A Novel PDZ Domain Containing Guanine Nucleotide Exchange Factor Links Heterotrimeric G Proteins to Rho*

Small GTP-binding proteins of the Rho family play a critical role in signal transduction. However, there is still very limited information on how they are activated by cell surface receptors. Here, we used a consensus sequence for Dbl domains of Rho guanine nucleotide exchange factors (GEFs) to search DNA data bases, and identified a novel human GEF for Rho-related GTPases harboring structural features indicative of its possible regulatory mechanism(s). This protein contained a tandem DH/PH domain closely related to those of Rho-specific GEFs, a PDZ domain, a proline-rich domain, and an area of homology to Lsc, p115-RhoGEF, and a Drosophila RhoGEF that was termed Lsc-homology (LH) domain. This novel molecule, designated PDZ-RhoGEF, activated biological and biochemical pathways specific for Rho, and activation of these pathways required an intact DH and PH domain. However, the PDZ domain was dispensable for these functions, and mutants lacking the LH domain were more active, suggesting a negative regulatory role for the LH domain. A search for additional molecules exhibiting an LH domain revealed a limited homology with the catalytic region of a newly identified GTPase-activating protein for heterotrimeric G proteins, RGS14. This prompted us to investigate whether PDZ-RhoGEF could interact with representative members of each G protein family. We found that PDZ-RhoGEF was able to form, in vivo, stable complexes with two members of the Gα12 family, Gα12 and Gα13, and that this interaction was mediated by the LH domain. Furthermore, we obtained evidence to suggest that PDZ-RhoGEF mediates the activation of Rho by Gα12 and Gα13. Together, these findings suggest the existence of a novel mechanism whereby the large family of cell surface receptors that transmit signals through heterotrimeric G proteins activate Rho-dependent pathways: by stimulating the activity of members of the Gα12 family which, in turn, activate an exchange factor acting on Rho.

proteins function in nuclear transport, and Ras plays a central role in cell proliferation and differentiation (2,4). In the case of Ras, recent studies have revealed how it works at the molecular level. This small GTP-binding protein exchanges GDP for GTP upon activation of Ras-specific guanine nucleotide exchange factors (GEFs) 1 (5), and in the GTP-bound state, Ras physically associates with Raf (6) thereby recruiting this serine threonine kinase to the plasma membrane. This causes the activation of Raf and initiates the activity of a sequential cascade of kinases leading to the stimulation of mitogen-activated protein kinases (MAPKs), p42 MAPK and p44 MAPK , also known as extracellular signal-regulated kinases-2 and -1, respectively, which, in turn, control the activity of nuclear transcription factors that are critical for cell growth (7,8).
The Rho family of GTP-binding proteins, which consists of the Rho, Rac, and Cdc42 subfamilies, has been shown to regulate several aspects of cytoskeleton function (4). For example, Rho participates in the formation of actin stress fibers and mediates the redistribution of cytoskeletal components (4,9). Rac is involved in the regulation of lamellipodia (pleat-shaped protrusions at the cell periphery) and membrane ruffling (10); and Cdc42 regulates the formation of thin finger-like cytoplasmic extensions known as filopodia (11). These proteins play an important role in the regulation of cell morphology, cell aggregation, tissue polarity, cytokinesis, cell motility, and also in smooth muscle contraction (12)(13)(14). However, recent evidence suggests that Rho proteins are also integral components of signaling pathways leading to transcriptional control. For example, Rac and Cdc42 regulate the activity of the c-Jun aminoterminal kinase (JNK) thereby affecting the transcriptional activity of c-Jun (15), and Rho has been recently shown to induce expression from the serum responsive element (SRE) through the transcriptional activation of the serum response factor (SRF) (16).
The functional activity of small GTP-binding proteins of the Ras superfamily is tightly regulated in vivo by proteins that control their GDP/GTP bound state. Whereas GEFs promote the exchange of GDP for GTP thus activating Ras-like proteins (2), GTPase-activating proteins increase the low intrinsic rate of GTP hydrolysis of small GTPases (reviewed in Ref. 3) and are negative modulators. The mechanisms of activation of GEFs for Ras by cell surface receptors have been intensely investigated. For example, the biochemical route connecting the epidermal growth factor-receptor tyrosine kinase to Ras has been recently identified (5,17), and includes the phosphorylation of the receptor itself on tyrosine residues, thus creat-ing docking sites for adapter molecules such as Grb2 and Crk. These adapter molecules, in turn, help recruit to the membrane Sos, a Ras-GEF, thereby inducing the exchange of GDP for GTP on Ras. In contrast, the signaling pathway regulating the activity of GEFs for Rho family members is still poorly understood.
Many GEFs for Rho, Rac, and Cdc42, including dbl, ost, lfc, lbc, vav, ect2, tim, and net (reviewed in Refs. 18 and 19) were discovered by virtue of their ability to transform NIH 3T3 cells when overexpressed or when activated by truncations. All these proteins share a 250-amino acid stretch of significant sequence similarity with Dbl, termed Dbl homology (DH) domain, adjacent to a pleckstrin-homology (PH) domain (18,19). The DH domain was shown to be responsible for nucleotide exchange activity toward GTPases of the Rho family (20,21). Of interest, one of these GEFs for Rho-like proteins, the protein product of the vav proto-oncogene, proto-Vav, also exhibits a Src homology (SH) 2 domain flanked by two SH3 domains (22,23), and we have recently shown that tyrosine phosphorylation of proto-Vav by hematopoietic specific tyrosine kinases can activate its GEF activity for Rac both in vitro and in vivo (24,25). However, no other GEF for Rho-like proteins has been found to be regulated by tyrosine phosphorylation, nor to contain a phosphotyrosine-binding domain such as an SH2 or PTB domains. Furthermore, the vast majority of the known GEFs for small GTP-binding proteins of the Rho family are expressed in a very restricted tissue-specific manner (18,19), and their mechanism of activation is still largely unknown.
