The Gbg Complex of the Yeast Pheromone Response Pathway

Genetic evidence suggests that the yeast STE4 and STE18 genes encode Gβ and Gγ subunits, respectively, that the Gβγ complex plays a positive role in the pheromone response pathway, and that its activity is subject to negative regulation by the Gα subunit (product of the GPA1 gene) and to positive regulation by cell-surface pheromone receptors. However, as yet there is no direct biochemical evidence for a Gβγ protein complex associated with the plasma membrane. We found that the products of the STE4 and STE18 genes are stably associated with plasma membrane as well as with internal membranes and that 30% of the protein pool is not tightly associated with either membrane fraction. A slower-migrating, presumably phosphorylated, form of Ste4p is enriched in the non-membrane fraction. The Ste4p and Ste18p proteins that had been extracted from plasma membranes with detergent were found to cosediment as an 8 S particle under low salt conditions and as a 6 S particle in the presence of 0.25 M NaCl; the Ste18p in these fractions was precipitated with anti-Ste4p antiserum. Under the conditions of our assay, Gpa1p was not associated with either particle. The levels of Ste4p and Ste18p accumulation in mutant cells provided additional evidence for a Gβγ complex. Ste18p failed to accumulate in ste4 mutant cells, and Ste4p showed reduced levels of accumulation and an increased rate of turnover in ste18 mutant cells. The gpa1 mutant blocked stable association of Ste4p with the plasma membrane, and the ste18 mutant blocked stable association of Ste4p with both plasma membranes and internal membranes. The membrane distribution of Ste4p was unaffected by the ste2 mutation or by down-regulation of the cell-surface receptors. These results indicate that at least 40% of Ste4p and Ste18p are part of a Gβγ complex at the plasma membrane and that stable association of this complex with the plasma membrane requires the presence of Gα.

Genetic evidence suggests that the yeast STE4 and STE18 genes encode G␤ and G␥ subunits, respectively, that the G␤␥ complex plays a positive role in the pheromone response pathway, and that its activity is subject to negative regulation by the G␣ subunit (product of the GPA1 gene) and to positive regulation by cell-surface pheromone receptors. However, as yet there is no direct biochemical evidence for a G␤␥ protein complex associated with the plasma membrane. We found that the products of the STE4 and STE18 genes are stably associated with plasma membrane as well as with internal membranes and that 30% of the protein pool is not tightly associated with either membrane fraction. A slower-migrating, presumably phosphorylated, form of Ste4p is enriched in the non-membrane fraction. The Ste4p and Ste18p proteins that had been extracted from plasma membranes with detergent were found to cosediment as an 8 S particle under low salt conditions and as a 6 S particle in the presence of 0.25 M NaCl; the Ste18p in these fractions was precipitated with anti-Ste4p antiserum. Under the conditions of our assay, Gpa1p was not associated with either particle. The levels of Ste4p and Ste18p accumulation in mutant cells provided additional evidence for a G␤␥ complex. Ste18p failed to accumulate in ste4 mutant cells, and Ste4p showed reduced levels of accumulation and an increased rate of turnover in ste18 mutant cells. The gpa1 mutant blocked stable association of Ste4p with the plasma membrane, and the ste18 mutant blocked stable association of Ste4p with both plasma membranes and internal membranes. The membrane distribution of Ste4p was unaffected by the ste2 mutation or by down-regulation of the cell-surface receptors. These results indicate that at least 40% of Ste4p and Ste18p are part of a G␤␥ complex at the plasma membrane and that stable association of this complex with the plasma membrane requires the presence of G␣.
