Gγ Subunit-selective G Protein β5Mutant Defines Regulators of G Protein Signaling Protein Binding Requirement for Nuclear Localization*

The signal transducing function of Gβ5 in brain is unknown. When studied in vitro Gβ5 is the only heterotrimeric Gβ subunit known to interact with both Gγ subunits and regulators of G protein signaling (RGS) proteins. When tested with Gγ, Gβ5interacts with other classical components of heterotrimeric G protein signaling pathways such as Gα and phospholipase C-β. We recently demonstrated nuclear expression of Gβ5 in neurons and brain (Zhang, J. H., Barr, V. A., Mo, Y., Rojkova, A. M., Liu, S., and Simonds, W. F. (2001) J. Biol. Chem. 276, 10284–10289). To gain further insight into the mechanism of Gβ5 nuclear localization, we generated a Gβ5 mutant deficient in its ability to interact with RGS7 while retaining its ability to bind Gγ, and we compared its properties to the wild-type Gβ5. In HEK-293 cells co-transfection of RGS7 but not Gγ2 supported expression in the nuclear fraction of transfected wild-type Gβ5. In contrast the Gγ-preferring Gβ5 mutant was not expressed in the HEK-293 cell nuclear fraction with either co-transfectant. The Gγ-selective Gβ5 mutant was also excluded from the cell nucleus of transfected PC12 cells analyzed by laser confocal microscopy. These results define a requirement for RGS protein binding for Gβ5 nuclear expression.

In eukaryotic cells seven transmembrane-spanning receptors regulate intracellular processes in response to extracellular signals through their interaction with signal-transducing heterotrimeric guanine nucleotide-binding regulatory proteins (G proteins) (1). Complementary DNAs from five G protein ␤ subunit genes (G␤ [1][2][3][4][5] ) have been identified by molecular cloning. The G␤ 5 isoform shares much less homology with other isoforms (ϳ50%) and is preferentially expressed in brain (2). A splice variant of G␤ 5 , G␤ 5 -long (G␤ 5 L), is present in retina that contains a 42-amino acid N-terminal extension (3). G␤ 5 is the only G␤ subunit known with the potential to assemble with either G␥ subunits or regulators of G protein signaling (RGS) 1 proteins. G␤ 5 can heterodimerize with G␥ (2, 4 -6) and interact with other classical components of heterotrimeric G proteins signaling pathways such as G␣ (5,6) and phospholipase C-␤ (2, 4, 6 -8) when tested in vitro. Other in vitro studies show that G␤ 5 and G␤ 5 L, but not the other G␤ isoforms, can form tight heterodimers with RGS proteins 6, 7, and 11, an interaction mediated by a G␥-like (GGL) domain present in a subfamily of RGS proteins (9 -11). In native tissues, however, G␤ 5 has been purified not as a heterodimer with G␥ but instead bound to RGS6 (12) and RGS7 (12)(13)(14)(15). The failure to demonstrate native G␤ 5 -G␥ complexes is complicated by the known instability of such complexes in detergent solution (10,16). Thus, purifications employing detergent may fail to identify potential G␤ 5 -G␥ complexes that could dissociate in the process of isolation (16). Alternative approaches to assess the biologic importance of the G␤ 5 -G␥ interaction demonstrated in vitro are needed.
We recently demonstrated the nuclear expression of G␤ 5 in neurons and brain (17). In this study chimeric protein constructs containing green fluorescent protein (GFP) fused to wild-type G␤ 5 demonstrated nuclear localization in transfected PC12 cells, but not GFP fusions with a mutant G␤ 5 with proline substitutions in its putative coiled-coil domain. Introduction of prolines into the putative ␣-helical N-terminal region of G␤ 5 would disrupt its ability to form a coiled-coil with both G␥ subunits and RGS proteins containing GGL domains. Such a G␤ 5 mutant therefore fails to provide mechanistic insight as to which of its two potential heterodimeric conformations is important for nuclear targeting. We describe here a G␥-selective G␤ 5 mutant that fails to undergo nuclear localization in transfected PC12 and HEK-293 cells. In the latter cells the nuclear expression of wild-type G␤ 5 is dependent on RGS7 co-transfection. These results suggest it is interaction with the GGL domain containing RGS proteins that directs nuclear localization of G␤ 5 . Additional studies of RGS7 do not support a role for a putative interaction with 14-3-3 proteins in G␤ 5 nuclear localization.

