. 6a-c). Previous crystal structures of p120RasGAP SH3 domain revealed divergent structural features at the putative xP binding sites compared to canonical SH3 domains 43 , and our co-crystal structure confirms this divergence. These structures suggest that p120RasGAP may not bind canonical polyproline-containing ligands since Leu-288 and Leu-335 replace aromatic residues in xP1, and Val-332 replaces a typical Proline in xP2 (Fig. 6a, b). Additionally, we observe that Asp-334, which is often an uncharged Proline or Asparagine residue in canonical SH3 domains (Fig. 6a, b), partially obscures the xP2 site (Fig. 6b, d). In support of these structural analyses that suggest a lack of a canonical binding site, no polyproline-containing ligand has been identified for p120RasGAP SH3 domain 30,31 . Thus, if a yet-unidentified atypical peptide ligand for p120RasGAP SH3 exists, the mode of binding would be expected to diverge significantly from typical SH3peptide interactions.
Nonetheless, despite the lack of conserved xP1 and xP2 sites, the p120RasGAP SH3 domain contains a conserved acidic residue in the RT loop, Glu-298, that typically forms the specificity pocket in SH3 domains (Fig. 6a, d) 31,35 . Interestingly, in the co-crystal structure of p120RasGAP with DLC1 we find that Glu-298, together with Tyr-290, which is a conserved aromatic residue in the xP2 binding pocket, directly coordinate the arginine finger residue Arg-1114 of RhoGAP (Fig. 2b, d). Trp-317, which can contribute to the xP2 pocket, specificity pocket, or both 35 , also directly contacts DLC1. This potentially suggests a degradation of polyproline binding function, but retention and repurposing of the specificity pocket to recognize the critical arginine finger of DLC RhoGAP proteins.

Discussion
DLC1 is a potent tumor suppressor which is deleted in many human tumors 10,53 . Its loss is associated with enhancement of both cell proliferation and migration, while its overexpression leads to reduction of cell growth and motility 9,54,55 , suggesting that intricate control of its activity is required for normal cell growth. Autoregulation of DLC1 involving the SAM and/or SMART domains has been demonstrated 9,18 , and both localization 56 and phosphorylation 57-59 may be similarly involved. Importantly, higher order protein complexes form 24 , and binding of partner proteins is implicated in DLC1 regulation (reviewed in ref. 60). It is the direct regulation of DLC1 by p120RasGAP that we assess at the molecular level in this study. We determined the cocrystal structure of the SH3 domain of p120RasGAP in complex with the RhoGAP domain of DLC1. The structure demonstrates a previously underappreciated role of SH3 domains as direct modulators RhoGAP proteins and illustrates a mode of binding that is competitive with the interactions of the RhoGAP domain with small GTPases. Our structure and analysis highlight several features that are distinct for the recognition of DLC1 by p120RhoGAP, provide insight into DLC1 regulation, and raise the possibility of alternate regulation mechanisms for small GTPases. The inhibition of DLC1 RhoGAP activity by p120RasGAP SH3 domain demonstrates a clear and direct mechanism by which crosstalk between the Rho and Ras pathways can occur via interaction of their cognate regulatory proteins 33 . The multiple modular domains in both the RhoGAP and RasGAP protein families have been shown to act as scaffolds that promote pathway crosstalk 8,61 , and the interaction of DLC1 with p120RasGAP provides one such intersection point that may affect the balance of Rho vs Ras signaling. In general, crosstalk between these critical small GTPase pathways can play key roles in modulating cell proliferation, spreading, migration and growth, and in cancer, and multiple nodes have been identified that impact both Ras and Rho pathways (reviewed in refs. 62, 63). Our co-crystal structure demonstrates a molecular level basis for direct Ras pathway impingement on Rho signaling cascades. To achieve this, the binding site of the p120RasGAP SH3 domain and its recognition of DLC1 is unique compared to other known typical and atypical SH3 binding sites and RhoGAP regulation mechanisms.
