The Rho guanine nucleotide exchange factor Lsc homo-oligomerizes and is negatively regulated through domains in its carboxyl terminus that are absent in novel splenic isoforms.

Rho GTPases control fundamental cellular processes, including cytoskeletal reorganization and transcription. Rho guanine nucleotide exchange factors (GEFs) compose a large (>65) and diverse family of related proteins that activate Rho GTPases. Lsc/p115-RhoGEF is a Rho-specific GEF required for normal B and T lymphocyte function. Despite its essential role in lymphocytes, Lsc/p115-RhoGEF signaling in vivo is not well understood. To define Lsc/p115-RhoGEF signaling pathways in vivo, we set out to identify proteins that interact with regulatory regions of Lsc. The 146-amino acid C terminus of Lsc contains a predicted coiled-coil domain, and we demonstrated that deletion of this C terminus confers a gain of function in vivo. Surprisingly, a yeast two-hybrid screen for proteins that interact with this regulatory C terminus isolated a larger C-terminal fragment of Lsc itself. Co-immunoprecipitation experiments in mammalian cells demonstrated that Lsc specifically homo-oligomerizes and that the coiled-coil domain in the C terminus is required for homo-oligomerization. Mutagenesis experiments revealed that homo-oligomerization and negative regulation are distinct functions of the C terminus. Two novel isoforms of Lsc found in the spleen lack portions of this C terminus, including the coiled-coil domain. Importantly, the C termini of both isoforms confer a gain of function and eliminate homo-oligomerization. These results define two important features of Lsc signaling. First, Lsc homo-oligomerizes and is negatively regulated through domains in its C terminus; and second, functionally distinct isoforms of Lsc lacking these domains are present in the spleen.

cling between active GTP-bound and inactive GDP-bound states (3). Rho guanine nucleotide exchange factors (GEFs) 1 compose a large and diverse family of proteins that activate Rho GTPases by catalyzing the release of GDP in exchange for GTP (4 -6). Rho GEFs are characterized by a catalytic Dbl homology (DH) domain and a nearly invariant adjacent pleckstrin homology (PH) domain (6,7). Many Rho GEFs contain additional conserved domains dedicated to functional associations with molecules other than their cognate GTPases (6,7).
Lsc/p115-RhoGEF (Arhgef1) is a Rho-specific GEF that is expressed in lymphoid and myeloid tissue and, to a lesser degree, in other organs (8 -10). Lsc/p115-RhoGEF contains conserved DH and PH domains, an N-terminal RGS (regulator of G-protein signaling)-like domain, and a C-terminal predicted coiled-coil domain (see Fig. 1). Lsc/p115-RhoGEF is both an effector and a regulator of the heterotrimeric GTPase subunit G␣ 13 . G␣ 13 activates Lsc/p115-RhoGEF exchange activity (11) and can be inhibited by the RGS-like domain of Lsc/p115-RhoGEF (12). This is of particular interest because it enables Lsc/p115-RhoGEF to couple ligand activation of G 13 -coupled receptors to Rho signaling pathways.
Recently, targeted disruption of Lsc demonstrated that it is required for normal B and T lymphocyte function. Lsc Ϫ/Ϫ mice have reduced populations of splenic marginal zone B cells, enhanced marginal zone B cell chemotaxis, impaired proliferation of stimulated splenic B cells, and a reduced population of splenic T cells (13). Lsc Ϫ/Ϫ mice have impaired thymus-dependent and type 2 thymus-independent immune responses (13).
Despite its essential role, Lsc/p115-RhoGEF signaling in lymphocytes in vivo is not well understood. Several ligands for G 13 -coupled receptors activate Rho and are known to affect lymphocyte function, including sphingosine 1-phosphate (14,15) and lysophosphatidic acid (16,17). However, it is not known which ligands activate Lsc/p115-RhoGEF or which effector pathways are activated by Lsc/p115-RhoGEF signaling in vivo. To define Lsc signaling pathways in vivo, we set out to identify proteins that interact with regulatory regions of Lsc. Previous work demonstrated that deletion of the C-terminal 110 or 152 amino acids (aa) C-terminal to the PH domain of p115-RhoGEF confers a gain of function in vivo (10,18). We hypothesized that the corresponding region might negatively regulate the murine ortholog Lsc. Deletion of the C-terminal 146 aa of Lsc conferred a 2-3-fold gain of function, suggesting that this region, contain-□ S The on-line version of this article (available at http://www.jbc.org) contains Supplemental Fig. 1.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM  ing a predicted coiled-coil domain, negatively regulates Lsc in vivo. We isolated a larger C-terminal fragment of Lsc itself in a screen for proteins that interact with this regulatory C terminus. We subsequently demonstrated that Lsc homo-oligomerizes and is negatively regulated through domains in its C terminus and that Lsc activity can be regulated by generating functionally distinct isoforms lacking these domains.

EXPERIMENTAL PROCEDURES
Reagents and Antibodies-o-Nitrophenyl-␤-D-galactopyranoside was obtained from ICN Biomedicals (Aurora, OH). Anti-FLAG monoclonal antibody M2, 3-amino-1,2,4-triazole, aprotinin, and phenylmethylsulfonyl fluoride were obtained from Sigma. Anti-hemagglutinin peptide (HA) monoclonal antibody 12CA5 was obtained from Roche Applied Science. Horseradish peroxidase-conjugated goat anti-mouse antibody was obtained from Bio-Rad. Protein A-Sepharose beads were obtained from Amersham Biosciences. All yeast media were obtained from Qbiogene (Carlsbad, CA). Grade 410 filter paper for yeast filter lifts was obtained from VWR (West Chester, PA). All other chemical reagents were obtained from Fisher unless noted otherwise.
