Gγ13 Interacts with PDZ Domain-containing Proteins*

The G protein γ13 subunit (Gγ13) is expressed in taste and retinal and neuronal tissues and plays a key role in taste transduction. We identified PSD95, Veli-2, and other PDZ domain-containing proteins as binding partners for Gγ13 by yeast two-hybrid and pull-down assays. In two-hybrid assays, Gγ13 interacted specifically with the third PDZ domain of PSD95, the sole PDZ domain of Veli-2, and the third PDZ domain of SAP97, a PSD95-related protein. Gγ13 did not interact with the other PDZ domains of PSD95. Coexpression of Gγ13 with its Gβ1 partner did not interfere with these two-hybrid interactions. The physical interaction of Gγ13 with PSD95 in the cellular milieu was confirmed in pull-down assays following heterologous expression in HEK293 cells. The interaction of Gγ13 with the PDZ domain of PSD95 was via the C-terminal CAAX tail of Gγ13 (where AA indicates the aliphatic amino acid); alanine substitution of the CTAL sequence at the C terminus of Gγ13 abolished its interactions with PSD95 in two-hybrid and pull-down assays. Veli-2 and SAP97 were identified in taste tissue and in Gγ13-expressing taste cells. Coimmunoprecipitation of Gγ13 and PSD95 from brain and of Gγ13 and SAP97 from taste tissue indicates that Gγ13 interacts with these proteins endogenously. This is the first demonstration that PDZ domain proteins interact with heterotrimeric G proteins via the CAAX tail of Gγ subunits. The interaction of Gγ13 with PDZ domain-containing proteins may provide a means to target particular Gβγ subunits to specific subcellular locations and/or macromolecular complexes involved in signaling pathways.

A sophisticated and ordered protein network is essential to the proper functioning of cells. Precise assembly of individual components, through targeting and anchoring of proteins within designated subcellular compartments, ensures the integrity of these networks (1,2). Specific protein-protein interactions are important for accomplishing this complex task. For example, PSD95 (postsynaptic density protein 95, also called SAP90), a member of the MAGUK (membrane-associated guanylate kinase) protein family (3), helps to assemble a complex postsynaptic protein network via its interactions with several different proteins. Some of these interactions rely on three PDZ domains (named after PSD95, Disc-large, and ZO-1) located in the N-terminal half of PSD95. PDZ domains function as protein-protein interaction modules and consist of about 90 amino acids (4 -6).
PDZ domains typically bind to the extreme C terminus of a target protein in a sequence-specific manner. The PDZ domains of PSD95 recognize a canonical -X(S/T)XA motif (where X represents any amino acid and A represents an aliphatic amino acid). PDZ domains have been identified in proteins in bacteria, yeast, Drosophila, metazoans, and plants and comprise the most common protein module identified in the sequenced genome (4,6). In addition to their affinity for C-terminal motifs, PDZ domains can also bind to internal sequences that mimic free C termini. Finally, PDZ domains can form both homo-and heterooligomers, enabling them to participate widely in the formation of vast and complex cellular networks (4,6).
The PDZ domains containing Lin/MALS/Veli proteins have been implicated in protein trafficking and assembly (7)(8)(9)(10). In mammals this group includes members that are expressed ubiquitously (e.g. Mint-3) or selectively in neurons (e.g. MALS-1). MALS, along with CASK and Mint-1, forms a tightly associated heterotrimeric protein complex that is evolutionarily highly conserved (cf. the Lin-2-Lin-7-Lin-10 complex in Caenorhabditis elegans). Both CASK and Mint-1 also contain more than one PDZ domain, as well as other protein interaction modules, enabling the CASK-Mint-1-Veli complex to interact with an array of proteins during vesicle trafficking and protein targeting. The Lin-2-Lin-7-Lin-10 complex (the invertebrate equivalent of CASK-Mint-1-Veli) is required for the basolateral localization of LET-23, a receptor tyrosine kinase essential for vulval development in Caenorhabditis elegans (11,12). The homologous tripartite complex in mammals, coordinated with the kinesin motor protein KIF17, has been shown to transport the N-methyl-D-aspartic acid receptors along the microtubules in neuronal dendrites (10).
