Human RGS6 gene structure, complex alternative splicing, and role of N terminus and G protein gamma-subunit-like (GGL) domain in subcellular localization of RGS6 splice variants.

RGS proteins are defined by the presence of a semiconserved RGS domain that confers the GTPase-activating activity of these proteins toward certain G alpha subunits. RGS6 is a member of a subfamily of RGS proteins distinguished by the presence of DEP and GGL domains, the latter a G beta 5-interacting domain. Here we report identification of 36 distinct transcripts of human RGS6 that arise by unusually complex processing of the RGS6 gene, which spans 630 kilobase pairs of genomic DNA in human chromosome 14 and is interrupted by 19 introns. These transcripts arise by use of two alternative transcription sites and complex alternative splicing mechanisms and encode proteins with long or short N-terminal domains, complete or incomplete GGL domains, 7 distinct C-terminal domains and a common internal domain where the RGS domain is found. The role of structural diversity in the N-terminal and GGL domains of RGS6 splice variants in their interaction with G beta 5 and subcellular localization and of G beta 5 on RGS6 protein localization was examined in COS-7 cells expressing various RGS6 splice variant proteins. RGS6 splice variants with complete GGL domains interacted with G beta 5, irrespective of the type of N-terminal domain, while those lacking a complete GGL domain did not. RGS6 protein variants displayed subcellular distribution patterns ranging from an exclusive cytoplasmic to exclusive nuclear/nucleolar localization, and co-expression of G beta 5 promoted nuclear localization of RGS6 proteins. Analysis of our results show that the long N-terminal and GGL domain sequences of RGS6 proteins function as cytoplasmic retention sequences to prevent their nuclear/nucleolar accumulation. These findings provide the first evidence for G beta 5-independent functions of the GGL domain and for a role of G beta 5 in RGS protein localization. This study reveals extraordinary complexity in processing of the human RGS6 gene and provides new insights into how structural diversity in the RGS6 protein family is involved in their localization and likely function(s) in cells.

RGS 1 proteins comprise a family of proteins, defined by the presence of a semiconserved region called the RGD, that have been implicated in the negative regulation of heterotrimeric G protein signaling (1)(2)(3). The existence of such proteins was first shown by genetic studies in yeast (4) and Caenorhabditis elegans (5) where the yeast pheromone desensitization factor Sst2p and the C. elegans Sst2p homolog Egl-10 were found to negatively regulate Gpa1 and GOA-1, respectively, both homologs of the mammalian heterotrimeric G protein G␣ o . The presence of a semiconserved domain of ϳ120 amino acids in Sst2p and Egl-10 (i.e. the RGD) enabled Koelle and Horvitz (5) to demonstrate the existence of a family of mammalian proteins with this domain. More than 20 mammalian genes encode proteins with this hallmark RGD or less related versions of this domain. Subsequent studies demonstrated that RGS proteins or their isolated RGDs display GTPase-activating protein activity toward G i and G q proteins (1-3), providing insight into how RGS proteins could function to turn off G proteins following their GTP-dependent activation by receptors. Some studies have suggested that RGS proteins may interact with the effectors adenylyl cyclase or phosphoinositide phospholipase C (6,7) or with receptors to block heterotrimeric G protein signaling (8). Yet, the physiological function of RGS proteins has been shown only for pheromone signaling in yeast (4), neuronal signaling in C. elegans (5), and, in mammals, phototransduction in the eye where RGS9 is required for the rapid inactivation of transducin (9,10). Moreover, recent evidence from this and other laboratories has shown that several RGS proteins are localized predominantly at intracellular sites other than the plasma membrane including the nucleus (11)(12)(13)(14)(15)(16), where G proteins as well as their effectors and activating receptors are not thought to localize. These findings raise the fascinating possibility that some members of the RGS protein family may have functions apart from, or in addition to, regulatory control of heterotrimeric G protein signaling.
Five subfamilies of RGS proteins have been proposed based upon sequence identities within the RGD of these proteins (3). Interestingly, this classification also groups proteins with similar structural domains outside of the RGD that may be involved in subcellular targeting, regulation, protein interaction(s) or specific functions of members of a given subfamily.
Indeed, the R4 subfamily proteins RGS4 and RGS16 possess NESs that function to transport these proteins from the nucleus to the cytoplasm (13) and the N-terminal domain of RGS3 has been implicated in its recruitment to the plasma membrane (14). RGS proteins in the RZ subfamily possess cysteine string motifs and several other protein domains have been identified that are unique to members of a given RGS protein subfamily (3). Of particular interest to the present study is the R7 subfamily of RGS proteins that includes RGS6 as well as RGS7, RGS9, and RGS11. Each of these RGS proteins has an N-terminal domain that contains a DEP (disheveled, Egl-10 and pleckstrin homology) and a GGL (G-protein gamma subunit-like) domain. Although the function of the DEP domain is unknown, the GGL domain has been shown to represent a binding site for G␤5 (17,18), an atypical G␤ subunit (19). Present evidence suggests that R7 family members may represent physiological binding partners for G␤5 (10,20), rather than G␥ as observed for other G␤ proteins. The precise function of this interaction is not yet clear although G␤5 binding to these RGS proteins has been implicated in RGS and G␤5 protein stability (10,18,20), in regulation of RGS protein signaling or GAP activity (21)(22)(23)(24)(25) and in localization of G␤5 (26).