Cell Culture-COS-7 and NIH 3T3 cells were maintained in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% bovine calf serum (BCS), 4.5 g/liter glucose, 100 IU/ml penicillin, and 100 g/ml streptomycin. Transfections were performed with Lipofect-AMINE and Plus reagents (Invitrogen) according to the manufacturer's instructions except where noted. Saccharomyces cerevisiae strain Y190 was obtained from Clontech (Palo Alto, CA) and maintained and manipulated according to standard protocols (19).
cDNA-NIH 3T3 cells (ATCC CRL1658) were obtained from the American Type Culture Collection (Manassas, VA), and total RNA was extracted with Trizol reagent (Invitrogen) according to the manufacturer's instructions. C57BL/6J mice received ϳ3 ϫ 10 8 sheep red blood cells by intraperitoneal injection; and 10 days later, total RNA was extracted from their spleens with Trizol reagent. Total RNA was used to generate first-strand cDNA using an oligo(dT) [12][13][14][15][16][17][18] primer with the Superscript first-strand synthesis kit (Invitrogen) according to the manufacturer's instructions.
Yeast Two-hybrid Screen-S. cerevisiae strain Y190 with integrated GAL1-UAS-HIS3 and GAL1-UAS-␤-galactosidase reporter genes (Clontech) was sequentially transformed with Lsc.CT-pAS2-1 and then an NIH 3T3 cDNA library in vector pACT2. pAS2-1 and pACT2 are Gal4binding domain (BD) and Gal4 activation domain (AD) fusion protein vectors, respectively. Cotransformants were grown on synthetic dropout medium lacking tryptophan, leucine, and histidine and supplemented with 25 mM 3-amino-1,2,4-triazole. Candidate clones were identified by selecting for activation of the HIS3 reporter gene and screening for activation of the ␤-galactosidase reporter gene. HIS3 reporter gene expression was selected for by retrieving colonies Ն3 mm in diameter after 8 days of incubation at 30°C. ␤-Galactosidase reporter activity was screened for by immersing a filter lift of cotransformed colonies in liquid nitrogen for 15 s and then exposing it to 5-bromo-4-chloro-3-

TABLE I Oligonucleotides
Below is a list of oligonucleotides used to construct the plasmids described under "Experimental Procedures."

C-terminal Regulatory and Homo-oligomerization Domains in Lsc 30977
indolyl-␤-D-galactopyranoside (X-gal) for 8 h at 30°C. Blue colonies were scored positive for ␤-galactosidase activity. A total of 2.5 ϫ 10 5 colonies were evaluated for expression of the HIS3 reporter gene; 20 of these had robust growth (Ն3 mm in diameter), and three of these demonstrated ␤-galactosidase reporter activation. An additional four cotransformants were Ͻ3 mm in diameter, but had particularly strong ␤-galactosidase reporter activity. The two groups were combined, and a total of seven cotransformants were selected for further workup. The library plasmid from each clone was isolated as described (23), and the library insert was sequenced.
Colony Filter Lift ␤-Galactosidase Assay-Single yeast colonies were streaked in a patch (10 ϫ 15 mm) on synthetic dropout medium lacking tryptophan and leucine and incubated for 72 h at 30°C. Filter lifts of the patches were processed as described for the yeast two-hybrid screen.
Yeast Liquid Culture ␤-Galactosidase Assay-This assay was conducted as described (23) with the modifications noted below. Three independent colonies of each cotransformed strain were grown separately to logarithmic phase in synthetic defined medium lacking tryptophan and leucine; the absorbance at 600 nm was measured; and the yeast cells in a 1.5-ml aliquot were pelleted by centrifugation. Each pellet was washed and resuspended in 200 l of buffer A (60 mM Na 2 HPO 4 ⅐7H 2 O, 40 mM NaH 2 PO 4 ⅐H 2 O, 10 mM KCl, and 1 mM MgSO 4 ⅐7H 2 O). 100-l aliquots of resuspended cells were subjected to three freeze-thaw cycles in liquid nitrogen, and then 750 l of 0.27% ␤-mercaptoethanol in buffer A and 150 l of 4 mg/ml o-nitrophenyl-␤-D-galactopyranoside in buffer A were added. When the supernatant turned pale yellow, 400 l of 1 M Na 2 CO 3 was added; the mixture was centrifuged; and the absorbance of the supernatant at 420 nm was recorded. ␤-Galactosidase units ϭ 1000 ϫ A 420 /(t ϫ V ϫ A 600 ), where t is the time in minutes from the addition of o-nitrophenyl-␤-D-galactopyranoside to the addition of Na 2 CO 3 , V is 0.1ϫ starting cell volume of 1.5 ml/resuspended cell volume of 0.10 ml, and A 600 is the value for each sample measured during logarithmic growth phase (23). The ␤-galactosidase units were normalized to the controls indicated in the figure legends.