Heterotrimeric G proteins, consisting of G␣ and G␤␥ subunits, function as signal transducers for the seven transmembrane helix G protein-coupled receptors (GPCRs). 2 The dynamic cycle of association and dissociation of G␣ and G␤␥ and its role as a switch in GPCR-based signaling pathways have been well defined. Specific interactions between GPCRs and heterotrimeric G proteins are important for maintaining the specificity and fidelity of signal relay (13,14). This is achieved in part by the membrane targeting of heterotrimeric G proteins and their specific interactions with particular receptors.
Specific lipid modification of G␣ and G␥ plays an important role in targeting and anchoring these proteins to the cell membrane (15). The G␤ subunit, although not modified by lipids, is membrane-bound by virtue of its tight interaction with G␥. The prenylation of the G␥ is dictated by the conserved CAAX motif at its C terminus (15), where A and X represent an aliphatic amino acid and any amino acid, respectively. Isoprenoids are covalently attached to the cysteine residue of the CAAX motif, whereas G␥ is still in the cytoplasm. Subsequently, the last three amino acids are proteolytically cleaved, and the free C terminus is carboxymethylated. This carboxymethylation is believed to be a membrane-associated event, and the insertion of the attached lipids into the cellular membrane completes the anchoring of G␤␥ on the inner side of the cellular membrane (16,17).
The G protein ␥13 subunit (G␥13) has a restricted and distinct pattern of expression (18,19). G␥13 has been shown to play a role in taste signal transduction (18). In this study, we demonstrate interactions between G␥13 and PDZ domain-containing proteins. These interactions provide a novel means of targeting and anchoring heterotrimeric G proteins to the cellular membrane and of assigning them to macromolecular complexes of receptors and channels.

EXPERIMENTAL PROCEDURES
Plasmids-PCR amplification was used to prepare DNA from wild type and mutant G␥13. G␥13T65A contains a single nucleotide substitution leading to a threonine to alanine substitution at amino acid 65 of G␥13. G␥13 DNAs were subcloned into the two-hybrid vector pGBDU-1 (a gift of E. A. Craig and P. J. James of the University of Wisconsin, Madison (20)) to create plasmids pBD-G␥13 and pBD-G␥13T65A. Plasmid pBD-G␤1G␥13 was created by subcloning PCRamplified G␤1 and G␥13 DNAs into the vector pBridge (Clontech). pGAD424 plasmids containing various PDZ domains from PSD95 (pGAD424-PDZ1, pGAD424-PDZ2, pGAD424-PDZ3, pGAD424-PDZ12, pGAD424-PDZ23, and pGAD424-PDZ123;, see Table 1) were the kind gifts of Dr. M. Noda and have been described previously (21). The third PDZ domain of SAP97 was amplified by PCR and subcloned into pGADT7 (Clontech). All PCRs were carried out according to the standard protocol accompanying the enzymes Taq (PerkinElmer Life Sciences) or Pfu (Stratagene, La Jolla, CA) DNA polymerases. All PCR products were sequenced to ensure no inadvertent errors. All two-hybrid plasmids used in this study, along with their derivation, are listed in Table 1. For example, plasmids pAD-PSD95⌬38, pAD-Veli2⌬3, and pAD-Veli2 were recovered from two-hybrid screening in this study (see details under "Results"). pAD-Veli2-L27 was created by cutting pAD-Veli2 with XhoI followed by self-ligation to remove 382-bp DNA. Partial digestion of pAD-Veli2 with EcoRI/XhoI, which removes 302-bp DNA, followed by blunt-end self-ligation, was used to create plasmid pAD-Veli⌬121. Similarly, plasmids pAD-PSD95-PDZs and pAD-PSD95-SHGUK were constructed by ApaI digestion and religation and XmaI/ PvuII digestion and blunt end ligation of plasmid pAD-PSD95⌬38. Plasmid GW1-myc-PSD95 was a gift of Dr. Morgan Sheng and was described previously (22). Hemagglutinin-tagged G␥13 (both wild type and mutantT65A versions) was cloned into the pcDNA3.1 ϩ vector between the EcoRI (5Ј) and XhoI (3Ј) sites.