We undertook studies to clone members of the RGS protein family to further our understanding of the structural diversity and complexity within this family. Recent evidence suggests that this complexity may arise not only from the more than 20 mammalian genes encoding RGS protein members, but also by the existence of multiple forms of a given RGS gene product. We identified two major transcripts for human RGS3 (27) and subsequently demonstrated the existence of twelve alternatively spliced forms of human RGS12 (12). RGS9 has also been found to exist in two splice variant forms (28). Here we report identification of 36 splice variant forms of RGS6 that arise by use of two alternative transcription start sites within the human RGS6 gene and by complex alternative splicing of the two primary RGS6 mRNAs. Identification of these transcripts enabled us to deduce the structure of the human RGS6 gene and the splicing mechanisms that generate these novel RGS6 transcripts. Interestingly, these RGS6 transcripts encode proteins with long or short N-terminal domains, complete or incomplete GGL domains, seven distinct C-terminal domains, and a common internal domain where the RGD is located. The existence of diversity outside of but not within the RGD of the RGS6 protein family raises complex questions in relation to a unifying hypothesis that these proteins have a single function dictated by this domain. We examined the role of structural diversity within the N-terminal and GGL domain of RGS6 splice variants in their interaction with G␤5 and subcellular localization patterns and assessed whether G␤5 alters the subcellular localization of RGS6 proteins. Our results demonstrate unique subcellular distribution patterns of RGS6 protein variants that can be ascribed to N-terminal and GGL domains of these proteins functioning as cytoplasmic retention sequences. The role of the GGL domain in cytoplasmic retention of RGS6 was observed in cells lacking G␤5, providing the first evidence for a function of this domain independent of its interaction with G␤5. Moreover, co-expression studies with G␤5 and RGS6 proteins provide new evidence that G␤5 interaction with RGS proteins promotes changes in their subcellular distribution. These results demonstrate extraordinary complexity in processing of the human RGS6 gene and provide new insight into how structural complexity in RGS6 proteins dictates their subcellular localization and possible functions.

EXPERIMENTAL PROCEDURES
Materials-5Ј-RACE-ready cDNA, marathon-ready cDNA, Quickscreen cDNA library panel and pEGFP vector were purchased from Clontech. pCR2.1 and pCR3.1 were from Invitrogen. Elongase was from Invitrogen. Antibody to and cDNA encoding mouse G␤5 was a generous gift of Dr. William Simonds (National Institutes of Health) and Dr. Vladen Slepak (University of Miami), respectively. Cell culture medium and serum was provided by the Diabetes Endocrinology Research Center (the University of Iowa). Oligonucleotide primers and other molecular biological reagents were obtained from the University of Iowa DNA Core. Polyclonal RGS6 antibodies were generated with a synthetic peptide immunogen corresponding to residues 1-19 of RGS6L by Biosynthesis Incorporated (Lewisville, TX).
PCR Amplification of RGS6 cDNAs-Full-length cDNAs encoding various forms of RGS6 were amplified using a PCR-based strategy we described previously (12). We utilized a 505-bp expressed sequence tag identified as RGS6 (GenBank TM accession number H09621) to design primers for use in 5Ј-and 3Ј-RACE to amplify overlapping segments of RGS6 cDNAs essentially as we described previously (12). Marathonready human brain cDNA (adapter sequences on both cDNA ends) were used as templates for 5Ј-RACE and 3Ј-RACE using adapter-specific forward or reverse primers in combination with appropriate RGS6 forward or reverse primers. This cDNA was synthesized from poly(A)containing mRNA. Resulting PCR products were cloned into pCR2.1, and sequence analysis of multiple clones revealed successful amplification of overlapping 5Ј-and 3Ј-cDNA fragments of RGS6 from brain cDNA. Sequence analysis of these clones revealed the existence of multiple splice variant forms of RGS6. cDNAs encoding these splice variant forms of RGS6 were amplified using forward and reverse primers encompassing the translational start and stop sites, respectively. cDNAs were cloned into pCR3.1 and double-stranded sequencing was performed by automated fluorescent dideoxynucleotide sequencing by the University of Iowa DNA Core Facility.
Preparation of EGFP Constructs of RGS6 -Various RGS6 protein cDNAs were PCR-amplified using gene-specific primers incorporating restriction sites to facilitate their cloning into EGFP vector. First, amplified RGS6 protein cDNAs were cloned in the T/A cloning vector pCR2.1 (Invitrogen). Then, restriction enzyme digestion and agarose gel purification of the cloned RGS6 protein cDNAs was performed. RGS6 protein cDNAs were ligated to EGFP vector in-frame with its C-terminal or N-terminal EGFP sequence. Double-stranded sequencing of all cloned RGS6 protein cDNAs was performed by automated fluorescent dideoxynucleotide sequencing by the University of Iowa DNA Core Facility.
Cell Culture and Transfection-COS-7 cells were grown in DMEM supplemented with 10% fetal bovine serum and gentamycin (50 g/ml) (complete DMEM) in a 5% CO 2 humidified atmosphere at 37°C. COS-7 cells were transiently transfected with vectors containing various RGS6 protein cDNAs and mouse G␤5 cDNA by electroporation using a BioRad Gene-Pulser. Typically, COS-7 cells (10 7 /ml) were transfected with 40 g of plasmid DNA at settings of 0.22 kV and 950 F. Cells were diluted in complete DMEM and plated in two-chambered slides (Nunc) at a density of ϳ10 6 cells/well. Transfected cells were used in experiments 40 h following transfection.
