Regulation of membrane and subunit interactions by N-myristoylation of a G protein alpha subunit in yeast.

Initiation of the mating process in yeast Saccharomyces cerevisiae requires the action of secreted pheromones and G protein-coupled receptors. As in other eukaryotes, the yeast G protein α subunit undergoes N-myristoylation (GPA1 gene product, Gpa1p). This modification appears to be essential for function, since a myristoylation site mutation exhibits the null phenotype in vivo (gpa1G2A). Here we examine how myristoylation affects Gpa1p activity in vitro. We show that the G2A mutant of Gpa1p, when fused with glutathione S-transferase, can still form a complex with the G protein βγ subunits. The complex is stabilized by GDP and is dissociated upon treatment with guanosine 5′-O-(thiotriphosphate). In addition, there is no apparent difference in the relative binding affinity of Gβγ for mutant and wild-type Gpa1p. Using sucrose density gradient fractionation of cell membranes, Gpa1p associates normally with the plasma membrane whereas Gpa1pG2A is mislocalized to a microsomal membrane fraction. A portion of Gβγ is also mislocalized in these cells, as it is in a gpa1Δ strain. In contrast, wild-type Gpa1p reaches the plasma membrane in cells that do not express Gβγ or cell surface receptors. These findings indicate that mislocalization of Gpa1pG2A is not caused by a redistribution of Gβγ, nor is it the result of any difference in Gβγ binding affinity. These data suggest that myristoylation is required for specific targeting of Gpa1p to the plasma membrane, where it is needed to interact with the receptor and to regulate the release of Gβγ.

The actions of many extracellular stimuli (hormones, neurotransmitters, odorants, light) are elicited by signaling cascades that involve a membrane-bound receptor, a regulatory G protein composed of three subunits (␣, ␤, ␥), and an effector that propagates the signal. Upon receptor activation, the G protein ␣ subunit releases GDP, binds GTP, and dissociates from the G ␤␥ subunit complex. The G protein remains in the active dissociated state until GTP is hydrolyzed. In yeast, most components of this signaling cascade were discovered genetically, through gene mutations that block pheromone signaling and mating. Disruption of genes encoding G ␤ (STE4) or G ␥ (STE18) result in a sterile phenotype. Loss of G ␣ (GPA1, also known as SCG1) leads to constitutive signaling and mating even in the absence of pheromone. Thus it appears that G ␤␥ activates the downstream signaling pathway, and the primary role of Gpa1p is to regulate the levels of free G ␤␥ in the cell (1)(2)(3).
In order to respond to extracellular signals, G proteins and receptors must associate with the plasma membrane of the cell. Since G proteins are not integral membrane proteins, however, it is not clear how they become attached to the lipid bilayer or how they are targeted specifically to the plasma membrane. One mechanism for controlling protein localization is posttranslational modification (4). Indeed, it has been shown that all G protein ␣ subunits undergo some form of fatty acylation. G s␣ , G i␣ , G q␣ , G 12␣ , G 13␣ , G o␣ , and G z␣ are palmitoylated (5-10); G o␣ , G i␣ , G z␣ , and Gpa1p are N-terminal myristoylated (10 -18); G t␣ is heterogeneously acylated (14,19,20).
In yeast, mutations in the N-myristoyltransferase (NMT1) gene (11,21,22), or mutations that replace the myristoylated Gly residue of Gpa1p (11), result in constitutive activation of the pheromone response pathway. Thus nonmyristoylated Gpa1p appears unable to bind G ␤␥ in vivo, even though it remains associated with the cell membrane (11). The ability of myristoylated or nonmyristoylated Gpa1p to bind G ␤␥ has never been documented, however. Moreover, while nonmyristoylated Gpa1p sediments with the membrane fraction, the type of membrane was never defined (11). Thus it is not known if myristoylation is necessary for high affinity binding to G ␤␥ , or for specific binding to the plasma membrane, or both. For example, Gpa1p could be recruited to the plasma membrane via myristoylation-dependent coupling to G ␤␥ . Conversely, Gpa1p might require myristoylation to reach the plasma membrane, and consequently is able to interact with other plasma membrane proteins such as G ␤␥ or the receptor.
Previous studies have revealed a possible role for G ␤␥ in targeting G␣ to the plasma membrane. Sternweis (23) has shown that binding of purified G o␣ and G i␣ to lipid vesicles requires co-reconstitution with G ␤␥ , suggesting that G ␤␥ can serve as a membrane anchor for the ␣ subunit in mammalian cells. However, other proteins must also be involved since expression of G ␣ in excess of G ␤␥ does not prevent membrane association (24), and disruption of the G protein ␤ and ␥ genes in yeast (STE4, STE18) does not lead to solubilization of Gpa1p (25). Again, the type of membrane was not defined in these experiments, so they do not reveal whether G ␤␥ can direct G ␣ to a specific cell membrane compartment.
Past studies have also revealed several functional changes associated with the myristoylation of G proteins. For example, myristoylation of G i␣ promotes binding to G ␤␥ , direct inhibition of adenylyl cyclase in vitro, as well as binding to cell membranes (16, 17, 26 -29). Myristoylated peptides corresponding to the N-terminal domain of G t␣ were found to inhibit binding to G ␤␥ . In these experiments, a random myristoylated peptide was also an effective inhibitor, suggesting that the lipid moiety can play an important role in G ␤␥ recognition (14,15). Less direct experiments (e.g. N-terminal deletions or proteolysis) have yielded similar conclusions (reviewed in Ref. 30).
More recently, Chabre and colleagues have proposed an alternative model for how G␣ myristoylation contributes to subunit and membrane interactions. They showed that tight association of purified subunits requires the presence of lipids or detergent, suggesting that acylation contributes indirectly to subunit interaction by restricting their mobility to the twodimensional plane of the membrane or micelle (31). These experiments were possible because of the unique ability of G t to be solubilized even in the absence of detergent. We reasoned that the G protein in yeast would also be useful in this regard, because of the unusual ability of Gpa1p to remain membrane associated even when myristoylation is blocked (11). Thus a direct role for myristoylation in subunit binding can be examined without the confounding effects of membrane dissociation. Moreover, strains that bear disruption mutations of G protein subunits are available, so a role for G ␤␥ expression in G ␣ trafficking can be tested directly.
In this report, we investigate the role of myristoylation in G protein subunit and membrane interactions. These experiments reveal that G ␤␥ binds to myristoylated and nonmyristoylated Gpa1p with similar affinity, but that myristoylation is required for Gpa1p to reach the plasma membrane where G ␤␥ is normally located. These findings reveal a role for myristoylation in G protein subcellular localization and highlight the importance of proper membrane localization for normal G protein function.
Plasmid Construction-Standard methods for the manipulation of DNA were used throughout (32). Enzymes and buffers were obtained from New England Biolabs. PCR reagents were from Perkin-Elmer. Sequencing reagents were purchased from U. S. Biochemical Corp. Expression plasmids used in this study were pRS316 (amp r , CEN/ARS, URA3), or pRS316-ADH (pRS316 with ADH1 promoter and terminator from pAD4M, described below) (33), or pRS316-GAL.
