Selective uncoupling of G alpha 12 from Rho-mediated signaling.

The heterotrimeric G protein G(12) has been implicated in such cellular regulatory processes as cytoskeletal rearrangement, cell-cell adhesion, and oncogenic transformation. Although the activated alpha-subunit of G(12) has been shown to interact directly with a number of protein effectors, the roles of many of these protein-protein interactions in G(12)-mediated cell physiology are poorly understood. To begin dissecting the specific cellular pathways engaged upon G(12) activation, we produced a series of substitution mutants in the regions of Galpha(12) predicted to play a role in effector binding. Here we report the identification and characterization of an altered form of Galpha(12) that is functionally uncoupled from signaling through the monomeric G protein Rho, a protein known to propagate several Galpha(12)-mediated signals. This mutant of Galpha(12) fails to bind the Rho-specific guanine nucleotide exchange factors p115RhoGEF and LARG (leukemia-associated RhoGEF), fails to stimulate Rho-dependent transcriptional activation, and fails to trigger activation of RhoA and the Rho-mediated cellular responses of cell rounding and c-jun N-terminal kinase activation. Importantly, this mutant of Galpha(12) retains coupling to the effector protein E-cadherin, as evidenced by its ability both to bind E-cadherin in vitro and to disrupt E-cadherin-mediated cell-cell adhesion. Furthermore, this mutant retains the ability to trigger beta-catenin release from the cytoplasmic domain of cadherin. This identification of a variant of Galpha(12) that is selectively uncoupled from one signaling pathway while retaining signaling capacity through a separate pathway will facilitate investigations into the mechanisms through which G(12) proteins mediate diverse biological responses.

Heterotrimeric guanine nucleotide-binding regulatory proteins (G proteins) 1 mediate cellular signaling from heptahelical cell surface receptors to a wide variety of downstream effector proteins that in turn propagate signals to elicit certain cellular responses and changes. The ␣-subunits of the G 12 subfamily of G proteins, G␣ 12 and G␣ 13 , have been linked to cellular events such as cytoskeletal rearrangements (1,2), cell proliferation (3), Na ϩ /H ϩ exchange (4,5), activation of phospholipase C-⑀ (6), activation of phospholipase D (7), down-regulation of cell-cell adhesion (8), activation of radixin (9), and modulation of tight junction-mediated paracellular permeability (10). Furthermore, mutationally activated G␣ 12 and G␣ 13 have been demonstrated to elicit oncogenic transformation of several cell lines (11,12). Although a number of proteins involved in these processes have been shown to bind directly to activated (GTPliganded) G␣ 12 and/or G␣ 13 (13), the roles of these effector proteins in mediating particular G 12 -dependent cellular events are still largely undefined.
Two subsets of G 12 effectors that are the best characterized are a class of Rho-specific guanine nucleotide exchange factors that includes p115RhoGEF, leukemia-associated RhoGEF (LARG), and PDZ-RhoGEF, as well as members of the cadherin superfamily of cell surface adhesion proteins (14 -19). Cell-free studies using purified proteins have demonstrated that G␣ 13 directly stimulates the ability of p115RhoGEF to enhance guanine nucleotide exchange on the monomeric G protein RhoA and, reciprocally, that p115RhoGEF binding accelerates the rate of GTP hydrolysis by G␣ 13 and G␣ 12 (16,18). Also, activated G␣ 12 and G␣ 13 have been demonstrated to bind the cytoplasmic domain of several cadherins in vitro, and expression of mutationally activated G␣ 12 and G␣ 13 in cells disrupts the extracellular adhesive function of epithelial cadherin (Ecadherin) in a manner that requires direct G␣ 12 -cadherin interaction (8,19). Furthermore, binding of activated G␣ 12 to cadherin results in release of the cytoplasmic protein ␤-catenin from cadherin, allowing ␤-catenin to act as a transcriptional activator of genes involved in cell proliferation, differentiation, and oncogenesis (17,19).
As the number of known G 12 binding partners has increased, so has the apparent complexity of the signaling networks that originate with the receptor-driven activation of G 12 proteins. To elucidate the biological significance of the interactions between G 12 proteins and their various effectors, reagents that selectively manipulate the interaction between G␣ 12/13 and individual target proteins would be of great value, as these tools could reveal the role of particular G 12 -effector interactions in specific signaling events. To this end, we have introduced a series of mutations into the primary sequence of G␣ 12 and from these have identified a mutant that is impaired both in binding Rho-specific guanine nucleotide exchange factors and in activating Rho-mediated signaling pathways. This mutant retains normal binding to E-cadherin as well as the ability to disrupt cadherin function when expressed in cells. This variant of G␣ 12 provides a novel reagent for dissecting the roles of distinct downstream effector pathways that are triggered following G␣ 12 activation and also provides important new structurefunction information regarding the nature of the interaction between G 12 proteins and Rho-specific guanine nucleotide exchange factors.