SRE-mediated Transcription Assay-The pSRE.L reporter plasmid has been described (21). The pRL-TK plasmid contains the Renilla reniformis (sea pansy) luciferase cDNA downstream of the herpes simplex virus thymidine kinase gene promoter. COS-7 cells were plated at a density of 5 ϫ 10 4 cells/well in Falcon Primaria 24-well plates (BD Biosciences), transfected in serum-free medium for 3 h, and incubated in medium with 0.5% BCS for 24 h; and then the firefly and Renilla luciferase activities of the cell lysate were measured using the dual luciferase assay reporter system (Promega) with a MicroLumat Plus LB96V luminometer (Berthold Technologies, Bad Wildbad, Germany). The firefly luciferase activities of cells transfected with Lsc and each Lsc mutant are presented as the mean Ϯ S.D. of four independent wells in each experiment, where the measured firefly luciferase activity of each well was adjusted for transfection efficiency by calculating the ratio of measured firefly luciferase activity to Renilla luciferase activity and then normalized to the control transfected with only pRL-TK, pSRE.L, and ␤gal-pBJ.KS. The firefly luciferase activities of Lsc and Lsc mutants were compared using a one-way analysis of variance; and if a difference existed, post hoc pairwise comparisons were made using the Student-Newman-Keuls test (24).
Immunoprecipitation and Immunoblotting-Immunoprecipitation and immunoblotting were performed using standard methods (25) with the modifications noted below. COS-7 cells were plated at a density of 2 ϫ 10 5 cells/well in Falcon Primaria 6-well plates (BD Biosciences), transiently transfected for 12-16 h, and then incubated in DMEM supplemented with 10% BCS for an additional 24 h. Cell lysates were harvested with 1 ml of cold lysis buffer (50 mM Hepes (pH 7.4), 150 mM NaCl, 2 mM EDTA, 1% Igepal, 10% glycerol, 10 mM NaF, 100 g/ml phenylmethylsulfonyl fluoride, 1 g/ml aprotinin, and one tablet of Complete protease inhibitor mixture (Roche Applied Bioscience)/50 ml of lysis buffer) and gentle scraping and spun at 14,000 ϫ g for 10 min, and the protein concentration of the supernatants was measured (DC protein assay, Bio-Rad) and equalized by dilution in lysis buffer. Anti-FLAG antibody M2 was added to 400 l of supernatant; the mixture was incubated on ice for 3 h; protein A-Sepharose beads (100 l of a 50:50 slurry) were added; and the samples were rocked for 1 h and then washed three times with cold lysis buffer. The beads were pelleted; the supernatant was removed; and 2ϫ Laemmli buffer with 10% dithiothreitol was added. The samples were boiled for 5 min, subjected to PAGE, transferred to a polyvinylidene difluoride membrane (Bio-Rad), and blotted sequentially with murine anti-HA monoclonal antibody 12CA5 and horseradish peroxidase-conjugated goat anti-mouse antibody. Crude lysate was removed prior to the addition of the M2 immunopre-cipitating antibody, processed similarly, and blotted with either the 12CA5 or M2 antibody. The immunoblots were developed with the ECL Plus detection system (Amersham Biosciences) according to the manufacturer's instructions, and images were collected with a Storm 860 imager (Amersham Biosciences).
Amplification of Lsc Partial cDNAs from Spleen-Lsc partial cDNA clones were amplified by PCR from mouse spleen first-strand cDNA using primers DL1026 and DL2080 with an annealing temperature of 60°C for 40 cycles on a RoboCycler Gradient 96 temperature cycler (Stratagene). The PCR reaction products were subjected to agarose gel electrophoresis; and amplicons smaller than the common isoform amplicon were isolated, subcloned into pCR2.1-TOPO (Invitrogen), and sequenced. The sequences of these clones were aligned with the cDNA sequence of the common isoform of Lsc (GenBank TM /EBI accession number U58203) to identify partial cDNAs lacking exonic sequence found in the common isoform. These clones were aligned with the Lsc genomic DNA sequence (GenBank TM /EBI accession number AC073679) to predict the 5Ј-and 3Ј-splice junction sequences corresponding to the absent exonic sequence. Because Ͼ98% of introns utilize the consensus 5Ј-GT/3Ј-AG intronic splice site dinucleotide pair (26), we considered only clones with this dinucleotide pair for further analysis.

RESULTS AND DISCUSSION
C-terminal Residues Negatively Regulate Lsc in Vivo-Before conducting a screen to identify proteins that interact with the C terminus of Lsc, we wanted to determine whether Lsc is negatively regulated by residues C-terminal to the PH domain. Activated Rho induces SRE-mediated transcription (27,28), and this can be used to measure GEF activity (21). We compared the ability of Lsc and a mutant form of Lsc lacking the C-terminal 146 aa (Fig. 1) to activate SRE-mediated transcription from a firefly luciferase reporter plasmid (pSRE.L) in transiently transfected COS-7 cells. Deletion of the C-terminal 146 aa conferred a 2-3-fold increase in firefly luciferase activity compared with full-length Lsc (Fig. 2). We obtained similar results in HeLa cells (data not shown). In contrast, deletion of the DH domain eliminated activation, demonstrating that the action of Lsc requires the nucleotide exchange domain (Fig. 2). Deletion of the C terminus has now been shown to confer a gain of function to Lsc/p115-RhoGEF in vivo by three different groups in three different cell types using two different endpoints for Rho signaling (10,18). Together, these results provide convincing evidence that Lsc/p115-RhoGEF is negatively regulated in vivo by residues C-terminal to the PH domain. Elegant work by one of these groups demonstrated that deletion of the C terminus confers a gain of function in vivo, but a loss of function in vitro, suggesting that C-terminal negative regulation in vivo may require interaction with another molecule present in the cell (18).