Furthermore, the question remains whether the balance between the Ras and Rho pathways via p120RasGAP may be dysregulated in transformed cells that lack DLC1. DLC1 is an important tumor suppressor that is mutated in many cancers, nearly as frequently as the commonly known p53 tumor suppressor 53 . Targeting of DLC1 activity via p120RasGAP in cancer cells has been the subject of at least one study 64 ; however, future studies will be necessary to help better define the spatio-temporal nature of the interaction between these proteins and whether cellular transformation by loss of DLC1 involves Ras signaling via p120RasGAP.
High sequence similarity between DLC1 and its family members DLC2 and DLC3 suggests similar mechanisms of regulation across the family: the full-length sequences display identities of 54 and 47%, for DLC1 vs DLC2 and DLC1 vs DLC3, respectively, and RhoGAP domain sequence identities of 81 and 70%, respectively. There are few sequence insertions in the RhoGAP domains, and the arginine finger residues are in a conserved location (Supplementary Fig. 3). Nonetheless, previous studies found that DLC2 and DLC3 are suppressed to a lesser extent by p120RasGAP 33 , implicating that DLC1 is unique in its ability to be potently inhibited by the SH3 of p120RasGAP. Analysis of the conservation of residues of DLC1 that bind p120RasGAP suggests that these are only partially conserved in DLC2 and DLC3 (Supplementary Fig. 3). Importantly, the two residues we mutated to abolish binding to p120RasGAP SH3, Thr-1223 and Leu-1267, are not well conserved in DLC2 (Met-808 and Leu-851) and DLC3 (Gly-716 and Met-760). Previous studies and structural analysis together imply that the modulation of RhoGAP activity by p120RasGAP may be favored toward DLC1, suggesting isoform-specific effects on signaling cascades downstream of different cellular stimuli. Thus, DLC1 may be unique compared to its related family members and likely among the large RhoGAP family in its binding and regulation by the SH3 domain of p120RasGAP.
Many RhoGAP proteins are multidomain proteins, and regulation of their RhoGAP activities may involve intramolecular inhibition via direct interaction in cis of these modular domains with the RhoGAP domain 65 . For example, it has been postulated that in the Graf RhoGAP family including Graf1 and oligophrenin, the N-terminal BAR (Bin/ amphiphysin/Rvs) domain autoinhibits the RhoGAP activity by binding and masking the RhoGAP active site 66,67 . Similarly, p50RhoGAP is thought to be autoinhibited by binding of its BCH domain to the RhoGAP active site 68 . However, molecular level details of these interactions are currently unobserved. Additionally, RhoGAPs may be inhibited in trans by a binding partner, like DLC1 by p120RasGAP, but the details of regulation are unknown. Our structure, therefore, provides a molecular level observation of inhibition of a RhoGAP domain by direct masking of the catalytic arginine finger. Whether other RhoGAP proteins are regulated by masking of the arginine finger by direct protein binding remains to be revealed.
Our structure also represents a noncanonical role for an SH3 domain binding site: to regulate a RhoGAP protein. Previous studies have demonstrated that SH3 domains can mediate non-canonical interactions (reviewed in refs. 31 and 35, 69). These atypical interactions can involve secondary structure elements that extend beyond the canonical polyproline peptide (e.g., Crk, Fig. 7a, PDB ID: 1CKA 70 ), or as observed for the p67phox SH3 binding p47phox via both polyproline and a unique helix-turn-helix structure that binds the specificity site (PDB ID: 1K4U 71 , Fig. 7b). Alternatively, the polyproline motif can present as part of a folded domain, as observed in the HIV-1 Nef protein core domain bound to the Fyn SH3 domain (PDB ID : 1EFN 72 , Fig. 7c). Additionally, there are SH3 interactions that are devoid of typical polyproline peptide contacts and instead are mediated solely by tertiary contacts with other protein domains. Examples of these proteinprotein interactions include the SH3 domain of SlaI (yeast), CIN85 and amphiphysin, which bind ubiquitin at the polyproline binding sites (PDB ID: 2JT4 73,74 , Fig. 7d), the SH3 domain of 53BP2, whose polyproline site is partially masked by its own ANK repeat region, binding the core domain of p53 (PDB ID: 1YCS 75 , Fig. 7e), and the Fyn SH3 domain which binds the SAP SH2 domain near the specificity pocket but away from the polyproline binding site (PDB ID: 1M27 34 , Fig. 7f). The atypical selectivity of these SH3 domains is generally associated with the variable lengths and sequences of n-Src and RT loops in the SH3 domain family 31 . We now add to this growing list the p120RasGAP SH3 domain, which uses an extended surface that includes the specificity site to bind to DLC1 RhoGAP; interestingly, the recognition and cradling of an SH3 domain by a catalytic domain is thus far unique to DLC1 and p120RasGAP (Fig. 7g).