EXPERIMENTAL PROCEDURES
Materials-The Myc-p115RGS plasmid and the plasmid containing a GST fusion to the RGS domain of LARG were gifts from Tohru Kozasa (University of Illinois, Chicago). The pGEX-2T plasmid containing a GST fusion to the rhotekin RhoA-binding domain (GST-RBD) was kindly provided by Robert Lefkowitz (Duke University, Durham, NC). The reporter plasmid SRE-L was a gift from Channing Der (University of North Carolina, Chapel Hill), and the internal control reporter plasmid pRL-TK and dual-luciferase system were purchased from Promega. Anti-G␣ 12 and anti-RhoA antibodies were purchased from Santa Cruz Biotechnology and anti-Myc antibody from Roche Applied Science. Anti-␤-catenin antibody was purchased from Zymed Laboratories Inc., and Cy3-conjugated secondary antibody from Jackson Immunoresearch (West Grove, PA). Mouse monoclonal antibody specific for phospho-SAPK (stress-activated protein kinase)/JNK (Thr 183 /Thr 185 ) was purchased from Cell Signaling Technology (Beverly, MA). Protease inhibitors were purchased from Sigma.
Construction of Plasmids-The Myc epitope tag (EQKLISEEDL) was introduced into mutationally activated G␣ 12 (G␣ 12 QL ) by first creating a silent AgeI restriction site within the cDNA for G␣ 12 QL using the QuikChange site-directed mutagenesis kit (Promega) according to the manufacturer's instructions. This site was positioned to allow insertion of the Myc tag between proline 139 and valine 140 of G␣ 12 QL . Next, the oligonucleotides 5Ј-ccggtatcaggaggtggtggatctgagcagaagctgatcagcgaggaggacctgtcaggtggaggaggttca-3Ј and 5Ј-ccggtgaacctcctccacctgacaggtcctcctcgctgatcagcttctgctcagatccaccacctcctgata-3Ј were synthesized as 5Ј-end phosphorylated forms, combined at 1 M each in 10 mM Tris, pH 7.5, 1 mM EDTA, and allowed to anneal by incubation at 95°C for 5 min with subsequent cooling from 85 to 30°C over a period of ϳ4 h. The resulting double-stranded oligonucleotide, containing the Myc tag flanked on each side by the flexible linker sequence SGGGGS (20) and harboring appropriate 4-base overhangs, was then ligated into AgeI-digested G␣ 12 QL . Correct orientation of the Myc tag was verified by sequencing. NAAIRS substitution mutants were generated in Myc-tagged G␣ 12 QL as follows. For each 6-amino acid sequence designated for replacement by the sextet Asn-Ala-Ala-Ile-Arg-Ser, an oligonucleotide was designed that contained the 15 nucleotides immediately upstream of the designated 6-codon sequence within G␣ 12 QL , followed by the nucleotide sequence 5Ј-aatgctgctatacgatcg-3Ј that encodes the amino acid sequence NAAIRS, followed by the 15 nucleotides within G␣ 12 QL immediately downstream of the 6-codon sequence. An antiparallel, precisely complementary oligonucleotide was also synthesized, and these two oligonucleotides were used in the QuikChange procedure to introduce the desired NAAIRS substitution. All constructs were verified by sequencing.
The construction and purification of the adenovirus harboring G␣ 12 QL has been described previously (8). The cDNA encoding the ⌬p115 variant of G␣ 12 QL (see "Results") was subcloned into the adenoviral shuttle vector pAdTrak-CMV, which also contains the green fluorescent protein (GFP) cDNA, with each cDNA positioned downstream of a separate cytomegalovirus promoter. Correct subcloning of the cDNA was confirmed by sequencing. The recombinant pAdTrak plasmid and the adenoviral backbone plasmid pAdEasy-1 were co-transformed by electroporation into competent BJ5183 Escherichia coli. Recombinant viral DNA was purified by cesium chloride density centrifugation. The recombinant adenovirus was amplified in HEK293 cells and purified using the Adeno-X virus purification kit (BD Biosciences). A recombinant adenovirus produced using the empty pAdTrack-CMV vector was purified by the same procedure and used as a control.