EXPERIMENTAL PROCEDURES
Identification of G␤ 5 -specific Residues in the Putative Coiled-coil Domain of G␤ 5 -The Pileup algorithm of the University of Wisconsin Genetics Computer Group (18) was used to align the coding sequences of human (GenBank TM accession number NM_006578) (19), tiger salamander (GenBank TM accession number AF369757) (20), Drosophila melanogaster (GenBank TM accession number AAF46336), and Caenorhabditis elegans (GenBank TM accession number AF291847) (21-23) G␤ 5 with mammalian G␤ subunits 1-4. Amino acids in the line up that were identical among the four G␤ 5 sequences, but not found in the * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. corresponding positions of G␤ [1][2][3][4] , and belonging to a different amino acid homology group from the corresponding positions of G␤ 1-4 were identified and considered to be conserved, G␤ 5 -specific residues. For the purposes of this analysis the amino acid homology groups were taken to be basic (His, Lys, and Arg), acidic (Asp and Glu), polar uncharged (Cys, Gly, Asn, Gln, Ser, Thr, and Tyr), and nonpolar (Ala, Phe, Ile, Leu, Met, Pro, Val, and Trp). The portion of this analysis including the N-terminal coiled-coil region of the G␤ subunits is shown in Fig. 1.
cDNAs and Mutagenesis of G␤ 5 and RGS7-Expression constructs encoding full-length murine G␤ 5 in pcDNA3 (4) and bovine RGS7 in pcDNA4/HisMax-C (17), which adds N-terminal His 6 and Xpress epitope tags, have been described previously. N-terminally HA epitopetagged G␤ 5 in pcDNA3 (coding sequence inserted between HindIII and BamHI sites) was created by PCR to give the following sequence, MAYPYDVPDYAEFKAA . . . , in which the starting methionine and the alanine corresponding to position 2 in the wild-type sequence are underlined, flanking 14 residues of inserted sequence including the nonapeptide HA epitope (24). The mutants HA-G␤ 5 -QAAC (G␤ 5 -Lys 25 3 Gln/Glu 29 3 Ala/Lys 32 3 Ala/Leu 33 3 Cys) and HA-G␤ 5 -YMIN (G␤ 5 -Phe 93 3 Tyr/Thr 338 3 Met/Val 351 3 Ile/Ala 353 3 Asn) were generated by PCR from a template of wild-type HA-G␤ 5 in pcDNA3 employing the Pwo polymerase (Roche Molecular Biochemicals) by overlap extension as described previously (25). N-terminally HA epitope-tagged G␤ 1 in pcDNA3 (coding sequence inserted between EcoRI and XbaI sites) was created by PCR to give the following sequence, MAYPYDVPDYAGS . . . , in which the starting methionine and the serine corresponding to position 2 in the wild-type sequence are underlined, flanking 11 residues of inserted sequence including the nonapeptide HA epitope (24). The point mutant RGS7-Ser 434 3 Ala (S434A) was generated by PCR from a template of wild-type bovine RGS7 in pcDNA4/HisMax-C (17) employing the Pwo polymerase by overlap extension (25). G␥ 2 in pcDNA4/HisMax-C vector (BamHI to NotI) was generated by Pwo polymerase PCR using bovine G␥ 2 as a template (26). The DNA sequence of all inserts was verified by dideoxy sequencing (27).
