Structural Basis for the Specific Recognition of RhoA by the Dual GTPase-activating Protein ARAP3*

ARAP3 (Arf-GAP with Rho-GAP domain, ANK repeat, and PH domain-containing protein 3) is unique for its dual specificity GAPs (GTPase-activating protein) activity for Arf6 (ADP-ribosylation factor 6) and RhoA (Ras homolog gene family member A) regulated by phosphatidylinositol 3,4,5-trisphosphate and a small GTPase Rap1-GTP and is involved in regulation of cell shape and adhesion. However, the molecular interface between the ARAP3-RhoGAP domain and RhoA is unknown, as is the substrates specificity of the RhoGAP domain. In this study, we solved the crystal structure of RhoA in complex with the RhoGAP domain of ARAP3. The structure of the complex presented a clear interface between the RhoGAP domain and RhoA. By analyzing the crystal structure and in combination with in vitro GTPase activity assays and isothermal titration calorimetry experiments, we identified the crucial residues affecting RhoGAP activity and substrates specificity among RhoA, Rac1 (Ras-related C3 botulinum toxin substrate 1), and Cdc42 (cell division control protein 42 homolog).

The small GTPase Ras superfamily in human consists of at least 154 members divided into five principal families: the Ras, Rho, Rab, Arf, and Ran families (1). These proteins act as molecular switches, cycling between an inactive GDP-bound form and an active GTP-bound form, transducing signals from plasma membrane receptors to downstream signal molecules. These small GTPases generally have low intrinsic GTPase hydrolytic activity, and efficient hydrolysis requires interaction with family-specific groups of GAPs (GTPase-activating proteins) which can accelerate the cleavage step by several orders of magnitude. Arf6, an Arf (ADP-ribosylation factor) family member, is involved in endocytic pathways, endosomal recycling, cell migration, exocytosis, actin reorganization, plasma membrane reorganization, and cytokinesis (2). RhoA belongs to the Rho (Ras homologous) family, which primarily regulates cell shape, polarity, and motility by inducing stress fibers in cells that provides intracellular contractility (3)(4)(5)(6). In addition, Arf6 plays a role in cancer cell invasion, and RhoA regulates cancer cell morphology and migration (7,8).
ARAP3, a multidomain protein, belongs to the ARAP superfamily proteins and contains both functional ArfGAP and RhoGAP domains, a sterile ␣ motif domain, five pleckstrin homology (PH) 3 domains and an Ras-associating (RA) domain (see Fig. 1a). ARAP3 was originally identified as a PI3K effector in a screening for PtdIns (3,4,5)P 3 binding proteins (9). ARAP proteins are regulated at different levels. It has been demonstrated that the ArfGAP activity of ARAP3 is induced by the phosphoinositols PtdIns(3,4,5)P 3 and phosphatidylinositol 3,4bisphosphate in vivo and in vitro, whereas its RhoGAP activity is unaffected in vitro (9,10). Moreover, the RhoGAP activity of ARAP3 is up-regulated by direct binding of Rap1-GTP to the RA domain in vivo and in vitro (10). ARAP3 is unique because of its two different functional GAPs, ArfGAP and RhoGAP, which act specifically toward Arf6 and RhoA, respectively (9,10). This property most likely provides an efficient way to coordinate the two signaling processes. However, the mechanisms underlying the specificity of ARAP3 for Rho and Arf GTPases have not been thoroughly characterized.
In this study, we focused on the structure and function of RhoGAP in ARAP3. It is known that ARAP3 regulates many cell functions through its RhoGAP activity, such as formation of lamellipodia upon growth factor stimulation (11), development of angiogenesis in PI3K signaling pathway (12), and inhibition of peritoneal dissemination of scirrhous gastric carcinoma cells (13). Because Rap1-GTP can up-regulate the RhoGAP activity of ARAP3, the role of ARAP3 in downstream Rap1-GTP cell processes is RhoGAP-dependent. In neutrophils, ARAP3 regulates chemotaxis and adhesion-dependent processes (14). In macrophages, following long term treatment by TGF-␤1, ARAP3 down-regulates the production of chemokines and slows cell migration (15). ARAP3 mediates the neurite outgrowth from PC12 cells in response to treatment by basic fibroblast growth factors, nerve growth factor, and cAMP, all of which can up-regulate Rap1-GTP level in PC12 cells (16 -18). However, the nature of the ARAP3-RhoGAP⅐RhoA interaction was still unknown.