Production of Recombinant Proteins-The glutathione S-transferase (GST) fusion of the N-terminal domain of p115RhoGEF (p115RGS) was produced by PCR amplification of this domain (residues 1-252) from the Myc-p115RGS plasmid with restriction sites incorporated into the amplified product to facilitate its subcloning into the plasmid pGEX-2T (Amersham Biosciences). This construct was verified by sequencing.
Fusions of GST to p115RGS, to the cytoplasmic domain of E-cadherin (19), and to the RGS domain of LARG were produced in BL21-DE3 Gold cells (Stratagene) and purified using glutathione-coated Sepharose (Amersham Biosciences) as described previously (19). GST-rhotekin RhoA-binding domain was purified as described previously (21).
In Vitro Binding Assays-HEK293 cells, grown in 6-well plates to ϳ70% confluency, were transfected with either Myc-tagged G␣ 12 QL , empty pcDNA3.1 plasmid, or various NAAIRS mutants of Myc-tagged G␣ 12 QL using Lipofectamine reagent (Invitrogen) according to the manufacturer's instructions. Approximately 48 h post-transfection, cells were washed twice with phosphate-buffered saline (PBS), harvested by scraping in PBS, pelleted by centrifugation at 800 ϫ g, and then resuspended in lysis buffer (50 mM HEPES, pH 8.0, 1 mM EDTA, 3 mM dithiothreitol, 10 mM MgSO 4 , 1% polyoxyethylene-10-lauryl ether) supplemented with the protease inhibitors TPCK (61 M), TLCK (58 M), and phenylmethylsulfonyl fluoride (267 M), and continually inverted at 4°C for 30 min. This lysate was centrifuged at 100,000 ϫ g for 45 min at 4°C, and the resulting supernatant was diluted 10-fold using lysis buffer lacking polyoxyethylene-10-lauryl ether. An aliquot of this solution was set aside, and the remainder was divided equally into tubes that received purified GST fusion proteins immobilized on glutathione-Sepharose (see above). Samples were continually inverted for 2 h at 4°C, and then glutathione-Sepharose was pelleted by centrifugation at 1300 ϫ g and washed four times with 1 ml of reduced-detergent lysis buffer (containing 0.1% polyoxyethylene-10-lauryl ether). Pelleted material was resuspended and subjected to SDS-PAGE and immunoblot analysis as described (17) in order to detect G␣ 12 QL or its NAAIRS variants.
For binding assays utilizing untagged 35 S-labeled variants of G␣ 12 QL , the proteins were produced using a TNT in vitro coupled transcription/ translation system (Promega) according to the manufacturer's instructions. Reactions were diluted into reduced detergent lysis buffer (see above) and incubated with GST fusion proteins as described above, and then proteins were separated by SDS-PAGE and gels fixed in 10% acetic acid, 1% glycerol, dried under vacuum, and analyzed by autoradiography.
Luciferase Reporter Assays-HEK293 cells were transfected with the SRE-L plasmid (containing the cDNA for firefly luciferase positioned downstream of serum response element) and the pRL-TK plasmid (containing the cDNA for Renilla luciferase positioned downstream of a thymidine kinase promoter) plus a plasmid encoding G␣ 12 QL or a mutant variant. Approximately 36 h post-transfection, cells were washed with serum-free Dulbecco's modified Eagle's medium and then incubated in the same medium for an additional 16 h. Cells were washed with PBS and then incubated in Passive ® lysis buffer (Promega) for 20 min. Lysates were cleared by centrifugation and then assayed by luminometry for firefly and Renilla luciferase activities using a dual-luciferase assay system (Promega). Firefly luciferase activity measurements were normalized for the corresponding Renilla luciferase values. Measurements were performed using a TD-20/20 luminometer (Turner Designs, Sunnyvale, CA).
Cell Rounding Assays-MDA-MB-231 cells were seeded at a density of 200,000 cells/dish on 35-mm glass-bottom Petri dishes (MatTek, Ashland, MA) and allowed to grow for 24 h. Cells were infected with adenovirus harboring a cDNA encoding GFP plus either G␣ 12 QL , a variant of G␣ 12 QL , or no cDNA. Infections were allowed to proceed for 4 h, and then cells were serum-starved for 15-16 h. Cell rounding phenotype was visualized using an Eclipse TE300 inverted microscope (Nikon).