Yeast Two-hybrid Screening-A yeast two-hybrid cDNA library from mouse brain was obtained (Clontech), and the host yeast strains (PJ69-4a and PJ69-4␣) used for the two-hybrid assays were as described previously (20). The yeast two-hybrid system TRAFO protocol (23) was used to maximize transformation efficiency; transformants were selected on solid synthetic medium lacking uracil, leucine, histidine, and adenine. The candidate genes recovered after two-hybrid selection were isolated, sequenced, and retested for interaction with G␥13 by an independent two-hybrid assay using diploid cells; all independent two-hybrid interaction assays were carried out in PJ69-4a/PJ69-4␣ diploid cells using the strategy described (24). ␤-Galactosidase activity of the lacZ reporter gene was measured using o-nitrophenyl ␤-D-galactopyranoside as a substrate according to the MatchMaker II manual (Yeast Protocols Handbook, Clontech).
Antibodies-The anti-HA monoclonal antibody was obtained from Sigma. The anti-Myc monoclonal antibody (9E10) was produced at the Hybridoma Center, Mount Sinai School of Medicine. The polyclonal antibody directed against PSD95 was obtained from Upstate Biotechnology, Inc. (Lake Placid, NY). Rabbit polyclonal antiserum against G␥13 was as described previously (18). Monoclonal SAP97 antibody was from Abcam (Cambridge, MA).
Coimmunoprecipitation and Western Blots-HEK293 cells were transfected by Effectene (Qiagen, Valencia, CA) with wild type or mutant forms of G␥13 alone or together with PSD95. 48 h post-transfection, the cells were rinsed with phosphate-buffered saline, placed at 4°C, and then lysed for 30 min in 200 ml of Lysis Buffer (50 mM Tris-HCl, pH 7.4, 1 mM EDTA, 1% Triton X-100 and 1ϫ complete protease inhibitor mixture (Roche Applied Science)). Unless otherwise indicated, all subsequent steps were performed at 4°C. Cultured cell lysates were clarified by microcentrifugation for 15 min. Protein concentrations were determined by the Bradford assay (Bio-Rad). Equal amounts of cultured cell lysate total protein were used in immunoprecipitation assays. Lingual and brain tissues from mice were homogenized in Lysis Buffer; homogenates were centrifuged at 13,000 ϫ g for 10 min to remove tissue debris, and the supernatants were used for immunoprecipitation pull-down assays. Antibodies to be used in pull-down assays were mixed with protein G-Sepharose (Amersham Biosciences), incubated at room temperature for 1 h to enable conjugates to form, and then washed three times with phosphate-buffered saline to remove excess antibody. Antibody-protein G-Sepharose conjugates were mixed with cell lysates, shaken for 2 h at 4°C, and then washed two times with RIPA buffer (20 mM Tris-HCl, pH 7.4, 137 mM NaCl, 10% glycerol, 1% Triton X-100, 0.5% deoxycholate, 0.1% SDS), two times with 0.1 M Tris-HCl containing 0.5 M LiCl, and then two times in 50 mM Tris-HCl, pH 7.5. Samples were incubated in SDS gel loading buffer for 3 min at 95°C, then separated by SDS-PAGE, and visualized by Western blotting. Immunocytochemistry-Immunocytochemistry using transgenic mice expressing green fluorescent protein (GFP) from the ␣-gustducin promoter was as described previously (18). Frozen sections (10 m thick and fixed in 4% paraformaldehyde and cryoprotected in 20% sucrose) of lingual tissue from GFP mice were incubated in Blocking Buffer (3% bovine serum albumin, 0.3% Triton X-100, 2% goat serum, and 0.1% sodium azide in phosphate-buffered saline) for 1 h at room temperature. The sections were then incubated overnight at 4°C with purified primary antibodies as follows: either polyclonal rabbit anti-G␥13 (18), rabbit anti-SAP97 (Affinity Bioreagent, Golden, CO), or goat anti-PSD95 (Santa Cruz Biotechnology, Santa Cruz, CA). The secondary antibodies were Cy3-conjugated goat anti-rabbit Ig for G␥13 and SAP97 and Cy3conjugated donkey anti-goat Ig for PSD95 (Molecular Probes, Eugene, OR). Immunofluorescence images were obtained with an Olympus AX70 fluorescence microscope in the presence of Vector antifade mounting medium (Molecular Probes, Eugene, OR). Incubation of sections with primary or secondary antibodies alone produced no observable immunofluorescence (data not shown).