Immunofluorescence and Immunohistochemistry-Cells were rinsed three times with DPBS before fixation for immunofluorescence. For visualization of GFP-tagged RGS6 proteins in COS-7 cells, cells were fixed by treatment with 4% paraformaldehyde for 20 min at room temperature followed by permeabilization with DPBS containing 0.1% Triton X-100 and 0.1% Nonidet P-40 for 10 min at room temperature. After permeabilization, cells were treated with DPBS containing 100 g/ml RNase A (Roche Applied Science) for 20 min at room temperature prior to staining with propidium iodide. For immunocytochemical detection of expressed G␤5 in COS-7 cells, cells were fixed and permeabilized by treatment with 50% methanol, 50% acetone for 1 h at 4°C prior to treatment with RNase A and incubation with anti-G␤5 (ϳ1 g/ml) in DPBS containing 5% bovine serum albumin for 1 h at room temperature. Cells then were rinsed three times with DPBS and incubated in FITC-conjugated secondary antibody (1 mg/ml) in DPBS for 20 min at room temperature. RNase A treatment and propidium iodide staining were performed as described above. Cells were air-dried and then mounted using Vecta Shield mounting solution. Confocal microscopy was performed as we described previously (13). Images shown are representative of a minimum of 1000 cells derived from four or more separate transfections.
Immunohistochemistry was performed by Molecular Histology (MD) with synthetic peptide immunogen affinity-purified (SulfoLink, Pierce) anti-RGS6L. Briefly, tissue slides were incubated with and without anti-RGS6L, followed by biotinylated goat anti-rabbit antibody and strepatavidin-horseradish peroxidase prior to incubation with substrate (DAB). Hematoxylin counterstain (stains nuclei blue) was used.
Co-immunoprecipitation of G␤5 and RGS6 -For co-immunoprecipitation studies, COS-7 cells were co-transfected with RGS6-GFP and mouse G␤5 cDNA and grown for 48 h in 6-well culture dishes. Cells were harvested by lysis with 1 ml of ice-cold RIPA buffer (150 mM NaCl, 50 mM Tris-HCl, pH 7.5, 0.5% deoxycholate, 1% Nonidet P-40, 6 mM MgCl 2 , 10 mM phenylmethylsulfonyl fluoride) followed by centrifugation at 16,000 ϫ g for 1 min at 4°C. Resulting supernatants were incubated with anti-GFP antibodies overnight at 4°C, followed by addition of protein A-Sepharose and incubation for an additional 3 h. Immunoprecipitates were collected by centrifugation and washed three times in RIPA buffer, suspended in Laemmli sample buffer and boiled for 3 min. Proteins were subjected to SDS-PAGE and immunoblotting as described previously (12).

RESULTS
Identification of Human RGS6 cDNAs-We used a PCRbased strategy to first amplify and clone two RGS6 cDNAs (Genbank TM AF073920, AF073921), the first identified RGS6 cDNAs, using sequence information from a 505-bp expressed sequence tag (GenBank TM H09621). The naming of these cDNAs was derived from the conceptual translation of rat brain PCR products of Koelle and Horvitz (5) in which degenerate primers were used to amplify a semiconserved region, now known as the RGS domain, of a family of mammalian proteins sharing sequence identify with EGL-10, the RGS protein in C. elegans. Inspection of the sequence of these two RGS6 cDNAs suggested the possibility of splicing of RGS6 transcripts at their 5Ј-ends; the larger form encoded a protein with an additional 139 N-terminal amino acids, and the cDNAs had different 5Ј-untranslated sequences. An apparent difference at the 3Ј-ends of these two cDNAs was subsequently resolved by identifying a 2-bp sequencing error in the longer form. Thus, the two RGS6 cDNAs appeared to represent long and short N-terminal forms of RGS6 so the short form was named an RGS6 variant to highlight this possibility.
Using 5Ј-and 3Ј-RACE and sequence information from these two RGS6 cDNAs, we were able to identify the existence of 36 novel RGS6 transcripts from oligo(dT)-primed cDNA from human brain. Sequence analysis of full or partial RGS6 cDNAs revealed that these variant RGS6 forms, although highly homologous, exhibit differences in their 5Ј-, 3Ј-, and internal sequences. Two different 5Ј-cDNA ends were identified that correspond to proteins with the long or short N-terminal sequences mentioned above. RGS6 cDNAs with each type of 5Ј-end existed in combination with seven different types of 3Ј-ends, encoding seven different C-terminal domains of the proteins. In addition, each of the RGS6 transcripts distinguished by different 5Ј-and 3Ј-ends existed in two forms that differed in the region encoding the G protein ␥-subunit-like (GGL) domain that defines the subfamily of RGS proteins that includes RGS6, RGS7, RGS9, and RGS11. Sequence differences in this region encode RGS6 proteins with complete or incomplete GGL domains. The coding sequences of the 36 different RGS6 cDNAs encode proteins ranging from 284 to 490 amino acids. Fig. 1 illustrates the structure and naming of RGS6 proteins encoded by the 36 distinct RGS6 cDNAs identified here. We devised a nomenclature based upon the differing N-and Cterminal sequences (Table I) and the presence or absence of a complete GGL domain of the encoded RGS6 proteins. Long and short N-terminal forms of RGS6 were designated as RGS6L and RGS6S, respectively, and the C-terminal forms were designated ␣, ␤, ␥, ␦, ⑀, , and . RGS6L proteins have a 139 N-terminal sequence not present in the otherwise homologous RGS6S proteins. Likewise the forms lacking a 37 amino acid sequence that includes the C-terminal 25 amino acids of the GGL domain, designated (-GGL), are otherwise homologous to RGS6 proteins with complete GGL domains. Moreover, ␣ and ␤ forms exist in two forms, with (␣1, ␤1) or without (␣2, ␤2) an 18 amino acid sequence located N-terminal to their common Cterminal sequence (Table I). Thus, using this nomenclature, long N-terminal forms of RGS6 with the ␣1 terminus with and without the GGL domain are designated RGS6L␣1 and RGS6L␣1(-GGL), respectively. Fig. 1 also illustrates that all forms of RGS6 have an RGS domain and that RGS6S forms lack the DEP domain, whose function is presently unknown, present in RGS6L proteins.