It remains to be seen if other SH3 domains can directly regulate RhoGAP activity, but a previous study suggested that the SH3 domains of Src, Crk, and Nck do not act as inhibitors of DLC1 activity 33 . Further work will be needed to identify other potential SH3-RhoGAP regulatory pairs. Interestingly, an SH3 domain in the RacGEF Asef binds and inhibits its Dbl homology domain and blocks the Rac1 binding site (PDB ID: 2DX1 76 ), and this regulation mechanism may be shared among other SH3-domain-containing GEFs like collybistin I and intersectin-1. Thus, the SH3 domain fold may play an expanded role in the regulation of Rho family GTPase pathways.
Molecular modeling of protein structures is becoming more important. Deep learning models of the molecular structure of DLC1 RhoGAP domain by Alphafold (AF-Q96QB1-F1-model_v2.pdb 77 ), while correctly predicting the overall fold have not adequately predicted the fine details of the structure. Alphafold presents helix αG in a position that is slightly extended away from the RhoGAP core, similar to other RhoGAP structures. In contrast, both experimentally determined crystal structures of DLC1 RhoGAP domain (our co-crystal structure here, and the previously unpublished apo structure (PDB ID: 3KUQ 39 )) reveal that this helix is packed closer against the RhoGAP domain. This compaction allows interaction with the SH3 domain, whereas the protruded position it creates steric clashes with the SH3 domain. (Supplementary Fig. 4). Perhaps future deep learning models might help reveal additional SH3-RhoGAP binding partners based on structural similarities with DLC1. Similarly, a previously reported triple mutant p120RasGAP SH3 domain (N311R/L313A/ W319G) was shown to impair both interaction with and inhibition of DLC1RhoGAP 33 . These mutants were based on a PatchDock docking model and comparison with the co-crystal structure reveals an approximately 180°rotation in the SH3 domain, explaining why this previous study did not achieve full loss of inhibition by point mutagenesis. While molecular modeling is advancing, experimental studies remain paramount.
In summary, the co-crystal structure of DLC1 RhoGAP domain in complex with the SH3 domain of p120RasGAP reveals a mechanism by which Ras and Rho pathways can crosstalk, and a previously unobserved molecular mechanism by which SH3 domains impinge on GTPase signaling cascades.