To determine the structure and function of the ARAP3-RhoGAP domain, we solved the x-ray crystal structures of the RhoGAP domain and of the complexed RhoA⅐RhoGAP domain with a substrate transition state analog GDP⅐AlF 4 Ϫ . Through these structures, we mapped the interface between RhoA and the ARAP3-RhoGAP domain. Using the complexed structure as a guide, we performed site-directed mutagenesis to analyze the role of interface residues on complex formation and on GTP hydrolysis. Meanwhile, we also found the residues responsible for the substrate specificity of ARAP3 among RhoA, Rac1, and Cdc42 in vitro.

Results
Overall Structure of the ARAP3-RhoGAP Domain in Complex with RhoA in the Transition State-To investigate the structure and function of the ARAP3-RhoGAP domain, we constructed different fragments according to the secondary structure prediction. The most promising construct is the fragment containing ARAP3 residues 906 -1107, which binds RhoA with a dis-sociation constant (K D ) of 2.45 M ( Fig. 1d; see Table 2). After extensive crystallization trials, we determined the 2.30 Å crystal structure of the ARAP3-RhoGAP domain (Table 1). Although we failed to obtain crystals by mixing the ARAP3-RhoGAP domain with different RhoA constructs, we successfully obtained crystals by expressing a fusion protein of the ARAP3-RhoGAP domain and RhoA, using a 28-residue flexible linker (19,20), which was purified in buffer containing GDP, Mg 2ϩ , Al 3ϩ , and F Ϫ to mimic the transition state (21,22).
Crystals of the ARAP3-RhoGAP domain belong to space group P4 3 2 1 2 with two molecules in the asymmetric unit. The two ARAP3-RhoGAP domain molecules in the asymmetric unit are similar to one another with a C␣ atom root mean square deviation (RMSD) of 0.375 Å calculated by PyMOL. The overall structure of the ARAP3-RhoGAP domain exhibits a canonical RhoGAP domain structure, which consists of nine long ␣-helices and two short ␣-helices packed in an anti-paral- ؊ . a, schematic diagram of the domain arrangement of ARAP3. b, overall structure of one molecular ARAP3-RhoGAP domain is represented in cylindrical helix cartoon and colored in slate. The 11 helical segments of the ARAP3-RhoGAP domain are labeled as ␣1-␣10. The ␣2-␣3 loop containing the arginine finger is colored in yellow. c, overall structure of one molecule of RhoA⅐GDP⅐AlF 4 Ϫ in complex with the ARAP3-RhoGAP domain (cyan) shown in cartoon representation. RhoA is colored in gray, and the switch I region, without interpretable electron density in our structure, is represented by a black dashed line. GDP and AlF 4 Ϫ are shown as sticks, and the Mg 2ϩ ion is shown as a green ball. d, ITC fitting curves of the ARAP3 RhoGAP domain to RhoA⅐GDP⅐MgF x . e, superimposition of the structure of the ARAP3-RhoGAP domain in apo (slate) and complexed state (cyan). f, the 2F o Ϫ Fc electron density map around the GDP, contoured at level 1.0 sigma, modeled as GDP⅐Mg 2ϩ ⅐AlF 4 Ϫ ⅐H 2 O. g and h, superimposition of the switch II region of RhoA in the ARAP3-RhoGAP⅐RhoA complex (gray) with that of RhoA-GTP␥S (PDB code 1A2B; colored in slate, g) and RhoA-GDP (PDB code 1FTN; colored in green, h). lel way, labeled as ␣1-␣10, with an extra ␣-helix at the C terminus (Fig. 1b). Most RhoGAP domains have a conserved motif in the loop between the ␣2 and the ␣3 helices. In this motif, a central arginine is the primary element required for GTPase rate enhancement and is called the arginine finger. The sequence of this motif in the ARAP3-RhoGAP domain is GVYRKGG (residues 939 -945; Fig. 1b). The ARAP3-RhoGAP domain structure exhibits structural similarity to many other RhoGAP structures with an RMSD from 1.4 to 2.4 Å and identity from 17 to 30%, analyzed by the Dali server (23).