Rho Activation Assays-MDA-MB-231 cells were seeded at a density of ϳ250,000 cells/well in 6-well plates and allowed to grow for ϳ24 h. Cells were infected with adenovirus harboring a cDNA encoding GFP plus either G␣ 12 QL , a variant thereof, or no cDNA. Infected cells were incubated for 4 h and then serum-starved for an additional 17-20 h. Rho activation assays were performed as described previously (21). Briefly, cells were washed quickly with PBS while on ice, and cold lysis buffer (25 mM HEPES, pH 8.0, 150 mM NaCl, 10 mM MgCl 2 , 1 mM EDTA, 1% Nonidet P-40, 2% glycerol, 10 g/ml leupeptin, 10 g/ml aprotinin) was added to the wells. Cells were lysed for 10 min and then centrifuged at 16,000 ϫ g for 5 min at 4°C. Supernatants were assayed for protein concentration. An aliquot was saved from each lysate to assay for total endogenous RhoA levels, and equal amounts of protein incubated with the GST fusion of the RhoA-binding domain of rhotekin immobilized on glutathione-Sepharose. Samples were mixed for 1 h at 4°C, and then the glutathione-Sepharose was pelleted by centrifugation at 700 ϫ g and washed three times with cold lysis buffer. Pelleted material was resuspended in SDS-PAGE sample buffer, separated by gel electrophoresis, and subjected to immunoblot analysis to detect RhoA.
c-jun N-terminal Kinase Activation Assays-MDA-MB-231 cells were seeded at ϳ250,000 cells/well in 6-well plates and allowed to grow for 24 h. Cells were infected with adenovirus harboring a cDNA encoding GFP plus either G␣ 12 QL , a variant thereof, or no cDNA. Infections were allowed to proceed for 4 h, and then cells were serum-starved for 24 h. Cells were then washed quickly twice with cold PBS while on ice, and cold c-jun N-terminal kinase (JNK) assay buffer (50 mM Tris, 1% Nonidet P-40, 0.5% sodium deoxycholate, 0.1% SDS, 2 mM EDTA, 50 mM NaF, 150 mM NaCl, 2 mM DTT, 0.2 mM sodium vanadate, 2.5 mM sodium pyrophosphate, 1 mM ␤-glycerophosphate, 3 g/ml leupeptin, 4 g/ml aprotinin, 30 M TPCK, 29 M TLCK, 133 mM phenylmethylsulfonyl fluoride) was added to the wells. Cells were lysed for 10 min and then centrifuged at 16,000 ϫ g for 5 min at 4°C. Supernatants were assayed for protein concentration, and then equal amounts of total protein from lysates were separated by SDS-PAGE and subjected to immunoblot analysis to detect levels of phospho-JNK.
Cell Aggregation Assays-Parental MDA-MB-435 cells and those stably expressing E-cadherin were infected with recombinant adenoviruses harboring G␣ 12 QL , a variant of G␣ 12 QL , or a control adenovirus. Three days post-infection, cells were subjected to fast-aggregation assays as described previously (8).
Indirect Immunofluorescence-DLD-1 and HEK293 cells grown on glass coverslips were washed twice in PHEM buffer (60 mM PIPES, pH 6.9, 25 mM HEPES pH 7.0, 10 mM EGTA, 4 mM MgSO 4 ) and then incubated at 37°C for 20 min in the same buffer containing 4% paraformaldehyde. Cells were then permeabilized by a 5-min incubation in PHEM buffer containing 0.5% Triton X-100 followed by three 5-min washes in PHEM containing 0.1% Triton X-100. A blocking solution of PHEM buffer containing 10% goat serum (Invitrogen) was added, cells were incubated for 30 min at 37°C, and then primary antibody to ␤-catenin was applied at a 1:100 dilution in PHEM buffer plus 5% goat serum. Following an overnight incubation at 4°C, cells were washed three times in PHEM plus 0.1% Triton X-100, and then Cy3-conjugated secondary antibody was applied at a 1:500 dilution in PHEM buffer plus 5% goat serum for 1 h. Cells were then washed three times as described above and incubated for 5 min in Hoechst stain (Molecular Probes, Eugene, OR) at a 1:1000 dilution in PHEM buffer, washed with distilled water, and mounted onto slides using ProLong antifade solution (Molecular Probes). Cells were visualized using an LSM-410 laser scanning confocal microscope (Carl Zeiss).

