Substrate Specificity and Recognition Is Conferred by the Pleckstrin Homology Domain of the Dbl Family Guanine Nucleotide Exchange Factor P-Rex2*

Dbl family guanine nucleotide exchange factors (GEFs) are characterized by the presence of a catalytic Dbl homology domain followed invariably by a lipid-binding pleckstrin homology (PH) domain. To date, substrate recognition and specificity of this family of GEFs has been reported to be mediated exclusively via the Dbl homology domain. Here we report the novel and unexpected finding that, in the Dbl family Rac-specific GEF P-Rex2, it is the PH domain that confers substrate specificity and recognition. Moreover, the β3β4 loop of the PH domain of P-Rex2 is the determinant for Rac1 recognition, as substitution of the β3β4 loop of the PH domain of Dbs (a RhoA- and Cdc42-specific GEF) with that of P-Rex2 confers Rac1-specific binding capability to the PH domain of Dbs. The contact interface between the PH domain of P-Rex2 and Rac1 involves the switch loop and helix 3 of Rac1. Moreover, substitution of helix 3 of Cdc42 with that of Rac1 now enables the PH domain of P-Rex2 to bind this Cdc42 chimera. Despite having the ability to recognize this chimeric Cdc42, P-Rex2 is unable to catalyze nucleotide exchange on Cdc42, suggesting that recognition of substrate and catalysis are two distinct events. Thus substrate recognition can now be added to the growing list of functions that are being attributed to the PH domain of Dbl family GEFs.

The Rho family of G proteins are members of the Ras superfamily (1,2). A hallmark of G proteins is their ability to undergo conformational changes upon binding to either GTP or GDP. These G proteins therefore function as binary switches, as different partner proteins within the cell recognize the GTPbound "on" state and the GDP-bound "off" state of the protein.
There are twenty Rho family members identified to date, with RhoA, Rac1, and Cdc42 being the most well characterized members (3). They carry out a wide and diverse range of functions, from regulating the actin cytoskeletal network to the neutrophil oxidative burst (4,5).
Diffuse B-cell lymphoma (Dbl) 1 family guanine nucleotide exchange factors (GEFs) catalyze nucleotide exchange on the Rho subfamily of small G proteins (6,7). Dbl was first isolated as an oncoprotein from a diffuse B-cell lymphoma (8). Since then, its characteristic 300-amino-acid sequence has been identified in over 60 mammalian proteins, making this one of the largest families of oncoproteins (6,9). This 300-amino-acid stretch is subdivided into a Dbl homology (DH) domain followed invariably by a pleckstrin homology (PH) domain. Beyond the DHPH domains, Dbl family members share little sequence similarity with each other (6,7,9).
The DH domain consists of ϳ200 amino acids and is the catalytic domain (10). Within the DH domain are three highly conserved regions referred to as conserved regions 1, 2, and 3 (CR1, CR2, and CR3) (7). The structure of several Dbl family members shows that the DH domain consists of an elongated helical bundle, with the CR1, CR2, and CR3 forming long helices that pack together to form the core of the module (11)(12)(13). PH domains are lipid-binding domains that are found in a wide variety of signaling molecules (14 -16). They consist of a seven-stranded ␤-sandwich that is capped on one end by a C-terminal ␣-helix (14). The invariant association of the PH domain with the DH domain in all Dbl family members suggests a functional interdependence between these modules. These PH domains have been shown to mediate membrane association and to modulate the activity of the DH domain (17)(18)(19)(20)(21).