Replacement of the N-terminal Gly2 codon was achieved by PCR using mismatched primers (32) with pBSH/E-GPA1 or pBSH/E-GPA1myc as templates. The forward primers contain a BamHI site and encode Met-Gly-Cys-Thr (5Ј-GGG GAT CCC ATG GGG TGT ACA-3Ј), or Met-Ala-Cys-Thr (5Ј-GGG GAT CCC ATG GCG TGT ACA-3Ј). The reverse primer (5Ј-ATC AGA ACC ACC GGC AA-3Ј) is complementary to nucleotides 398 -414 of the GPA1 open reading frame. The PCR products were digested with BamHI and HindIII. pRS316-GPA1 and pRS316-GPA1 G2A (Gly 2 3 Ala substitution) were constructed by ligation of the BamHI-HindIII PCR products into the corresponding sites of pBB194 (from J. Gordon, Washington University). pBB194 consists of pRS316 and the full-length XbaI-EcoRI-EcoRI GPA1 product, including an engineered BamHI site at position Ϫ10 relative to the initiator ATG (prepared by S. Reed, Scripps Institute).
Tagged and untagged versions of GPA1 were also placed under the control of a heterologous constitutive promoter from ADH1. pRS316-ADH was constructed by moving the BamHI-BamHI-BamHI region containing the ADH1 promoter and terminator from pAD4M (from P. McCabe, Onyx Pharmaceuticals) (40,41) into a derivative of pRS316 in which the KpnI-SacI polylinker had been replaced with annealed complementary synthetic oligonucleotides (5Ј-CAA GCT TAG ATC TAA GCT TAG ATC TAG CT-3Ј, and 5Ј-AGA TCT AAG CTT AGA TCT AAG CTT GGT AC-3Ј), containing KpnI, HindIII, BglII (destroyed upon insertion of the ADH1 cassette), HindIII, BglII, and a SacI-compatible 3Ј end. Plasmids pRS316-ADH-GPA1 and pRS316-ADH-GPA1 G2A were constructed by ligation of the 1750-base pair BamHI-EcoRI fragment of GPA1 into the corresponding sites of pRS316-ADH. For overexpression of GPA1 or gpa1 G2A , the 1750-base pair SalI-EcoRI fragment of GPA1 from pRS316-ADH-GPA1 was transferred into pBluescript, then digested with SalI-SacI and ligated to the corresponding sites in pAD4M.
The GPA1-GST 1 fusion was obtained from K. Blumer (Washington University) and consists of the full-length GPA1 coding region plus an artificial BamHI site at the 3Ј end of the open reading frame (prepared by PCR using the oligonucleotide 5Ј-GCG GAT CCT ATA ATA CCA ATT TTT TTA AGG TTT TGC-3Ј), ligated to a corresponding in frame BamHI site in GST (prepared by PCR using the oligonucleotides 5Ј-GCG GGA TCC ATC GAA GGT CGT GGG ATG TCC CCT ATA CTA GGT TAT TGG-3Ј, containing a BamHI site; and 5Ј-GGG AAT TCT TAT TTT GGA GGA TGG TCG CCA CC-3Ј containing an EcoRI site). pRS316-GPA1-GST fusions were constructed by ligation of the 380-base pair SalI-HindIII fragment from pRS316-ADH-GPA1 or pRS316-ADH-GPA1 G2A and the 1680-base pair HindIII-EcoRI fragment from GPA1-GST into the SalI and EcoRI sites of pRS316. pAD4M-GPA1-GST was constructed by ligation of the 2060-base pair SalI-SacI fragment from pRS316-GPA1-GST into the SalI and SacI sites of pAD4M.
pRS316-GAL-STE4/18 was constructed by ligation of the EcoRI-SalI fragment from pL19 (containing GAL1/10 and STE4, from M. Whiteway, National Research Council of Canada) (42) into pRS316. STE18 was PCR-amplified using M57p4 as a template (from M. Whiteway) and oligonucleotides that flank the open reading frame (GGG AAT TCT AGG ATA GTA GCA ATC GCA, 5Ј-oligo; GAG GCT CTA CGT AGC AAG, 3Ј-oligo), digested with EcoRI and SacI, and ligated to the corresponding sites in pRS316-GAL-STE4. The resulting plasmid allows expression of both genes from the bidirectional, galactose-inducible GAL1/10 promoter.
Growth Inhibition Assay-The growth inhibition assay (halo assay) was performed as described previously (43). Briefly, 100 l from an overnight culture grown at 34°C were diluted with 2 ml of SCD-uracil, followed by the addition of an equal volume of 1% (w/v) dissolved agar (55°C) and poured onto a culture dish of the same medium. Sterile filter discs were spotted with 15 g of synthetic ␣-factor and placed on the nascent lawn. The resulting halo of the G 1 -arrested cells closest to the source of ␣-factor was documented after 2 days at 24°C.
Metabolic Labeling and Purification of Gpa1p-GST-BJ2168 cells expressing pAD4M-GST, pAD4M-GPA1-GST, or pAD4M-GPA1 G2A -GST were grown to mid logarithmic phase in selective medium. Approximately 1.3 ϫ 10 8 cells were harvested by centrifugation at 1000 ϫ g for 10 min. The cells were washed twice with 10 ml of YPD plus 1% ethanol and resuspended at 2.5 ϫ 10 7 cells/ml in YPD plus 1% ethanol. After 10 min of growth at 30°C, cells were treated with cerulenin (2 g/ml) to inhibit endogenous fatty acid synthesis 15 min before the addition of 30 Ci/ml [ 3 H]myristic acid (DuPont NEN, ϳ39 Ci/mmol) and grown for 1.5 h. Growth was stopped by the addition of NaN 3 to 10 mM. All subsequent manipulations were carried out at 0 -4°C. Cells were harvested by centrifugation and washed once with 10 mM NaN 3 and once with lysis buffer A (40 mM triethanolamine, pH 7.2, 2 mM EDTA, 2 mM dithiothreitol, 0.15 M NaCl, 2 M 4-(2-aminoethyl)-benzenesulfonylfluoride-HCl (AEBSF), and 10 g/ml each of leupeptin, pepstatin, benzamidine, aprotinin). Cell pellets were resuspended in 300 l of lysis buffer A and subjected to glass bead vortex homogenization for 4 min. The lysate was centrifuged twice at 500 ϫ g for 10 min, and the resulting supernatant was treated with 1% Triton X-100 (final concentration) at 4°C for 1 h to solubilize membrane proteins. 100 l of glutathione-Sepharose 4B resin (Pharmacia Biotech Inc.) (20% slurry) was added to the soluble material and mixed at 4°C for 2 h. The resin were washed three times with 20 mM sodium P i , pH 7.3, 350 mM NaCl. The bound proteins were eluted by boiling in sodium dodecyl sulfatepolyacrylamide gel electrophoresis (SDS-PAGE) sample buffer (62.5 mM Tris⅐HCl, pH 6.8, 10% glycerol, 2% SDS, 1% ␤-mercaptoethanol, 0.0005% bromphenol blue) for 10 min. Samples were subjected to a linear gradient of 7-15% SDS-PAGE, and fixed in H 2 O:isopropanol: acetic acid (65:25:10) (v/v/v) for 30 min. To cleave possible thioesterlinked fatty acids, identical gels were soaked either with 1 M hydroxylamine, pH 7.0, at room temperature for 18 h, or with 1 M Tris⅐HCl, pH 7.0, as a control. The gels were then fixed again and treated with Amplify (Amersham Corp.) for 30 min, vacuum dried, and exposed to x-ray film (Eastman Kodak Co.) (44).