RESULTS
To identify structural determinants of G␣ 12 necessary for its coupling to downstream effector proteins, we designed a series of substitution mutants within the mutationally activated (QL) form of G␣ 12 . To allow for proper post-translational modification (e.g. acylation) of these variants of G␣ 12 QL , the proteins were expressed in HEK293 cells. To distinguish the ectopically expressed G␣ 12 QL variants from endogenous, wild-type G␣ 12 , a Myc epitope tag was inserted into G␣ 12 QL at the ␣B/␣C loop within the helical domain of the protein. Structural analyses of other G␣ subunit proteins have revealed that this highly conserved region is spatially removed from the GTP-binding and known effector-binding domains of the ␣-subunits, and this region was used to insert GFP into G␣ q without disrupting interaction between G␣ q and its effector protein, phospholipase C-␤ (20). Myc-tagged G␣ 12 QL expressed in HEK293 cells exhibited binding to the G␣ 12 effectors E-cadherin and p115RhoGEF (Fig. 1A), which was essentially identical to that of untagged G␣ 12 QL (data not shown). We first attempted to uncouple G␣ 12 from the G 12 -specific RGS protein p115RhoGEF by converting a highly conserved glycine within the "Switch I" region of G␣ 12 to a serine, because other G␣ subunits in mammalian cells and yeast have been uncoupled from RGS proteins by this alteration (22,23). We generated this Gly-to-Ser mutation within Myc-tagged G␣ 12 QL . This variant, denoted as G208S-G␣ 12 QL , was produced in HEK293 cells and extracted from membrane preparations of the cells, and then binding to GST fusions of either the cytoplasmic C-terminal domain of E-cadherin or the N-terminal RGS domain of p115RhoGEF was assessed. As shown in Fig.  1A, Myc-tagged G␣ 12 QL bound strongly to both of these effector proteins, and the apparent affinity of G208S-G␣ 12 QL for both effector proteins was not significantly changed.
Because the single point mutant of G␣ 12 did not yield the type of altered function we sought, we embarked on a more global strategy to identify mutant forms of G␣ 12 impaired in effector binding. To this end, we produced a series of substitution mutants within the region of G␣ 12 that encompasses all three of the "Switch regions" that are known to play a critical role in other G␣-effector interactions (24,25). To produce each mutant, a sextet of consecutive amino acids in the primary sequence of Myc-tagged G␣ 12 QL was replaced by the sequence Asn-Ala-Ala-Ile-Arg-Ser (NAAIRS), which is believed to be a well tolerated substitution in proteins because of its appearance in both ␤-sheet and ␣-helical secondary structures (26). This NAAIRS mutagenesis strategy has been successfully employed to dissect functional domains in the retinoblastoma protein (27) and in telomerase (28).
The panel of G␣ 12 NAAIRS mutants was expressed in HEK293 cells and then extracted from membranes. The amount of plasmid DNA used for transfecting cells was varied to achieve similar levels of ectopic G␣ 12 expression, as determined by immunoblot analysis (data not shown). These G␣ 12 variants were screened for binding to p115RhoGEF and Ecadherin as described above. In all, 13 variants were analyzed that covered the primary sequence of G␣ 12 from just upstream of the Switch I region to just downstream of the Switch III region (see Fig. 1B). None of these proteins bound to immobilized GST lacking a protein adduct (Fig. 1A, and data not shown). The majority of these variants exhibited binding to the GST-p115RGS and GST-E-cadherin proteins that was not markedly different from that of parental G␣ 12 QL (for an example, see results for the ⌬238 -243 variant in Fig. 1A; others are not shown). A few additional variants showed a marked decrease in binding to both effector proteins; an example of this is  Fig. 1A. This may reflect a nonspecific, global effect of the mutation on the structure of G␣ 12 , and therefore these variants were not pursued further. However, one NAAIRS variant, designated ⌬244 -249 in Fig.  1A, showed a nearly complete loss of binding to p115RhoGEF while retaining normal binding to E-cadherin. This variant also bound to GST fusions of the cytoplasmic domains of neural cadherin and cadherin-14 (data not shown). The region of G␣ 12 altered in the ⌬244 -249 variant lies immediately downstream of the Switch II region of G␣ 12 (Fig. 1B). To ensure that the Myc epitope tag did not influence binding of G␣ 12 QL or its variants to these effector proteins, several of the G␣ proteins, including G␣ 12 QL , G208S-G␣ 12 QL , and the ⌬244 -249 variant, were produced without the Myc tag in a cell-free transcription/translation system (see "Experimental Procedures"). Pull-down experiments testing these proteins for binding to GST-E-cadherin and GST-p115RGS yielded essentially the same results observed in Fig. 1A (data not shown). Based on these initial findings, we designated this variant of G␣ 12 QL as ⌬p115-G␣ 12 QL to denote loss of its interaction with p115RhoGEF; this designation is used for the remainder of this report.