Dbl family members vary widely in their substrate specificity. GEFs like Tiam1 and Trio catalyze exchange only on Rac1 (22). Dbs and Dbl are able to activate both Cdc42 and RhoA (23). Vav1 and Vav2 catalyze exchange on RhoA, Rac1, and Cdc42 (24). The structural basis for this recognition process is only beginning to be understood (25). The crystal structures of the Rac1-Tiam1, Cdc42-Dbs, Dbs-RhoA, and Cdc42-Intersectin complexes show that the CR1 and CR3 regions of the DH domain of the GEF make extensive contact with the ␤2␤3 strands of the GTPases (25)(26)(27)(28). These regions were therefore predicted to determine the specificity of these interactions. Based on these crystal structure complexes, Trp 56 of Rac1 was identified as one of the critical residues for recognition by Rac-specific GEFs (27). Mutation of this residue to Phe, the corresponding residue in Cdc42, was sufficient to abolish binding by Tiam1. Moreover, this mutant Rac1 could now be recognized by Intersectin, a Cdc42-specific GEF. Although this region clearly determines the specificity of some of these interactions, it cannot explain the specificity of all GEF-GTPase interactions (25). Moreover, the contribution of the PH domain in these interactions has remained unexplored. P-Rex1 was isolated from neutrophils based on its activity as a Rac-specific GEF (29). Its activity was conditional on the presence of phosphatidylinositol 3,4,5-trisphosphate or G␤␥ subunits. P-Rex1 plays a crucial role in stimulating the neutrophil-oxidative burst (29). Our biochemical characterization of P-Rex2, a homologue of P-Rex1, shows that it also is a Rac-specific GEF. 2 In this report, P-Rex2 is shown to recognize Rac1 in an unusual way. The specificity of this interaction is conferred by the PH domain of P-Rex2 and not by the DH domain. Specificity of substrate interaction has never before been attributed to the PH domain of a Dbl family GEF. This report demonstrates that, within the PH domain, the ␤3␤4 loop is necessary and sufficient to confer this specificity. In addition, the PH domain of P-Rex2 appears to make contact with the switch loop and helix 3 of Rac1. However, although helix 3 of Rac1 is sufficient to confer binding to a chimeric GTPase, the helix by itself is not sufficient to allow P-Rex2 to catalyze nucleotide exchange. This suggests that substrate recognition and catalysis by the P-Rex2 GEF have distinct requirements.

EXPERIMENTAL PROCEDURES
Materials-The wild-type bacterial expression construct for RhoA was created by PCR amplification and cloned into the SmaI/NotI sites of pGEX 4T-1 (Amersham Biosciences). Bacterial expression constructs for Rac1 and Cdc42 were kind gifts from Prof. Gary Bokoch (The Scripps Research Institute, La Jolla, CA). The DHPH domains of Dbl and the PH domain of Dbs were PCR-amplified from the human mammary cDNA library (Clontech) and cloned into the pcDNA3.1D V5-His-TOPO vector (Invitrogen). The DH domain (residues 22-231), the PH domain (residues 241-370), and the DHPH domains of P-Rex2 (residues 22-370) were cloned into pcDNA3.1D V5-His-TOPO vector by PCR. The ␤3␤4 loop deletion (amino acid residues 277-296), mutant of the PH domain of P-Rex2, and the Rac1W56F mutant were created using the QuikChange TM site-directed mutagenesis kit from Stratagene, according to the manufacturer's protocol. The Dbs/P-Rex2 ␤3␤4 loop chimeric PH domain was created by PCR as follows. The 5Ј end of the chimera was created by PCR using oligos 5Ј-CACCATGCTGCTGATGCAGGG-CTCATTCAGC-3Ј (sense) and 5Ј-CATCTGTAGATGCCTTGCTGTTCT-TCAACCGTCTGTGTTTTCTCTCCCTCTTCTTGCAGAAGAG-3Ј (antisense). The 3Ј end of the chimeric gene was amplified using oligos 5Ј-GAACAGCAAGGCATCTACAGATGGACATCGGTACCTTTTTTCC-TACAGCTACAAGCAGTCC-3Ј (sense) and 5Ј-CAGGCTCTGTGACTG-CTCCAG-3Ј (antisense). The 5Ј and 3Ј ends of the gene obtained by PCR were purified and were merged together by PCR using oligos 5Ј-CAC-CATGCTGCTGATGCAGGGCTCATTCAGC-3Ј (sense) and 5Ј-CAGGC-TCTGTGACTGCTCCAGCGC-3Ј (antisense). The full-length PCR product was subsequently purified and cloned into the pcDNA3.1D V5-His-TOPO vector by TOPO cloning. The bacterial expression construct for the Cdc42Rac1helix3 chimera was created by PCR as follows. The 5Ј end of the chimera was created by PCR using oligos 5Ј-GATCTGGTT-CCGCGTGGATCCAAAATTATTTCAGC-3Ј (sense) and 5Ј-CCACTTTT-CTTTCACGTTTTCAAATGAAGATGGACTCACAAGGG-3Ј (antisense). The 3Ј end of the chimera was created by PCR using oligos 5Ј-GTGA-AAGAAAAGTGGGTGCCTGAGATAACTCACCACTGTCCCAAC-3Ј (sense) and 5Ј-TCAGTCAGTCACGATGAATTCATCTGTCGACTG-3Ј (antisense). The 5Ј and 3Ј end PCR products were purified, mixed together, and the full-length chimera was created by PCR using oligos 5Ј-GATCTGGTTCCGCGTGGATCCAAAATTATTTCAGC-3Ј (sense) and 5Ј-TCAGTCAGTCACGATGAATTCATCTGTCGACTG-3Ј (antisense). The PCR product was digested with BamHI and EcoRI and was ligated into the pGEX-2T vector (Amersham Biosciences) cut with the same enzymes.