Structure of the Human RGS6 Gene-The finding that the identified RGS6 transcripts encode highly homologous proteins  Table I. that differ only by the presence or absence of sequence cassettes or in their C-terminal sequences suggested these transcripts might arise by alternative splicing mechanisms. Using BLAST nucleotide sequence analysis of identified RGS6 cDNAs with human gene data banks we were able to deduce the complete structure of the human RGS6 gene. The RGS6 gene spans 629,635 bp of DNA of human chromosome 14 and is interrupted by 19 introns, and the 20 RGS6 gene exons range in size from 51 to Ͼ332 bp. The intron-exon organization of the hRGS6 gene in relation to RGS6L␣1 mRNA is shown in Fig. 2. Table II shows the sizes of introns and the intron-exon splice junction sequences in RGS6 transcripts. As shown, introns vary in size from 155 bp to 387 kb, and all of the splice acceptor and donor sequences agree with the GT/AG consensus sequence (29). The RGS6 gene intron phasing is type 0 (the intron occurs between codons) for introns 2, 5, 7, 9 -11, 15-19; type 1 (the intron interrupts the first and second bases of the codon) for introns 3, 4, and 6; and type 2 (the intron interrupts the second and third codon) for introns 8 and 12-14.
Splicing of Human RGS6 Transcripts-We examined the relationship between exon and intron locations of the RGS6 gene to the structure of the 36 RGS6 cDNA variants we identified to gain insight into how these unique RGS6 variants arise. Fig. 3 illustrates the deduced splicing mechanisms involved in generating these 36 RGS6 transcripts. Transcripts encoding the two N-terminal forms of RGS6 arise by use of different transcriptional start sites. Exons 1-7 encode the 5Јuntranslated region and unique N-terminal domain of RGS6L proteins while an alternate transcription start site within intron 7 generates RGS6S transcripts whose 5Ј-untranslated sequence and translation start site are encoded within intron 7 (non-coding exons A, B, C, and D) and exon 8. The translational start site for RGS6S proteins begins at nucleotide 24 of exon 8, resulting in no reading frameshift compared with RGS6L proteins. The shared common internal domain found in all RGS6 proteins is encoded by exons 8 to 16 with splicing out of exon 13 in transcripts encoding -GGL forms of these proteins. The conserved RGS domain present in RGS6 proteins is encoded by exons 14 -16. Complex alternative splicing at the 3Ј-end of RGS6 transcripts generates 7 different C termini of RGS6 proteins. The C terminus of ␣1 forms of RGS6 is generated by splicing together exons 16, 18, and 20. ␤1 forms of RGS6 are formed the same way, except exon 18 is spliced to an alternate splice site (␤ splice site) within exon 20 located 298 bp 3Ј to the ␣ splice site (Table II). ␣2 and ␤2 forms of RGS6 transcripts differ from their ␣1 and ␤1 counterparts by splicing out of exon 18, resulting in deletion of 18 amino acids encoded by this exon with no change in reading frame in the encoded proteins. The unique C terminus present in ␥, ␦, and ⑀ forms of RGS6 is encoded by exon 19, these transcripts unique in retention of this exon. Indeed, alternative splicing within exon 19 generates transcripts encoding the ␦ and ⑀ forms, utilizing splice sites located 9 and 12 bp, respectively, 3Ј to the splice site used for ␥ RGS6 transcripts (Table II), with splicing out of exon 17 in all three forms as found in ␣ and ␤ splice forms of RGS6. The unique C terminus of forms of RGS6 arise from transcripts in which exon 16 is spliced to exon 17 while that of forms of RGS6 arise from transcripts in which exon 16 is spliced to an alternate splice site in exon 18 (Table II). These 3Ј splicing mechanisms thus generate transcripts encoding seven different C-terminal tails of RGS6, with two of these forms existing with or without exon 18-encoded sequences. Thus, use of two alternate transcription start sites to generate two different N-terminal domains of RGS6 and alternative splicing to generate RGS6 proteins with complete or incomplete GGL domains and nine different domains C-terminal to the conserved RGS domain leads to unusual complexity in the RGS6 protein family.
Interaction of RGS6 proteins with G␤5-The subfamily of RGS proteins that contain GGL domains are unique in their ability to interact with the atypical G␤ subunit G␤5, indeed, the GGL domain of these RGS proteins mediates interactions with G␤5 (17,18). The precise function of G␤5 interaction with members of this RGS protein subfamily is unclear although recent evidence suggests that such interactions may have a role in G␤5 protein stability (10,18,20), in regulation of RGS protein signaling or GAP activity (21)(22)(23)(24)(25) and in localization of G␤5 (26). The existence of natural splice variants of RGS6 differing by the presence of a complete or incomplete GGL domain immediately raised the question of whether these proteins differed in their ability to interact with G␤5. Therefore, we examined the ability of both long and short forms of RGS6 and their -GGL splice variant forms to interact with G␤5 in COS-7 cells by co-immunoprecipitation studies. We focused our attention on the ␣2 splice variant forms of RGS6 for these studies. COS-7 cells were co-transfected with G␤5 and GFPtagged forms of RGS6 proteins, and cell lysates were subjected to immunoprecipitation with anti-GFP and immunoblotting with anti-GFP and anti-G␤5. Fig. 4A shows that G␤5 co-precipitated with RGS6S␣2 but not RGS6S␣2(-GGL) or GFP alone and, in a similar fashion, RGS6L␣2 but not RGS6L␣2(-GGL) effectively co-precipitated G␤5 from cell lysates (Fig. 4B). These results demonstrate that G␤5 interacts with both long and short splice variant forms of RGS6 with complete GGL domains but does not interact with the corresponding splice variant forms lacking the approximate C-terminal half of the GGL domain in these proteins.