The pheromone response pathway of Saccharomyces cerevisiae provides a microbial model for studying function of heterotrimeric G proteins. ␣-Factor and a-factor pheromones are peptides synthesized by haploid cells of ␣ and a mating type, respectively. Each pheromone binds specific receptors located on cells of the opposite mating type; receptor activation causes cell division arrest at G 1 and expression of genes controlling the conjugation of the two mating types. Both ␣-factor receptors (encoded by the STE2 gene) and a-factor receptors (encoded by STE3) are members of the rhodopsin/␤-adrenergic receptor family in that they contain seven transmembrane segments, and they require homologs of mammalian G␣, G␤, and G␥ protein subunits (encoded by GPA1, STE4, and STE18, respectively) for signal transduction. The receptor appears to form a direct physical association with the G-protein subunits since ␣-factor binds ste4 mutant cells more weakly (1) and since it dissociates more rapidly from membranes assayed in the presence of GTP analogs or when the membranes are prepared from mutants with defects in GPA1, STE4, or STE18 (2). Genetic evidence indicates that G␤␥, rather than G␣, activates subsequent events in the response pathway, that is loss of function mutations in STE4 and STE18 block the signal (3)(4)(5) whereas loss of function mutations in GPA1 activate signaling (6 -8). Moreover, gain of function mutations affecting STE4 (designated STE4 Hpl ) (9, 10) and overexpression of wild-type Ste4p (11)(12)(13) cause constitutive activation of the pathway. Gpa1p and Ste18p are modified covalently with myristoyl and farnesyl moieties, respectively (14,15).
Genetic studies also suggest that G␤ associates with the G␣ and G␥ homologs in vivo as well as with other proteins thought to play roles in the pheromone response pathway. However, biochemical characterization of these potential interactions has not been reported. Interactions between Gpa1p and Ste4p were inferred from two-hybrid genetic analysis (10,16). Moreover, overexpression of Gpa1p compensates for overexpression of Ste4p (11)(12)(13); STE4 Hpl mutations are suppressed by mutations in GPA1 (10), and one allele is suppressed by overexpression of the wild-type Gpa1p (17). Two-hybrid genetic analysis also predicts that Ste4p binds to Ste18p (16). Mutations in the STE18 gene have been identified that either suppress or enhance partial defects in STE4 (16,18). Some ste4 alleles are suppressed by overexpression of Ste18p (16,19), while certain ste18 alleles are suppressed by overexpression of Ste4p (20). Genetic tests also suggest that Ste4p interacts with the products of the STE5 (21,22), STE20 (22)(23)(24)(25), CDC24 (26), SYG1 (27), and AKR1 (28,29) genes; the interaction with Ste5p has been confirmed by demonstrating that epitope-tagged Ste4p coimmunoprecipitated with Ste5p (21). Thus, these proteins may function as either effectors or regulators of the G␤␥ complex.
Although heterotrimeric G proteins are present among plant, animal, and fungal kingdoms, our current understanding of G protein structure and subcellular localization is based largely on studies of mammalian proteins (reviewed in Refs. 30 -32). Mammals are known to contain over 20 different isoforms of G␣ representing four different classes; there are 5 and 11 known isoforms of G␤ and G␥, respectively. X-ray crystallographic structures have been solved for two G protein heterotrimers (33,34), as well as for the transducin ␣-subunit bound to GDP and to GTP analogs (35) and for the transducin ␤␥ dimer (36). G proteins are bound to the plasma membrane where they mediate a variety of hormonal responses; they also associate with internal membrane compartments where they play a role in membrane trafficking. Stable association of G proteins with membranes involves cooperative interactions among the three subunits as well as lipid modifications of G␣ and G␥ (see Ref. 37). Tight binding of G␣ to the plasma membrane is thought to occur only when it contains myristoyl and palmitoyl modifications and when it is complexed with G␤␥ (37). It is unknown whether G␤␥ requires G␣ for membrane association.
The object of the present study was to initiate biochemical characterization of G␤␥ from an organism that is amenable to genetic analysis. We sought to determine whether the products of the STE4 and STE18 genes form a G␤␥ protein complex that is associated with plasma membranes. We found that approximately 40% of both proteins was bound tightly to plasma membranes, whereas the remainder was associated with internal membranes or was not tightly bound to membranes. Stable association of the G␤␥ complex with plasma membranes required G␣ but did not require the receptor. After detergent extraction of plasma membranes, Ste4p and Ste18p were found to cosediment and to coimmunoprecipitate. The size of the resulting protein complex depended on salt concentration, suggesting that Ste4p and Ste18p interact with other proteins in a salt-dependent manner.