Purification of Proteins-GST and GST-tagged Rac1, RhoA, and Cdc42 were expressed and purified from Escherichia coli as described elsewhere (30).
Complex Formation Assay-The complex formation assay was performed essentially as described previously (31). The DHPH domains of P-Rex2, Dbl, or Dbs were labeled with [ 35 S]Met using the rabbit reticulocyte lysate-coupled transcription translation system (Promega) for 90 min at 30°C in a total reaction volume of 10 l, according to the manufacturer's instructions. 2 l of this reaction was then incubated with 5 g of purified GST, GST-Rac1, GST-RhoA, or GST-Cdc42 in 500 l of complex formation assay buffer (20 mM Tris, pH 7.5, 100 mM NaCl, 2.0 mM EDTA, and 0.5% Triton X-100) with 25 l of glutathione-Sepharose beads (Amersham Biosciences) for 2 h at 4°C. The beads were spun down and washed five times with the same buffer. The beads were then boiled in SDS-PAGE sample buffer. The samples were analyzed by SDS-PAGE, transferred onto a nitrocellulose membrane, and detected by autoradiography. All experiments were repeated at least three times to ensure the reproducibility of the experiments. The data shown are representative of each of these data sets.
Rac1 was loaded with GDP or GTP␥S, as described earlier (31). Briefly, 12.5 M purified GST-Rac1 was incubated with 20 M GDP or GTP␥S in loading buffer (10 mM Hepes, pH 7.5, 100 mM NaCl, 7.5 mM EDTA) at 23°C for 25 min. The complexes were stabilized using 20 mM MgCl 2 . 5 g of the loaded protein was then used for the complex formation assay in a modified complex formation assay buffer with 20 mM MgCl 2 instead of 2.0 mM EDTA.
Guanine Nucleotide Exchange Assay-The in vitro GEF assay was performed on a Mithras LB 940 (Berthold Technologies) fluorescence spectrophotometer by modifying a basic protocol described elsewhere (32). 2 M purified GST-tagged Rho family G protein was incubated with 100 nM purified GST-tagged P-Rex2 and 400 nM N-methylanthranioyl-GDP (Molecular Probes) in 50 l of exchange buffer (20 mM Tris, pH 7.5, 50 mM NaCl, 10 mM MgCl 2 , 1 mM dithiothreitol, 50 g/l bovine serum albumin, and 10% glycerol). Excitation and emission wavelengths were 355 nm and 460 nm, respectively.