Role of N Terminus and GGL Domain in Subcellular Localization of Human RGS6 Proteins-Different splice variant forms of RGS6 were tagged with GFP to examine their subcellular distribution by confocal microscopy. We focused our experimental attention on RGS6 proteins possessing a common C terminus so we could examine the role of N-terminal and GGL domain variants in subcellular localization. For these experiments, we studied the ␣2 C-terminal variants because of their good expression in COS-7 cells and their characterized interaction with G␤5 (Fig. 4). Fig. 5 shows confocal microscope images of COS-7 cells expressing GFP-tagged forms of RGS6L␣2, RGS6L␣2(-GGL), RGS6S␣2, and RGS6Sa2(-GGL). Green represents GFP fluorescence from expressed RGS6 proteins, red represents fluorescence from propidium iodide staining of nuclei in these cells and yellow (in the overlay) shows overlapping red and green fluorescence. RGS6L␣2 was localized exclusively in the cytoplasm of cells with a fine granular distribution pattern (Fig. 5A). In contrast, RGS6S␣2 was variably localized in the cytoplasm, nucleus, and nucleoli of COS-7 cells. Fig. 5B shows the three major patterns of intracellular distribution of RGS6S␣2 that were observed. Approximately 60% of cells exhibited a predominant nuclear pattern of RGS6S␣2 (middle panel) while ϳ30% of cells exhibited a punc- tate pattern of distribution throughout the cytoplasm and nucleus (left panel). This latter pattern is most similar to that of RGS6L␣2 but differs by its localization in the nucleus and its punctate appearance. Nucleolar localization of RGS6S␣ was observed in 2-5% of transfectants (right panel). These three distinct patterns of expression of RGS6S␣2 were reproducibly observed and did not appear to be related to the levels of expression of RGS6S␣2 in cells. Because RGS6L␣2 differs from RGS6S␣2 only by its unique N-terminal domain, these results suggest that the N-terminal domain of RGS6L␣2 is responsible for its exclusive cytoplasmic localization.
To examine the role of the GGL domain in RGS6 protein localization, we compared the intracellular distribution of RGS6L␣2 and RGS6S␣2 to their splice forms lacking a complete GGL domain. Fig. 5A shows that RGS6L␣2(-GGL) exhibited a cytoplasmic pattern of distribution like that observed for its GGL domain-containing counterpart. However, RGS6S␣2 (-GGL) was localized exclusively in the nucleus and nucleolus  Leu Tyr Ser Thr with no perinuclear or cytoplasmic accumulation of the protein (Fig. 5C). In contrast, the GGL domain-containing form of this protein exhibited a punctate distribution throughout the cytoplasm and nucleus in most transfectants while nuclear or nucleolar localization of RGS6S␣2 was observed in a small fraction of transfectants and was always accompanied by considerable perinuclear and cytoplasmic protein localization (Fig. 5B). These results show that the GGL domain is required for cytoplasmic localization of RGS6S␣2 but does not play a similar role in cytoplasmic targeting of RGS6L␣2.
Recently, we showed that certain RGS proteins localize in the cytoplasm or nucleus based upon structural differences in the proteins (13). The RGS domain was identified as the molecular determinant for nuclear localization of these RGS proteins, and N-terminal sequences were identified that functioned as nuclear export sequences in RGS4 and RGS16 and as a cytoplasmic retention sequence in RGSZ. Thus, the requirement of the GGL domain for cytoplasmic localization of RGS6S␣2 could reflect its role as a cytoplasmic retention sequence, as a nuclear export sequence or as a sequence that prevents nuclear import of the protein by unknown mechanisms. This latter possibility seems unlikely, in part because of the observed nuclear and nucleolar localization of the GGL domain-containing RGS6S␣2 in some cells (Fig. 5B). In addition, sequence analysis of the GGL domain of RGS6 proteins revealed the absence of prototypical nuclear export sequences like those we described in RGS4 and RGS16.
Recently, Zhang et al. (26) reported that endogenous RGS7 and G␤5 are present in both the cytoplasm and nucleus in mouse brain and PC12 cells. Ectopically expressed G␤5, like native G␤5, localized to the cytoplasm and nucleus of PC12 cells, while a G␤5 mutant impaired in binding to GGL domains of RGS proteins was localized mainly in the cytoplasm when expressed in PC12 cells. Thus, it was suggested that G␤5 undergoes nuclear translocation in neurons by an RGS proteindependent mechanism. Indeed, several previous studies have shown that G␤5 exists in complex with GGL domain-containing RGS proteins in native tissues and in cells expressing both proteins (17,18,20,23,30,31). Therefore, it is of particular interest that we observed a different pattern of subcellular localization of RGS6S␣2 and RGS6S␣2(-GGL) in COS-7 cells, which do not express G␤5 endogenously. That is, any difference in subcellular localization of these two proteins is due to the presence or absence of a complete GGL domain but is unrelated to interactions with G␤5. In view of these observations and the interaction of RGS6 proteins with G␤5 (Fig. 4), it seemed essential to determine whether the patterns of localization of G␤5 and RGS6 proteins containing GGL domains are influenced by the expression of each other.