Crystals of RhoA⅐GDP⅐AlF 4 Ϫ complexed with the ARAP3-RhoGAP domain belong to space group P2 1 2 1 2 1 , and the asymmetric unit contains two molecules. The structure was solved by molecule replacement, refined to 2.10 Å (Table 1), and the overall structure of one complex in the asymmetric unit is shown in Fig. 1c. The two complex molecules in the asymmetric unit are similar to one another with a C␣ atom RMSD of 0.4 Å calculated by PyMOL. Residues 28 -40 in the switch I region of RhoA, which cannot be observed because of missing electron density, are shown by the dotted line in Fig. 1c. In this final refined complex structure, the electron density map next to the GDP in the active center is almost square planar, modeled as AlF 4 Ϫ (Fig. 1f). The structures of the ARAP3-RhoGAP domain in the apo and the complexed state show no significant difference with a C␣ atom RMSD of 0.523 Å calculated by PyMOL. The main differences between the apo and the complexed state structures of the ARAP3-RhoGAP domain are found in the ␣2-␣3 loop and the ␣3 helix (Fig. 1e). The Arg-942 residue in the ␣2-␣3 loop is moved closer to interact with and stabilize the orientation of GDP⅐AlF 4 Ϫ . The ␣3 helix of ARAP3 RhoGAP domain has a 6.5°rotation between the apo and the complexed state, driving Arg-949 ARAP3 closer to Asp-90 RhoA to form tight salt bridges (Fig. 1e). RhoA⅐GDP⅐AlF 4 Ϫ superimposes well with the structure of RhoA-GTP␥S (PDB code 1A2B; Fig. 1g) but has a conformational difference at the switch II region compared with RhoA-GDP (PDB code 1FTN; Fig. 1h). However, we could not compare the conformation of switch I from these structures because the electron density map of this region was missing from the structure we solved.
Interface between ARAP3-RhoGAP Domain and RhoA-ARAP3-RhoGAP domain complexed with RhoA in the transition state buries 2092 Å 2 of solvent-accessible surface. The interaction surface of the ARAP3-RhoGAP domain involves the ␣2-␣3 loop and the ␣3, ␣4, and ␣8 helices. Residues contributing to the interaction interface from RhoA are located at the switch II region, the ␣3 helix, and the insert region (Fig.  2a). The complexed structure shows the conserved active center as reported previously. In the complexed structure, residue Arg-942 is pointed into the nucleotide binding pocket; the guanidinium group of Arg-942 interacts with the ␣-oxygen of GDP and two fluorides of AlF 4 Ϫ , and the main chain oxygen of Arg-942 forms a hydrogen bond with the side chain amide group of the RhoA catalytic residue Gln-63. Gln-63 RhoA is thus stabilized and forms hydrogen bonds with a fluoride of AlF 4 Ϫ and a water molecule through its side chain (Fig. 2b). Through these conserved interactions, GTP is positioned in line with the water molecule, because the aluminum ion of the ␥-phosphate analog AlF 4 Ϫ is located in line with the water molecule and the ␤-oxygen of GDP, with distances of 2.0 and 1.9 Å (Fig. 2b), respectively. In addition, the positively charged arginine finger of the ARAP3-RhoGAP neutralizes negative charge at the ␥-phosphate, which makes the water molecule easier to conduct a  AUGUST 5, 2016 • VOLUME 291 • NUMBER 32

Structure of the ARAP3-RhoGAP⅐RhoA Complex
nucleophilic attack on the ␥-phosphate of GTP; thus the GTP hydrolysis rate of RhoA is accelerated. The RhoGAP domain of ARAP3 interacts with RhoA mainly through electrostatic interactions. Negatively charged GDP⅐AlF 4 Ϫ and negatively charged residues Glu-64 and Asp-65 of the switch II region and Asp-90 of the ␣3 helix of RhoA interact tightly with positively charged residues of the ARAP3-RhoGAP domain (Arg-942 of the ␣2-␣3 loop, Arg-949 of the ␣3 helix, and Arg-982 and Arg-985 of the ␣4 helix) through salt bridges (Fig. 2b). We can observe the electrostatic interactions more clearly from the electrostatic representation of the complexed structure (Fig. 2c); residues Glu-64, Asp-65, and Asp-90 and GDP⅐AlF 4 Ϫ forming the negatively charged surface of RhoA directly bind to the positively charged surface of the ARAP3-RhoGAP domain formed by residues Arg-942, Arg-949, Arg-982, and