The impaired binding of the ⌬p115-G␣ 12 QL variant to p115RhoGEF suggested that this protein would be functionally uncoupled from Rho signaling. To test this hypothesis, we first examined the ability of ⌬p115-G␣ 12 QL to activate serum response factor (SRF)-mediated transcription in cultured cells. G␣ 12 signaling to SRF has been well characterized as a Rhodependent pathway because of its sensitivity to specific inhibitors of Rho (29). HEK293 cells were transfected with a reporter plasmid harboring a luciferase cDNA downstream of an SRFresponsive element. Co-transfection with a plasmid encoding G␣ 12 QL elicited an approximately 6-fold increase in SRF activity ( Fig. 2A). This effect was blunted by the co-expression of p115RGS, which contains the G protein-interacting domain of p115RhoGEF and acts as a dominant negative by sequestering activated G␣ 12 (18,30). In contrast to the results obtained with G␣ 12 QL , expression of ⌬p115-G␣ 12 QL failed to stimulate SRFmediated transcriptional activation ( Fig. 2A). Immunoblot analysis of cell lysates verified that ⌬p115-G␣ 12 QL was expressed at levels comparable with G␣ 12 QL in these experiments (Fig. 2B).
The G 12 subfamily ␣-subunit G␣ 13 stimulates the ability of p115RhoGEF to trigger guanine nucleotide exchange on the small monomeric G protein Rho. Although G␣ 12 interacts strongly with the RGS domain of p115RhoGEF in vitro (16) (also see Fig. 1A), there are little data available on the ability of G␣ 12 to stimulate Rho activation specifically through p115RhoGEF. However, evidence has emerged that a closely related Rho-specific guanine nucleotide exchange factor, LARG, is functionally activated by G␣ 12 in cells (31). Therefore, we next tested G␣ 12 QL and the ⌬p115-G␣ 12 QL variant for interaction with LARG. As shown in Fig. 2C, G␣ 12 QL was clearly bound by an immobilized GST fusion of the RGS domain of LARG, but ⌬p115-G␣ 12 QL exhibited essentially no binding to this LARG fusion protein, even though analysis of cell extracts confirmed similar expression levels of these G␣ proteins (Fig.  2C). Hence, the ⌬p115-G␣ 12 QL variant is compromised in binding to both p115RhoGEF and the closely related LARG.
Two additional, well documented readouts of G␣ 12 -dependent cellular signaling mediated through RhoA are cell rounding (8,32,33) and the activation of JNK (34 -36). Therefore, as additional tests of whether the ⌬p115 variant is uncoupled from Rho-mediated signaling pathways, we examined the abilities of G␣ 12 QL and ⌬p115-G␣ 12 QL to stimulate these responses when ectopically expressed in cells. As shown in Fig. 3, A and B, essentially all MDA-MB-231 cells infected with a recombinant adenovirus encoding G␣ 12 QL exhibited a marked change from a flattened, splayed morphology to a distinctly rounded morphology, whereas a control adenovirus caused no such effect. However, cells infected with an adenovirus encoding the ⌬p115-G␣ 12 QL variant retained the normal, flattened appearance that most closely resembled the control cells (Fig. 3C), even though cells expressed similar levels of the variant as of G␣ 12 QL (Fig. 3D). 2. A and B, ability of G␣ 12 QL and the ⌬p115-G␣ 12 QL variant to stimulate serum response element-mediated transcriptional activation. HEK293 cells were transfected with the SRE-L reporter plasmid, the pRL-TK reporter control plasmid, and additional plasmids as described below. Approximately 40 h post-transfection, cells were washed and then grown in serum-free medium for an additional 15 h. Cells were then lysed and cleared by centrifugation, and the lysates were analyzed by luminometry as described under "Experimental Procedures." A, the ratio of firefly luciferase activity (responsive to SRF) to Renilla luciferase activity (not