GFP-tagged RGS6 proteins and G␤5 were co-transfected into COS-7 cells, and their subcellular distribution was examined by GFP fluorescence and immunocytochemistry (G␤5). In cells transfected with G␤5 alone, G␤5 was expressed homogenously throughout the cytoplasm and nucleus as has been reported previously (26). Fig. 6 illustrates the subcellular patterns of distribution observed during co-expression of G␤5 with RGS6L␣2 and RGS6S␣2. As shown, co-expression of G␤5 with RGS6L␣2 or RGS6S␣2 promoted nuclear localization of both RGS6 proteins (Fig. 6A) as well as G␤5 and their resulting co-localization in the nucleus (Fig. 6B). While it is clear that RGS6 expression with G␤5 promotes nuclear localization of G␤5, some G␤5 is still found in the cytoplasm (Fig. 6). However, the effect of G␤5 co-expression on RGS6 protein localization in the nucleus is particularly striking. RGS6S␣2 is localized exclusively in the nucleus when co-expressed with G␤5 quite in contrast to its patterns of expression when expressed alone (Fig. 5). Moreover, the exclusive cytoplasmic distribution observed during expression of RGS6L␣2 alone (Fig. 5) contrasts with the predominant nuclear pattern of expression of this protein when expressed with G␤5 (Fig. 6). Some RGS6L␣2 is still present in the cytoplasm of these G␤5 expressing cells, again illustrating that the N-terminal domain of this protein is important in its cytoplasmic targeting (i.e. by comparison to RGS6S␣2) as noted above.
Expression of RGS6L-The 36 identified RGS6 transcripts are highly homologous and differ only in sequences encoding the long N terminus of RGS6L, the GGL domain and the extreme C terminus of splice forms. Sequence overlap among transcripts encoding the C-terminal forms of RGS6 makes it problematic to determine the expression pattern of one transcript independent of others. For example, exon 19 splice forms differ by only 9 -12 bp and splice forms with exon 18 and/or 20 sequences share overlapping sequence, e.g. coding in some forms and non-coding in others (Table II, Fig. 3). In addition, the sizes of the coding sequences of transcripts for different C-terminal forms of RGS6 are very similar although differences may exist in their 3Ј-untranslated regions (i.e. which were not determined for each splice form). However, splicing in or out of exon 13 sequences in RGS6 transcripts enabled us to demonstrate expression of transcripts encoding both RGS6L␣2 and RGS6L␣2(-GGL) in human brain (Fig.  7A). In addition, we raised an antibody to a synthetic peptide corresponding to residues 1-19 of RGS6L proteins, a sequence present only in the unique N-terminal domain of RGS6L. Fig. 7B shows that this polyclonal anti-RGS6L antibody recognized ectopically expressed RGS6L␣2 as well as RGS6L protein(s) in mouse brain lysates. Immunohistochemistry of adult mouse brain demonstrated RGS6L immunoreactivity in several brain regions including the hippocampus and cerebellum (Fig. 7C). We found patterns of both cytoplasmic-and nuclear-localized RGS6L immunoreactivity in brain regions. It is possible that these patterns of localization in native brain could recapitulate the patterns of localization we observed during expression of RGS6L␣2 alone (cytoplasmic) or together with G␤5 (nuclear), as G␤5 is expressed in cerebellum (32). However, it is unclear which C-terminal forms of RGS6L are present in these tissues and whether RGS6L proteins with different C-terminal domains localize differ- Exons are shown as filled boxes and noncoding sequences as empty boxes. Two primary transcripts encode two 5Ј-splice forms of RGS6; one is processed to RGS6L transcripts and the other is processed to RGS6S transcripts. Retention or skipping of exon 13 generates transcripts encoding proteins with or without a complete GGL domain, respectively. 3Ј-splicing generates seven different 3Ј-ends from each primary RGS6 mRNA. Use of alternate splice sites in exon 20 (Table II) generates ␣ and ␤ forms of RGS6 while ␥, ␦, and ⑀ forms arise from use of three alternate splice sites within exon 19 (Table II). Splicing of exons 16 to 17 or to an alternate splice site in exon 18 (Table II) generates and forms of RGS6, respectively. RGS6 ␣ and ␤ transcripts each exist in two splice variant forms arising by retention (␣1 and ␤1) or skipping (␣2 and ␤2) of exon 17. ently. It is also possible that signaling-mediated changes regulate RGS6L localization in brain, as we provide evidence for nuclear and nucleolar localization of RGS6 proteins in response to activation of stress signaling pathways in the accompanying article (37). DISCUSSION The present study has revealed extraordinary complexity in the processing of the human RGS6 gene to generate 36 distinct forms of RGS6 differing in N-and C-terminal and GGL domain sequences. The possibility of such processing was evident when we first reported the sequence of two RGS6 cDNAs that differed by the presence or absence of 5Ј-sequence encoding the N-terminal domain of one form. However, the identification of 36 RGS6 cDNAs enabled us to determine the structure of the human RGS6 gene and to deduce how these distinct transcripts arise. While it is possible that related transcripts could arise from distinct genes, our analysis shows that all 36 RGS6 mRNAs arise from a single human RGS6 gene by use of two distinct transcription start sites and by complex alternative splicing mechanisms. Although all forms of RGS6 described here retained the hallmark RGS domain characteristic of this protein family, our results show that RGS6 protein variants with structural differences outside of this domain exhibit differences in their subcellular localization patterns and interactions with G␤5. The observed nuclear or nucleolar localization of certain RGS6 protein variants raises the possibility that their function(s) might be very different from that of other RGS proteins or other RGS6 protein variants that are not localized to these organelles.