Cell Culture-Rat pheochromocytoma PC12 cells were grown in 75cm 2 flasks grown at 37°C and 5% CO 2 containing DMEM supplemented with 10% horse serum, 5% fetal bovine serum, 4 mM L-glutamine, 1ϫ penicillin/streptomycin (BioFluids) (supplemented DMEM). HEK-293 cells were grown as above except that 10% fetal bovine serum and no horse serum were used in the medium.
Transient Transfection of PC12 and HEK-293 Cells-One day prior to transient transfection, PC12 cells at 90% confluence were harvested and plated in 2 ml of supplemented DMEM into the chambers of poly-D-lysine-coated chamber slides. The number of cells was adjusted to 70% of their density at harvest. After overnight incubation of the cells, 1.5 g of plasmid DNA and 4 l of LipofectAMINE 2000 reagent (Invitrogen), diluted in 100 l of Opti-MEMI medium, were mixed, incubated at room temperature for 20 min, and added directly to each chamber. After mixing gently, the cultures were incubated for 24 -48 h prior to confocal microscopic analysis. HEK-293 cells were transfected in 75-cm 2 culture flasks using either LipofectAMINE 2000 or Superfect transfection reagent (Qiagen) according to the manufacturer's recommendations. Typically 10 g of total plasmid DNA and 60 l of Lipo-fectAMINE 2000 reagent or 15 g of total plasmid DNA and 40 l of Superfect reagent were used per 75-cm 2 culture flask.
Inositide-specific Phospholipase C Activity Determination-To determine the G␥-dependent activation of inositide phospholipid-specific phospholipase C (PLC) by wild-type and mutant G␤ 5 constructs, HEK-293 cells metabolically labeled with [ 3 H]inositol and co-transfected with human PLC-␤2 were employed, and the [ 3 H]inositol phosphates isolated by the method of Berridge et al. (29) with modifications as described previously (4). The PLC activity in cells co-transfected with vector only was compared with that in cells transfected with G␤ 5 construct without or with G␥ 2 .
Subcellular Fractionation of HEK-293 Cells-48 h following transfection, HEK-293 cells were harvested from a 75-cm 2 flask, split into two fractions, and pelleted by centrifugation at 500 ϫ g for 5 min at 4°C. One cell pellet was used to isolate cell nuclei and generate nuclear extracts using a commercially available cell lysis kit according to the manufacturer's instructions (NE-PER Reagents, Pierce). The other cell pellet was resuspended in ice-cold lysis buffer (25 mM HEPES, pH 7.5, 0.3 M NaCl, 0.2 mM EDTA, 1.5 mM MgCl 2 , 1% Triton X-100, 0.1% SDS, and 0.5% sodium deoxycholate containing 1ϫ protease inhibitor mixture (Calbiochem Set III)). After removal of the insoluble material by centrifugation, the supernatant fraction was employed as whole cell lysate.
Immunocytochemistry and Confocal Laser Microscopy-PC12 cells were processed for immunofluorescent staining as described previously (30). Briefly, PC12 cells were plated onto poly-D-lysine pre-coated covered chamber slides (Lab-Tek II, Nalge) and grown in supplemented DMEM at 37°C for 16 h. The medium was discarded, and the cells were washed and then fixed in 2% (v/v) formaldehyde in Dulbecco's phosphate-buffered saline (Biofluids, Rockville, MD). The slides were then incubated with one or more primary antibodies in Dulbecco's phosphate-buffered saline, 10% fetal calf serum (v/v), and 0.075% (w/v) saponin for 1 h, washed, and then incubated with appropriate labeled secondary