Arg-985. In addition to the interactions shown above, there are several hydrogen bonds formed between the ARAP3-RhoGAP domain and RhoA: the main chain oxygen of residue Gly-945 ARAP3 and the main chain nitrogen of residue Arg-947 ARAP3 form hydrogen bonds with residue Asn-94 RhoA (see Fig. 4b); the main chain nitrogen of residue Ala-948 ARAP3 forms a hydrogen bond with residue Glu-97 RhoA (see Fig. 4b); and the main chain oxygen of residue Leu-1057 ARAP3 forms a hydrogen bond with residue Tyr-66 RhoA . There are also water-mediated contacts between the main chain nitrogen of residue Gly-944 ARAP3 and Gly-14 RhoA and between the main chain oxygen of residue Gly-944 ARAP3 and Asp-90 RhoA . In addition to electrostatic and hydrogen bonding interactions, hydrophobic interactions are also observed; residue Phe-972 ARAP3 of the ␣3Ј-␣4 loop and residue Pro-1061 ARAP3 of the ␣8 helix make hydrophobic contacts with residue Met-134 RhoA and residue Tyr-66 RhoA in the switch II region.
Based on the complexed structure, we made site-directed mutations in ARAP3-RhoGAP to verify the importance of residues for the interaction between the ARAP3-RhoGAP domain and RhoA. We used ITC experiments to measure the binding  Table 2). Moreover, a single mis-sense mutation of each of the residues Arg-949, Arg-982, or Arg-985 in the RhoGAP domain to glutamic acid was sufficient to abolish complex formation ( Fig. 2d and Table 2). The GTPase assay results were consistent with the ITC findings. The GTP hydrolysis reaction rate, using 25 nM wild type ARAP3-RhoGAP domain and 35 M RhoA-GTP, was 10-fold greater than that of the RhoA alone (Fig. 2e). A single point mutation of each of the residues R942A, R949A, R949E, R982E, and R985E in the RhoGAP domain dramatically reduced the GAP activity of the ARAP3-RhoGAP domain under the same conditions (Fig. 2e). Mutants R982A and R985A still retained slight GAP activity so they could weakly accelerate the GTP hydrolysis reaction rate of RhoA by ϳ3-fold (Fig. 2e). Intriguingly, mutant R949A also dramatically reduced the GAP activity of the ARAP3-RhoGAP domain (Fig. 2e), indicating that Arg-949 ARAP3 is important for the ARAP3-RhoGAP domain to perform its GAP activity. Indeed, Arg-949 ARAP3 interacts tightly with Phe-972 ARAP3 and Asp-975 ARAP3 through cation-and electrostatic interactions (Fig. 2f), thus stabilizing the RhoGAP structure. All of these results demonstrate that both the arginine finger and the residues responsible for binding to RhoA are crucial to maintain the GAP activity of the ARAP3-RhoGAP domain.
Comparison with Other RhoGAP⅐RhoA Complexes in the Transition State-Although there are over 70 different RhoGAPs, only two other RhoGAPs in complex with RhoA in the transition state have been solved (Fig. 3, b and c): ARHGAP1, which is also called p50RhoGAP or Cdc42GAP (PDB codes 1TX4 and 1OW3), and ARHGAP20 (PDB code 3MSX). The sequence of the RhoGAP domain of ARAP3 shows identity with ARHGAP1 (26%) and ARHGAP20 (25%). The most remarkable difference was found at the interface involving the switch I region of RhoA; in the p50RhoGAP⅐RhoA complex residues Tyr-34 RhoA , Pro-36 RhoA , Val-38 RhoA , and Phe-39 RhoA make extensive hydrophobic contacts with RhoGAP of ARHGAP1 (24), whereas in the ARHGAP20⅐RhoA complex residues Tyr-34 RhoA , Val-35 RhoA , Pro-36 RhoA , and Phe-39 RhoA make extensive hydrophobic interactions with RhoGAP of ARHGAP20. However, we could not observe these hydrophobic interactions between RhoA and RhoGAP of ARAP3, because the switch I region in our structure is poorly ordered, and the electron density map of residues 28 -40 is missing. The missing electron density in the switch I region of RhoA indicates that the switch I region may not be concerned with the ARAP3-RhoGAP binding. Another notable difference was found at the N terminus of the ␣9 helix. The ␣9 helix of ARAP3 RhoGAP is shorter than ARHGAP1 and ARHGAP20 (Fig. 3d), and the ␣8-␣9 loop (residues 1067-1071) is poorly ordered, losing the hydrophobic interactions with residues Leu-69 and Leu-72 in the ␣2 helix of RhoA (24).