Deletion of the C terminus of p115-RhoGEF did not confer a gain of function in vivo in one report that compared the activities of p115-RhoGEF and a mutant form of p115-RhoGEF lacking both N-and C-terminal residues (29). Deletion of the C terminus can confer a gain of function in the absence of these N-terminal residues (10), so it is unlikely that this explains the absence of an effect.
The Regulatory C Terminus of Lsc Homo-oligomerizes-To identify proteins that interact with the regulatory C terminus of Lsc, we conducted a yeast two-hybrid screen of an NIH 3T3 cDNA library using a peptide fragment corresponding to the C-terminal 146 aa of Lsc as bait. We screened a library derived from NIH 3T3 cells because they express endogenous Lsc (8) and could be expected to express proteins that interact with Lsc. The library was cloned into the pACT2 vector expressing translated library products as fusion proteins with an N-terminal Gal4-AD. A total of seven candidate clones were isolated from 2 ϫ 10 5 screened cotransformants. One of these clones, GIP9 (GEF-interacting protein-9), contained a partial cDNA encoding the 315-aa C terminus of Lsc itself (nt 1813-2760), including the PH domain and regulatory C terminus. Cotransformation with bait and GIP9 plasmids was necessary and sufficient to specifically activate the ␤-galactosidase reporter (Supplemental Fig. 1), but sequencing revealed that the Lsc partial cDNA in GIP9 was not in the same translation frame as the upstream Gal4-AD cDNA. Yeast two-hybrid screens utilizing the pACT2 library vector have isolated clones in which the presumed reading frame of the library inserts did not match the frame of the upstream Gal4-AD cDNA (30 -32), and functional complementation (30) and Western blotting (31) were used to demonstrate that alternate frame translation products of the insert were expressed from the pACT2 vector. Cotransformation of yeast with the bait plasmid and a version of GIP9 with the Lsc partial cDNA converted to the same translation frame as the Gal4-AD (Lsc.PHϩCT-pACT2) activated the reporter compared with the Lsc.CT-BD/no insert-AD control (Fig.  3A), whereas cotransformation with the bait plasmid and a version of GIP9 lacking 792 bp of the Lsc partial cDNA downstream of the first stop codon predicted in the Gal4-AD reading frame (GIP9⌬LscORF-pACT2) did not (Supplemental Fig. 1). Deletion of these 792 nucleotides did not affect the cDNA encoding the predicted 22-aa GIP9 translation product in the Gal4-AD reading frame, but did eliminate Lsc cDNA nucleotides that would be required to generate alternate frame translation products in the Lsc reading frame. These results strongly suggest that the GIP9 library clone expressed one or more alternate frame translation products (corresponding to fragments of Lsc) that interact with the bait peptide. This prompted us to validate this interaction in mammalian cells using an independent method to assess intermolecular association. We demonstrated that these two C-terminal fragments of Lsc also interact in mammalian cells by co-immunoprecipitating differentially epitope-tagged forms of the 146-and 315-aa fragments in transiently transfected COS-7 cells (Fig.  3B). Together, the results in yeast and mammalian cells indicate that Lsc self-associates with its regulatory C terminus.
To determine whether residues within the 146-aa regulatory C terminus are sufficient for Lsc self-association, we first compared the ability of the 146-aa regulatory C terminus to interact with peptide fragments corresponding to the entire 315-aa C terminus, the 146-aa C terminus alone, and the PH domain alone using the yeast two-hybrid system. Qualitative assessment of ␤-galactosidase reporter activation by filter lift assay indicated that the 146-aa C terminus could homo-oligomerize, but did not interact with the PH domain (Fig. 3A). Although streaked patches of yeast expressing two fusion proteins con- taining the 146-aa C terminus (Lsc.CT-BD and Lsc.CT-AD) had much less ␤-galactosidase activity than yeast expressing the 146-aa C terminus and the 315-aa C terminus, they also grew much more slowly (Fig. 3A). This raised the possibility that the difference in reporter activation between the two strains could be the result of a difference in cell number rather than evidence that the PH domain is required for Lsc self-association. To address this, we compared ␤-galactosidase reporter activation using a quantitative assay that partially accounts for different growth rates by normalizing reporter activity to assayed cell number (33). The quantitative ␤-galactosidase assay indicated that yeast expressing the two 146-aa fragments activated the reporter at least 4-fold more than the control, but less than half as much as yeast expressing the 146-and 315-aa fragments (Fig. 3A). These results demonstrate that the 146-aa regulatory C terminus can homo-oligomerize, but we cannot rule out the possibility that the PH domain directly or indirectly enhances this self-association. To demonstrate that the regulatory 146-aa C terminus can homo-oligomerize in mammalian cells, we co-immunoprecipitated differentially epitope-tagged forms of the 146-aa C-terminal fragment in transiently transfected COS-7 cells (Fig. 3B).