In this study, we explored the existence of novel GEFs for Rho-like proteins possessing structural domains that might suggest a role in signal transduction. Here, we report the identification of a novel, ubiquitously expressed GEF for Rholike proteins containing a PDZ domain. This protein, termed PDZ-RhoGEF, was found to activate biochemical pathways specific for Rho, in a Rho-dependent manner. Interestingly, PDZ-RhoGEF was found to be closely related to the Drosophila DRhoGEF2, and recent genetic analysis suggests that DRho-GEF2 acts downstream of the concertina gene, a Drosophila G␣ 12 homolog. Here, we found that PDZ-RhoGEF physically associates in vivo with activated ␣ subunits of heterotrimeric G proteins of the G 12 family, G␣ 12 and G␣ 13 . Association was found to occur through a novel structural domain, termed Lsc homology (LH) domain, located between the PDZ and the DH domain, and also present in the NH 2 -terminal, regulatory domain of the lsc proto-oncogene product and its human homolog, p115-RhoGEF. This LH domain is distantly related to the G protein-binding region of a family of proteins known as regulators of G protein signaling (RGSs) (26). Together, our present findings suggest the existence of a novel pathway by which the large family of G protein-coupled receptors communicates to Rho through the activation of G 12 /G 13 and the physical association between G␣ 12 or G␣ 13 with LH containing GEFs for Rho, thereby stimulating Rho-dependent pathways.

EXPERIMENTAL PROCEDURES
Expression Plasmids-KIAA0380 (human PDZ-RhoGEF), kindly provided by T. Nagase, Kazusa DNA Research Institute, Japan, was subcloned into the pCEFL vector as an Asp718-NotI fragment, thus generating the pCEFL-PDZ-RhoGEF expression plasmid. Then, the coding sequence for the AU1 hexapeptide (DTYRYI) was cloned inframe with the open reading frame of PDZ-RhoGEF, immediately upstream of the termination codon, thus generating a carboxyl-terminal AU1-tagged PDZ-RhoGEF. cDNAs encoding deletion mutants of PDZ-RhoGEF, as indicated in the corresponding figures, were generated by restriction enzyme digestion or polymerase chain reaction amplification using pCEFL-PDZ-RhoGEF-AU1 as a template. Sequences of mutagenic oligonucleotides will be made available upon request.
Plasmids expressing epitope-tagged MAPK and JNK, pcDNA3 HA-MAPK and pcDNA3 HA-JNK, respectively, as well as expression plas-mids for constitutively activated forms of Ras, RhoA, G ␣q , G ␣i2 , G ␣s , G ␣12 , and G ␣13 , ␤ 1 and ␥ 2 subunits of G proteins and RhoI41, a RhoA mutant insensitive to the C3 toxin activity, were described previously (15,27). Reporter plasmids that express the chloramphenicol acetyltransferase (CAT) gene under the control of the wild-type serum response element from the c-fos promoter (SREwt), or a mutant lacking the ternary complex factor-binding site (SREmutL) as well as an expression vector for the C3 toxin were kindly provided by R. Treisman (16).
Northern Blot Analysis-Human multiple tissue Northern blots, each lane containing 2 g of poly(A) ϩ RNA, were purchased from CLONTECH. Total RNA was isolated from several cell lines by RNeasy kit (QIAGEN) according to the manufacturer's instructions, then separated by electrophoresis on a 2% denaturing formaldehyde-agarose gel (20 g of RNA/lane), and transferred to Hybond TM -N nylon membrane (Amersham Life Science). The cDNA probe used for analysis of the PDZ-RhoGEF mRNA was prepared using as a template a 1832-base pair SacI-XbaI fragment derived from pCEFL-PDZ-RhoGEF, containing both 3Ј-translated and -untranslated regions of PDZ-RhoGEF cDNA. Human ␤-actin cDNA (2.0 kilobase pairs) was also used as a control probe. Probes were labeled using a Random Primer DNA labeling kit (Boehringer-Mannheim) with [␣-32 P]dCTP, and RNA hybridization performed as described (28).
Cell Lines and Transfection-Human kidney 293T cells and COS-7 cells were maintained in Dulbecco's modified Eagle's medium (Life Technologies, Inc.) supplemented with 10% fetal bovine serum. NIH 3T3 fibroblasts were maintained in Dulbecco's modified Eagle's medium supplemented with 10% bovine calf serum. 293T cells were transfected by the calcium-phosphate precipitation technique to examine the expression of PDZ-RhoGEF and its mutants or using LipofectAMINE Plus TM reagent (Life Technologies, Inc.) according to the manufacturer's protocol for the co-immunoprecipitation study. Transfection into NIH 3T3 cells were carried out by the calcium-phosphate precipitation technique, and COS-7 cells were transfected by the DEAE-dextran method. In each experiment, total amount of DNA was adjusted to 3-10 g/plate with vector alone without insert. For the transfection into 293T cells, tissue culture plates were treated with phosphate-buffered saline containing 20 g/ml poly-D-lysine for 10 min before seeding the cells, to prevent them from detaching from the plates.
Kinase Assays-MAPK activity in cells transfected with an epitopetagged MAPK (HA-ERK2, referred in here as HA-MAPK) was determined as described previously (15), using myelin basic protein (Sigma) as a substrate. JNK assays in cells transfected with an epitope-tagged JNK (HA-JNK) was also determined as described previously (15), using purified, bacterially expressed GST-ATF2(96) fusion protein as a substrate. Samples were separated by SDS-gel electrophoresis on 12% acrylamide gels. Autoradiography was performed with the aid of an intensifying screen. Parallel lysates of cells transfected with the HA-MAPK or HA-JNK expression plasmids were processed for Western blot analysis using an antibody against the HA epitope.
Reporter Gene Assays-NIH 3T3 cells were transfected with different expression plasmids together with 1.0 g of pCMV-␤gal, a plasmid expressing the enzyme ␤-galactosidase, and 1.0 g of pSREmutL, the reporter plasmid expressing the CAT gene. 293T cells were transfected with expression vectors for PDZ-RhoGEF or its mutants together with 0.5 g of pCMV-␤gal and 0.5 g of pSREwt, the reporter plasmid expressing a luciferase gene under the control of the wild-type SRE. After overnight incubation, NIH 3T3 cells and 293T cells were washed twice with phosphate-buffered saline, and kept for approximately 24 h in Dulbecco's modified Eagle's medium supplemented with 0.5% calf serum or 0.5% fetal bovine serum, respectively. Cells were then lysed using reporter lysis buffer (Promega). Additional DNAs were added to the transfection mixtures as indicated in each figure. CAT activity was assayed in the cell extracts by incubation at 37°C for 10 -16 h in the presence of 0.25 Ci of [ 14 C]chloramphenicol (100 mCi/mmol) (ICN) and 200 g/ml butyryl-CoA (Sigma) in 0.25 M Tris-HCl, pH 7.4. Labeled butyrylated products were extracted using a mixture of Xylenes (Aldrich) and counted as described (15). Luciferase activity in cell extracts was measured using Luciferase assay system (Promega). ␤-Galactosidase activity present in each sample was assayed by a colorimetric method, and used to normalize for transfection efficiency.