Despite these differences, all three RhoGAP⅐RhoA complexes exhibit two interfaces in the same location. The first interface consists of the ␣3 helix of RhoGAP and the ␣3 helix of RhoA, and the second interface consists of the ␣4 helix of RhoGAP and the switch II region of RhoA. Residues in the first interface are less conserved than those in the second interface among ARAP3, ARHGAP1, and ARHGAP20 (Fig. 4a). There is a conserved alanine residue at the end of the ␣2-␣3 loop of the RhoGAP domain (Fig. 4a), which interacts with Asn-94 RhoA in ARHGAP1 and ARHGAP20 (Fig. 4b). However, in ARAP3, it is replaced by the Gly-945 APAP3 (Fig. 4b). There is also a conserved arginine residue at the ␣3 helix (Fig. 4a), which interacts with Glu-97 RhoA only in ARHGAP20 (Fig. 4b). In the first interface, ARHGAP1 only has interactions with Asn-94 RhoA and Glu-97 RhoA , whereas the ARAP3-RhoGAP domain has extra interactions with Asp-90 RhoA through salt bridges, and ARHGAP20 has interactions with Asp-90 RhoA , Glu-93 RhoA , Asn-94 RhoA , and Glu-97 RhoA (Fig. 4b). In the ␣4 helix, there are two conserved residues, lysine and arginine (Fig. 4a). The conserved arginine in the ARAP3-RhoGAP domain (Arg-985) interacts with Asp-65 RhoA , whereas in both ARHGAP1 and ARHGAP20, it interacts with both Glu-64 RhoA and Asp-65 RhoA (Fig. 4c). In ARHGAP1 and ARHGAP20, the conserved lysine interacts with Asp-65 RhoA (Fig. 4c). In ARAP3, however, the conserved lysine is not on the interface and is replaced by Arg-982 that interacts tightly with Glu-64 RhoA (Fig. 4c).

ARAP3-RhoGAP Domain Specifically Recognizes RhoA through Negatively Charged Residues Asp-90 and Glu-97 in the
␣3 Helix-All of the residues involved in interacting with the ARAP3-RhoGAP domain are conserved among RhoA, Cdc42, and Rac1, except for the negatively charged residues Asp-90 and Glu-97 in the ␣3 helix (Fig. 5a). In Cdc42, Ser-88 replaces Asp-90, whereas in Rac1, Ala-88 and Ala-95 replace Asp-90 and Glu-97, respectively. We speculated that residues Asp-90 and Glu-97 of RhoA confer the substrate specificity of the ARAP3-RhoGAP domain over Cdc42 and Rac1. To test this hypothesis, we generated RhoA-like mutations in Cdc42 and Rac1 and tested the effects on binding by ITC and GTPase activity assays. The ITC results showed that the ARAP3-RhoGAP domain binds wild type Cdc42 and the RhoA-like Cdc42 mutant (S88D) with K D values of 59.88 and 5.41 M, respectively ( Fig. 5b and Table 2). The RhoA-like Cdc42 mutant (S88D) shows ϳ11-fold stronger binding affinity than wild type Cdc42 to the RhoGAP domain of ARAP3. The ARAP3-RhoGAP domain shows very weak binding affinity to wild type Rac1 that could not be detected by ITC. Meanwhile, the  (Fig. 5c and Table 2). Additionally, in GTPase activity experiments, the ARAP3-RhoGAP domain had greater efficiency in the presence of the RhoA-like Cdc42 or Rac1 mutants than with wild type Cdc42 and Rac1 (Fig. 5, d and e,

Discussion
ARAP3 was first reported in 2002 (9), and since then the function of ARAP3 in different cells and tissues has been extensively investigated, and most of the ARAP3 functions are related to its RhoGAP activity. However, the molecular mechanism of the RhoA-specific GAP activity of the ARAP3-RhoGAP domain remained unclear. In this study, we solved the crystal structure of the ARAP3-RhoGAP domain in complex with RhoA⅐GDP⅐AlF 4 Ϫ . Based on the interaction surface, we found that single point mutations of Arg-942, Arg-949, Arg-982, or Arg-985 can dramatically reduce the binding affinity and the RhoGAP activity of the ARAP3-RhoGAP domain toward RhoA in vitro (Fig. 2, d and e, and Table 2). Over the