Previous studies of RGS6 protein function have been limited to analysis of interaction of one form of RGS6 with G␤5 and G␣ subunits. This form, corresponding to what we have named RGS6L␣2, was amplified by Posner et al. (25) using our RGS6 sequence information and Snow et al. (18) isolated the same RGS6L␣2 cDNA and one with a 3Ј-cDNA end insertion (that corresponds to RGS6L␣1). Snow et al. (18) showed that RGS6L specifically bound to G␤5 during co-translation in reticulocyte lysates and that G␤5 was co-precipitated from COS-7 cells transfected with both cDNAs. G␤5 did not bind to RGS6L in reticulocyte assays when its GGL domain was deleted or mutated. Subsequent studies by Posner et al. (25) demonstrated that RGS6L⅐G␤5 complexes purified from Sf9 cells accelerated the GTPase activity of G␣ o , but not that of other G␣ subunits. These studies suggest that the GGL domain of RGS6L␣2 is a binding partner for G␤5, as observed for other GGL domaincontaining RGS proteins (17,20,31), and that purified heteromeric complexes of RGS6L and G␤5 enhance the GTPase activity of G␣ o specifically in vitro. The identification of RGS6 protein splice variants lacking a complete GGL domain and shown here not to interact with G␤5, obviously raises the possibility of different functional activities of these proteins in vivo. However, it must be acknowledged that the role of RGS6 proteins and of G␤5 binding to them in cells is not known and will require further cell-based studies. A, immunoblots showing that G␤5 co-precipitates with RGS6S␣2 but not RGS6S␣2(-GGL). B, immunoblots showing that G␤5 co-precipitates with RGS6L␣2 but not RGS6L␣2(-GGL). COS-7 cells were transfected and immunoprecipitation and immunoblotting was performed as described under "Experimental Procedures." Indeed, the precise role of G␤5 interactions with other members of the GGL domain-containing family of RGS proteins is the subject of some controversy. Studies by Slepak and coworkers (20,23) have shown that G␤5 interaction with RGS7 prevents interaction of RGS7 with G␣ in vitro, yet purified RGS7⅐G␤5 complexes enhance the GTPase activity of G␣ o in vitro (25). Also, while purified RGS7⅐G␤5 complexes selectively enhanced the GTPase activity of G␣ o in vitro, expression of RGS7 and G␤5 in cells inhibited M3 muscarinic receptor-induced calcium mobilization mediated by G␣ q (20). The possibility that G␤5 interaction with RGS7 determines its specificity toward G␣ o is not supported by studies showing that the G␣ o specificity of RGS7 resides within its RGS domain (33). A role for RGS protein-G␤5 interactions in effector regulation of G proteins was demonstrated by findings that G␤5 interaction with RGS9 enhances the GAP activity of cGMP phosphodiesterase toward transducin, although part of this G␤5 effect was independent of the GGL domain of RGS9 (22). Finally, G␤5 binding to RGS proteins has been suggested to increase stability of both proteins by post-transcriptional mechanisms (20), likely accounting for the lack of G␤5L in retinas of RGS9 Ϫ/Ϫ mice (10).
Our analysis of subcellular localization of RGS6 proteins provides new insights into where and how RGS6 proteins are targeted in the cell. A novel role of the GGL domain and of co-expressed G␤5 in RGS6 protein localization is suggested from our studies. We limited our studies and the following discussion of the subcellular distribution of RGS6 proteins to variants with a common C-terminal ␣2 domain. This enabled us to focus on the role of structural differences in the N terminus and GGL domains of RGS6 protein variants and of the GGL domain-binding protein G␤5 on subcellular targeting of RGS6 proteins. Initially, we found that RGS6L␣2 localized exclusively in the cytoplasm while RGS6S␣2 localized in puncta throughout the cytoplasm and nucleus or, less commonly, localized primarily in the nucleus or nucleoli with reduced cytoplasmic localization. The long N terminus of RGS6L proteins, therefore, is responsible for their exclusive cytoplasmic targeting. A similar role of the GGL domain was suggested by the finding that RGS6S␣2(-GGL) localized exclusively in the nucleus or nucleoli, in contrast to the pattern of localization observed for its GGL domain-containing splice variant counterpart. Although these results show that the GGL domain is required for cytoplasmic targeting of RGS6S proteins, RGS6L␣2(-GGL) had the same subcellular distribution pattern as its GGL domain-containing splice variant. These findings suggest that the N terminus common to RGS6L protein variants and the GGL domain found in both RGS6L and RGS6S proteins function as cytoplasmic retention sequences. The GGL domain of RGS6 proteins may contribute to but is not required for the cytoplasmic localization of RGS6L proteins, i.e. the N terminus of these proteins overcomes the need for the GGL domain. This differential requirement of the GGL domain in cytoplasmic targeting of RGS6S and RGS6L proteins was observed in cells not expressing G␤5, providing new evidence for G␤5-independent roles of this protein domain.