Metabolic Labeling and Immunoprecipitation-YGS5 cells expressing pRS316-ADH, pRS316-ADH-GPA1-myc, or pRS316-ADH-GPA1 G2Amyc were grown and metabolically labeled with [ 3 H]myristic acid as described above. All subsequent manipulations were carried out at 0 -4°C. Approximately 2 ϫ 10 8 cells were harvested by centrifugation and washed once with 10 mM NaN 3 and once with radioimmune precipitation buffer (150 mM NaCl, 50 mM Tris⅐HCl, pH 7.4, 5 mM EDTA, 1% Nonidet P-40, 0.5% deoxycholate, 0.1% SDS, 2 M AEBSF, and 10 g/ml each of leupeptin, pepstatin, benzamidine, aprotinin). Cells were then resuspended in 150 l of radioimmune precipitation buffer. Acidwashed glass beads were added and the suspension was vortexed at high speed for 4 min. The lysates were centrifuged twice at 500 ϫ g for 10 min. The supernatant was precleared by mixing with 10% protein A-Sepharose CL-4B (Pharmacia) for 30 min, and then immunoprecipitated by mixing with a monoclonal antibody 9E10 for 1 h, followed by the addition of 10% protein A-Sepharose CL-4B for 30 min. The protein A resin was washed four times with radioimmune precipitation buffer, and the bound proteins were eluted by boiling in SDS-PAGE sample buffer for 10 min. Samples were subjected to 8% SDS-PAGE, fixed in H 2 O:isopropanol:acetic acid (65:25:10) (v/v/v) for 30 min, treated with Amplify (Amersham) for 30 min, and vacuum dried. To cleave possible thioester-linked fatty acids, gels were soaked with 1 M hydroxylamine, pH 7.0, at room temperature for 18 h, or with 1 M Tris⅐HCl, pH 7.0, as a control.
Co-purification of Gpa1p-GST with G ␤␥ -BJ2168 cells expressing pAD4M-GST, pAD4M-GPA1-GST, or pAD4M-GPA1 G2A -GST were grown to mid-logarithmic phase in SCD-leucine medium. Growth was stopped by the addition of NaN 3 to 10 mM. All subsequent manipulations were performed at 0 -4°C. 5 ϫ 10 8 cells were harvested by centrifugation at 1000 ϫ g for 10 min and washed in lysis buffer A. Cell pellets were resuspended in 350 l of lysis buffer A, 3 mM MgCl 2 , 10 M GDP and subjected to glass bead vortex homogenization for 4 min. The lysate was centrifuged twice at 500 ϫ g for 10 min, and the resulting supernatant was treated with 1% Triton X-100 (final concentration) at 4°C for 1 h to solubilize membrane proteins. 100 l of glutathione-Sepharose 4B resin (Pharmacia) (20% slurry) was added to the lysate and mixed at 4°C for 2 h. The resin was resuspended in either buffer A containing 3 mM MgCl 2 , 10 M GDP, 1% Triton X-100 (GDP buffer), or 50 mM MgCl 2 , 20 M GTP␥S, 1% Triton X-100 (GTP␥S buffer) and mixed at 4°C for 2 h. (Higher magnesium concentrations (50 mM) were initially used to promote subunit dissociation, since these conditions have previously been shown to accelerate guanine nucleotide exchange and binding of GTP␥S. Magnesium forms a complex with the bound GTP and is essential for the conformational change leading to G ␤␥ release (45). The effect of raising the magnesium concentration varies somewhat depending on the G ␣ subtype, however, and can even retard exchange in some cases (46). Although all of our initial experiments were carried out at high magnesium, in subsequent experiments we found no difference in the extent of G ␤␥ binding or dissociation at 3 mM versus 50 mM magnesium.) The resin was washed three times with 20 mM sodium P i , pH 7.3, 350 mM NaCl. The bound proteins were eluted by boiling in 100 l of SDS-PAGE sample buffer for 10 min. The purified proteins were resolved by 12.5% SDS-PAGE, transferred to nitrocellulose, and probed with antibodies against GST (from J. Steitz, Yale University), Ste4p, or Ste18p (from D. Jenness, University Massachusetts). 2 Antibody detection was achieved using horseradish peroxidaseconjugated goat anti rabbit IgG (Bio-Rad) and the ECL chemiluminescence system (Amersham), according to the manufacturer's instructions.
G ␤␥ Binding in Solution and Determination of Relative G ␤␥ Binding Affinity-BJ2168 cells expressing pAD4M-GST, pAD4M-GPA1-GST, or pAD4M-GPA1 G2A -GST, were grown to mid logarithmic phase in SCDleucine medium. Growth was stopped by the addition of NaN 3 to 10 mM. All subsequent manipulations were performed at 0 -4°C. 7.5 ϫ 10 7 cells were harvested by centrifugation at 1000 ϫ g for 10 min and washed in lysis buffer A. Cell pellets were resuspended in lysis buffer A containing 50 mM MgCl 2 , 100 M GDP, 10 mM NaF, and 30 M AlCl 3 to promote G ␤␥ dissociation and subjected to glass bead vortex homogenization for 4 min. The lysate was centrifuged twice at 500 ϫ g for 10 min, and the resulting supernatant was treated with 1% Triton X-100 (final concentration) at 4°C for 1 h. 100 l of glutathione-Sepharose 4B resin (Pharmacia) (20% slurry) was added to the lysate and mixed at 4°C for 2 h. After removing the supernatant, the resin was used for G ␤␥ binding in solution and determination of relative G ␤␥ binding affinity.
For G ␤␥ binding in solution, the resin was mixed with lysates from JTY2117 cells expressing G ␤␥ , prepared as described below. After 2 h, the resin was centrifuged and resuspended in buffer A containing 50 mM MgCl 2 and either 100 M GDP or 100 M GTP␥S. The resin was washed three times with 20 mM sodium P i , pH 7.3, 350 mM NaCl. The bound proteins were eluted and analyzed by SDS-PAGE and immunobloting with antibodies against GST, Ste4p or Ste18p, as described above.
For determination of relative G ␤␥ binding affinity, the resin was mixed with lysates from cells expressing G ␤␥ and 100 M GDP for 2 h at 4°C. Lysates from cells expressing G ␤␥ were mixed with lysates from cells expressing no G ␤␥ to correct for any difference in the concentration of other proteins. The resin was washed three times with 20 mM sodium P i , pH 7.3, 350 mM NaCl. The bound proteins were analyzed by SDS-PAGE and immunobloting as described above.