Protein purification
Cells expressing His 6 -DLC1, His 6 -H-Ras, His 6 -p120RasGAP SH2-SH3-SH2 or GAP domain protein were resuspended in lysis buffer containing 50 mM HEPES pH 7.3, 500 mM NaCl and 20 mM imidazole, and cells expressing GST-SH3 or GST-RhoA were resuspended in lysis buffer containing 20 mM Tris pH 7.4, 150 mM NaCl and 1 mM dithiothreitol. Cells were lysed by three freeze/thaw cycles in the presence of 0.1 mg/ml lysozyme followed by sonication (4 min total, 25% amplitude, QSonica). Total cellular lysate was clarified by centrifugation at 5000 × g at 4°C for 1 h. His 6 -tagged DLC1 RhoGAP, HRas, p120RasGAP SH2-SH3-SH2 or GAP domain proteins were purified by applying clarified lysate to nickel-nitrilotriacetic acid (Ni-NTA) agarose beads (Qiagen) for 1 h, which were then washed with lysis buffer prior to step-wise protein elution with lysis buffer containing increasing concentrations of imidazole (40 mM-100 mM-250 mM-500 mM). Fractions containing protein of interest were pooled and applied to size exclusion chromatography (Superdex75 Increase) column equilibrated in buffer containing 20 mM Tris pH 8.0, 150 mM NaCl. The final protein was concentrated in a spin concentrator (Amicon, MilliporeSigma). H-Ras and p120RasGAP GAP domain proteins were additionally purified by anion exchange chromatography (Mono Q 5/50 GL Cytiva) in 20 mM Tris pH 8 and eluted by a gradient of NaCl from 0.05 to 1 M. DLC1 RhoGAP was used in its His 6 -tagged form for both crystallization and biochemical assays; likewise, His 6 -HRas and His 6 -p120RasGAP GAP domain were used in biochemical assays.
GST-SH3 or GST-RhoA proteins were purified by affinity chromatography by incubating clarified lysate with Glutathione Sepharose beads (Cytiva) in a gravity flow column at 4°C for 2 h. Beads were washed with 20 column volumes of lysis buffer, and resuspended in lysis buffer at a 1:1 bead:buffer ratio by volume. GST-fusion proteins were proteolytically cleaved by addition of GST-Prescision protease on-bead. Tag-free SH3 or RhoA proteins were collected as flowthrough and further purified over size exclusion chromatography (Superdex 75 prep grade, Cytiva) in buffer containing 20 mM Tris pH 8.0 and 150 mM NaCl. Proteins were concentrated in spin concentrators (Amicon, MilliporeSigma).
For co-crystallization protein samples, cells expressing His 6 -DLC1 RhoGAP and GST-p120RasGAP SH3 were mixed prior to lysis, which was carried out as described above. The complex, which formed in solution, was purified over glutathione affinity chromatography (Cytiva), liberated from beads by PreScission protease cleavage of the GST tag, and purified by size exclusion chromatography in 20 mM Tris pH 8.0, 150 mM NaCl to yield the protein complex and efficiently separate the excess p120RasGAP SH3 alone.

X-ray crystallography
Crystallization trials were carried out with a protein sample mixture consisting of a copurified complex of DLC1 RhoGAP with p120RasGAP SH3 at 18 mg/ml. Screen setup was performed using a TTP Labtech Mosquito in sitting-drop vapor diffusion setups at ambient temperature (∼22°C). Initial co-crystals were obtained in PEG/Ion Screen (Hampton Research) condition B4, comprised of 0.2 M Magnesium nitrate, 20% PEG 3350, pH 5.9. Optimal crystals, which grew as rods with approximate dimension 100 μm × 20 μm × 20 μm, were obtained in 0.2 M Magnesium nitrate, 16.5% PEG 3350, in 4 μl hanging drops (2 μl of protein complex plus 2 μl reservoir solution) over 1 ml reservoir solution. Crystals were harvested and cryopreserved in reservoir solution supplemented with 20% ethylene glycol and flash cooled in liquid nitrogen.
X-ray diffraction data were collected from a single crystal at the Northeastern Collaborative Access Team (NECAT) Beamline 24-ID-E at Argonne National Laboratory Advanced Photon Source. X-ray data were processed and scaled in HKL2000 78 to 3.2 Å resolution in spacegroup H3 with unit cell dimensions a = b = 143.7 Å, c = 152.8 Å, α = β = 90°, γ = 120°. XTriage analysis predicts four copies of the 1:1 complex per asymmetric unit with a solvent content of 46.5%.