Lsc Homo-oligomerizes in Mammalian Cells through a Predicted Coiled-coil Domain in the Regulatory C Terminus-We
hypothesized that homo-oligomerization of the regulatory C terminus might support full-length Lsc homo-oligomerization in vivo. We demonstrated that Lsc can homo-oligomerize in vivo by co-immunoprecipitating differentially epitope-tagged forms of Lsc from COS-7 cells: HA epitope-tagged Lsc co-immunoprecipitated with FLAG epitope-tagged Lsc (Fig. 4). In contrast, we were not able to co-immunoprecipitate an HA epitope-tagged form of the GTPase dynamin-1 containing a predicted coiled-coil domain (34,35) with FLAG epitope-tagged Lsc, indicating that Lsc homo-oligomerization is specific (Fig.  4). Comparison of the Lsc band density on the anti-HA blots of the lysate and immunoprecipitate indicated that at least 20% of the heterologously expressed Lsc was homo-oligomerized. This is likely an underestimate of the extent of homo-oligomerization of the epitope-tagged proteins because it assumes 100% efficiency of the immunoprecipitating antibody and does not account for homo-oligomers dissociated during co-immunoprecipitation or formed with the endogenous primate ortholog of Lsc. To determine whether the regulatory C terminus of Lsc is necessary for Lsc homo-oligomerization, we attempted to co-immunoprecipitate differentially epitope-tagged Lsc and a mutant form of Lsc lacking the 146-aa regulatory C terminus from transiently transfected COS-7 cells. Strikingly, deletion of the 146-aa regulatory C terminus completely eliminated the ability of Lsc to homo-oligomerize: FLAG epitope-tagged Lsc was unable to co-immunoprecipitate HA epitope-tagged Lsc lacking the regulatory C terminus (Fig. 4). This demonstrates that residues in the regulatory C terminus are required for Lsc homo-oligomerization.
Several Rho GEFs are known to homo-oligomerize and/or hetero-oligomerize; and in at least some cases, homo-oligomerization appears to confer a gain of function. Heterologously expressed Dbl (36) and RasGRF1 (37) homo-oligomerize, and point mutations in their DH domains disrupt homo-oligomerization and confer a loss of function. Endogenous RasGRF1 and the closely related RasGRF2 also hetero-oligomerize (37). ␤1PIX (Arhgef7/p85 Cool-1 ) homo-oligomerizes, and this requires a predicted coiled-coil domain C-terminal to the conserved PH domain (38,39). Deletion of this coiled-coil domain abrogates homo-oligomerization and the capacity to generate membrane ruffles, suggesting that ␤1PIX signaling can be facilitated by homo-oligomerization (38,39). Heterologously expressed ␤1PIX and the closely related ␣PIX (Arhgef6/Cool-2) also hetero-oligomerize (39). Bcr contains an N-terminal coiled-coil oligomerization domain required for the transforming activity of the oncogenic Bcr-Abl fusion protein (40,41), but the effect of oligomerization on native Bcr GEF activity has not been reported. To determine whether the DH domain is necessary for Lsc homo-oligomerization, we co-immunoprecipitated differentially epitope-tagged Lsc and a mutant form of Lsc lacking the DH domain from COS-7 cells. In contrast to Dbl (36) and RasGRF1 (37), disruption of the DH domain had no effect on Lsc homo-oligomerization: HA epitope-tagged Lsc lacking the DH domain co-immunoprecipitated with FLAG epitope-tagged Lsc (Fig. 4). This demonstrates that Lsc does not homo-oligomerize through its DH domain.
Lsc now joins a small group of Rho GEFs known to homooligomerize and, to our knowledge, is the first member that is a Rho-specific GEF (6,9). The Rho GEFs studied to date appear to homo-oligomerize through one of three mechanisms involving the conserved DH domain, an N-terminal coiled-coil domain, or a C-terminal coiled-coil domain as with Lsc.
To identify C-terminal residues required for Lsc homo-oligomerization, we first compared the effect of progressively smaller deletions in the 146-aa regulatory C terminus on its ability to interact with the 315-aa C terminus of Lsc in the yeast two-hybrid system. Deletion of the C-terminal 83 aa (Lsc.CT (774 -836)) almost completely eliminated activation of the reporter compared with the control (Fig. 5A). Deletion of just the 31-aa coiled-coil domain (Lsc.CT(⌬CC)) also conferred a substantial reduction in reporter activation, suggesting that the coiled-coil domain is required for Lsc homo-oligomerization (Fig. 5A) autonomously activated the reporter and were not informative (Fig. 5A).
To test whether the coiled-coil domain is required for homooligomerization of Lsc in mammalian cells, we attempted to co-immunoprecipitate differentially epitope-tagged Lsc and a mutant form of Lsc lacking the coiled-coil domain from transiently transfected COS-7 cells: deletion of the coiled-coil do-main almost completely abrogated homo-oligomerization, indicating that the coiled-coil domain is required for homooligomerization (Fig. 5C).