For isolation of G ␤␥ , JTY2117 cells expressing pRS316-GAL or pRS316-GAL-STE4/18 were grown to mid logarithmic phase in SCDuracil. Cells were collected by centrifugation at 1000 ϫ g for 10 min and resuspended in SCG-uracil medium at 0.4 A 600 nm /ml. Cells were grown for another 6 h at 30°C and lysed in buffer A and solubilized with 1% Triton X-100 (final concentration) at 4°C for 1 h as described above.
Cell Disruption and Membrane Fractionation-YGS5 cells expressing pRS316-ADH-GPA1, pRS316-ADH-GPA1 G2A , pAD4M-GPA1, or pAD4M-GPA1 G2A were grown to mid logarithmic phase in selective media. Approximately 5 ϫ 10 8 cells were harvested and washed in lysis buffer B (same as lysis buffer A but lacking NaCl). All subsequent manipulations were carried out at 0 -4°C. Cell pellets were resuspended in 200 l of lysis buffer B, and subjected to glass bead vortex homogenization for 4 min. The lysate was centrifuged twice at 500 ϫ g for 10 min, and the resulting supernatant was centrifuged at 140,000 ϫ g for 30 min. A portion of the supernatant ("S" fraction) was diluted with an equal volume of 2 ϫ SDS-PAGE sample buffer and boiled for 10 min. The pellet ("P" fraction) was resuspended in lysis buffer B at the original sample volume, diluted with an equal volume of 2 ϫ SDS-PAGE sample buffer, and boiled for 10 min. The extracts were resolved by 8% SDS-PAGE, and transferred to nitrocellulose (47). Blots were probed with antiserum against Gpa1p (35), as described above.
Sucrose Gradient Fractionation-Methods for cell membrane fractionation have been described in detail elsewhere (48). 3 Briefly, cells were grown in selective media to mid-logarithmic phase, centrifuged, and resuspended in YPD at 0.5 A 600 nm /ml. After one doubling, growth was stopped by addition of NaN 3 to 10 mM. Approximately 3 ϫ 10 9 cells were harvested by centrifugation at 500 ϫ g for 10 min and washed twice with SK buffer (1.2 M sorbitol, 0.1 M potassium P i , pH 7.5).
Spheroplasts were prepared by resuspending cells in 10 ml of SK, 1 mg of zymolyase 100T (Kirin Brewery), 28.8 mM ␤-mercaptoethanol for 45 min at 30°C. All subsequent manipulations were performed at 0 -4°C. Spheroplasts were centrifuged at 500 ϫ g for 10 min, washed once with SK, and once with lysis buffer C (0.8 M sucrose, 20 mM triethanolamine, pH 7.2, 1 mM EDTA, 1 mM dithiothreitol, 2 M AEBSF, 10 g/ml each of leupeptin, pepstatin, benzamidine, and aprotinin). Cell pellets were resuspended in 1.0 ml of lysis buffer C and disrupted with 25 strokes of a motor-driven Potter-Elvehjem homogenizer. The lysate was cleared of unbroken cells and debris by centrifuging twice at 500 ϫ g for 10 min. 606 mg of sucrose were added to 650 l of the supernatant and dissolved (final sucrose concentration, 70% (w/v)). The sample was transferred to a Beckman thin walled polypropylene tube and overlaid with 1 ml of sucrose solutions of 60%, 50%, 40%, and 30% (w/v), respectively. The samples were then centrifuged in a Beckman SW50Ti swinging bucket rotor for 16 h at 170,000 ϫ g in a Beckman L80 ultracentrifuge. 16 samples of 300 l each were collected from the bottom of the gradient, diluted 1:1 with 2 ϫ SDS-PAGE sample buffer, and boiled for 10 min. Fractions 1 through 13 or 14 (fractions 14, 15, and 16 do not contain Gpa1p, Ste4p, or Pma1p) were resolved by 8% SDS-PAGE and immunoblotting as described above. Blots were probed with antibodies to Gpa1p and Ste4p (see above) as well as Pma1p (raised against the plasma membrane H ϩ ATPase from Neurospora crassa and specifically recognizes Pma1p in Saccharomyces cerevisiae, from C. Slayman, Yale University) (49).
In general, Gpa1p was expressed under the control of a heterologous gene promoter from ADH1. This promoter was used to normalize expression of GPA1, which is significantly reduced in ste11 and ste18 mutants (35) and is absent in a/␣ diploid cells (36,50). As shown in Fig. 1A, expression of GPA1 under control of the ADH1 promoter only modestly reduces pheromone responsiveness, as demonstrated by the growth inhibition (halo) assay (see "Experimental Procedures").
Epitope Tagging and Expression of Gpa1p-GST-For immunoprecipitation experiments, an epitope-tagged version of Gpa1p was prepared using a well defined antigenic domain of the myc oncogene product. The monoclonal antibody 9E10 binds to this epitope with high affinity and specificity, and does not recognize any endogenous yeast proteins (39,51). The epitope tag does not interfere with Gpa1p function in vivo. As shown in Fig. 1A, Gpa1p-myc can restore growth to the gpa1⌬ strain, and confers a normal pheromone response. Moreover, tagged and untagged versions of Gpa1p are expressed at equal levels in the cell and fractionate to the same membrane compartment following sucrose density gradient fractionation of membranes (see below, Figs. 7 and 8). Thus, epitope-tagging provides a convenient method for the sensitive and selective detection of functional Gpa1p.
For G ␤␥ binding experiments, we prepared a full-length version of Gpa1p fused to glutathione S-transferase (Gpa1p-GST). We placed the GST domain at the C terminus of Gpa1p so as not to interfere with myristoylation and G ␤␥ binding (functions requiring an intact N terminus) (30). Indeed, the Gpa1p-GST fusion was able to complement the growth defect of the gpa1⌬ mutant, indicating that it can bind G ␤␥ in vivo (Fig. 1B). However, the fusion protein did not produce a halo when expressed in this mutant strain (data not shown). This is presumably because an intact C-terminal domain of G ␣ is required for binding to the receptor (30). As expected, the gpa1⌬ mutation is not complemented by either Gpa1p G2A or the Gpa1p G2A -GST fusion (Fig. 1B) (11).
Gpa1p and Gpa1p G2A Bind Reversibly to G ␤␥ -As discussed above, it is not known why Gpa1p G2A cannot function in vivo. One possibility is that myristoylation is required for proper folding or binding to G ␤␥ . To test this hypothesis, we examined whether the G ␤␥ subunits can bind to Gpa1p G2A in vitro. Because Gpa1p is not available in purified form we devised a new strategy to detect subunit interactions using the Gpa1p-GST fusion. GST was chosen because it is a well characterized soluble protein for which antibodies and a glutathione-based affinity resin are available. Thus Gpa1p-GST (and any bound G ␤␥ ) can be easily purified using this resin. Moreover, G ␤␥ binding can be easily tracked using G ␤ and G ␥ antibodies. Functional interaction of G ␤␥ with Gpa1p can be demonstrated by showing specific binding in the presence of GDP, but is reversed by GTP␥S which promotes subunit dissociation.