RESULTS
Interaction of G␥13 with PSD95 and Veli-2-The yeast two-hybrid system is a valuable tool for studying protein-protein interactions and for identifying interacting partners of known proteins (25), and it has been used to characterize the interaction between the G protein ␤ and ␥ subunits (26). In a two-hybrid assay, the G␥13 subunit can interact weakly with all five known G␤ subunits and shows no distinct preference among them (27). 3 Because the signaling specificity of the G␤␥ dimer is largely dictated by G␥, we searched for proteins that could interact directly with G␥13. For this purpose we subcloned G␥13 into vector pGBDU-1 and used it as "bait" to screen a mouse brain cDNA library in the GAL4-based yeast two-hybrid system, where positive interactions of G␥13 with unknown proteins would activate several reporter genes (HIS3, ADE2, and lacZ) of the host yeast cells. The activation of the HIS3 or ADE2 genes was monitored by yeast cell growth on media lacking the amino acids histidine or adenine. Activation of the lacZ reporter rendered yeast colonies blue in the presence of 5-bromo-4-chloro-3-indolyl-␤-D-galactopyranoside in the growth media and was quantitatively measured using o-nitrophenyl ␤-D-galactopyranoside as a substrate (Matchmaker II; Clontech). After screening approximately 10 million independent colonies (ϳ3 times the complexity of the library), we obtained four clones that activated the HIS3, ADE2, and lacZ reporter genes. Two of these were identical clones encoding a part of the PSD95 protein starting at amino acid 39 (pAD-PSD95⌬38). The other two clones were derivatives of the Veli-2 gene. One of these clones encoded the full-length Veli-2 protein (pAD-Veli2), although the other encoded a truncated form lacking the first three amino acids (pAD-Veli2⌬3).
Both PSD95 and Veli-2 were retested in an independent two-hybrid assay using diploid cells to confirm their interaction with G␥13 (Fig.  1A). PSD95 and Veli-2 failed to activate the reporter genes when combined with numerous unrelated genes in control two-hybrid assays (data not shown). Therefore, the interactions of G␥13 with PSD95 and Veli-2 appeared to be specific. The strength of interaction of G␥13 with PSD95 and Veli-2 was examined quantitatively using ␤-galactosidase activity in a yeast two-hybrid reporter assay. By this measure, PSD95 had higher affinity for G␥13 than did Veli-2 (Fig. 1B).
In these studies we relied on the ability of G␥13 to interact with other proteins in the absence of G␤. In vivo, however, G␤␥ subunits comprise a nondissociating unit. Some G␥13 domains exposed during our screening procedure might ordinarily be inaccessible within the context of the G␤␥ dimer. To investigate this, we coexpressed G␥13 with G␤1 in our twohybrid assay. Coexpression of G␥13 with G␤1 did not prevent PSD95 (Fig.  1C) or Veli-2 (data not shown) from activating the HIS3, ADE2, or lacZ reporter genes, suggesting that the G␥13 domain(s) responsible for interaction with PSD95 or Veli-2 is not blocked by the presence of G␤1. 3 Z. Li and R. F. Margolskee, unpublished observations. FIGURE 1. Interaction of the heterotrimeric G protein ␥13 subunit (G␥13) with PSD95 and Veli-2 proteins in the yeast two-hybrid system. A, activation of the HIS3 and ADE2 reporter genes allows the growth of yeast cells on plates with synthetic medium lacking the amino acids histidine and adenine. Yeast two-hybrid vectors pAS2-1 and pACT2 were used as the negative control (Ϫ), whereas the same pair of plasmids harboring p53 and large T-antigen of SV40, respectively (Matchmaker II, Clontech), was used as the positive control (ϩ). B, quantitation of ␤-galactosidase activity in yeast cells resulting from the interaction of G␥13 with PSD95 and Veli-2 proteins. The ␤-galactosidase activity from the p53/T-antigen interaction (p53/T-ag), the positive control, was defined as 100. C, interaction of the G␤1␥13 with PSD95 in the two-hybrid assay allows the yeast cells to grow on plates lacking histidine and adenine. A pACT2 vector with G␤1␥13 was used as a negative control.