Immunoprecipitation and Western Blot Analysis-To confirm the expression of PDZ-RhoGEF and its mutants, 293T cells were transfected with vector or expression vector for each PDZ-RhoGEF-AU1 DNA construct. Then, cells were cultured for 48 h, washed twice with phosphate-buffered saline, and lysed at 4°C in a buffer containing 25 mM HEPES, pH 7.5, 0.3 M NaCl, 1.5 mM MgCl 2 , 0.2 mM EDTA, 0.5 mM dithiothreitol, 1% Triton X-100, 0.5% sodium deoxycholate, 0.1% SDS, FIG. 1. PDZ-RhoGEF contains several domains involved in signal transduction and is widely expressed in many human tissues. A, amino acid sequence of PDZ-RhoGEF. The expected translational product of PDZ-RhoGEF, accession number AB002378, is depicted using the 20 mM ␤-glycerophosphate, 1 mM sodium vanadate, 1 mM phenylmethylsulfonyl fluoride, 20 g/ml aprotinin, and 20 g/ml leupeptin. For the co-immunoprecipitation study of PDZ-RhoGEF and heterotrimeric G protein subunits, 293T cells were transfected with vector or expression vector for PDZ-RhoGEF-AU1 or its mutants together with vector or expression vectors carrying cDNAs for the constitutively activated forms of G␣ s G␣ i2 , G␣ q , G␣ 12 , and G␣ 13 (G␣ s QL, G␣ i2 QL, G␣ q QL, G␣ 12 QL, and G␣ 13 QL, respectively) as well as plasmids expressing ␤ 1 and ␥ 2 cDNAs or epitope-tagged G␣ 12 QL and G␣ 13 QL (HA-G␣ 12 QL and HA-G␣ 13 QL, respectively). After culture for 48 h, the cells were washed twice with phosphate-buffered saline, then lysed at 4°C in 900 l of lysis buffer (20 mM Tris-HCl, pH 7.5, 150 mM NaCl, 20 mM MgCl 2 , 1 mM sodium vanadate, 1% Triton X-100, 10 M AlCl 3 , 10 mM sodium fluoride, 20 g/ml leupeptin, 20 g/ml aprotinin, and 1 mM phenylmethylsulfonyl fluoride), and centrifuged at 15,000 rpm for 10 min at 4°C. The epitopetagged PDZ-RhoGEF and its mutants or HA-G␣ 12 QL and HA-G␣ 13 QL were immunoprecipitated from the cleared lysates by incubation for 1 h at 4°C with the specific antibody against AU1 or HA, respectively. Immuno complexes were recovered with the aid of Gamma-bind Sepharose beads (Pharmacia).
Lysates and anti-AU1 and anti-HA immunoprecipitates were analyzed by Western blotting after SDS-polyacrylamide gel electrophoresis, transferred to Immobilon TM -P Transfer Membranes (Amersham Life Science) and immunoblotted with the corresponding rabbit antisera or mouse monoclonal antibody as indicated in each figure. Immuno complexes were visualized by enhanced chemiluminescence detection (Amersham Life Science) using goat anti-rabbit or anti-mouse coupled to horseradish peroxidase as a secondary antibody (Cappel). In this study, we used mouse monoclonal antibodies anti-HA epitope (12CA5, BABCO) and anti-AU1 epitope (BABCO). Polyclonal antiserum to ␤ subunits of heterotrimeric G proteins was kindly provided by William Simonds, and anti-G␣ 12/13 rabbit polyclonal serum was described before (29). Rabbit polyclonal antisera to G␣ q and to G␣ s were purchased from Santa Cruz Laboratories and to G␣ i2 from Upstate Biotechnology.

Identification of a Novel PDZ Domain Containing Guanine
Nucleotide Exchange Factor for Rho Proteins-We set out to identify candidate GEFs for small GTP-binding proteins of the Rho family which contain domains previously implicated in signal transduction. To that end, we searched available DNA data bases using a consensus amino acid sequence derived from the DH domain of known GEFs for Rho-related GTPases (2). This search revealed the existence of a number of yet uncharacterized proteins exhibiting DH-like domains (data not shown). Subsequent analysis of their DNA sequences and their expected translational products suggested that many of them encode putative GEFs for Rho-like proteins. One of them, KIAA0380, accession AB002378, was of particular interest and was further characterized. The corresponding plasmid DNA was obtained from Dr. T. Nagase, Kazusa DNA Research Institute, Japan, and its nucleotide sequence confirmed. Of interest, its open reading frame encodes a protein of 1522 amino acids, possessing areas of high homology to other signaling molecules (Fig. 1A). As depicted in Fig. 1B, this molecule contains a tandem of DH and PH domains, being the DH domain closely related to those of p115-RhoGEF (30) (53% identity, 74% similarity) and Lsc (31) (53% identity, 72% similarity), and to the recently identified DRhoGEF2 (32, 33) (39% identity, 64% similarity). Furthermore, this molecule displays 35% of identity and 45% homology when analyzed for global homology with p115-RhoGEF/Lsc. However, the DH and PH domains were more distantly related to the DH domain of Dbl (20) and PH domain of pleckstrin (34), respectively. This molecule also contains an NH 2 -terminal region exhibiting extensive homology to a recently identified structural domain termed PDZ (35), which is involved in protein-protein interactions, and is also present in DRhoGEF2. As discussed above, the DH domain is believed to be responsible for the nucleotide exchange activity of GEFs, and both p115-RhoGEF and Lsc have been shown to behave as Rho-specific guanine nucleotide releasing factors (30,36). Thus, the newly identified molecule, which was tentatively named PDZ-RhoGEF, might represent a novel exchange factor for Rho.