years, researchers have made loss of function mutants of the ARAP3-RhoGAP domain by point mutation of Arg-942 or Arg-982 (9,13,16,17); here we proved that residue Arg-942 is the catalytic arginine finger of RhoGAP and that residue Arg-982 is critical for the interaction between RhoGAP and RhoA (Fig.  2b). Only two other RhoGAP⅐RhoA complexed structures (ARHGAP1, PDB code ITX4 and 1OW3; ARHGAP20, PDB code 3MSX) have been solved (Fig. 3, b and c). The ARAP3-RhoGAP domain shows ϳ25% sequence identity with ARHGAP1 and ARHGAP20, and the complexed structures show differences at the binding interfaces, especially at the first interface (Fig. 4, a and b). Our complexed structure is an important contribution toward understanding the interactions between RhoGAPs and RhoA.
There are ϳ22 mammalian genes encoding Rho family small GTPases, but ϳ70 protein members of RhoGAPs have been found in mammals (25,26). Most of the RhoGAPs show selec- tivity among the different Rho family small GTPases. ARAP1 and ARAP3 selectively use RhoA as their substrate (9,27), but the ARAP2-RhoGAP domain lacks RhoGAP activity, because the catalytic arginine finger is replaced by glutamine. However, ARAP2 still has a direct interaction with RhoA-GTP (28). ARAP1 is an Arf1/Arf5 ArfGAP, whereas ARAP2 and ARAP3 are specific for Arf6 (9,27,28). The mechanism for the specificity of ARAP3 among Rho and Arf GTPases remains elusive. In this study, we have found that residues Asp-90 and Glu-97 in the ␣3 helix of RhoA are responsible for the substrate-specific GAP activity of the ARAP3-RhoGAP domain. In Cdc42, the related residues are Ser-88 and Glu-95, whereas in Rac1 they are Ala-88 and Ala-95 (Fig. 5a). The ARAP3-RhoGAP domain has interactions with both Asp-90 and Glu-97 (Fig. 4b) and exhibits higher binding affinity and GAP activity against a RhoA-like mutant Cdc42 (S88D) or Rac1 (A88D,A95E) than wild type Cdc42 and Rac1 (Fig. 5, b-e, and Tables 2 and 3). It is notable that the K cat values are comparable (the biggest difference is 2-fold) and the K m values are the main factor for the differences of the catalytic efficiencies, further proving that ARAP3-RhoGAP chooses its specific target mainly through binding affinity. The same molecular mechanism for the specificity of RhoGAP activity was reported for the Myo9b-RhoGAP domain (29). As shown in Fig. 4b, ARHGAP1 interacts only via Glu-97 and ARHGAP20 interacts with both Asp-90 and Glu-97. Functionally, ARHGAP1 is a GAP with a preference for Cdc42 over Rho and Rac1 (30,31), and ARHGAP20 is a GAP with a preference for RhoA (32), supporting our idea that RhoGAPs choose their specific target by recognizing differences in the ␣3 helix among RhoA, Cdc42, and Rac1. In addition, ␤2-chimaerin contains a Rac1-specific GAP domain, its specificity is conferred by interactions of Phe-315 and Glu-317 with Ala-88 and Ala-95 in the ␣3 helix of Rac1 (33). Glu-95 in Cdc42 and Glu-97 in RhoA would block that interaction and make them poor targets for ␤2-chimaerin. Thus, RhoGAPs may choose their specific target by the differences in the ␣3 helix of RhoA, Rac1, and Cdc42. Intriguingly, the GTPase substrate preference of p190-RhoGAP can be switched by phospholipids from RhoA and Rac1 to only Rac1 (34) and to only RhoA by phosphorylation (35). Phosphorylation can also switch the FIGURE 5. The ARAP3-RhoGAP domain shows higher binding affinity and GAP activity toward RhoA-like Cdc42, and Rac1 mutations in vitro. a, partial sequence alignment of RhoA, Cdc42, and Rac1. Interface residues that form hydrogen bonds with the ARAP3-RhoGAP domain based on the complexed crystal structure are marked with triangles below. The