Coiled coils form when amphipathic ␣-helical coiled-coil domains assemble into a superhelix (45). Coiled-coil domains consist of amino acid heptad repeats with the first and fourth positions occupied by hydrophobic residues that align to form a hydrophobic interface in the assembled superhelix (45). To test whether the coiled-coil structure itself is required for homooligomerization, we attempted to co-immunoprecipitate Lsc and a mutant form of Lsc in which the coiled-coil domain was disrupted rather than deleted. Replacement of leucine by proline in the fourth position of two successive heptad repeats in the coiled-coil domain (L875P and L882P) is predicted to disrupt the ␣-helix required to form the hydrophobic face that permits coiled-coil formation. The COILS2.2 algorithm (42) estimates that the L875P/L882P double substitution reduces the probability that a 28-aa region of the C terminus will adopt a coiled-coil configuration from Ͼ95 to 0% (Fig. 5B). This L875P/L882P double substitution almost completely abrogated Lsc homo-oligomerization as detected by co-immunoprecipitation (Fig. 5C). This indicates that the predicted coiled-coil structure itself is required for Lsc homo-oligomerization. Homo-oligomerization was not completely eliminated as it was with deletion of the entire C-terminal 146 aa. This suggests that another region of the C terminus may be capable of participating in Lsc homo-oligomerization, but to a much smaller degree than the coiled-coil domain.
Both deletion and disruption of the coiled-coil domain virtually eliminated homo-oligomerization. In contrast to deleting the C terminus, disruption with the L875P/L882P double substitution leaves all but two residues intact. This enabled us to examine the relationship between Lsc homo-oligomerization and negative regulation more specifically. Importantly, although deletion of the 31-aa predicted coiled-coil domain conferred a small gain in function, disruption of the coiled-coil domain with the L875P/L882P double substitution conferred no significant gain of function (Fig. 5D). Together, these results demonstrate that disruption of the coiled coil substantially impairs homo-oligomerization (Fig. 5C) without conferring a gain of function (Fig. 5D). This indicates that homo-oligomerization and negative regulation are distinct functions of the C terminus and suggests that additional proteins interact with the C terminus to regulate Lsc. Deletion of the coiled-coil domain likely confers a gain of function by directly or indirectly affecting interaction with one of these proteins, whereas disrupting the coiled-coil domain does not.

Isoforms of Lsc Lacking the C-terminal Coiled-coil and Regulatory Domains Are Present in the Spleen-Several
Rho GEFs, including ␤PIX (38), obscurin (46), and collybistin (47), undergo alternative splicing. ␤PIX is of particular interest because the absence of the C-terminal homo-oligomerization domain in the ␤2PIX splice variant confers important functional differences compared with ␤1PIX (38). Alignment of the Lsc cDNA and genomic locus DNA sequences revealed that the regulatory C terminus of Lsc is encoded by six exons (Fig. 6A). We hypothesized that Lsc splice variants lacking portions of the 146-aa C-terminal regulatory domain might exist and have important functional differences compared with the common isoform. Because Lsc is required for normal B and T lymphocyte function, we attempted to identify C-terminal splice variants from the spleens of adult mice. We isolated two partial cDNAs, c237 and c249, lacking nucleotides encoding portions of the regulatory C terminus of the common isoform, including the coiled-coil domain (Fig. 6, A-C). c237 lacks nucleotides corresponding to translated exon 26 of the common isoform (nt 2486 -2649) (Fig.   FIG. 5. Disruption of a predicted coiled-coil domain impairs homo-oligomerization, but confers no gain of function. A, deletion of the coiled-coil domain in the regulatory C terminus impairs Lsc self-association in yeast. Yeast cells were cotransformed with a BD vector expressing the indicated peptide fragment and an AD vector expressing the C-terminal 315 aa of Lsc (Lsc.PHϩCT-AD). Liquid culture assay for ␤-galactosidase activity was performed as described in the Fig. 3 legend. A schematic diagram of the peptide fragments expressed in the BD vector is shown. Regulatory C-terminal residues are shown as gray boxes; absent residues are indicated by solid lines; and amino acid numbers corresponding to regulatory C-terminal boundaries are indicated above. Activity is presented as the mean Ϯ S.D. ␤-Galactosidase (␤-gal) units were normalized to the Lsc.PHϩCT-AD/no insert-BD control (1.0 Ϯ 0.9; not shown). Data from a representative experiment are shown (n ϭ 2). AA, autonomous activation of reporter. A schematic diagram of the boundaries of the predicted coiledcoil domain (CC) and the ␣-helices (H1 and H2) is also shown below. B, the L875P/L882P double substitution disrupts the predicted coiled-coil domain. Shown is the probability that 28-aa segments of Lsc (left) and Lsc with the L875P/L882P double substitution (right) form coiled coils as a function of Lsc amino acid number. Probabilities were derived by the COILS2.2 algorithm (42). C, deletion or disruption of the coiled-coil domain nearly eliminates Lsc homo-oligomerization. Shown are Western blots of the immunoprecipitate (IP; upper) and lysates (lower) derived from COS-7 cells transiently transfected with the indicated pBJ1 plasmids and prepared as described in the Fig. 4 legend. The ladder is shown in kilodaltons. A representative blot is shown (n ϭ 2). D, disruption of the coiled-coil domain confers no significant gain of function. SRE reporter assay was performed, and data are presented as the mean Ϯ S.D. of firefly activity as described in the Fig. 2 legend. *, p ϭ 0.01 for the indicated pairwise comparison; NS, not significant (p Ͼ 0.05). Data from a representative experiment are shown (n ϭ 3). H, N-terminal HA epitope tag; F, N-terminal FLAG epitope tag. 6, A-C). The predicted 5Ј-and 3Ј-intronic splice sites corresponding to the absent exon are the same sites used to join coding exons 25 and 26 and coding exons 26 and 27 in the common isoform, respectively (Fig. 6, A-C). The absence of this exon generates a frameshift so that aa 829 -919 of the common isoform are replaced by 27 aa not present in the common isoform (Fig. 6, A-C). c249 lacks nucleotides corresponding to exons 24 -27 and part of exon 28 (nt 2331-2819) of the common isoform (Fig. 6, A-C). The predicted 5Ј-intronic splice site corresponding to the absent nucleotides is the same site used to join exons 23 and 24 in the common isoform, and the predicted alternative 3Ј-intronic splice site (CCCCCACCCCCACAGCT-GCCACAG/C) corresponds to a suitable consensus sequence with a pyrimidine-rich region followed by the sequence NCAG/C, where G/C is the intron/exon border (48). The absence of this sequence also generates a frameshift so that aa 778 -919 of the common isoform, almost the entire regulatory C terminus, are replaced by 13 aa not present in the common isoform (Fig. 6, A-C).