PDZ-RhoGEF exhibits additional structural features, a proline-rich region (amino acids 149 to 160) and another area (amino acid 290 -486) showing a high degree of homology (ϳ34% identity and ϳ53% similarity) to Lsc, p115-RhoGEF, and DRhoGEF2 (Fig. 1B), that was termed Lsc homology (LH) domain. Thus PDZ-RhoGEF exhibits a number of characteristics that suggest a function in signal transduction: a DH-PH domain highly related to those of Rho GEFs, a PDZ domain, and an area of homology to three other GEFs.
PDZ-RhoGEF Is Widely Expressed-The majority of the known GEFs for small GTP-binding proteins of the Rho family are expressed in a restricted tissue-specific manner. To investigate the pattern of expression of PDZ-RhoGEF, we performed Northern blot analysis of RNAs from a broad range of human tissues using a non-conserved region including both 3Ј-translated and -untranslated sequences, nucleotides 3959 -5790, of the PDZ-RhoGEF cDNA as a probe. As shown in Fig. 1C, a prominent RNA transcript of approximately 7 kilobases was readily detected in many human tissues, albeit to a different extent. PDZ-RhoGEF is highly expressed in the brain, testis, heart, ovary, and placenta, and to lower levels in kidney, pancreas, spleen, prostate, colon, skeletal muscle, lung, and liver. Whereas no expression was detected in the thymus and small intestine, two RNA species were expressed at comparable levels in peripheral blood leukocytes. The nature of the smaller transcripts detected in leukocytes, as well as that of the additional minor RNA species detected in placenta, spleen, brain, and heart is still unknown, and under current investigation. Similarly, transcripts for PDZ-RhoGEF were detectable in fre-1-letter amino acid code. B, sequence comparison of PDZ-RhoGEF with proteins possessing PDZ, LH, DH, and PH domains. Accession numbers: DRhoGEF2, AF032870; Rhophilin, U43194; GRIP, U88572; Lsc, U58203; p115-RhoGEF, U64105; proto-Dbl, P10911; pleckstrin, P08567. C, Northern blot analysis. A radiolabeled probe encompassing nucleotide sequence corresponding to both 3Ј-translated and -untranslated regions of PDZ-RhoGEF cDNA (1832 base pairs) was hybridized with poly(A) ϩ RNA blots prepared from the indicated human tissues. The probe for actin mRNA was hybridized with the same membrane as a control. The arrow indicates the band of actin mRNA. quently used mammalian cell lines, such as HeLa, 293T, COS-7, and normal human keratinocytes, but limited expression was observed in NIH 3T3 cells (data not shown). Thus, we can conclude that PDZ-RhoGEF is expressed in a large variety of human tissues, rather than in a tissue-restricted manner.
Expression of Wild-type and Truncated Mutants of PDZ-RhoGEF-To begin exploring the biochemical specificity of PDZ-RhoGEF and the relative contribution of each structural domain, we engineered expression plasmids for epitope-tagged forms of wild-type and truncated PDZ-RhoGEF mutants. Initially, PDZ-RhoGEF was subcloned in an expression vector, pCEFL (37), and then the coding sequence for the hexapeptide DTYRYI was cloned in-frame with the open reading frame of PDZ-RhoGEF immediately upstream of the termination codon, thus generating a carboxyl-terminal AU1-tagged PDZ-Rho-GEF. As shown in Fig. 2A, when transfected into 293T cells, wild-type PDZ-RhoGEF was readily detectable with an anti-AU1 epitope-specific antibody. Similarly, each sequential NH 2terminal deletion mutant, ⌬-127, ⌬-238, ⌬-702, and ⌬-956, lacking progressively, the PDZ domain, the proline-rich region, the LH domain, and the DH domain, as depicted in Fig. 2A, were also detected in transfected 293T cells.
PDZ-RhoGEF Fails to Induce MAPK Cascades-Whereas Ras controls the MAPK cascade, recent data suggest that Rac and Cdc42 can regulate the activity of the JNK pathway (15). Thus, as an approach to investigate whether PDZ-RhoGEF activates small GTPases of the Ras, Rac, or Cdc42 class, we investigated the ability of PDZ-RhoGEF and its deletion mutants to activate MAPK and JNK in COS-7 cells. As shown in Fig. 2B, none of the PDZ-RhoGEF expression plasmids enhanced the activity of co-transfected HA-MAPK or HA-JNK, although, 12-O-tetradecanoylphorbol-13-acetate addition and activated Ras potently stimulated MAPK, and anisomycin and activated Dbl strongly activated JNK when used as controls under identical experimental conditions. Thus, these data strongly suggest that PDZ-RhoGEF cannot activate Ras and Rac/Cdc42 regulated pathways when expressed in COS-7 cells.
PDZ-RhoGEF Activates SRE in a Rho-dependent Manner-Recently, Rho proteins have been shown to signal to the SRE through a pathway affecting the transcriptional activity of SRF, independent of any MAPK described to date (16). Thus, we next asked whether PDZ-RhoGEF could induce expression from a reporter plasmid containing a mutated SRE which eliminated the ternary complex factor-binding site and which was shown to be potently activated by Rho (16). As shown in Fig.  2C, an activated form of Rho potently induced expression from this reporter system when used as a control and, under identical experimental conditions, PDZ-RhoGEF caused a remarkable, nearly 15-fold, elevation in CAT activity. Furthermore, whereas deletion of the PDZ and the proline-rich region did not have any demonstrable effect, deletion of the entire NH 2 -ter-minal regulatory domain enhanced the ability of PDZ-RhoGEF to stimulate SRF-dependent transcription when expressed at comparable levels. In contrast, as was expected, the deletion mutant lacking also the DH domain failed to elevate CAT expression. Thus, PDZ-RhoGEF potently stimulates SRF-dependent transcription, and that can be enhanced upon deletion of a putative negative regulatory region located between the proline-rich region and the catalytically active DH domain.