antibody (fluorescein isothiocyanate-conjugated goat antirabbit IgG (H ϩ L) (Jackson ImmunoResearch, West Grove, PA)) and/or Rhodamine Red TM -X goat anti-mouse IgG (H ϩ L) (Molecular Probes)) in the same buffer for 45 min. After staining, 1-2 drops of Permount (Vector Laboratories) was added to the sample surface. Confocal images were collected using a Zeiss LSM 510 laser scanning confocal microscope with a 100ϫ/1.3 N.A. Plan-Neofluar objective. Scans were performed sequentially using 488 and 543 nm excitation and bandpass emission filters of 505-550 and 560 -615 nm, respectively, for Rhodamine Red and fluorescein to eliminate spectral bleed through of fluorescence between the red and green channels. The pinhole in all cases was adjusted to produce a 1.5-m optical slice thickness. Transmitted light differential interference contrast images were collected simultaneously with the green dye using 488 nm excitation. Final images were processed using Zeiss LSM software. 5 Mutants-In order to assess the potential biologic importance of the G␤ 5 -G␥ interaction for the nuclear localization of G␤ 5 , a G␤ 5 mutant that retained demonstrable G␥ interaction with selective loss of RGS protein binding was desired. By analogy with the crystal structures of G protein heterotrimers (31,32) and free G␤ 1 ␥ 1 complex (33), it was expected that the interaction of G␥ or an RGS protein GGL domain with G␤ 5 involved many of the interactions observed in G␤ 1 ␥. These interactions included a two-stranded parallel coiled-coil involving N-terminal portions of G␤ 5 and G␥/GGL domain and hydrophobic interactions between more C-terminal regions of G␥/GGL domain and the ␤-propeller structure encoded by WD-40 repeats of the G␤ (10,(31)(32)(33).

Design and Expression of Putative G␥-selective G␤
Analysis of the coding sequences of G␤ 5 of mammalian (2,19), amphibian (20), insect (GenBank TM accession number AAF46336), or roundworm (21-23) origin in comparison to mammalian G␤ subunits 1-4 reveals four conserved residues in the putative coiled-coil region of the G␤ 5 s that are G␤ 5specific (Fig. 1). These G␤ 5 -specific residues occupy the a, d, and g positions of the G␤ 5 heptad repeat (Fig. 1), positions that may determine the specificity of coiled-coil dimerization (reviewed in Ref. 34) and that might contribute to specific interaction of G␤ 5 with RGS protein GGL domains (9). In order to create a G␥-preferring G␤ 5 mutant, these four residues in HA epitope-tagged mammalian G␤ 5 were mutated to their counterparts in G␤ 1 to produce mutant G␤ 5 -QAAC (G␤ 5 -K25Q/ E29A/K32A/L33C).
Modeling of a G␤ 5 -GGL domain heterodimer by Snow et al. (10) suggested that G␤ 5 ␤-propeller residues comprising a hydrophobic pocket able to accommodate a conserved Trp residue in the GGL domain (corresponding to Phe64 of G␥ 1 ) might also contribute to the selectivity of G␤ 5 -RGS protein assembly. Based on the model by Snow et al. (10), four ␤-propeller residues forming the hydrophobic pocket in HA-tagged G␤ 5 were mutated to their more bulky G␤ 1 counterparts. The aim of this mutagenesis was to reduce the size of the putative G␤ 5 ␤-propeller hydrophobic pocket to exclude the conserved Trp residue in the GGL domain while still accommodating the Phe residue present in G␥ corresponding to residue 64 of G␥ 1 (10). These alterations produced mutant G␤ 5 -YMIN (G␤ 5 -F93Y/T338M/V351I/A353N).