One PDZ Domain of PSD95 Interacts with the C Terminus of G␥13-A common feature of PSD95 and Veli-2 proteins is the presence of one or more PDZ domains. To determine whether these PDZ domains underlie the interaction of PSD95 and Veli-2 with G␥13, several deletion mutants of PSD95 and Veli-2 were generated and tested in a yeast twohybrid assay (Fig. 2). Splitting the Veli-2 gene within the PDZ motif yielded the following two fragments: the N-terminal amino acids 1-121, containing the L27 domain (28), and the C-terminal amino acids 122-207. Neither of these fragments bound G␥13 in a two-hybrid assay ( Fig.  2A), suggesting that the PDZ domain of Veli-2 must be intact for this protein to interact with G␥13. For PSD95, a deletion mutant (PDZ123, amino acids 39 -435) containing all three PDZ domains, but lacking the C terminus, bound G␥13 (Fig. 2B), indicating a role for one or more of PDZ domains of PSD95 in the interaction with G␥13.
To determine whether a particular PDZ domain of PSD95 was responsible for the interaction with G␥13, individual PDZ domains of rat PSD95 The ability of deletion constructs to interact with G␥13 in a yeast two-hybrid growth assay is indicated (ϩ or Ϫ). Interactions with G␥13 were detected only with constructs containing the third PDZ domain of PSD95. C, a single mutation (G␥13T65A) at the C terminus of G␥13 abolishes its interaction with PSD95. (PDZ1, PDZ2, and PDZ3; Fig. 2B) (21), their combinations (PDZ123, Fig.  2B; PDZ12 and PDZ23; data not shown), and the SH3-GU kinase-like domains were subcloned and tested for their interaction with G␥13. Plasmids containing these domains were expressed in PJ69-4a/PJ69-4a diploid cells (24) and examined for activation of the two-hybrid reporter genes HIS3 and ADE2. As shown in Fig. 2B, only the third PDZ domain of PSD95 interacted with G␥13; the first two PDZ domains alone or in combination and the SH3-GU kinase domains failed to do so. In this regard it should be noted that the third PDZ domain of PSD95 is the most similar to the PDZ domain of Veli-2 ( Fig. 3; see "Discussion").
The PDZ domains of PSD95 and Veli recognize the canonical X(S/ T)XA motif at the C terminus of proteins (where X represents any amino acid; S/T is serine or threonine; A is any aliphatic amino acid). A serine or threonine at the Ϫ3 position is critical for this interaction. Most interestingly, the C terminus of G␥13 is CTAL, which is predicted to be a target of the PDZ domains of PSD95 and Veli-2. If this CTAL tail is indeed the target of PDZ domain binding, then mutation of the threonine at the Ϫ3 position should disrupt this interaction. As predicted, a  11 and 12). HA-tagged wild type or mutant G␥13 proteins were immunoprecipitated (IP) by an anti-HA antibody conjugated to Sepharose beads. PSD95-myc was detected in immunoblots by an anti-Myc antibody. Wild type and mutant G␥13-HA were detected in immunoblots by an anti-HA antibody. The immunoblotted (IB) G␥13 and PSD95 bands are at 7 and 95 kDa, respectively. A and B, immunoblots from anti-HA immunoprecipitates (IP: HA). C and D, immunoblots from cell lysates. Representative immunoblots from a single experiment that was replicated three times with equivalent results are shown. Note that PSD95 was pulled down by wild type G␥13 but not by mutant G␥13T65A (A, lanes 9 and 10 versus lanes 11 and 12). The right margin indicates lysate or anti-HA immunoprecipitate (IP: HA) and the antibody used in the immunoblot (IB: HA or IB: myc).