variable residues at the interface are colored in red. b and c, ITC fitting curves of the ARAP3-RhoGAP domain to Cdc42⅐GDP⅐MgF x (green triangles), Rac1⅐GDP⅐MgF x (cyan inverted triangles), RhoA-like mutants Cdc42 S88D ⅐GDP⅐MgF x (red squares), and Rac1 A88D,A95E ⅐GDP⅐MgFx (blue circles). d and e, the K m and K cat /K m values of the ARAP3-RhoGAP domain toward wild type RhoA, Cdc42, and RhoA-like mutants of Cdc42 and Rac1, measured by MESG/PNP system. The error bars are standard errors generated by nonlinear fitted using Origin, same as errors shown in Table 3.
GTPase substrate preference of MgcRacGAP (male germ cell RacGAP, also named Rac GTPase-activating protein 1) from Rac1 and Cdc42 to RhoA (36). It is reasonable that there are other factors that regulate the substrate specificity of different RhoGAPs, and the detailed mechanisms need further investigation.
The ARAP3-ArfGAP activity can be activated by PtdIns(3,4,5)P 3 (9); two tandem PH domains in the N terminus, especially the first PH domain, are required for the interaction with PtdIns(3,4,5)P 3 (37). The mechanism of the regulation between PtdIns(3,4,5)P 3 , PH domains, and the ArfGAP remains unclear. ␤2-Chimaerin, a Rac1 specific GAP, contains SH2-C1-RacGAP tandem domain. The N-terminal peptide binds the DAG (diacylglycerol)-binding site in the C1 domain and partially occupies the GAP active site. By binding to DAG through the C1 domain, the N-terminal peptide is released from the GAP active site, and GAP activity is enhanced (33). Whether the activation of ARAP3-ArfGAP by binding to PtdIns(3,4,5)P 3 adopts the same mechanism as ␤2-chimaerin requires further investigation. ARAP3 also contains an RA domain, and ARAP3 RhoGAP activity can be enhanced by direct binding of Rap1-GTP to the RA domain (10). ARHGAP20 has PH-RA-RhoGAP tandem domains; the RA domain can autoinhibit RhoGAP activity, and the autoinhibitory effect can be released by binding to Rap1-GTP (32). However, the mechanism of how Rap1-GTP regulates the ARAP3-RhoGAP activity remains obscure. ARAP3 has five PH domains, and the function of PH3 to PH5 needs further investigation. In summary, our work demonstrates the interface between the ARAP3-RhoGAP domain and RhoA at atomic resolution and is an important contribution to our understanding of the regulation and specific recognition of RhoA by ARAP3.
Protein Expression and Purification-All proteins were expressed in Escherichia coli Bl21 DE3 (Gold) cells (Novagen) that were cultured in LB medium at 310 K to an A 600 of 0.8 -1.0, subsequently shifted to 289 K, and induced with 0.3-0.4 mM isopropyl ␤-D-thiogalactopyranoside for 24 h. RhoGAPs (wild type or mutant) were purified by nickel-nitrilotriacetic acid affinity chromatography after lysing cells in 20 mM Tris (pH 7.5) and 500 mM NaCl and eluted in buffer containing 30 -500 mM imidazole (pH 7.5). The eluted His tag proteins were further purified using a Superdex 200 column (GE Healthcare) in buffer containing 20 mM Tris (pH 7.5) and 200 mM NaCl. The fusion protein was also purified by nickel-nitrilotriacetic acid affinity chromatography after lysing of cells in 20 mM Tris (pH 7.5), 500 mM NaCl, 5 mM MgCl 2 , and 1 mM GDP, and eluting with buffer containing 30 -500 mM imidazole (pH 7.5). Further purification used a Superdex 200 column (GE Healthcare) in buffer containing 20 mM Tris (pH 7.5), 200 mM NaCl, 5 mM MgCl 2 , 1 mM AlCl 3 , and 10 mM NaF. For ITC experiments, small GTPases and RhoGAPs were purified as described for the fusion protein (without AlCl 3 ). All purified proteins were concentrated and stored at 193 K. GDP and GTP were purchased from Sangon Biotech.