Both G␤ 5 mutants could be expressed to levels comparable with wild-type G␤ 5 when transiently transfected into HEK-293 cells (Fig. 2). The expression of the G␤ 5 -QAAC mutant, like that of wild-type G␤ 5 , was greater in cells co-transfected with RGS7 or G␥ 2 than in cells co-transfected with vector only (Fig.  2A, lanes 2-5, cf. lanes 6 and 7). The ability of G␥ or GGL domain-containing RGS protein co-expression to stabilize G␤ 5 expression has been observed previously (15,35) in several systems. Unlike the wild-type G␤ 5 , the expression of the G␤ 5 - FIG. 1. Identification of G␤ 5 -specific residues in the putative coiled-coil domain of G␤ 5 . An alignment of the coding sequences of human, salamander (Sal), D. melanogaster (Dm), and C. elegans (Ce) G␤ 5 with mammalian G␤ subunits 1-4 was performed as described under "Experimental Procedures." The portion of this analysis that includes the N-terminal region coiled-coil region of the G␤ subunits is shown. Amino acids that were identical among the 4 G␤ 5 sequences, but not found in the corresponding positions of mammalian G␤ 1-4 and belonging to a different amino acid homology group from the corresponding positions of G␤ 1-4 , were identified and considered to be conserved, G␤ 5 -specific residues (yellow highlights). The position of the residues of G␤ 1 in the coiled-coil heptad repeats, taken from the crystal structures of G␤ 1 complexes (31)(32)(33), are indicated as a-g above the G␤ 1 sequence. YMIN mutant was greater in cells co-transfected with G␥ 2 than in cells co-transfected with RGS7 (Fig. 2B, lanes 6 and 7, cf.  lanes 2-5). Nevertheless, the expression of the G␤ 5 -YMIN mutant was greater in cells co-transfected with RGS7 than with vector only (Fig. 2B, cf. lanes 5 and 7). Also wild-type G␤ 5 but not G␤ 5 -YMIN mutant co-transfection enhanced the expression of RGS7 (Fig. 2B, cf. lanes 2 and 7). As expected the C-terminally directed G␤ 5 antibody SGS failed to recognize the G␤ 5 -YMIN mutant containing two mutations within the SGS epitope (4) (Fig. 2B, lane 6) but reacted with the G␤ 5 -QAAC mutant as well as with the wild type (data not shown).
Interactions of Putative G␥-selective G␤ 5 Mutants with RGS7, G␥, and G␥-dependent Effectors-The ability of the putative G␥-selective G␤ 5 mutants to bind RGS7 and G␥ subunits was tested in co-immunoprecipitation assays. In transiently transfected HEK-293 cells, immunoprecipitation of the G␤ 5 -QAAC mutant was able to co-immunoprecipitate RGS7 from cell lysates as well as wild-type G␤ 5 (Fig. 3A, cf. lanes 2 and 3). In contrast only trace RGS7 was evident in immunoprecipitates of the G␤ 5 -YMIN mutant even though the mutant G␤ 5 itself was precipitated and wild-type G␤ 5 was able to fully co-precipitate RGS7 in the same experiment (Fig. 3B, cf. lanes  2 and 3). However, because the expression of RGS7 when co-transfected with the G␤ 5 -YMIN mutant is poor (Fig. 3B,  lane 3 of lysates), the absence of RGS7 in the G␤ 5 -YMIN mutant immunoprecipitates cannot be interpreted.
The ability of the putative G␥-selective G␤ 5 mutants to bind G␥ subunits was separately tested in co-immunoprecipitation assays from lysates of transiently transfected HEK-293 cells and compared with wild-type G␤ 5 and G␤ 1 (Fig. 4). Analysis of the lysates demonstrated the co-expression of all transfected G␤ constructs with G␥ 2 (Fig. 4A). Neither the wild-type G␤ 5 nor the G␤ 5 -QAAC mutant was able to co-immunoprecipitate G␥ 2 from cell lysates, however (Fig. 4B, lanes 2 and 4), a finding consistent with the known instability of G␤ 5 -G␥ 2 binding in detergent solution (10,16). Unlike wild-type G␤ 5 and G␤ 5 -QAAC, however, both the G␤ 5 -YMIN mutant and G␤ 1 clearly co-immunoprecipitated with G␥ 2 under the same conditions (Fig. 4B, lanes 3 and 5).
The ability of the putative G␥-selective G␤ 5 mutants to activate inositide-specific PLC-␤ in co-transfected HEK-293 cells in a G␥-dependent fashion was also tested. Wild-type G␤ 5 activated PLC-␤ 2 in a G␥-dependent fashion in COS-7 cells as described previously (2,4). Both mutants G␤ 5 -QAAC and G␤ 5 -YMIN activated PLC-␤ 2 in a G␥-dependent fashion indistinguishable from the wild-type G␤ 5 (data not shown).