The Rac1W56F Mutation Does Not Affect P-Rex2 Binding-
Comparison of the crystal structure complexes of Tiam1 and Rac1 with other Rho GTPase-GEF complexes has identified the Trp 56 residue of Rac1 as a key determinant of specificity for Rac-specific GEFs (27). This residue of Rac1 was shown to interact with a conserved Ile in the DH domain of Rac-specific GEFs. However, sequence alignment of the DH domain of P-Rex2 with that of other GEFs shows that this conserved Ile is not present in the DH domain of P-Rex2 or P-Rex1 (Fig. 1A). Because mutation of the Trp 56 residue of Rac1 to Phe is sufficient to abolish binding by the DHPH domain of the Rac1specific GEF Tiam1, we decided to test whether this mutation had any effect on the binding ability of P-Rex2 (27). The DHPH of P-Rex2 was expressed and labeled with [ 35 S]Met using a rabbit reticulocyte lysate system and tested for its ability to bind a GST-Rac1W56F mutant in a pull-down assay. Unexpectedly, the Rac1W56F mutation did not decrease the binding of Rac1 to the DHPH domain of P-Rex2 (Fig. 1B). This sug-2 R. E. Joseph and F. A. Norris, manuscript submitted for publication. gested that the mode of interaction of P-Rex2 with Rac1 was different from that of other Rac-specific GEFs.
Substrate Recognition Occurs via the PH Domain of P-Rex2-Substrate recognition by the Dbl family of GEFs has been thought to be mediated exclusively by the DH domain (25). The above result indicates that P-Rex2 might be an exception. To determine the domain of P-Rex2 that recognizes its Rac1 substrate, the DH, PH, and DHPH domains of P-Rex2 were expressed and labeled with [ 35 S]Met using a reticulocyte lysate and tested for their ability to interact with GST-Rac1 in a pull-down assay. The DHPH domain of Dbl, which interacts with Rac1 as well as RhoA and Cdc42, was used as a positive control. As shown in Fig. 2A, surprisingly, the PH domain alone of P-Rex2 was able to interact with Rac1, whereas the DH domain alone of P-Rex2 was unable to do so. Moreover, The PH domain of P-Rex2 was able to discriminate between Rac1, RhoA, and Cdc42 and bind specifically to Rac1. Increasing amounts of GST-tagged Rac1 were able to associate with increasing amounts of the PH domain, indicating that the levels of GTPase being used in the pull-downs were not saturating (Fig. 2B). This is the first demonstration of the ability of a PH domain of a Dbl family GEF to recognize its G protein substrate.
The ␤3␤4 Loop Region of the PH Domain of P-Rex2 Is Required for Rac1 Interaction-The crystal structure of the Dbl family GEF Dbs in complex with its substrate Cdc42 shows that, in addition to interaction of the DH domain with Cdc42, the PH domain of Dbs has an extended ␤3␤4 loop, which makes contact with the switch loop 2 and helix 3b of Cdc42 (28). Alignment of the PH domain of P-Rex2 with other PH domains shows that the P-Rex2 PH domain also has an extended ␤3␤4 loop (Fig. 3A). To determine whether this loop of P-Rex2 makes contact with Rac1, a ␤3␤4 loop deletion mutant of P-Rex2 PH domain was tested for its ability to bind Rac1. As shown in Fig.  3B, the ␤3␤4 loop deletion caused a Ͼ80% reduction in the ability of the P-Rex2 PH domain to bind Rac1. This suggests that the ␤3␤4 loop of the PH domain of P-Rex2 is required for the interaction with Rac1.
The ␤3␤4 Loop Region of the PH Domain of P-Rex2 Is the Determinant of Rac Interaction-Although deletion of the ␤3␤4 loop of the PH domain of P-Rex2 caused a dramatic decrease in Rac1 binding, it was possible that this mutation affected the overall structure of the PH domain. The Dbl family GEF Dbs catalyzes exchange specifically on RhoA and Cdc42 and not on Rac1. Binding studies with the PH domain alone of Dbs showed that it did not interact with RhoA, Rac1, or Cdc42 (Fig. 4). To determine whether the ␤3␤4 loop of the PH domain of P-Rex2 would be sufficient to confer Rac1 recognition, we created a chimeric PH domain in which the ␤3␤4 loop region of the Dbs PH domain was replaced with that of P-Rex2. As shown in Fig.  4, this chimeric PH domain acquired the ability to specifically interact with Rac1 and not RhoA and Cdc42. This demonstrated that the ␤3␤4 loop of the PH domain of P-Rex2 is sufficient to produce Rac1 interaction.