Crystallization and Structure Solution-RhoGAP was concentrated to ϳ10 mg/ml, and the buffer was exchanged to 20 mM KH 2 PO 4 (pH 6.5), 200 mM NaCl, and 10 mM DTT. Crystals of RhoGAP formed in mother liquor containing 100 mM HEPES (pH 6.9) and 10% PEG8000. The fusion protein was concentrated to ϳ10 mg/ml, and the buffer was exchanged to 20 mM Tris (pH 7.5), 150 mM NaCl, 5 mM DTT, 5 mM MgCl 2 , 1 mM AlCl 3 , 10 mM NaF, and 2 mM GDP. Crystals of the fusion protein formed in the mother liquor containing 100 mM MES (pH 6.0), 15% (v/v) polyethylene glycol monomethyl ether 550, and 4% (v/v) acetone. All crystals were grown at 293 K via the hanging drop method. X-ray diffraction data of the crystals were collected on Beamline 17U1 of the Shanghai Synchrotron Radiation Facility. The data were processed with HKL2000 and programs in the CCP4 suite. The structure of the RhoGAP domain and the fusion protein (complex) were solved by molecular replacement using the program PHASER (38). All the structural models were subsequently refined by programs REF-MAC5 (39), PHENIX (40), and COOT (41), and ligands (GDP, AlF 4 Ϫ , and Mg 2ϩ ) were added manually. Crystallographic parameters are listed in Table 1. All structure figures were prepared with PyMOL (42). The transition state complex interface was calculated in PDBePISA (43).

TABLE 3 Kinetic parameters of the ARAP3-RhoGAP domain to different small GTPases
The GTPase activity assays were carried out at 293 K, using the MESG/PNP system. For the measurement of the kinetic parameters of the ARAP3-RhoGAP domain toward the RhoA, Cdc42 (S88D), Cdc42, and Rac1 (A88D,A95E), the concentrations of the ARAP3-RhoGAP domain were 20, 25, 60, and 30 nM, respectively. ND, not detected because of the very low binding affinity. Errors in the Isothermal Titration Calorimetry-ITC assays were carried out on a MicroCal iTC200 calorimeter (GE Healthcare) at 293 K. The buffer for RhoGAPs (wild type or mutant) and fulllength GTPases was 20 mM Tris (pH 7.5), 200 mM NaCl, 5 mM MgCl 2 , and 10 mM NaF. The titration protocol consisted of 20 injections of 0.8 -1.0 mM RhoGAPs (wild type or mutant) into 0.06 -0.08 mM GTPases. Curve fitting to a one-binding-site model was performed by ITC data analysis module of Origin7.5 (OriginLab) provided by the manufacturer. The parameters are summarized including the standard errors in Table 2.
GTPase Activity Assay-The initial rates and kinetic measurements of GAP-stimulated GTP hydrolysis were measured using the MESG and PNP system with a constant amount of RhoGAP and increasing concentrations of GTPases (44). MESG and PNP were purchased as part of the EnzChek phosphate assay kit (Life Technologies, E6646). Briefly, the indicated amount of GTPase was incubated in buffer: 50 mM HEPES (pH 7.6), 0.1 mM EDTA, 400 M GTP, 0.2 mM MESG, and 1 unit of PNP. The reactions were started by adding 10 l a solution containing the indicated amounts of RhoGAP and MgCl 2 to the mixture. The final concentration of MgCl 2 was 5 mM. The time courses of absorbance change at 360 nm were recorded by DU 800 (Beckman Coulter) at 293 K. The kinetic parameters were nonlinear fitted by Origin 8.5.1 (OriginLab) using the modified Michaelis-Menten equation as described (44) and are summarized including the standard errors in Table 3.