To confirm that the fusion protein is still myristoylated, we purified Gpa1p-GST as well as Gpa1p G2A -GST and GST alone from cells metabolically labeled with [ 3 H]myristic acid. As A, the growth inhibition assay ("halo assay") was used to evaluate ␣-factor responsiveness in gpa1⌬ cells (strain YGS5) expressing GPA1, myc epitope-tagged GPA1 (GPA1-myc), and epitope-tagged GPA1 under the control of the ADH1 promoter (ADHp:GPA1-myc) (plasmid pRS316). Approximately 10 7 cells embedded in agar were exposed to 15 g of ␣-factor. Cells were maintained at 24°C to allow functional expression of the temperature sensitive ste11 gene product. The zone of growth inhibition is comparable for all three conditions, demonstrating that pheromone sensitivity is not significantly altered by use of the heterologous promoter and/or the myc-epitope tag. B, to determine if Gpa1p-GST can bind G ␤␥ in vivo, YGS5 cells containing GPA1-GST, GPA1, gpa1 G2A -GST, or gpa1 G2A were streaked out at 34°C, or at 24°C to allow functional expression of the temperature sensitive ste11 gene product. Growth at 24°C indicates complementation of the gpa1⌬ mutation.
shown in Fig. 2A, only the wild-type fusion protein was labeled. Labeling was abolished by the G2A mutation, and was absent from GST. Moreover, labeling was resistant to hydroxylamine treatment, indicating that the incorporated label had not been converted to palmitic acid (myristic acid forms an amide linkage that is hydroxylamine resistant, whereas palmitic acid forms a thioester linkage that is hydroxylamine labile) (44). Essentially identical results were obtained when we immunoprecipitated the epitope-tagged (fully functional) version of Gpa1p (Fig. 2B) (11). Thus, while immunoprecipitation provides a convenient way to detect lipid modification of functional Gpa1p, the Gpa1p-GST fusion is similarly modified and can also be used to measure G ␤␥ binding in vitro.
To determine if Gpa1p G2A retains the ability to bind G ␤␥ , we purified Gpa1p-GST and examined the effects of guanine nucleotides on G ␤␥ subunit interaction. Detergent solubilized lysates from equal numbers of cells expressing Gpa1p-GST, Gpa1p G2A -GST, or GST alone were prepared in the presence of GDP, and mixed with the glutathione resin. The washed resin was then treated with GTP␥S to promote subunit dissociation, or with GDP to preserve the complex. Equal levels of Gpa1p and Gpa1p G2A fusions remained bound to the resin after these washes, as indicated by blotting with anti-GST antiserum (Fig.  3, top). By probing the same extracts with Ste4p antiserum, we   FIG. 2. Gpa1p-GST is myristoylated. A, to determine if the Gpa1p-GST fusion is myristoylated, BJ2168 cells expressing Gpa1p-GST, Gpa1p G2A -GST, or GST alone (plasmid pAD4M) were metabolically labeled with [ 3 H]myristic acid. All three proteins were purified from solubilized whole cell lysates using a glutathione-Sepharose 4B resin, and resolved by SDS-PAGE. Gels were treated with either 1 M hydroxylamine to cleave thiol-ester bonded fatty acids (top) or 1 M Tris⅐HCl (bottom), and analyzed by fluorography. B, to demonstrate that epitopetagged Gpa1p is myristoylated and that myristoylation is abolished by the G2A mutation, YGS5 cells bearing the vector alone, GPA1-myc, or gpa1 G2A -myc (pRS316-ADH) were metabolically labeled with [ 3 H]myristic acid. Proteins were immunoprecipitated from solubilized cell lysates with the myc monoclonal antibody 9E10. The immunoprecipitates were subjected to SDS-PAGE and treated with hydroxylamine (top) or Tris⅐HCl (bottom) as described above.

FIG. 3. Ste4p copurifies with Gpa1p and Gpa1p G2A .
To determine if nonmyristoylated Gpa1p is able to bind to G ␤␥ in vitro, Gpa1p-GST, Gpa1p G2A -GST, or GST alone were expressed (pAD4M) in BJ2168 and purified using the glutathione affinity resin. Equal numbers of cells were disrupted in buffer containing GDP. The lysates were solubilized with Triton X-100, incubated with the glutathione-Sepharose resin, then centrifuged and resuspended in buffer containing either GDP or GTP␥S. The resin was subsequently washed in high salt buffer, and the bound protein was eluted and analyzed by immunobloting with antibodies against GST (top) or Ste4p (bottom), as indicated.
found that G ␤ (presumably as the G ␤␥ complex (52), see below, Fig. 5) co-purified with Gpa1p-GST and Gpa1p G2A -GST, and not at all with GST (Fig. 3, bottom). For both the wild-type and mutant fusion proteins, addition of GTP␥S led to dissociation of G ␤␥ . These data reveal that G ␤␥ can form a complex with wild-type Gpa1p as well as with the Gpa1p G2A mutant, and that either complex is stabilized in the presence of GDP and dissociated in the presence of GTP␥S.
To determine if Gpa1p G2A -GST can bind to G ␤␥ in solution, we prepared Gpa1p-GST and G ␤␥ separately and tested their ability to interact in a guanine nucleotide dependent manner. To obtain Gpa1p without G ␤␥ , we purified Gpa1p-GST and Gpa1p G2A -GST in the presence of GDP and AlF 4 Ϫ . AlF 4 Ϫ was used because it mimics the ␥-phosphate of GTP when bound to GDP, and promotes complete subunit dissociation when added to the lysis buffer (53,54). To obtain G ␤␥ without Gpa1p, we expressed G ␤␥ in a MATa/␣ strain, where G protein subunits are normally not expressed (36,50). The expression of G ␤␥ was confirmed by immunobloting with Ste4p and Ste18p antibodies (Fig. 4). The fusion proteins Gpa1p-GST, Gpa1p G2A -GST, or GST alone were immobilized on the glutathione resin, washed extensively, and mixed with lysates from diploid cells expressing G ␤␥ . After removing the supernatant, the resin was treated with buffer containing GDP or GTP␥S. The bound proteins remaining were analyzed by immunobloting with GST, Ste4p, and Ste18p antisera. As shown in Fig. 5, both Ste4p and Ste18p bound well to Gpa1p-GST and Gpa1p G2A -GST, but not to GST alone in the presence of GDP. For both the wild-type and mutant fusion proteins, treatment with GTP␥S led to dissociation of Ste4p and Ste18p. In both binding experiments (Figs. 3 and 5), however, GTP␥S-dependent dissociation of G ␤␥ from the mutant was more pronounced than from the wild-type Gpa1p fusion. This could be due to a number of differences in G protein function, as discussed below. Since our primary concern was whether there was a myristoylation-dependent change in subunit affinity, we examined this parameter directly.