Nucleotide Binding of Rac Influences Interaction with the PH Domain of P-Rex2-The crystal structure of Dbs in complex with Cdc42 shows that, in addition to contacts between the DH domain and Cdc42, the PH domain makes contact with the GTP-sensitive switch loop2 of Cdc42 (28). To test whether the PH domain of P-Rex2 can also sense the nucleotide status of Rac1, purified GST-tagged Rac1 was loaded with GDP or GTP␥S and tested along with nucleotide-free Rac1 for the ability to bind to the PH domain. As shown in Fig. 5, the nucleotide-bound forms of Rac1 showed a dramatic reduction in the ability to bind the PH domain. This suggests that the PH domain of P-Rex2 makes nucleotide-sensitive contacts with Rac1, possibly in the switch 2 loop region, by analogy to what was seen in the Dbs/Cdc42 crystal structure contacts.
Helix 3 of Rac Is the Region That Determines the Rac1 Specificity of Binding of the PH Domain of P-Rex2-The crystal structure of Cdc42 with Dbs shows that the Dbs PH domain also makes contact with helix 3b of Cdc42 (28). This region is adjacent to the switch loop 2 of Cdc42. Importantly, the primary sequence of helix 3 differs between Rac1 and Cdc42 (Fig.   FIG. 2. Substrate recognition and specificity is  6A). To determine whether this helix in Rac1 plays a role in the Rac-specific interactions with the PH domain of P-Rex2, helix 3 of Cdc42 was replaced with that of Rac1, and this chimeric Cdc42 was tested for its ability to interact with the P-Rex2 PH domain. As shown in Fig. 6B, the PH domain of P-Rex2 was now able to recognize the chimeric Cdc42 with helix 3 of Rac1. This demonstrated that the determinant for Rac1 recognition by the PH domain of P-Rex2 lies in helix 3.
P-Rex2 Does Not Catalyze Exchange on Chimeric Cdc42 with Helix 3-The previous binding experiments demonstrated that the P-Rex2 PH domain by itself can interact with a chimeric Cdc42 with a helix 3 from Rac1. To determine whether binding is sufficient for catalysis, we tested the ability of P-Rex2 to catalyze exchange on this chimeric Cdc42 in an in vitro exchange assay. As shown in Fig. 7, P-Rex2 is unable to catalyze exchange on the chimeric Cdc42. This chimera did not differ from wild-type Cdc42 in its ability to load N-methylantranioyl-GDP (data not shown). This suggests that binding ability does not necessarily predict catalytic ability, as the two events seem to be distinct from each other.

DISCUSSION
Activation of Dbl family GEFs in vivo lead to the activation of specific G protein signaling pathways (6). This specificity hinges on the ability of the GEFs to discriminate among the different G proteins present within the cell. Determination of the determinants of specificity of interaction between Dbl family GEFs and their cognate Rho GTPases is therefore necessary to understand signaling specificity. The crystal structures of Tiam1, Dbs, and Intersectin with Rac1, Cdc42/RhoA, or Cdc42, respectively, have provided a major breakthrough for identifying critical residues that determine the specificity of these interactions (25)(26)(27)(28). Attention has been focused on the catalytic DH domain in particular, as it was found to make major contact with the GTPase. Alignment of the primary sequence of the DH domain of Rac1-specific GEFs has isolated a conserved Ile residue that makes contact with Trp 56 of Rac1. Discrimination between Cdc42 and Rac1 was thought to be mediated through this Trp residue, as the corresponding residue in Cdc42 is a Phe. However, the Rac1W56F mutation had no effect on the binding ability of the DHPH domain of P-Rex2. Furthermore, the DH domain alone was incapable of binding Rac1. Because the DH domain is the catalytic domain and P-Rex2 is a Rac1-specific GEF, this result suggests that the DH domain may have to undergo novel conformational rearrangements in order for it to catalyze exchange on Rac1.