responsive to SRF) is shown for cells transfected with either pcDNA3.1 (vector), Myc-tagged G␣ 12 QL (12QL), or Myc-tagged ⌬p115-G␣ 12 QL (⌬p115). These cell samples were co-transfected with either empty pcDNA3.1 vector (solid bars) or a plasmid harboring Myc-tagged p115RGS (dotted bars). Results shown are from a representative experiment of four separate experiments. B, expression of G␣ 12 QL proteins and Myc-p115RGS in cell samples analyzed in A were assessed by SDS-PAGE/immunoblot analysis of lysates using, respectively, anti-G␣ 12 and anti-Myc antibodies (see "Experimental Procedures"). C, interaction of G␣ 12 QL and ⌬p115-G␣ 12 QL with LARG. HEK293T cells were transfected with pcDNA3.1 (vector), Myc-tagged G␣ 12 QL (12QL), or Myc-tagged ⌬p115-G␣ 12 QL (⌬p115) and then lysed 46 h post-transfection. Lysates were subjected to pull-down assays (see "Experimental Procedures") using a GST fusion of the RGS domain of LARG (middle panel) or GST alone (top panel) and were visualized by immunoblotting using anti-G␣ 12 antibody. Levels of G␣ 12  To assay for JNK activation, MDA-MB-231 cells were infected with the recombinant adenoviruses described above, and cells were lysed and then subjected to immunoblot analysis using an antibody that specifically recognizes the activated, phosphorylated form of JNK. Although G␣ 12 QL significantly increased the level of activated JNK in these cells, expression of the ⌬p115-G␣ 12 QL variant did not lead to an increase in the phospho-JNK epitope above basal levels (Fig. 3E). Finally, we also directly examined the ability of ⌬p115-G␣ 12 QL to activate RhoA in MDA-MB-231 cells through pull-down assays using a GST fusion of the RhoA-binding domain of rhotekin. Although expression of G␣ 12 QL led to an increase in the level of activated RhoA as expected, the level of activated RhoA in cells expressing ⌬p115-G␣ 12 QL was similar to that seen in control cells (Fig. 3F).
The finding that the ⌬p115-G␣ 12 QL variant retained normal binding to the cytoplasmic domain of E-cadherin (see Fig. 1) suggested that even though signaling to Rho by this G␣ 12 mutant was abrogated, its ability to regulate cadherin function would be preserved. To test this hypothesis in a cellular context, we performed cell-cell adhesion assays using breast cancer cells stably expressing E-cadherin. These cells have been shown to form large aggregates, in a Ca 2ϩ -dependent manner, that can be reversed upon introduction of activated G␣ 12 QL via a process that involves the ability of the activated G␣ protein to trigger release of ␤-catenin from the cytoplasmic domain of E-cadherin (8). Consistent with previous results, breast cancer cells expressing E-cadherin (termed 435-E-cad cells) formed large, tightly clumped aggregates in a so-called "fast-aggregation" assay performed in the presence of Ca 2ϩ (Fig. 4A). Formation of these large aggregates was not observed in cells lacking E-cadherin (termed 435-puro cells, Fig. 4C). Although infection of 435-E-cad cells with a control adenovirus prior to the assay did not affect aggregate formation (Fig. 4A), infection with an adenovirus expressing G␣ 12 QL caused a dramatic reduction in aggregate size (Fig. 4D), as reported previously (8). Importantly, very similar results were obtained when the ⌬p115-G␣ 12 QL variant was expressed in these cells, i.e. this protein caused disruption of aggregate formation to essentially the same degree as G␣ 12 QL (Fig. 4G) indicating that the ⌬p115-G␣ 12 QL mutant retained the ability to modulate cadherin function in an intact cell. In all cases, inclusion of the Ca 2ϩ chelator EGTA in the assay disrupted aggregate formation (Fig. 4, B, E, and H) confirming that the cell-cell interactions observed were mediated by cadherins.