FIGURE 5. Expression of PDZ-containing proteins in taste tissue.
A, immunoreactivity to PSD95 and SAP97 was detected in taste cells by indirect immunofluorescence. Taste bud containing sections of the circumvallate papillae were from a transgenic mouse (GUS-GFP) expressing GFP in the gustducin-positive taste cells (a subset of the G␥13expressing taste cells). B, RT-PCR amplification identified SAP97, but not PSD95, from taste tissue. In positive controls PSD95 and SAP97 were amplified from brain. In negative controls PSD95 and SAP97 were not amplified from the no DNA controls. threonine to alanine substitution mutation of G␥13 (G␥13T65A) eliminated the two-hybrid interaction of G␥13 with PSD95 (Fig. 2C) and Veli-2 (data not shown).
Heterologously Expressed G␥13 and PSD95 Interact Physically-The above studies indicate that the C-terminal CAAX motif of G␥13 interacts with specific PDZ domains within the setting of a yeast two-hybrid assay. To determine whether PSD95 and G␥13 could interact in a less artificial cellular setting, we used a pull-down assay with heterologously expressed tagged proteins. HA-tagged wild type or mutant G␥13, along with Myc-tagged PSD95, were coexpressed in transfected HEK293 cells; following lysis, proteins were immunoprecipitated by an anti-HA antibody and probed on Western blots (Fig. 4). Probing the blots with an anti-Myc antibody showed that HA-tagged wild type G␥13 robustly pulled down PSD95 (Fig. 4A, lanes 9 and 10). Carrying out this same experiment with mutant HA-tagged G␥13 (G␥13T65A) lacking the CAAX motif, we find that PSD95 is not pulled down at all (Fig. 4A, lanes 11 and 12). Multiple controls were done to ensure that the pull-down interaction of G␥13 and PSD95 was specific. No pull-down of PSD95 was seen with the following: (a) untransfected cell lysates (Fig. 4A, lanes 1 and 2); (b) lysates from cells transfected with wild type or mutant G␥13 and vector (Fig. 4A, lanes 3-6); (c) lysates from cells transfected with PSD95-myc and vector (Fig. 4A, lanes 7 and  8). Immunoblots of lysates confirmed reproducible expression and detection of HA-tagged G␥13 and Myc-tagged PSD95 (Fig 4, B-D). In sum, in the cellular environment of transfected HEK293 cells there is a physical association of the C-terminal CAAX tail of G␥13 with the PDZ domain of PSD95.
Expression of PDZ Domain-containing Proteins in Taste Cells-Although first identified in taste receptor cells, G␥13 is also expressed in brain and retina (18,19) two tissues in which expression of PSD95 has been well documented (29,30). We have shown previously that in response to bitter compounds G␥13 activates phospholipase C␤2 to elevate inositol 1,4,5trisphosphate levels in taste receptor cells (18). Given the coexpression of PSD95 and G␥13 in brain and retina, we tested taste receptor cells for PSD95 expression. Taste bud-containing sections from transgenic mice expressing GFP from the gustducin promoter (18) were examined by indirect immunofluorescence with an anti-PSD95 antibody. PSD95 immunoreactivity was detected in taste cells (Fig. 5A, PSD95); however, amplification by RT-PCR failed to detect expression of PSD95 in taste tissue cDNA (Fig. 5B). To determine whether cross-reactivity with other PDZ domaincontaining proteins explained this result, we carried out additional immunostaining and RT-PCR experiments. SAP97, like PSD95, is a MAGUK protein. Furthermore, the epitope recognized by the anti-PSD95 antibody is very similar to SAP97, and this antibody detects a 97-kDa protein (the size of SAP97) on a Western blot. 4 When we used a more specific affinitypurified anti-SAP97 polyclonal antibody, we observed a pattern of immunostaining of mouse taste cells (Fig. 5A, SAP97) similar to that obtained with the anti-PSD95 antibody. In addition, expression of SAP97 in taste tissue cDNA was readily detected by PCR amplification (Fig. 5B).