RGS7 Requirement for Nuclear Expression of G␤ 5 in HEK-293 Cells-Because neither endogenous G␤ 5 (36) nor GGL domain-containing RGS proteins such as RGS7 is expressed in HEK-293 cells (Fig. 5, A and B, lane 1), they offer a useful model system to study the requirements for the expression and subcellular localization of these subunits upon transfection. When transfected alone, wild-type G␤ 5 can be detected in the whole cell lysate but not in the 293 cell nuclear extract (Fig. 5A,  lane 2). Co-transfection of G␤ 5 with either RGS7 or G␥ 2 greatly enhances the G␤ 5 signal in the cell lysate (Fig. 5A, lanes 3 and  4, cf. lane 2), presumably due to subunit stabilization by heterodimer formation (15,35), but only RGS7 co-transfection supported expression of G␤ 5 in the nuclear extract (Fig. 5A,  lane 3).
The expression and subcellular localization of HA epitopetagged G␤ 5 , G␤ 5 -QAAC, and G␤ 5 -YMIN were also studied in transfected HEK-293 cells in similar experiments (Fig. 5B). As expected HA epitope-tagged wild-type G␤ 5 was expressed in nuclear extracts upon co-transfection with RGS7 (Fig. 5B, lane 2) but not with G␥ 2 (data not shown). Like wild-type G␤ 5 , mutant G␤ 5 -QAAC also demonstrated RGS7 dependence for nuclear localization (Fig. 5B, cf. lanes 3 and 4). In contrast the G␥-selective mutant G␤ 5 -YMIN failed to appear in the nuclear extract with either RGS7 or G␥ 2 co-transfection (Fig. 5B, lanes  5 and 6). The expression of RGS7 in G␤ 5 -YMIN co-transfected cells could be demonstrated in whole cell lysates with immunoblots employing the polyclonal antibody 7RC-1 such as that shown in Fig. 5B (anti-RGS7 panel). Nevertheless, the expression of RGS7 in lysates of G␤ 5 -YMIN co-transfected cells was much lower than in wild-type G␤ 5 and G␤ 5 -QAAC co-transfected cells (Fig. 5B, lane 6, cf. lanes 2 and 4). Indeed, blots of lysates from RGS7 and G␤ 5  Cell lysates prepared 2 days after transient transfection were incubated with anti-HA monoclonal antibody, and after precipitation, the starting lysates (A) and washed immunoprecipitates (IP) (B) were analyzed in parallel for epitope-tagged G␤ (HA antibody, upper panels) and G␥ 2 (lower panels) immunoreactivity by immunoblotting as described under "Experimental Procedures." The relative mobility of the specific immunoreactive bands (in kDa) is indicated to the right of each panel as is the mobility of the immunoglobulin G heavy chain (HC) in B. Cells were transfected with either vector alone (vec) or with cDNAs encoding G␥ 2 , wild-type (wt), mutant HA-G␤ 5 , or wild-type HA-G␤ 1 as indicated below the lanes. The G␤ 5 mutants employed were G␤ 5 -QAAC (G␤ 5 -K25Q/E29A/K32A/L33C) and G␤ 5 -YMIN (G␤ 5 -F93Y/T338M/V351I/A353N). A representative result is shown in an experiment repeated two more times with similar findings.