To determine if there is any difference in the relative G ␤␥ binding affinity for mutant and wild-type Gpa1p, equal amounts of Gpa1p-GST and Gpa1p G2A -GST were purified in the presence of GDP and AlF 4 Ϫ . The purified fusion proteins did not contain endogenous G ␤␥ as shown by immunobloting with Ste4p (Fig. 6) and Ste18p antisera (data not shown). The resin was subsequently mixed with increasing amounts of lysates from cells expressing G ␤␥ in the presence of GDP. Lysates from cells expressing G ␤␥ were mixed with lysates from cells that do not express G ␤␥ to correct for any difference in the concentration of other proteins. The binding of G ␤␥ was analyzed by SDS-PAGE and immunobloting with Ste4p antiserum. Each lane contained equal amounts of fusion protein. As shown in Fig. 6, both the wild-type and G2A mutant fusions bind to G ␤␥ in a saturable and concentration-dependent manner, and with no significant difference in affinity.
Gpa1p G2A Does Not Bind to Plasma Membranes-The data presented above reveal that Gpa1p G2A retains the ability to bind G ␤␥ . Thus, we considered an alternative possibility that the loss of function exhibited by this mutant is due to mislocalization within the cell. It was shown previously that both Gpa1p and Gpa1p G2A bind to cell membranes (11). However, those studies did not examine if either protein was specifically associated with the plasma membrane (11). Thus, signaling could be disrupted because Gpa1p G2A does not reach the plasma membrane where G ␤␥ is normally located. To test this hypothesis, we used differential and sucrose density gradient centrifugation methods to determine if the subcellular distribution of Gpa1p G2A is altered in any way.
Cells expressing Gpa1p were homogenized, centrifuged at low speed to remove unbroken cells and nuclei, and then separated into soluble (S) and membrane (pelleted, P) fractions by centrifugation at 140,000 ϫ g. As shown in Fig. 7A, wild-type Gpa1p and Gpa1p G2A both fractionate exclusively with the 140,000 ϫ g membrane fraction. Moreover, the antibodies recognize a single product for Gpa1p G2A (migrating at 56 kDa) but two products for wild-type Gpa1p, corresponding to the myristoylated (54 kDa) and nonmyristoylated (56 kDa) forms of the To demonstrate that Gpa1p G2A binds to both ␤ and ␥ subunits in solution, Gpa1p-GST, Gpa1p G2A -GST, or GST alone were expressed (pAD4M) in BJ2168 and purified using the glutathione affinity resin. Equal number of cells were disrupted in buffer containing GDP and AlF 4 Ϫ , conditions which dissociate all bound G ␤␥ . The lysates were solubilized with Triton X-100, and mixed with the glutathione-Sepharose resin. The resin was then centrifuged and mixed with lysates from JTY2117 cells expressing G ␤␥ (Fig. 4). After 2 h, the resin was centrifuged and resuspended in buffers containing either GDP or GTP␥S. The resin was subsequently washed in high salt buffer, and the bound protein was eluted and analyzed by immunobloting with antibodies against GST, Ste4p, or Ste18p, as indicated.
protein (11,35). Identical results were obtained using myctagged Gpa1p (data not shown). These results confirm that myristoylated and nonmyristoylated pools of wild-type Gpa1p, as well the nonmyristoylated Gpa1p mutant, are all predominantly associated with cell membranes (11).
To determine if nonmyristoylated Gpa1p still associates specifically with the plasma membrane, precleared cell lysates were subjected to high speed centrifugation through a 70 -30% sucrose flotation gradient. Fractions were collected from the bottom of the gradient, prepared for immunoblotting, and probed with antisera against Gpa1p and Pma1p (an integral plasma membrane protein marker) (49). Immunoblots for cells expressing myc-tagged Gpa1p are shown, but similar results were obtained using untagged Gpa1p (see Fig. 7E). As presented in Fig. 7B, wild-type Gpa1p associates predominantly with the plasma membrane fractions (containing Pma1p), while Gpa1p G2A accumulates in earlier fractions that contain Golgi and other microsomal membrane proteins (Fig. 7C). 3 To determine if mislocalization of Gpa1p G2A results in a concomitant redistribution of the G ␤␥ complex, we probed the same fractions with polyclonal antibodies to Ste4p. As shown in Fig. 7B, Ste4p is normally associated with the plasma membrane, but is partially mislocalized in cells expressing Gpa1p G2A (Fig. 7C). This redistribution of Ste4p could be the result of binding to the mislocalized population of Gpa1p G2A . Alternatively, functional Gpa1p may be required for proper targeting of G ␤␥ to the plasma membrane. To clarify the relationship between the localization of Gpa1p and Ste4p, we also examined the distribution of Ste4p in a gpa1⌬ strain. As shown in Fig. 7D, Ste4p localization is similar to that seen in cells expressing Gpa1p G2A . These results suggest that Gpa1p G2A is largely excluded from the plasma membrane and that any loss of functional Gpa1p expression leads to partial mislocalization of the G protein G ␤␥ subunits.
We also attempted to compare the distribution of the 54-kDa myristoylated and 56-kDa nonmyristoylated forms of the wildtype protein, since this would have provided the best possible internal control for the influence of myristoylation on localization. However, the lower abundance of the nonmyristoylated species in these preparations (Fig. 7A), the high concentrations of sucrose in the gradient samples (which can alter protein mobility) and the long centrifugation times (which can lead to sample degradation) were all confounding factors that made it impossible to reliably quantify myristoylated versus nonmyristoylated wild-type Gpa1p by this method.
The data in Fig. 7A indicate that the overall expression levels of wild-type Gpa1p are higher than the mutant. This difference is not surprising since a greater fraction of wild-type Gpa1p is myristoylated than nonmyristoylated in these cells. Nonetheless, to rule out the possibility that the lower expression of Gpa1p G2A might be responsible for the difference in membrane localization, we examined the distribution of Gpa1p G2A overexpressed from a multicopy plasmid (pAD4M). As shown in Fig.  7F, expression of both the mutant and wild-type protein is much higher than that with the single copy plasmid (using the same ADH1 promoter), yet they still remain completely associated with the 140,000 ϫ g membrane fraction. We then used the sucrose gradient method to determine if overexpression of the protein alters its subcellular localization. As shown in Fig.  7, G and H, the distribution of overexpressed mutant and wild-type Gpa1p is indistinguishable from that seen with the single copy plasmid. However, overexpression of the mutant appears to partially restore normal G ␤␥ localization. This is unexpected, particularly since the distribution of Gpa1p G2A is unaltered in this case. Nevertheless, these data indicate that even large differences in Gpa1p expression have no apparent effects on its subcellular distribution.