SAP97 expression was greatest at the taste pore, where the highest levels of taste receptors are found (Fig. 6A). SAP97 is the only member of MAGUK family of proteins that is not exclusively expressed in neuronal cells. It has the highest similarity to PSD95 (Fig. 3), and perhaps it can assume the function of PSD95 in taste tissue or in non-neuronal cells.
To determine whether SAP97 could indeed interact with G␥13, we cloned the third PDZ domain of SAP97 into the two-hybrid vector pGADT7, and we tested its ability to interact with pBD-G␥13 in a twohybrid assay. As expected, the third PDZ domain of SAP97 did interact with G␥13 (data not shown). By using an antibody against Veli-2 we detected immunoreactivity in taste cells (Fig. 6B). In contrast to SAP97, taste-expressed Veli-2 immunoreactivity was more evenly placed throughout the taste bud.
Endogenously Expressed G␥13 and PDZ Domain Proteins Interact Physically-The observations that 1) heterologously expressed G␥13 interacts with heterologously expressed PDZ domain proteins in HEK293 cells (Fig. 4) and that 2) G␥13 is coexpressed in native tissues with PSD95 and SAP97 (Figs. 5 and 6) suggest that G␥13 may very well interact with these and other PDZ domain proteins in vivo. To determine directly if endogenously expressed G␥13 and PDZ domain proteins interact in native tissues, we carried out pull-down assays with lysates from mouse brain and taste tissue (Fig. 7). First, we confirmed the ability to detect PSD95 and G␥13 in brain extracts and SAP97 and G␥13 in taste tissue extracts on immunoblots (Fig. 7, A and B, right  panels). Next, we determined that an anti-G␥13-specific antibody efficiently pulled down PSD95 from brain lysates (Fig. 7A) and SAP97 from 4 Santa Cruz Biotechnology, personal communication. . Endogenously expressed G␥13 and PDZ proteins in brain and taste tissue interact in pull-down assays. A, left panel, G␥13 was immunoprecipitated (IP) from mouse brain lysates with preimmune serum (lane 1) or antibody to G␥13 (lanes 2 and 3), and then immunoprecipitates were resolved by SDS-PAGE and immunoblotted with antibodies directed against PSD95 (IB: PSD95) or G␥13 (IB: G␥13). Right panel, brain lysate was shown to contain G␥13 and PSD95 by immunoblotting (IB). B, left panel, G␥13 was immunoprecipitated from mouse taste tissue lysates with preimmune serum (lane 1)or antibody to G␥13 (lanes 2 and 3), and then immunoprecipitates were resolved by SDS-PAGE and immunoblotted with antibodies directed against SAP97 (IB: SAP97) or G␥13 (IB: G␥13). Right panel, taste tissue lysate was shown to contain G␥13 and SAP97 by immunoblotting.
taste tissue lysates (Fig. 7B). As a control for the specificity of these interactions, we tested preimmune serum for the ability to pull down PDZ domain proteins; neither PSD95 nor SAP97 was pulled down from brain or taste tissue by the preimmune serum (Fig. 7, A and B). Thus, endogenously expressed G␥13 interacts with endogenously expressed PSD95 and SAP97 in brain and taste tissues, respectively. These results suggest that these specific interactions may have functional importance in vivo (see "Discussion").

DISCUSSION
We have identified an interaction between the C terminus of G␥13 and three closely related PDZ domains present in PSD95, SAP97, and Veli-2 proteins. The C-terminal region of G␥13 contains the canonical binding motif for class I PDZ domains (4,6). The third PDZ domain of PSD95 interacted with G␥13, but the other two PDZ domains of PSD95 did not, demonstrating the specificity of this interaction. One common structural feature of the three PDZ domains that interacted with G␥13 is the short sequence between the ␤B and ␤C regions; we speculate that this may be necessary to accommodate the prenyl moiety of G␥13.