As PDZ-RhoGEF induces the expression from an SRE-regulated reporter plasmid, we next explored whether Rho mediates this effect. As an approach, we took advantage of the finding that the botulinum toxin C3 ADP-ribosylates Rho thus preventing its activation (38,39). As shown in Fig. 3, cotransfection with a C3 toxin expression plasmid abolished the stimulation of CAT activity by the activated, NH 2 -terminal truncated form of PDZ-RhoGEF. Similar results were obtained using wild-type PDZ-RhoGEF (data not shown). To confirm a role for Rho, we then cotransfected PDZ-RhoGEF and C3 with an expression plasmid for a Rho mutant that is insensitive to the effects of C3, RhoI41 (38,39). As shown in Fig. 3, RhoI41 restored the ability of PDZ-RhoGEF to stimulate SRE-dependent expression. Taken together, we can conclude that PDZ-RhoGEF potently induces SRE, and that transcriptional stimulation requires a functional Rho protein. In line with these results, we have also found that PDZ-RhoGEF can potently induce the formation of actin stress fibers when expressed in of PDZ-RhoGEF (PDZ-RhoGEF-AU1) and its mutants were immunoprecipitated with anti-AU1 antibody and subjected to Western blot analysis with the antibody to AU1. Bands were visualized by the enhanced chemiluminescence technique using the appropriate horseradish peroxidaseconjugated goat antiserum. B, effects of wild-type and mutant PDZ-RhoGEF on MAPK and JNK activity. COS-7 cells were transfected with pcDNA3-HA-MAPK (1 g/plate) or pcDNA3-HA-JNK1 (2 g/plate) for MAPK and JNK assays, respectively, together with pCEFL vector or expression vectors carrying cDNAs for wild-type and NH 2 -terminal truncated mutants of PDZ-RhoGEF, the mutationally activated Ras (Ras-V12) and onco-Dbl, as indicated (PDZ-RhoGEF and its mutants: 3 g/plate, Ras-V12 and onco-Dbl: 1 g/plate). Treatments of cells with 10 g/ml anisomycin for 20 min or 100 ng/ml 12-O-tetradecanoylphorbol-13-acetate (TPA) for 10 min were used as a control. Kinase reactions were performed in anti-HA immunoprecipitates from the corresponding cellular lysates as described under "Experimental Procedures." Data represent the mean Ϯ S.E. of three independent experiments, expressed as fold increase with respect to vector-transfected cells (vector). Autoradiograms correspond to representative experiments. Western blot (WB) analysis was performed in the corresponding cellular lysates and immunodetected with the antibody to HA. C, effects of wild-type and mutants of PDZ-RhoGEF on the activity of SRE. NIH 3T3 cells were cotransfected with pSREmutL (1 g/plate) and pCMV-␤-gal (1 g/plate) plasmid DNAs together with expression vectors carrying cDNAs for wild-type and NH 2 -terminal truncated mutants of PDZ-RhoGEF (2 g/plate) or the constitutively activated mutants of Ras and RhoA (1.0 g/plate), as indicated. Cells were processed as described under "Experimental Procedures." The data represent CAT activity normalized by the ␤-galactosidase activity present in each cellular lysate, expressed as fold induction with respect to control cells, and are the mean Ϯ S.E. of triplicate samples from a typical experiment. Similar results were obtained in three separate experiments.
Madin-Darby canine kidney cells, a typical Rho-dependent effect (data not shown). Collectively, these data indicate that PDZ-RhoGEF can effectively activate Rho-dependent pathways.
PDZ-RhoGEF Associates Physically with Heterotrimeric G Protein ␣ Subunits of the G 12 Family-A genetic screen for components of the Rho signaling pathway in Drosophila led to the identification of DRhoGEF2, the Drosophila homolog of PDZ-RhoGEF, as an essential molecule for directing cell shape changes associated with gastrulation during early embryo development (32,33). Of interest, two other molecules were previously identified as critical for gastrulation, folded gastrulation, a secreted protein (40), and concertina, a G protein ␣ subunit related to G␣ 12 (41). Although a direct link between the folded gastrulation receptor, concertina and DRhoGEF2 is yet to be established, these studies suggested that this Rho-GEF might act downstream from heterotrimeric G proteins. Furthermore, a computer assisted search for proteins sharing areas of homology to the LH domain of PDZ-RhoGEF revealed that this domain exhibits limited similarity to a region within the putative catalytic domain of RGS14 (26), a newly discovered RGS for G protein ␣ subunits (Fig. 4). Together, these findings prompted us to explore whether PDZ-RhoGEF could interact directly with a heterotrimeric G proteins.
As an approach, we co-expressed in 293T cells the AU1tagged full-length PDZ-RhoGEF together with ␤ 1 ␥ 2 dimers or activated forms of representative members of each G protein ␣ subunit family, G␣ q , G␣ s , G␣ i2 , and G␣ 12 and G␣ 13 in 293T cells. In each case, we used G␣ proteins rendered constitutively active by replacing a critical glutamine residue within the C3 region that is essential for GTP hydrolysis for leucine (QL mutants) (1). Detailed biochemical characterization of each G protein subunit has been previously reported in our laboratory (29,(42)(43)(44)(45). As shown in Fig. 5A, the tagged PDZ-RhoGEF and all transfected G protein ␣ subunits and ␤␥ heterodimers were expressed, as judged by Western blot analysis using antiepitope monoclonal antibodies and G protein subtype-specific antisera. Surprisingly, we found that when PDZ-RhoGEF was immunoprecipitated, both members of the G␣ 12 family, G␣ 12 and G␣ 13 , co-immunoprecipitated with this putative nucleotide exchange factor for Rho. No other G protein subunit was found to co-immunoprecipitate with PDZ-RhoGEF, nor were G␣ 12 and G␣ 13 detectable in immunoprecipitates from control samples (Fig. 5A). To confirm these findings, we co-expressed the AU1-tagged PDZ-RhoGEF with NH 2 -terminal HA-tagged forms of G␣ 12 and G␣ 13 , and performed anti-AU1 and anti-HA Western blots on both anti-HA and anti-AU1 immunoprecipitates. As shown in Fig. 5B, the activated forms of G␣ 13 effi-ciently co-immunoprecipitated PDZ-RhoGEF. Similar results were obtained with G␣ 12 , although the co-immunoprecipitation of PDZ-RhoGEF was less efficient because the HA-tagged G␣ 12 was poorly expressed (data not shown). We concluded that both members of the G␣ 12 family physically associate in vivo with PDZ-RhoGEF, thus providing a direct link between heterotrimeric G proteins and small GTP-binding proteins of the Ras superfamily.