Although Dbl family GEFs are characterized by the presence of a DHPH domain, it is becoming increasingly clear that the PH domains of these GEFs are quite unlike PH domains found in other signaling proteins. In addition to the canonical membrane-targeting function of many PH domains, the PH domains of Dbl family GEFs have been shown to have several unique properties (28). The catalytic activity of the DH domain of Trio is increased 100-fold by the presence of the PH domain (12). Lipid binding by the PH domains of Vav, Sos, and Tiam1 have been shown to modulate the activity of the DH domain both in vitro and in vivo (33,34). Mutations in the PH domain that affect lipid binding has a big impact on the activity of the GEF in vivo in a manner that cannot be explained solely by differences in localization of the protein (35). The PH domain of Dbs makes contact with the switch loop 2 and helix 3b of Cdc42. Disruption of these interactions by the PH domain has a significant impact on the ability of Dbs to catalyze exchange on Cdc42 (28). The PH domain of Dbs was also recently shown to be an effector for Rac1, as it could bind to the GTP-bound form of Rac1 (36). Our own binding studies with the PH domain of P-Rex2 add new duties to the function of PH domains. The PH domain of P-Rex2 is responsible for substrate selectivity and recognition. This is the first instance where this function has been attributed to the PH domain of a Dbl family GEF. The ␤3␤4 loop region of the PH domain of Dbs makes contact with Cdc42 in the Dbs-Cdc42 complex. Interestingly, the same loop of the PH domain of P-Rex2 was shown to be necessary and sufficient for the interaction with Rac1. Although the PH domain of Dbs does not independently bind to its cognate GTPases, additional experiments will be required to test the PH domains of other Dbl family GEFs. P-Rex2 could select for Rac1 via its PH domain. The PH FIG. 5. The PH domain of P-Rex2 interacts with the switch loops of Rac1. GST-tagged Rac1 was loaded with GTP␥S or GDP and tested for their ability to interact with the PH domain of P-Rex2 as before. The PH domain of P-Rex2 showed diminished binding to the GTP␥S-or GDP-loaded form of Rac1 as compared with the nucleotidefree form of Rac1 (Rac1NF). Ponceau S staining of the nitrocellulose membrane shows protein levels.
FIG. 6. Helix 3 of Rac1 confers specificity to the interaction between Rac1 and the PH domain of P-Rex2. A, alignment of helix 3 of Rac1 with that of Cdc42. B, the PH domain of P-Rex2 was tested for its ability to bind to GST or GST-tagged forms of Rac1, Cdc42, and the Cdc42 chimera as before. Ponceau S staining of the nitrocellulose membrane shows protein levels. domain recognizes Rac1 by interacting with the switch loop and helix 3. Substitution of helix 3 of Cdc42 with that of Rac1 enabled the PH domain to interact with this chimeric protein. Moreover, it is likely that the PH domain of P-Rex2 interacts with switch loop 2 of Rac1 rather than switch loop1 because of its proximity with helix 3b. Even though P-Rex2 could interact with the Cdc42 chimera, it was unable to catalyze exchange on it. This suggests that substrate recognition and catalysis are two distinct events. Although the PH domain may recognize the substrate, it is the contacts being made by the catalytic DH domain that determine enzymatic activity. Other Dbl family GEFs have been previously shown to interact with many GTPases; however, they could catalyze exchange on a limited subset of these GTPases. Dbl, for example, can bind to RhoA, Rac1, and Cdc42 but will catalyze exchange on only RhoA and Cdc42 (28,37). Further studies will be required to delineate the contacts that are being made in these two distinct processes.
The structure of the DHPH domain of Dbl family GEFs shows that the orientation of the PH domain with respect to the DH domain can vary widely (28). The region in between the DH and PH domains vary greatly between different GEFs. However, it is this intervening region that determines the placement of the PH domain relative to the DH domain. Moreover, the NMR structure of Sos shows that this linker region is extremely flexible (13). This suggests that the orientation of the PH domain with the DH domain may be more flexible, and contacts with the cognate GTPase may be a more common occurrence than currently appreciated. This feature may not be easily discerned from the rigid crystal structure complexes. Given that the structure of Rho family GTPases are so conserved, Dbl family GEFs need multiple ways to specifically pair with the correct substrate. Future experiments will explore the details of these interactions.