Plasma Membrane Localization of Gpa1p Does Not Require G ␤␥ or the Receptor-The data presented in Figs. 3 and 5 indicate that nonmyristoylated Gpa1p can still bind to G ␤␥ . The results shown in Fig. 7 indicate that myristoylation is essential for proper localization of Gpa1p to the plasma membrane. Ste4p was also partially mislocalized in these experiments. To determine if Ste4p mislocalization was the result or the cause of mislocalization of Gpa1p, we examined the distribution of Gpa1p in cells that do not express G ␤␥ . To determine if the ␣-factor receptor is needed for Gpa1p targeting, we also examined Gpa1p localization in a ste2 mutant. As shown in Fig. 8, Gpa1p reaches the plasma membrane in a haploid mutant lacking STE18 (G ␥ ), STE2, as well as in wild-type diploid cells. Diploid cells normally lack the receptor, all three G protein subunits, and a number of other proteins required for signaling (including Ste4p, Fus3p, Ste12p, Ste5p, and Sst2p) (1,3). In all three strains, Gpa1p fractionates with the plasma membrane marker Pma1p, in a manner indistinguishable from wild-type haploid cells (Fig. 8, and data not shown). Thus Gpa1p is properly targeted to the plasma membrane even in the absence of G ␤␥ , the pheromone receptor, or any other proteins expressed exclusively in haploids. Collectively, these data indicate that Gpa1p G2A is mislocalized because it is not myristoylated and not because of a defect in subunit interaction.

DISCUSSION
Lipid modification of proteins can have profound effects on their function and subcellular localization (4). Although only a small number of cellular proteins are N-myristoylated, these include almost all G protein ␣ subunits (10 -18). In S. cerevisiae, it has been documented that Gpa1p is myristoylated and that a myristoylation site mutant exhibits the null phenotype (11). However, it is not clear why this modification is essential for G protein function.
Past studies have revealed a likely role for myristoylation in G protein subunit interaction. Jones et al. (17) have shown that nonmyristoylated G i1␣ requires higher concentrations of G ␤␥ to support ADP-ribosylation by pertussis toxin (which only recognizes the heterotrimer), suggesting that subunit binding affinity is reduced by the lack of myristoylation. Linder et al. (26) observed an even more dramatic reduction of toxin labeling for nonmyristoylated G o␣ . They also showed that nonmyristoylated G o␣ fails to bind a G ␤␥ affinity matrix and is insensitive to G ␤␥ -dependent inhibition of GTP␥S binding.
Previous studies have also shown that myristoylation of G ␣ FIG. 6. Gpa1p and Gpa1p G2A bind to G ␤␥ with similar affinity. To compare the relative G ␤␥ binding affinities of Gpa1p G2A and wildtype Gpa1p, GST fusion proteins were purified, and subsequently mixed with various ratios of amount of lysates from JTY2117 cells expressing either pRS316-GAL or pRS316-GAL-STE4/18 in the presence of GDP. The bound protein was equilibrated, washed, and analyzed by immunobloting with antiserum against Ste4p. Relative concentration, A 600 nm equivalents of G ␤␥ -expressing cells.
FIG. 7. Membrane fractionation of Gpa1p and Gpa1p G2A . Membrane fractionation and immunoblot analysis were used to determine the subcellular localization of Gpa1p and Gpa1p G2A . A, YGS5 cells expressing GPA1 or gpa1 G2A were lysed, and the precleared homogenates promotes membrane association. Blocking myristoylation leads to dissociation of G o␣ , G i␣ , and G z␣ (but not Gpa1p) from the membrane pellet to the cytosol (10,16,17), and to a marked decrease in the detergent-phase partitioning of G t␣ (55).
These findings are consistent with at least two models of how myristoylation could affect G protein subunit interaction. In one scenario, the myristoyl moiety binds directly to G ␤␥ . The lack of a myristoyl group in G ␣ leads to a reduction in subunit binding affinity and consequently to dissociation of G ␣ from the membrane. Alternatively, the myristoyl group could bind directly to the lipid bilayer. In this case, the lack of a myristoyl group would lead to reduced membrane association and only indirectly to a loss of subunit binding. That most G protein subtypes require myristoylation for membrane attachment has made it difficult to distinguish between such direct and indirect effects on subunit binding. Transducin is an exception in this regard, since it can be solubilized from membranes without the use of detergents. Indeed, Chabre and colleagues (31) have recently shown that transducin subunits form a tight complex in the presence of lipids or detergent, but are readily dissociated in aqueous solution. They proposed that lipid modifications of both G␣ and G␥ contribute indirectly to protein-protein interactions, by restricting their relative mobility to the twodimensional plane of a lipid bilayer or detergent micelle. They also suggested that direct lipid-lipid interactions might play a role in subunit association, a model that is supported by the recent finding that the modified residues of G ␣ and G ␥ are in close proximity in the crystal structure of the heterotrimer (53,54). However, none of these reports attempted to quantify the relative hydrophobicity or G ␤␥ binding affinity of myristoylated versus nonmyristoylated G ␣ .
The G protein in yeast is another attractive system for investigating how myristoylation affects subunit interaction, since Gpa1p remains associated with the lipid bilayer even in the absence of myristoylation. Thus a direct role for myristoylation in subunit binding can be examined without the confounding effects of membrane dissociation. In the experiments described above, we examined the ability of myristoylated and nonmyristoylated Gpa1p to bind G ␤␥ , and to localize specifically to the plasma membrane.
Initially we asked if the Gpa1p G2A mutant was still able to bind G ␤␥ . Previous studies have used the "2-hybrid" protein association assay to monitor G protein subunit interactions in yeast (52). This method was not appropriate in our case since it requires that the Gal4p DNA binding domain or nuclear localization sequence be fused to the N terminus of Gpa1p, an arrangement that would block myristoylation. In addition, this approach cannot be used to detect guanine-nucleotide dependent effects on subunit binding and dissociation. Thus we developed an alternative method in which we purified Gpa1p fused to GST, and examined its ability to bind G ␤␥ . The N terminus is unaltered, so the functional role of N-myristoylation can be tested explicitly. To carry out these experiments, we first showed that the fusion and non-fusion variants behave similarly. Wild-type Gpa1p and Gpa1p-GST are myristoylated and can bind G ␤␥ in vivo, whereas a G2A mutation blocks myristoylation and fails to complement gpa1⌬. Both wild-type and mutant fusions can bind to Ste4p/Ste18p in vitro, and binding is reversed by treatment with GTP␥S. Thus Gpa1p-GST is myristoylated, can bind guanine nucleotides, and can undergo the conformational change needed for subunit dissociation (Figs. 3 and 5). By probing the same samples with Ste4p and Ste18p antibodies, we confirmed that these proteins form a complex that associates with G ␣ . This is consistent with the situation in higher eukaryotes, where ␤ and ␥ subunits are known to function as a single unit (56), and corroborates in vivo experiments in yeast demonstrating that both subunits must be expressed in order to activate the downstream signaling pathway (57) and to interact with the presumptive G protein target, Ste5p (2).