PDZ-RhoGEF Binds Heterotrimeric G Proteins through a Novel Structural Domain Designated LH-To investigate further the identity of the structural domains of PDZ-RhoGEF involved in binding to G␣ 12 and G␣ 13 , we used the progressive truncated forms of PDZ-RhoGEF described above, lacking the PDZ domain (⌬-127), the PDZ domain, and the proline-rich domain (⌬-238), and the entire NH 2 -terminal regulatory region (⌬-702). We also engineered additional deletion mutants, lacking the LH domain (⌬LH), the PH domain (⌬PH), and the DH/PH domain (⌬DH/PH), as depicted in Fig. 6A. All deletion mutants were efficiently expressed when transfected into 293T cells (see above and Fig. 6A). However, when co-expressed with an activated epitope-tagged form of G␣ 12 and G␣ 13 , we found that all PDZ-RhoGEF deletion mutants lacking the LH domain were unable to associate in vivo with these G protein ␣ subunits (Fig. 6B). These data indicate that the ability of PDZ-RhoGEF to bind heterotrimeric G proteins of the G␣ 12 family is strictly dependent upon the structural integrity of its LH domain.
To further study the role of the LH domain in signaling, we investigated the ability of each deletion mutant to activate SRE, a Rho-dependent event. As shown in Fig. 6C, the PH and DH domains are strictly required to induce SRE, and as described above, deletion of the PDZ and PDZ ϩ proline-rich domains did not affect the ability of PDZ-RhoGEF to stimulate SRE. However, deletion of the entire NH 2 -terminal regulatory region or the LH domain enhanced the activity of PDZ-Rho-GEF. Thus, these data strongly suggest that the NH 2 -terminal regulatory region exerts a negative regulatory activity on the catalytic DH/PH domains, and that the LH domain may be responsible for this inhibitory activity.

A DH and PH Deletion Mutant of PDZ-RhoGEF Prevents Signaling from G␣ 12 and G␣ 13 to Rho-dependent Pathways-
The availability of biochemically inactive PDZ-RhoGEF mutants prompted us to ask whether they can affect signaling from G␣ 12 and G␣ 13 to SRE, a Rho-dependent event (27). For these experiments, we chose to use 293T cells as the transfection efficiency in these cells is greater than that in NIH 3T3 cells, thus allowing us to control the expression of each transfected DNA construct. As shown in Fig. 7, expression in 293T   FIG. 4. Comparison of the amino acid sequences of PDZ-RhoGEF, p115-RhoGEF, and RGS14. The sequences between residues 290 and 486 of PDZ-RhoGEF, residues 25 and 232 of p115-RhoGEF and residues 57 and 235 of RGS14 were optimally aligned on the basis of residue identity (black) and similarity (gray). cells of PDZ-RhoGEF lacking the DH/PH domains or the entire NH 2 -terminal region ϩ DH domain (⌬-956) did not affect SRE activation by RhoQL, when used as a control. In contrast, the PDZ-RhoGEF DH/PH deletion mutant specifically blocked SRE activation by G␣ 12 and G␣ 13 . Similarly, this PDZ-RhoGEF mutant diminished the activation of SRE when mediated by lysophosphatidic acid receptors (data not shown). Thus, PDZ-Rho-GEF ⌬DH/PH behaves as a dominant negative mutant, probably by preventing the coupling of G␣ 12 and G␣ 13 to the endogenously expressed PDZ-RhoGEF or to other related GEFs. DISCUSSION Although small GTP-binding proteins of the Rho family play a critical role in a variety of cellular functions, including the organization of the actin cytoskeleton and the activity of biochemical routes regulating gene expression and cell growth, FIG. 5. Direct interaction of PDZ-RhoGEF with heterotrimeric G protein ␣ subunits of the G 12 family. A, 293T cells were transfected with vector or expression vector (C) for PDZ-RhoGEF-AU1 (P-AU1) together with vector or expression plasmids carrying cDNAs for the constitutively activated mutants of G␣ s , G␣ i2 , G␣ q , G␣ 12 , and G␣ 13 as well as plasmids expressing ␤ 1 and ␥ 2 cDNAs, as indicated. Lysates, prepared as described under "Experimental Procedures," were immunoprecipitated (IP) with anti-AU1 antibody and subjected to Western blot (WB) analysis using anti-AU1 antibody and G protein subtype-specific antisera. In addition, total cellular lysates (TCL) were subjected to Western blot analysis with G protein subtype-specific antisera. B, lysates from cells transfected with vector or expression plasmids for PDZ-RhoGEF-AU1 together with vector or expression vectors carrying cDNAs for an epitope-tagged and activated form of G␣ 13 (HA-G␣ 13 QL) were immunoprecipitated with anti-AU1 and anti-HA antibody, and were subjected to Western blot analysis with antibodies to HA and AU1, respectively. In addition, total cellular lysates were used in Western blot analysis with anti-AU1 and anti-HA antibodies. Bands were visualized by the enhanced chemiluminescence technique using the appropriate horseradish peroxidase-conjugated goat antiserum. In this figure, G␣, AU1, and HA are used for abbreviations anti-G␣ antiserum, anti-AU1 antibody, and anti-HA antibody, respectively. how these GTPases are activated by cell surface receptors is still largely unknown. Thus, we investigated whether novel exchange factors for Rho-related GTPases might exist, exhibiting functional domains suggestive of a role in signal transduction. For this, we took advantage of the observation that all known GEFs for Rho proteins exhibit a DH domain, a 250amino acid stretch of significant sequence similarity to Dbl, a transforming protein that was originally isolated from a diffuse B cell lymphoma (46). Using a consensus sequence for DH domains of Rho-exchange factors (2) to search DNA data bases, we identified a yet uncharacterized molecule exhibiting a putative DH domain. Detailed analysis of its primary sequence revealed that this molecule contained additional areas of sim-ilarity with known modular domains, including a PH domain, a PDZ domain, a proline-rich region, and an area of homology to p115-RhoGEF, DRhoGEF2 and Lsc, that was not found in any other GEF described so far, and that was termed LH domain. As many of these protein regions are likely candidates to participate in signal transmission, we decided to investigate further this novel putative exchange factor.