As noted in the Introduction, activated G␣ 12 disrupts the interaction between cadherin and its associated cytoplasmic protein ␤-catenin in a manner that requires direct G␣ 12 -cadherin interaction (17,19). This effect of G␣ 12 QL binding to cadherins can be observed directly in cells by following the shift in subcellular localization of ␤-catenin from a peripheral, plasma membrane-associated pattern to a more diffuse staining throughout the cell when activated G␣ 12 is introduced (19). In DLD-1 cells (Fig. 5A) and HEK293 cells (data not shown) infected with a control adenovirus, ␤-catenin was observed to localize most prominently at the cell periphery, whereas in cells expressing G␣ 12 QL this perimeter staining was much less prominent relative to the staining in the cytoplasm and nucleus (Fig. 5B). In cells expressing ⌬p115-G␣ 12 QL , the staining pattern of ␤-catenin most closely resembled that of cells expressing G␣ 12 QL (Fig. 5C), providing additional evidence that the ⌬p115-G␣ 12 QL variant retains the ability to interact with the cadherin cytoplasmic domain in cells and thereby trigger a redistribution of ␤-catenin.

DISCUSSION
The purpose of this study was to identify determinants within the heterotrimeric G protein G␣ 12 that are necessary for its interaction with certain downstream effector proteins. To this end, a series of variants of constitutively active G␣ 12 (G␣ 12 QL ) were constructed by replacing sextets of consecutive amino acids with the well tolerated amino acid motif NAAIRS. This approach has proven successful in identifying domains of the retinoblastoma protein involved in interaction with the transcriptional activator E2F and in mapping important functional domains of telomerase (27,28). From this panel of proteins, we identified a mutant that is impaired in binding to the well characterized G 12 effector proteins p115RhoGEF and LARG and was thus termed ⌬p115-G␣ 12 QL . This mutant is physically and functionally uncoupled from Rho-mediated signaling, as it lacks the ability not only to bind to these Rho exchange factors but also to stimulate Rho activation in cells as well as several known G␣ 12 -mediated, Rho-mediated cellular responses. The mutation that uncoupled this G␣ 12 variant from p115RhoGEF is a replacement of six amino acids just downstream of the Switch II region of G␣ 12 QL . Because the three Switch regions of G␣ proteins are well documented as being important for binding to many effectors (24,25), it is not surprising that such an alteration in close proximity to one of these regions would impact effector binding. Indeed, a mutant of G␣ 13 in which a key lysine residue within the Switch I region was mutated to an alanine was recently reported to be functionally uncoupled from Rho-mediated signaling pathways (37). Importantly, however, the ⌬p115-G␣ 12 QL variant retains normal, functional interaction with at least one of its additional effector proteins, namely E-cadherin. This finding suggests that the binding site for p115RhoGEF is distinct from, or at a minimum does not completely overlap with, the E-cadherin binding site on G␣ 12 . Thus, the ⌬p115-G␣ 12 QL variant should prove a very useful molecular tool for dissecting the roles of Rho activation in particular G␣ 12 -mediated cellular events such as cell proliferation, oncogenic transformation, and activation of phospholipase C-⑀.
Another point of interest in this study is our finding that the Gly-to-Ser mutant produced within the Switch I region of G␣ 12 QL did not impact the ability of G␣ 12 QL to bind the p115RGS domain (see Fig. 1A). This amino acid substitution corresponds to a mutation that has been shown to uncouple mammalian G␣ i1 and G␣ o from RGS4 (23) and also a mutation that disconnects the budding yeast G␣ protein, Gpa1, from its RGS protein, Sst2p (22). This RGS domain of p115RhoGEF has recognizable similarity to other RGS proteins, and also binds to G␣ 12 and G␣ 13 and accelerates GTP hydrolysis of these proteins (16,18); hence p115RhoGEF is considered a functional RGS protein (38). Our finding that mutation of this conserved glycine residue in the Switch I region of G␣ 12 to a serine does not alter G␣ 12 -p115RhoGEF interaction suggests that the nature of this G␣-RGS pairing is fundamentally different from the canonical G␣-RGS pairing between the G␣ i/o subfamily proteins and their RGS proteins.
It is becoming increasingly clear that proteins of the G 12 subfamily are involved in a range of cell biological processes, including proliferative signaling and cancer progression, embryonic development, and cytoskeletal reorganization. Furthermore, these proteins have been shown to interact with an expanding list of intracellular signaling proteins. A full understanding of the signaling networks emanating from the activation of G␣ 12 will require reagents that allow selective uncoupling of this G protein from particular effector proteins. Hence, mutants of G␣ 12 that have lost the ability to interact with one or more effectors while retaining normal affinity for other effectors will provide valuable tools for such studies. The work reported here represents a significant advance toward this goal, in the form of a mutant of G␣ 12 that is uncoupled from a major, well characterized signaling arm, namely Rho-mediated pathways, while retaining the ability to signal through the cadherin/␤-catenin pathway.