Both c237 and c249 lack the coding region for the C-terminal coiled-coil domain. Strikingly, replacing the regulatory C terminus of the common isoform with the corresponding region of FIG. 6. The C termini of two novel Lsc isoforms abrogate homo-oligomerization and confer a gain of function. A, the partial cDNAs c237 and c249, corresponding to isoforms of Lsc, were isolated from spleen. Shown is a schematic diagram of the predicted exons for the common isoform of Lsc and the corresponding regions of c237, c249, and EST BG148369. The boundaries were predicted by alignment of cDNA and Lsc genomic (GenBank TM /EBI accession number AC073679) sequences. Exons encoding amino acids present in the common isoform of Lsc are indicated by gray boxes, and exons encoding amino acids not in the common isoform are represented by black boxes. Introns are shown as angled lines, and stop codons are indicated by asterisks. The common isoform cDNA nucleotide numbers corresponding to the beginning and end of the regulatory C terminus and exon boundaries are indicated above. B, shown is an alignment of the predicted amino acid sequences of the common isoform regulatory C terminus (aa 774 -919) and corresponding regions of c237, c249, and EST. Absent residues that are present in the common isoform are shown as dashes, and stop codons are represented by asterisks. C, shown are aligned schematic diagrams of the predicted translation products of the common isoform C terminus and the corresponding regions of c237, c249, and EST. A schematic diagram of the boundaries of the predicted coiled-coil domain (CC) and ␣-helices (H1 and H2) in the common isoform is shown below. Amino acids present in the common isoform of Lsc are show as gray boxes, and amino acids not in the common isoform are represented by black boxes. Absent residues present in the common isoform are indicated as solid lines. D and E, replacement of the regulatory C terminus of the common isoform of Lsc with the corresponding regions of c237, c249, and EST abrogated homo-oligomerization and conferred a gain of function. Shown are Western blots of the immunoprecipitate (IP; upper) and lysates (lower) derived from COS-7 cells transiently transfected with the indicated pBJ1 plasmids and prepared as described in the Fig.  4 legend. The ladder is shown in kilodaltons. A redundant lane was removed from the immunoprecipitate and lysate blots between the HA.Lsc.EST and HA.Lsc.237 lanes. A representative blot is shown (n ϭ 2). SRE reporter assay were performed, and data are presented as the mean Ϯ S.D. of firefly activity as described in the Fig. 2 legend. *, p Ͻ 0.05; **, p Ͻ 0.01. Data from a representative SRE experiment are shown (n ϭ 3). H, N-terminal HA epitope tag; F, N-terminal FLAG epitope tag. c237 and c249 completely abrogated homo-oligomerization and conferred a 2-3-fold gain of function (Fig. 6, D and E).
Identification of these novel isoforms of Lsc is important for several reasons. First, it indicates that functionally distinct forms of Lsc are present in the spleen, suggesting that this may be a mechanism to regulate the cellular activity of Lsc. Although transcripts for both isoforms were substantially less abundant than the transcript for the common isoform in whole spleen (data not shown), they may represent a larger fraction of the total Lsc transcript pool in homogeneous subpopulations of splenic cells such as subsets of B or T lymphocytes. It will be important to determine the cell of origin and the biological role of these splice variants. Second, the concomitant loss of homooligomerization and gain of function conferred by these alternate C termini reinforce the results from experiments described above demonstrating that Lsc homo-oligomerizes and is negatively regulated through domains in its C terminus. It is likely that the c237 and c249 isoforms represent splice variants of Lsc; the most compelling evidence is that the splice junction sites corresponding to the gaps in their cDNA sequences are identical to junctions used for splicing of the common isoform or conform to consensus splice junction sequences.