When the DH domain of this new molecule was compared with those of other DH containing proteins, we found that it was highly related to that of two Rho-specific GEFs, p115-RhoGEF (30) and Lsc (31), and to a recently described Drosophila Rho GEF, DRhoGEF2 (33). However, PDZ-RhoGEF was more distantly related to the DH domain of exchange factors  Fig.  2A). ⌬-LH, ⌬-PH, and ⌬-DH/PH constructs, lacking LH, PH, and both DH and PH domains, respectively, codes for amino acid sequences in which residues 171-484, 956 -1054, and 738 -1054 were deleted from PDZ-RhoGEF, respectively. Lysates from cells transfected with vector or with expression plasmid carrying an epitope-tagged PDZ-RhoGEF (PDZ-RhoGEF-AU1) and its deletion mutants were immunoprecipitated with anti-AU1 antibody and these immunoprecipitates were subjected to Western blot analysis with the antibody against AU1. B, 293T cells were transfected with vector or expression plamids for G␣ 12 QL or G␣ 13 QL together with vector or expression vectors carrying cDNAs for wild type and deletion mutants of PDZ-RhoGEF, as indicated. Lysates, prepared as described under "Experimental Procedures," were immunoprecipitated with anti-AU1 antibody and subjected to Western blot (WB) analysis using specific antiserum for heterotrimeric G proteins of the G 12 family. Total cellular lysates (TCL) were also subjected to Western blot analysis using the same antiserum. In A and B, bands were visualized by the enhanced chemiluminescence technique using the appropriate horseradish peroxidase-conjugated goat antiserum. C, effects of wild-type and deletion mutants of PDZ-RhoGEF on the activity of the SRE. NIH-3T3 cells were co-transfected with pSREmutL (1 g/plate) and pCMV-␤-gal (1 g/plate) plasmid DNAs together with expression vectors carrying cDNAs for wild-type and deletion mutants of PDZ-RhoGEF, as indicated. Cells were processed as described under "Experimental Procedures." The data represent CAT activity normalized by the ␤-galactosidase activity present in each cellular lysate, expressed as fold induction with respect to control cells, and are the mean Ϯ S.E. of triplicate samples from a typical experiment. Similar results were obtained in three independent experiments. activating Rac1 and/or Cdc42 such as Tiam1 (47), Vav (48), Ost (49), and Dbl (46), and to those acting on Ras, including Sos (50) and Ras-GRF (51). Consistent with this observation, expression of epitope-tagged forms of this novel GEF did not elevate the activity of co-transfected HA-tagged forms of MAPK and JNK, but potently stimulated the transcriptional activity of SRF, as judged by the use of a reporter plasmid under the control of a mutated SRE (16). Furthermore, experiments with the use of botulinum C3 exoenzyme, which ADP-ribosylates and inactivates Rho, and an inactivation resistant form of Rho, RhoI41, indicated that the enhanced expression from the SREdriven plasmid was dependent on the availability of a functional Rho. In addition, we have recently observed that expression of this molecule in Madin-Darby canine kidney cells potently induces the formation of actin stress fibers, to an extent comparable to that caused by expressing activated forms of Rho. 2 Thus, taken together, the primary sequence similarity with Rho GEFs and these biochemical profiles strongly suggest that this novel DH-containing molecule, that was designated PDZ-RhoGEF, can stimulate in vivo Rho-specific pathways.
In addition to the DH domain, a PH domain is present in all GEFs for Rho-related proteins described so far, located adjacent to the carboxyl end of the DH domain (18). PH domains are found in a wide variety of signaling molecules (52) and have been implicated in both protein-protein and protein-lipid interactions (53). Although a DH domain is necessary and sufficient for the exchange activity on Rho proteins in vitro (20), the integrity of the PH domain is required for the activity in vivo of this family of exchange factors (31) most likely by facilitating membrane translocation (31). Consistent with those observations, deletion of the DH domain, PH domain, or DH/PH domains abolished the biochemical activity of PDZ-RhoGEF. Similarly, PDZ-RhoGEF readily induced the appearance of foci of transformation when expressed in NIH 3T3 cells, and this activity required the presence of an intact DH and PH domain (data not shown), supporting a critical role for the DH and PH domains for the functional activity of PDZ-RhoGEF.
The most striking feature of this novel exchange factor was the presence of a PDZ domain, a protein-protein interaction domain originally identified as an area of homology between the product of the Drosophila dlg tumor suppresor gene and the synaptic protein PSD-95 (54), currently found in more than 60 distinct gene products (35). These domains can either bind specific recognition sequences such as the (S/T)XV motif at the carboxyl termini of certain proteins, or they can form hetero-or homodimers, suggesting that this modular protein-binding domain can participate in the formation of macromolecular complexes (35). Thus, the PDZ domain was expected to contribute to PDZ-RhoGEF function. Surprisingly, however, when this domain was deleted, we did not observe any demonstrable effect on the ability of PDZ-RhoGEF to induce Rho-dependent pathways. Similarly, deletion of the PDZ domain was shown not to affect the biological activities of the Rac1 exchange factor Tiam1 (55). Although, based on these results, we cannot rule out the possibility that the PDZ domain facilitates the interaction of PDZ-RhoGEF and other exchange factors with yet to be identified signaling molecules (see below), we therefore decided to focus our efforts on other putative functional domains. One such interesting domain is a stretch of 197 amino acids located 2 M. Zohar and J. S. Gutkind, unpublished data.

FIG. 7. Inhibition of G␣ 12 -and G␣ 13 -mediated SRE activation by the DH and PH deletion mutant of PDZ-RhoGEF.
293T cells were cotransfected with pSREwt (0.5 g/plate) and pCMV-␤-gal (0.5 g/plate) plasmid DNAs together with expression vectors carrying cDNAs for the constitutively activated mutants of G␣ 12 , G␣ 13 , and RhoA, ⌬-DH/PH, ⌬-956, and green fluorescent protein (GFP) (control), as indicated. Cells were processed as described under "Experimental Procedures." The data represent luciferase activity normalized by the ␤-galactosidase activity present in each cellular lysate, expressed as a percentage relative to that observed in control, and are the mean Ϯ S.E. of triplicate samples from a typical experiment. Similar results were obtained in three independent experiments.