naive and transfected PC12 cells demonstrated G␤ 5 in the cell nuclei and cytosol (17). The subcellular localization of the G␤ 5 -QAAC and G␤ 5 -YMIN mutants in transfected PC12 cells was therefore examined by antibody staining and confocal microscopy and compared with wild-type G␤ 5 (Fig. 6). In these experiments no GGL domain containing RGS proteins or G␥ subunits were co-transfected in order to allow the transfected G␤ 5 constructs to "choose" binding partners from the endogenous pool of RGS proteins and G␥ subunits present in PC12 cells (17,37). Antibodies to the G␤ 5 N terminus and the HA epitope tag both demonstrated the distribution of transfected wild-type G␤ 5 throughout the PC12 cell including the cell nucleus (Fig. 6) as described previously (17). The subcellular localization of the G␤ 5 -QAAC mutant closely resembled that of wild-type G␤ 5 (Fig. 6). In contrast the G␤ 5 -YMIN mutant was excluded from the PC12 cell nucleus when probed by either antibodies to the G␤ 5 N terminus or to the HA epitope tag. The G␤ 5 -YMIN mutant-transfected cells were still capable of G␤ 5 nuclear localization, because when non-epitope-tagged G␤ 5 was co-transfected with the G␤ 5 -YMIN mutant, nuclear localization of non-tagged G␤ 5 but not the HA-tagged mutant was evident in the same cells (Fig. 6).

DISCUSSION
The unique ability of G␤ 5 to interact selectively with G␣-and G␤␥-regulated effectors as a G␤ 5 -G␥ complex and with RGS proteins 6, 7, 9, and 11 as a G␤ 5 -RGS heterodimer may be critical to the role of G␤ 5 in signal transduction. New approaches are needed to assess the biological importance of the G␤ 5 -G␥ interactions so well documented in vitro since detergent-based purifications might miss or underestimate the presence of native G␤ 5 -G␥ complexes. Besides the mutational strategy employed here, other potential analytical tools such as conformation-dependent antibodies or selective proteolysis specific for the G␥-bound conformation of G␤ 5 would be valuable in this regard. Although well documented purifications from defined subcellular fractions of adult tissues have demonstrated only the presence of G␤ 5 -RGS complexes and not G␤ 5 -G␥ complexes (15), the potential importance of G␤ 5 -G␥ complex formation during ontogeny or in highly localized brain regions remains to be addressed.
Several observations in this study suggest G␤ 5 and its mutants are capable of a wider range of interactions with G␥ 2 and/or RGS7 than the most stringent assays, such as co-immunoprecipitation, would seem to suggest. Both wild-type G␤ 5 and the G␤ 5 -QAAC mutant stimulated PLC-␤ in a G␥ 2 -dependent fashion, although neither co-immunoprecipitated G␥ 2 . Cotransfection of RGS7 was observed to enhance the expression of the G␤ 5 -YMIN mutant even though the pair could not be immunoprecipitated. Several laboratories (15,35) have documented the ability of G␤ 5 -RGS and G␤ 5 -G␥ protein interaction to mutually stabilize the heterodimeric subunits. Since the G␤ 5 -YMIN mutant was designed to impair interaction between a hydrophobic pocket on the side of the G␤ 5 ␤-propeller and a critical Trp residue in the RGS protein GGL domain (10), the ability of RGS7 to partially stabilize the G␤ 5 -YMIN mutant may reflect an RGS7-G␤ 5 interaction outside this vicinity. The potential for such an interaction was recently demonstrated in FIG. 5. Expression and nuclear targeting of wild-type and mutant G␤ subunits in transfected HEK-293 cells. Cell harvested 2 days after transient transfection with either wild-type G␤ 5 (A) or HA epitope-tagged G␤ 5 and G␤ 5 mutants (B) were processed for either nuclear isolation and extraction (upper panels) or whole cell lysates (lower panels) as described under "Experimental Procedures." The relative mobility of the specific immunoreactive bands (in kDa), including that of TBP, is indicated to the right of each panel. The antibody used for immunodetection is indicated to the left of each panel. Cells were transfected with either vector alone (vec) or with cDNAs encoding RGS7, G␥ 2 , wild-type untagged, or HA epitope-tagged G␤ 5 or mutant HA-G␤ 5 , as indicated below the lanes. A representative result is shown in an experiment performed three times with similar findings.