The binding of Ste4p/Ste18p to wild-type and mutant Gpa1p-GST appears very similar. However, one consistent difference is that G ␤␥ dissociates more readily from the mutant when treated with GTP␥S. The mechanism underlying this difference is obscure, and could result from any one of a number of changes in G protein structure or function. Some possibilities (as yet untested) include changes in magnesium or guanine nucleotide binding affinity, alterations in protein stability or modification, differences in the rate of GTP hydrolysis or exchange, or competition with other G protein-binding proteins. Our primary concern was to rule out a difference in subunit binding affinity. Purified G ␤␥ subunits are not available from yeast however, so a direct binding assay could not be performed. Thus we used a method that relies only on the available G ␤ antibodies. Specifically, we equilibrated mutant and centrifuged at 140,000 ϫ g for 30 min. The cytosolic (S, supernatant) and membrane (P, pellet) fractions were analyzed by immunoblotting with antiserum against Gpa1p. The arrows indicate the position of nonmyristoylated (upper band, 56 kDa) and myristoylated (lower band, 54 kDa) forms of Gpa1p. Cells expressing GPA1 (B), gpa1 G2A (C and E), or vector alone (D) were lysed, and the precleared homogenates were resolved by 30 -70% sucrose density gradient centrifugation. 16 Fractions were collected from the bottom of gradient and analyzed by SDS-PAGE and immunoblotting using antibodies against Gpa1p, Ste4p, or Pma1p (H ϩ ATPase, plasma membrane marker). Fractions 14 -16 are not shown and did not contain Gpa1p, Ste4p, or Pma1p. Immunoblots from cells expressing myc-tagged Gpa1p are shown in the panels B and C, but similar results were obtained with untagged protein (panel E and data not shown). The top band of the doublet observed with the Gpa1p G2A mutant reflects nonspecific staining, since it is also seen in cells that do not express Gpa1p and its mobility is not altered by epitope-tagging, in contrast to authentic Gpa1 (compare panels C and E, containing tagged and untagged protein). F, G, and H, membranes were prepared and analyzed exactly as described for panels A-E, except that a multicopy plasmid (pAD4M) was used for Gpa1p overexpression.
FIG. 8. Gpa1p is normally localized in cells that lack G ␤␥ or receptor. Membrane fractionation and immunoblot analysis were used to determine if Gpa1p binding to the plasma membrane requires expression of the receptor or G ␤␥ . Plasmid pRS316-ADH-GPA1 was transformed into strains YPH499 (MATa, wild type), MHY6 (G ␥ mutant, ste18⌬), YDK102 (receptor mutant, ste2⌬), or YPH501 (MATa/␣, diploid wild type). Cell membranes were resolved by 30 -70% sucrose density gradient centrifugation, and fractions across the gradient were analyzed by SDS-PAGE and immunoblotting as described in Fig. 7.
wild-type Gpa1p-GST with a range of G ␤␥ concentrations (generated by mixing lysates from G ␤␥ -expressing and -nonexpressing cells), and performed immunoblots to detect the bound protein. In this case, we could not detect any difference in binding of G ␤␥ to the wild-type and mutant proteins (Fig. 6).
These results appear to be at odds with previous studies showing reduced binding of G ␤␥ to nonmyristoylated G␣ in mammals. However, another way that myristoylation could contribute to G protein subunit interaction is to promote colocalization within the cell. Loss of subunit binding in mammals could stem from a reduction in membrane or detergent micelle partitioning, rather than a direct loss of subunit binding affinity. In yeast, previous studies have shown that myristoylation of Gpa1p is not needed for membrane attachment, so the possibility remained that Gpa1p G2A is not at the plasma membrane and is therefore unable to bind G ␤␥ . Thus we examined the subcellular distribution of wild-type and mutant Gpa1p, using a sucrose density gradient centrifugation method. We found that the distribution of the two proteins differs markedly, with wild type mostly at the plasma membrane and Gpa1p G2A absent from these fractions. These experiments suggest that myristoylation can indeed play a role in subcellular localization of Gpa1p (Fig. 7). This method was also used to examine the distribution of wild-type Gpa1p in cells that lack specific components of the signal transduction apparatus. In these experiments, Gpa1p is properly localized to the plasma membrane even in cells that do not express G ␤␥ (Fig. 8). Thus, even if myristoylation contributes to the affinity of subunit interaction in vivo, this cannot explain the mislocalization of Gpa1p G2A since Gpa1p is still able to reach the plasma membrane even in the absence of G ␤␥ expression.
It is interesting that Ste4p is mislocalized in cells expressing Gpa1p G2A , as well as in cells that lack GPA1 altogether. The mislocalization is not complete, however, since substantial amounts of the protein still associate with the plasma membrane fractions that are devoid of Gpa1p G2A . This is in marked contrast to our finding that Gpa1p localization is unaltered in cells that lack G ␤␥ . Thus it appears that functional Gpa1p expression is required for proper localization of G ␤␥ , but G ␤␥ is not required for proper localization of Gpa1p. It is also interesting that the mislocalization of G ␤␥ appears less severe in cells that overexpress Gpa1p G2A . Although the mechanism for this difference is not clear, it suggests that the mutant form of the protein retains some ability to direct G ␤␥ to the proper membrane compartment. Perhaps a small (undetectable) amount of the mutant reaches the plasma membrane and helps to recruit G ␤␥ as well. Alternatively, the mutant could compete with G ␤␥ for a common binding target in the microsomal membrane compartment, and this competition allows G ␤␥ to disengage and reach the plasma membrane. In any case, a substantial pool of G ␤␥ is detached from Gpa1p G2A .
Finally, we also considered a third possibility, that reduced expression of Gpa1p G2A could be responsible for the difference in localization or function. The expression levels of both the mutant and nonmyristoylated wild-type forms of Gpa1p are similar (Fig. 7A), so this difference appears to reflect the in vivo situation. Nevertheless, differences in expression cannot explain the change in localization since increasing the expression of Gpa1p G2A does not change its subcellular distribution and still fails to complement the gpa1⌬ mutation (Fig. 7).
In summary, we have characterized the role of myristoylation in G protein subunit interaction and subcellular localization. The central findings of this study are that myristoylation of Gpa1p is required for its plasma membrane localization and is not required for its interaction with Ste4p/Ste18p. These conclusions are based on the observations that the Gpa1p G2A mutant fails to reach the plasma membrane, but is still able to interact with G ␤␥ subunits in vitro. Moreover, plasma membrane localization of Gpa1p does not require expression of G ␤␥ . Collectively, these data suggest that the loss of function exhibited by Gpa1p G2A is most likely due to a difference in G ␣ and G ␤␥ localization. The resulting pool of unsequestered G ␤␥ is then free to activate the downstream effector, leading to constitutive signaling in vivo.
These results may have significance with regard to the regulation of G protein signaling. Once thought to be co-translational and irreversible, there is emerging evidence that myristoylation can occur post-translationally and is reversed in a stimulus-dependent manner (58 -60). Several groups have also described proteins (including Gpa1p) that undergo myristoylation in response to hormone stimulation (35,(61)(62)(63)(64). The functional significance of hormone or pheromone-regulated myristoylation is not clear. However, one possibility, suggested by the results described here, is that myristoylation regulates the recruitment or retention of Gpa1p to the plasma membrane where it is available to transmit a signal from the receptor to G ␤␥ . A challenge for the future will be to determine how pheromone-regulated changes in myristoylation affect G protein signaling efficiency, and what additional proteins or modifications are required to recruit myristoylated Gpa1p to the plasma membrane.