FIG. 8. Proposed mechanism whereby G-protein coupled receptors stimulate Rho-dependent pathways. The stimulation of certain G protein-coupled receptors activates G␣ 12 and G␣ 13 which, in turn, interact with PDZ-RhoGEF through its LH domain, thereby causing the activation of PDZ-RhoGEF by a still unclear mechanism. Subsequently, the activated PDZ-RhoGEF catalyzes the exchange of GDP for GTP on Rho through its DH and PH domains, leading to the activation of Rho-dependent pathways. Additional possibilities are also discussed in the text. upstream from the DH domain, that was termed LH domain for Lsc homology domain, and that is also found in Lsc, p115-RhoGEF, and DRhoGEF2 but not in any other GEF. This suggests that the LH domain represents a distinctive feature of this subgroup of exchange factors, which might bear functional relevance. Indeed, deletion of the LH domain enhanced the ability of PDZ-RhoGEF to stimulate SRE, to an extent comparable to that of a truncation mutant lacking the entire NH 2terminal regulatory region. Similarly, we found that NH 2 -terminal truncated or LH-deleted forms of PDZ-RhoGEF exhibit enhanced focus forming activity in NIH 3T3 cells when compared with the wild-type form. 3 Thus, PDZ-RhoGEF, like other GEFs, appears to be negatively regulated by inhibitory sequences within the non-catalytic region (18), and this inhibitory function most likely resides in the LH domain.
Because of the possibility that the LH domain might have a regulatory function, we searched data bases for molecules displaying sequences related to the LH domain. Surprisingly, we found that the catalytic region of a recently described GTPaseactivating proteins for heterotrimeric G proteins, RGS14, exhibited a limited sequence similarity to the LH domain. RGSs were initially identified as homologues of Sst2 proteins, which are negative regulators of pheromone signaling in yeast (56). This protein family, currently with 19 members, shares a 120amino acid core, termed GH domain, which is essential for accelerating the rate of GTP hydrolysis on G␣ proteins (56). RGS14, together with RGS12, represent novel members of this family, characterized for being substantially larger (ϳ60 and ϳ140 kDa, respectively) than the majority of the other known RGSs (ϳ25 kDa) (26). This data suggested that the LH domain might confer to PDZ-RhoGEF the ability to interact with G protein ␣ subunits. Indeed, we found that PDZ-RhoGEF could form stable complexes in vivo specifically with members of the G 12 family of G protein ␣ subunits, G␣ 12 and G␣ 13 , and that this interaction required the presence of an intact LH domain. Taken together, we can conclude that PDZ-RhoGEF can interact physically with a particular subset of G␣ proteins, thereby providing a direct link between hetereotrimeric G proteins and small GTP-binding proteins of the Rho family.
These findings may have important implications regarding the function of G proteins of the G 12 family, G␣ 12 and G␣ 13 . These ubiquitously expressed G proteins were discovered by M. Simon's group upon amplification of mouse brain cDNA by polymerase chain reaction using degenerated oligonucleotides corresponding to regions highly conserved among G proteins (57). G␣ 12 and G␣ 13 exhibit 67% amino acid identity with each other, but only 35-44% of amino acid identity to ␣ subunits of other classes, such as G q, G i , and G s (57). Furthermore, whereas members of the G␣ q family of G proteins activate phosphatidylinositol-specific phospholipases, the G␣ s family stimulate adenylyl cyclases, and G␣ i inhibits adenylyl cyclases and activate certain phosphodiesterases and promote the opening of several ion channels (58 -60), members of the G␣ 12 family of GTPases appear not to affect any of these second messenger-generating systems (60). In this regard, the finding that concertina (cta), a Drosophila gene involved in embryogenesis (41), and that G␣ 12 and G␣ 13 can behave as remarkably potent oncogenes (29,42), provided early indications that this G protein class might be involved in growth regulation, albeit through poorly defined mechanisms. Intense investigation in many laboratories has recently generated a wealth of information on how G␣ 12 and G␣ 13 may act (see Ref. 61, for a recent review). In particular, one such study (62) demonstrated that activated G␣ 12 and G␣ 13 , but not G␣ i2 and G␣ q or different combinations of ␤ and ␥ subunits, mimicked the effect of activated RhoA on stress fibers and focal adhesion assembly, and we have recently provided evidence that G␣ 12 stimulates nuclear responses and cellular transformation through Rho (27). Furthermore, several studies have now provided evidence that members of the G 12 family of G proteins link many G proteincoupled receptors, including receptors for lysophosphatidic acid, thrombin, thromboxane A 2 , and acetylcholine to the activation of Rho-dependent pathways and, in many cases, cell growth control (27,(63)(64)(65)(66).
However, the nature of the molecules linking G␣ 12 and G␣ 13 to Rho remained largely unknown. In this regard, our present results suggest that PDZ-RhoGEF or other LH-containing Rho-GEF might act downstream from these G protein ␣ subunits in a biochemical route leading to Rho activation (see Fig. 8). Moreover, while our study was in the process of submission, it was reported that G␣ 13 could enhance the in vitro Rho-GEF activity of p115-RhoGEF, a distinct LH domain-containing exchange factor (67,68). However, how the interaction between RhoGEFs and G␣ 12 and/or G␣ 13 leads to Rho activation in vivo is still unclear. For PDZ-RhoGEF, it is possible that binding of G 12 proteins to its LH domain results in the translocation of PDZ-RhoGEF to the membrane where it can act on Rho. Alternatively, binding of G 12 to the LH domain might de-repress the functional activity of PDZ-RhoGEF by preventing its negative modulatory effect. In addition, although the functional significance of the PDZ domain is still unclear, it is noticeable that some receptors known to induce G␣ 12 and activate Rho exhibit a COOH-terminal PDZ-binding motif, such as SVV in the case of both human and murine lysophosphatidic acid receptors, EDG-2 and VZG-1, respectively (69,70). Thus, it is also possible that certain G 12 -coupled receptors might facilitate the recruitment of PDZ-RhoGEF by binding to its PDZ domain and, simultaneously, activating G 12 proteins. These, as well as other possibilities, are under current investigation.
We can conclude that our present findings support the existence of a novel mechanism whereby the large family of G protein-coupled cell surface receptors can stimulate Rho-dependent pathways (Fig. 8). This pathway involves the activation of G␣ 12 and/or G␣ 13 , which, in turn, will interact directly with Rho-exchange factors containing a G␣ 12 /G␣ 13 -binding region, such as an LH domain. This would result in the activation of Rho by a still unclear mechanism, thereby stimulating the activity of Rho-dependent pathways that, ultimately, would affect the cytoskeletal structure, nuclear gene expression, and cellular growth.