Binding of G␤ to G␥13 did not interfere with the ability of G␥13 to interact with Veli-2 or PSD95 in a two-hybrid assay. Likewise, we observed interactions in the presence of G␤ between heterologously expressed G␥13 or endogenous G␥13 and PSD95 or SAP97 in pulldown assays. This is consistent with the observation that the C-terminal sequences of G␥ subunits are neither required for nor part of their interaction with G␤ subunits (31). Of more than 13 G␥ subunits identified to date, only a few of them (G␥2, G␥5, G␥8olf, G␥12, and G␥13) have the class I type PDZ binding C-terminal sequence (-C(T/S)XX tail). The other G␥ subunits are unlikely to be recognized by the conventional PDZ domain because of the lack of Thr/Ser at the critical Ϫ3 position (they have a -C(V/A)XX tail). Either these G␥ subunits are not recognized by PDZ domain proteins or they may be recognized by proteins containing another type of PDZ domain.
G␥ subunits are post-translationally modified by the attachment of isoprenoid lipids to the invariant cysteine residue at the Ϫ4 position at their C terminus (15). Prenylation of the G␥ subunit, although not required for its dimerization with G␤, plays an important role in the interactions of G␤␥ with G␣ subunits, receptors, and effectors (32)(33)(34)(35)(36). It is generally believed that prenylation of G␥ is important for anchoring the G␤␥ subunit to the membrane, presumably by direct insertion of the isoprenoid into the lipid bilayer (35,(37)(38)(39). However, whether the lipid attachment is the only signal required for membrane targeting of G␤␥ remains unclear, because isoprenylated proteins are found in the cytoplasm as well as on cellular membranes.
The interaction of G␥13 with PDZ-containing proteins may provide a means to specifically and efficiently target certain G␤␥ subunits to particular subcellular locations. This interaction would provide a much more efficient signal than the C-terminal prenylation required by the kinetic membrane trapping/two-signal hypothesis models for the membrane association of lipid-modified proteins (17). The higher affinity of G␥13 for PSD95 over Veli-2 would facilitate unloading of G␥13 from Veli-2-containing transport vesicles. Although the timing of prenylation relative to binding of G␥ by PSD95/SAP97/Veli-2 remains unclear, we think it unlikely that prenylation of the CAAX motif of G␥13 would interfere with its binding to the PDZ domains. First, the binding of PSD95, SAP97, or Veli-2 to G␥13 could occur temporally before its prenylation. Second, the short ␤B to ␤C region of the PDZ domains of PSD95, SAP97, and Veli-2 can probably provide enough space for the lipid group of G␥ subunits. Third, post-translational modification occurs in both the two-hybrid system and in HEK293 cells, making it likely that at least some of the G␥13 is prenylated, and G␥13 expressed in these systems interacted with PSD95, SAP97, and Veli-2.
Ultimately, the binding site of G␥13 for PDZ domain-containing proteins, its CAAX motif, would be eliminated during maturation. It is generally the case that the last three amino acids of the CAAX motif are proteolytically cleaved, followed by carboxylation of the cysteine residue. Removal of the last three amino acids has been implicated as important for the localization of some CAAX motif proteins (40). Because the cleavage and carboxylation of the CAAX motif are membrane-associated events, they would probably occur after the delivery and anchoring of G␥13 to its subcellular location by PSD95, SAP97, and Veli-2, providing an additional means to regulate the interaction of these PDZ proteins with G␥13.
Our study provides the first evidence that proteins with PDZ domains can recognize the heterotrimeric G protein through the CAAX tail of G␥. The three PDZ-containing proteins identified in this study as interacting with G␥13 are known participants in protein trafficking and macromolecular signal complex assembly. Interaction of G␥ with PSD95, SAP97, or Veli-2 could efficiently and specifically target heterotrimeric G proteins to appropriate receptors and channels in neurons and taste cells. Because PSD95 is a member of the MAGUK protein family and there is broad expression of Veli proteins in both neuronal and nonneuronal cells, we propose that similar interactions may be widespread and functionally important to various signaling pathways. For example, the enhanced level of SAP97 at the taste pore could be responsible for the high local concentration of G␥13 with implications for taste receptor-mediated signaling.