The structure was solved by molecular replacement using Phenix Phaser 79 . Four copies of the DLC1 RhoGAP domain were placed using the previously unpublished DLC1A structure (PDB ID: 3KUQ 39 ) as a search model, with final TFZ score of 34.5 and LLG 1831. Next, placement of two of four expected SH3 domains was achieved using the crystal structure of p120RasGAPSH3 domain (PDB ID: 2J05 43 ) as search model with TFZ scores of 18.0 and 20.5; placement of the third and fourth copies failed. Each SH3 domain formed an identical interface with an individual DLC1RhoGAP copy in the asymmetric unit. Thus, the final two copies of SH3 were placed manually based on this interface with the remaining two DLC1RhoGAP copies and validated by refinement. Phenix Autobuild 80 , and manual building in Coot 81 , were performed. Refinement was carried out in Phenix 82 . A total of 1010 protein residues were modeled, with 5 N-terminal vector-derived residues in each DLC1 protein chain. All crystallography software was compiled by SBGrid 83 . Raw diffraction images have been deposited to the SBGRID Data Bank: https://doi.org/10.15785/SBGRID/876. Protein model coordinates and structure factors have been deposited to the Protein Data Bank with accession number 7TPB and prepared using PDB Extract version 3.27. Crystallization figures were generated in ccp4mg 84 . Protein interface calculations were performed on the PISA server (Protein Interfaces, Surfaces and Assemblies) at the European Bioinformatics Institute 44 , and structure-based alignments were run on the Dali server 85 .

Malachite green assays
Hydrolysis of GTP by RhoA was monitored in an endpoint assay using the malachite green reagent (Malachite Green Kit (Sigma Millipore) or BIOMOL Green (Enzo Life Sciences)). All assays were performed with purified RhoA at 5 μM in assay buffer with a final composition of 20 mM Tris-HCl pH 8.0, 250 mM NaCl, 1 mM MgCl 2 and 2 mM Ethylenediaminetetraacetic acid in final reaction volume of 80 μl. The reactions were initiated by addition of GTP at 50 μM. To monitor the effect of p120RasGAP SH3 domain, DLC1 RhoGAP protein at 0.05 μM and a titration series from 0 to 20 μM of wild-type or mutant SH3 domain protein were included. To assess the activity of wild-type and mutant DLC1, 0.1 μM DLC1 proteins were mixed with 0, 1, or 10 μM SH3 domain protein. The reactions were carried out at room temperature for 40 min in clear bottom 96-well plates (Corning) and terminated by addition of 20 μl of the malachite green working reagent, which was prepared as per the manufacturer's instructions (Sigma Millipore), followed by color development for 30 min. A single absorbance reading at 620 nm for each sample was measured at room temperature in a Synergy H1 plate reader (BioTek) using Gen5 3.11 software. A standard phosphate curve with concentrations from 4 to 40 μM was used to generate a linear standard curve and was used to calculate the phosphate generated in each sample well. Reactions in the absence of added GTP were included to control for background phosphate content signal, which was found to be negligible. The IC 50 of SH3 domain or SH2-SH3-SH2 domain inhibition were calculated in GraphPad Prism by plotting the mean phosphate signal (n = 4 or n = 7, respectively) versus log10 SH3 concentration using nonlinear regression using the model: One-Site -Fit LogIC50 and constraining the top and bottom values to the maximum (zero SH3 inhibition) and minimum (zero DLC1). Hydrolysis of GTP by Ras was performed similarly, using 5 μM purified H-Ras and 0.075 μM p120RasGAP GAP domain, in the presence of p120RasGAP SH3 domains at 0, 1, and 10 μM. For experiments using p120RasGAP SH2-SH3-SH2 domain, the hydrolysis reactions were performed using 200 μM GTP for 30 min, and color development for 20 min using the BIOMOL green malachite reagent (Enzo Life Sciences).

Reporting summary
Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
The data that support this study are available from the corresponding authors upon reasonable request. Coordinates have been deposited in the RSCB Protein Data Bank (PDB) under accession code 7TPB (DLC1 RhoGAP + p120RasGAP SH3). X-ray diffraction images are available online at SBGrid Data Bank: https://data.sbgrid.org/dataset/876/. Pre The source data underlying Figs. 4a-d and 5b, c is provided as a Source Data file. Source data are provided with this paper.