We identified the partial cDNA for another potential isoform of Lsc lacking a portion of the regulatory C terminus by searching the NCBI Murine EST Database. This EST (GenBank TM / EBI accession number BG148369) was of interest because it was isolated from resting germinal B lymphocytes and lacks portions of exons 26 and 27 (nt 2488 -2706) encoding aa 830 -902 of the common isoform, including the coiled-coil domain (Fig. 6, A-C). As expected from the results of experiments above, replacement of the regulatory C terminus of the common isoform with the corresponding region from EST BG148369 substantially reduced Lsc self-association in the yeast twohybrid system (Fig. 5A), abrogated Lsc homo-oligomerization (Fig. 6D), and conferred a gain of function in mammalian cells (Fig. 6E). Alignment of the EST and Lsc genomic DNA sequences reveals an acceptable predicted alternative consensus 5Ј-splice site, but no compelling consensus 3Ј-splice site corresponding to the absent exonic sequence (data not shown). This EST could therefore represent an incorrect or incomplete splicing event, a correctly spliced product with an atypical 3Ј-splice site, or an artifact of library preparation. It is also possible that a consensus 3Ј-splice sequence cannot be located because the genomic sequence acquired from the NCBI Database contains a sequencing error. CONCLUSIONS We have demonstrated that Lsc homo-oligomerizes and that homo-oligomerization and negative regulation are distinct functions of the C terminus. Our results suggest a model in which a portion of the C terminus negatively regulates Lsc by inhibiting interaction with an activating protein or by facilitating interaction with an inhibitory protein independent of the coiled-coil domain required for homo-oligomerization (Fig. 7). In this model, the splenic isoforms c237 and c249 have enhanced basal activity and are unable to homo-oligomerize because they lack the C-terminal regulatory and homo-oligomerization domains (Fig. 7). One potential activating protein in this model, whose effect may be inhibited by the C terminus, is the heterotrimeric GTPase subunit G␣ 13 . G␣ 13 activates Lsc/ p115-RhoGEF exchange activity (10,49) by an incompletely understood mechanism. It physically associates with the N terminus of Lsc/p115-RhoGEF and a second site between p115-RhoGEF aa 288 and 760, including the DH domain (49). Deletion of the C terminus enhances the affinity of Lsc/p115-Rho-GEF for G␣ 13 in vitro (49), consistent with a model in which the C terminus inhibits G␣ 13 interaction with the DH domain of Lsc/p115-RhoGEF.
We used the COILS2.2 algorithm (42) to analyze 61 human Rho GEF amino acid sequences and identified six Rho GEFs, in addition to Lsc and ␤1PIX, that have predicted coiled-coil domains C-terminal to their conserved PH domain: Lbc, GEF-H1 (Arhgef2), p190-RhoGEF, ␣PIX (Arhgef6), p114-RhoGEF, and KIAA720. This raises the possibility that one or more of these Rho GEFs homo-oligomerize in a manner similar to Lsc and potentially hetero-oligomerize with one another. Previous work has identified molecules that bind the coiled-coil domain-containing C termini of several of these Rho GEFs. The cateninlike protein ␣-catulin binds a portion of the C terminus of Lbc and augments GEF activity (50). A portion of the GEF-H1 C terminus is necessary for microtubule association (51,52). Portions of the C terminus of p190-RhoGEF are required for association with microtubules (53), 14-3-3 (54), and a destabilizing element of the 3Ј-untranslated region of the light neurofilament subunit mRNA (55). Despite the lack of direct proof that the coiled-coil domains in the C termini of these Rho GEFs are required for these interactions, these interactions with the C termini suggest that it is possible that these or other molecules could bind the C terminus of Lsc to regulate Lsc function.
The role of homo-oligomerization in Lsc function is not clear. Other Rho GEFs, including ␤1PIX (38, 39), Dbl (36), RasGRF1 and RasGRF2 (37), and Bcr (40,41), can homo-oligomerize, and homo-oligomerization appears to facilitate signaling in vivo. The mechanism of Lsc homo-oligomerization most closely resembles that of ␤1PIX, which also homo-oligomerizes through a coiled-coil domain C-terminal to the PH domain (38,39). In contrast to ␤1PIX, the C terminus of Lsc is neither sufficient 2 nor necessary (18) for localization to the plasma membrane. This suggests that, although ␤1PIX and Lsc may share a similar mechanism of homo-oligomerization, it serves different functions for each.
Homo-oligomerization offers theoretical advantages, including signal amplification and scaffold formation (56). Heterooligomerization offers additional advantages, including facilitating cross-talk between the signaling pathways associated with the participating Rho GEFs. Interestingly, the COILS2.2 algorithm does not predict that the other Rho GEFs containing RGS-like domains (LARG, PDZ-Rho GEF, and GTRAP48) are 2 D. J. Lerner, unpublished data. FIG. 7. Homo-oligomerization and negative regulation are distinct functions of the C terminus. Lsc homo-oligomerizes through a predicted coiled-coil domain in the C terminus. Two models of negative regulation are shown: the C terminus promotes the action of the Lsc inhibitor (A), and the C terminus prevents the action of the Lsc activator (B). The splenic isoforms c237 and c249 lack the C-terminal domains required for homo-oligomerization and negative regulation. The regulatory C terminus of Lsc containing the coiled-coil domain is indicated by thick lines. The Lsc DH and PH domains are indicated by white boxes. The remainder of Lsc is represented by thin lines. The Lsc inhibitor is shown as a gray diamond, and the Lsc activator is indicated by the black T shapes.
likely to form coiled coils through the region C-terminal to their PH domains. This suggests that homo-oligomerization may serve a role in Lsc independent of the RGS-like domain.