FIG. 6. Confocal dual immunofluorescence analysis of PC-12 cells transiently transfected with HA epitope-tagged and/or wild-type G␤ 5 . Two days following transfection with the cDNAs indicated above the columns, cells were analyzed by laser confocal microscopy after dual staining with affinity-purified ATDG anti-G␤ 5 antibody (green signal) and anti-HA antibody (red signal) as described under "Experimental Procedures." Images of the cells illuminated by transmitted light only and the corresponding immunofluorescence signal (monitored singly or in combination (merge)) are indicated. Yelloworange signal indicates co-localization of probes. Results of one experiment are shown as representative of four independent experiments with similar findings.
C. elegans between an isolated N-terminal fragment of the RGS7 homolog EGL-10 and the G␤ 5 homolog GPB-2 (38). The finding that only the G␤ 5 -YMIN mutant and G␤ 1 were able to co-immunoprecipitate G␥ 2 from detergent lysates complements an experiment performed by Snow et al. (10) in which the strength of G␤ 5 -G␥ 2 interaction in detergent solution was greatly enhanced by replacing the conserved Phe-61 in G␥ 2 with a corresponding Trp residue characteristic of GGL domains, and further supports their model of the G␤ 5 -GGL/G␥ interface.
In previous studies of G␤ 5 nuclear localization the possible requirement for G␤ 5 -G␥ interaction for nuclear targeting was not fully assessed (17). Nuclear localization of G protein heterotrimers including apparent G␤␥ complexes has been reported in rat liver (39) and thrombin-and phorbol ester-treated Swiss 3T3 cells (40). In the previous study of G␤ 5 , a mutant GFP-G␤ 5 fusion protein that failed to undergo nuclear localization was likely deficient in both G␥ and GGL domain interaction due to double proline insertions in the G␤ 5 ␣-helical Nterminal region (17). This approach thus provided little insight into the heterodimerization requirements for G␤ 5 nuclear targeting.
The studies described herein strongly suggest G␥ interaction is insufficient for G␤ 5 nuclear targeting and that heterodimerization with a GGL domain-containing RGS protein such as RGS6 or -7 is critical. In HEK-293 cells, which do not express endogenous G␤ 5 or RGS7, the nuclear targeting of wild-type G␤ 5 (or the RGS7-interacting G␤ 5 -QAAC mutant) could only be reconstituted by RGS7 and not G␥ 2 co-transfection. Next, because the G␤ 5 -YMIN mutant failed to localize to the nucleus in either 293 or PC12 cells, despite its tight binding to G␥ demonstrated in vitro, the key molecular determinants of nuclear targeting appear to be intrinsic to the RGS protein, not to the G␤ 5 or G␥ subunits. However, the approach utilized here cannot exclude that mutation of the 4 residues in the G␤ 5 -YMIN mutant abrogated a nuclear localization signal (NLS), but this is unlikely for two reasons. No recognizable NLS motifs could be discerned in the native G␤ 5 sequence (41). Furthermore, when tested in HEK-293 cells neither untagged or HA epitopetagged wild-type G␤ 5 demonstrated nuclear expression when co-transfected with G␥ 2 making the presence of an occult NLS on G␤ 5 unlikely.
Demonstration of a requirement for RGS protein binding is consistent with several reports demonstrating the nuclear targeting of RGS2 (42)(43)(44), RGS10 (42,45), splice variants of RGS3 (46), and RGS12 (47). Mutation of a key serine residue in a conserved 14-3-3-binding site in the core RGS domain of RGS7 (Ser 434 3 Ala), a site present in many RGS proteins including RGS2 and RGS3 (48), made no difference in the steady-state distribution of RGS7 in resting PC12 cells (data not shown). This would seem to argue against a major role for 14-3-3 binding in the regulation of nucleocytoplasmic distribution of G␤ 5 -RGS complexes in resting cells. Whether phosphorylation or other post-translational modification of G␤ 5 -RGS complexes might govern their nuclear localization under stimulated conditions, and how potential nucleocytoplasmic shuttling of G␤ 5 -RGS heterodimers might mediate information transfer in the cell in response to extracellular signals remain two central questions for future study.