However, co-expression of the GGL domain binding protein G␤5 with RGS6L promoted the exclusive nuclear and nucleolar targeting of RGS6L. Thus, co-expression of G␤5 unveiled a cryptic role of the GGL domain of RGS6L proteins in cytoplasmic targeting that is overshadowed by its N-terminal domain in the absence of G␤5. Our results showed that G␤5 binding to RGS6L, as well as to RGS6S, is mediated by binding to the GGL domain because no binding was observed in RGS6 splice variants lacking a complete GGL domain. Thus our studies with naturally occurring splice variants of RGS6 confirm the importance of the GGL domain in G␤5 binding determined by mutational analysis of this domain (18). The equivalent binding of G␤5 to RGS6L and RGS6S that we observed further suggests that G␤5 interaction with the GGL domains of these proteins is not hindered or enhanced by the presence of the long N-terminal domain of RGS6L. However the complete conversion of RGS6L␣2 from an exclusively cytoplasmic to exclusively nuclear and nucleolar protein by co-expression of G␤5 suggests that G␤5 interaction with RGS6L diminishes or neutralizes the role of both its N terminus and GGL domain in cytoplasmic targeting. It is interesting to speculate that G␤5 could produce such an effect if its binding to the GGL domain of RGS6L produced both conformational changes in its N terminus that disrupted its role in cytoplasmic targeting and prevented interaction of its GGL domain with protein(s) required for the cytoplasmic targeting function of the GGL domain revealed in our studies. The ability of G␤5 to enhance the stability of certain GGL-containing RGS proteins might reflect such conformational changes, which will need to be established by structural studies of G␤5⅐RGS protein complexes. It is also possible that: 1) G␤5 possesses NLSs that confer nuclear targeting of the RGS6L⅐G␤5 complex, or 2) formation of the RGS6L⅐G␤5 complex facilitates its interaction with NLS-containing proteins resulting in a "piggyback" mechanism of nu- clear/nucleolar transport. G␤5 does not possess prototypical NLSs although it does exhibit some nuclear localization when expressed alone. However, we did identify, in the accompanying article (37), functional NoLSs that target RGS6 proteins to the nucleus and nucleolus during activation of stress signaling (i.e. NoLSs also represent bona fide NLSs). Deletion mutagenesis identified these motifs in regions corresponding to amino acids 40 -121 (DEP domain), 121-182 and the RGD of RGS6L. Indeed, we previously demonstrated that isolated RGS domains possess constitutive NLSs (13). Therefore, it is possible that one or more of these sequences may account for nuclear/ nucleolar localization observed with RGS6S and with RGS6L co-expressed with G␤5. Therefore, we favor the hypothesis that G␤5 neutralizes the cytoplasmic retention functions of the GGL domain and N terminus of RGS6L to promote its localization to nuclear and nucleolar sites.
Indeed, our results provide the first evidence that G␤5 inter-action with RGS proteins has an effect on RGS protein localization. Co-expression of G␤5 resulted in the localization of RGS6L␣2 in the nucleus and nucleoli in contrast to the exclusive cytoplasmic localization of RGS6L␣2 in cells not expressing G␤5. An experiment opposite to ours showed the lack of nuclear localization of a G␤5 mutant in PC12 cells that was impaired in binding to GGL domains, suggesting RGS-dependent nuclear localization of G␤5 (26). Indeed, it is possible that RGS6 localization to the nucleus brings along the bound G␤5 in our studies. The remaining G␤5 in the cytoplasm of cells in which RGS6S␣2 and RGS6L␣2 were exclusively localized in the nucleus or nucleoli, could reflect saturation of RGS6 GGL domains with G␤5 due to higher levels of expression of G␤5 relative to RGS6. It is unknown whether RGS6 is an obligate heteromer with G␤5 in cells expressing both proteins. It is of note that native RGS6L exhibited both cytoplasmic and nuclear patterns of subcellular localization in mouse cerebellum. Evaluation of this finding in light of the present studies in COS-7 cells raises the very interesting possibility that the cytoplasmic and nuclear/nucleolar forms of RGS6L may represent monomeric RGS6L and RGS6L⅐G␤5 complexes, respectively, as G␤5 is expressed throughout the brain and in cell lines of neuronal origin (32). Alternatively, it is possible that transport of RGS6L⅐G␤5 complexes to nuclear and nucleolar sites is subject to regulation or that, in the absence of G␤5, other mechanisms exist for nuclear/nucleolar transport of monomeric RGS6L. Indeed, we now show the existence of this latter possibility in the accompanying article.
The nucleolar localization of RGS6 is unique among all RGS proteins studied to date. Studies in our laboratory first showed that some RGS proteins are nuclear proteins (RGS2, RGS10, and RGS12) and that others are nucleocytoplasmic shuttle proteins (RGS4 and RGS16) (12,13). Studies in other laboratories supported these observations and showed that RGS3 and RGS8 also are nuclear RGS proteins (11,15,16,34). Obviously, nuclear RGS proteins would be expected to have functions quite distinct from regulation of cell surface G protein-coupled receptor signaling. Recent findings provide the first insights into possible nuclear roles of RGS proteins. We found that RGS12 is a nuclear matrix protein and possesses cell cycle and transcriptional regulatory effects (35) and a recent study demonstrated that RGS2 promotes adipocyte differentiation (36).
It seems likely that the complex splicing of RGS6 mRNAs to generate 36 distinct forms of RGS6 proteins may be important in not only the localization of RGS6 proteins, but also in their tissue-specific patterns of expression, stability, and cellular functions. We speculate that differences in C-terminal domains of RGS6 protein splice variants, like those in N-terminal and GGL domains, may impact the subcellular localization or function of these proteins. Our results suggest new functions for the GGL domain, that are independent of its interaction with G␤5, and of the GGL domain-binding protein G␤5 and the N terminus unique to RGS6L proteins in the subcellular targeting of RGS6 proteins. Thus, splicing of RGS6 transcripts has important implications in generating proteins with different modular functional domains. It seems clear that some forms of RGS6 may be in a position to exert regulatory effects on G protein signaling by virtue of their localization in the cytoplasm or near plasma membrane localized G proteins, while localization of other splice variants or the same (i.e. as a heteromer with G␤5) RGS6 proteins in the nucleus likely would preclude such effects and/or perhaps position these proteins for specific nuclear or nucleolar functions. Hopefully the present work will facilitate studies to reveal the likely complex roles of the RGS6 protein family.