EXPERIMENTAL PROCEDURES
Strains and Plasmids-Strains used in this study are described in Table I. All strains are congenic to strain 381G except for strain JH59. Strain JH59 was constructed by transforming strain W303 (6) with three plasmids (pPgk-Scg, pEL37, and pBH21) that lead to high level expression of all three G protein subunits. High copy plasmid pPgk-Scg (38) contains the URA3 gene; the PGK1 promoter directs synthesis of GPA1 mRNA. The single copy YCp plasmid pEL37 (provided by E. Leberer) contains the HIS3 gene; the divergent GAL1,10 promoter directs synthesis of both GPAl and STE4 mRNA. High copy plasmids pBH21 and M91p1 contain the LEU2 and URA3 genes, respectively; the ADH1 promoter on both plasmids directs synthesis of STE18 mRNA. Plasmid pBH21 was constructed by replacing a 1.2-kilobase BglII fragment of plasmid M91p1 (provided by M. Whiteway) with a 3.5-kilobase BglII fragment from YEp13 carrying LEU2.
Culture Media-YM-1 medium is a rich liquid medium (39). Minimal galactose medium is yeast nitrogen base medium (Difco) buffered to pH 5.8 with 85 mM sodium succinate and supplemented with ammonium sulfate (5 mg/ml) as the nitrogen source, galactose (2%) as the carbon source, and adenine (20 g/ml) and tryptophan (40 g/ml).
Anti-Ste4p antiserum was prepared by immunizing rabbits with full-length Ste4p expressed from plasmid pBH19 in E. coli strain BL21(DE3) (46). Plasmid pBH19 was constructed by cloning the 1.3kilobase NcoI/SalI fragment (containing the STE4 gene) from plasmid pL19 (11) into plasmid pET-8c (46). Antibodies were affinity purified by using a resin that had been coupled to bacterially expressed Ste4p. Anti-Ste18p was prepared by injecting rabbits with the peptide, TS-VQNSPRLQQPQEamide, conjugated with glutaraldehyde to keyhole limpet hemocyanin.
Immunoblotting Methods and Quantitation-Protein samples were diluted 1:3 with sample buffer (1 g of urea dissolved in 1 ml of 17.5 mM Tris-HCl, pH 6.8, 1.75% sodium dodecyl sulfate, 1% ␤-mercaptoethanol, bromphenol blue) and heated to 37°C for 10 min. Proteins were resolved by using SDS-PAGE (47) and transferred (48) to an Immobilon membrane (Millipore Corp., Bedford, MA). Membranes were blocked with 20 mM Tris-Cl, pH 7.5, 0.5 M NaCl, 0.05% Tween 20 containing 5% nonfat dried milk; they were probed with primary antiserum and secondary antibodies diluted in blocking buffer. Secondary antibodies were either goat anti-rabbit or goat anti-mouse immunoglobulins conjugated with horseradish peroxidase (Kirkegaard and Perry Laboratories, Inc., Gaithersburg, MD). Conjugates were visualized by using chemiluminescence reagents (Renaissance kit, DuPont NEN; or SuperSignal kit, Pierce). Autoradiographic results were quantified by using a densitom- Preparation of Cleared Lysates-Unless otherwise indicated, a 150-ml culture was grown overnight at 30°C in YM1 medium (39) to 10 7 cells/ml. Cells were collected by centrifugation, washed twice with ice-cold membrane buffer (10 mM Tris acetate, pH 7.6, 1 mM magnesium acetate, 0.1 mM EDTA, 8% glycerol) containing 1 ϫ protease inhibitors (100 g/ml phenylmethylsulfonyl fluoride, 10 g/ml pepstatin A), and resuspended in 0.5 ml of the same buffer. Glass beads were added, and cells were lysed by mechanical disruption. Unbroken cells were removed from the lysate by centrifugation for 5 min at 330 ϫ g.
Renografin Density Gradients-Membranes from the cleared lysates were resolved on Renografin gradients as described (45).
Plasma Membrane Purification-Strain JH59 was cultured to approximately 5 ϫ 10 7 cells/ml in minimal galactose medium. 12.5 g (wet weight) of cells were rinsed twice with membrane buffer containing 0.1 mM dithiothreitol and 1 ϫ protease inhibitors, suspended in 7.5 ml of the same buffer, and lysed by mechanical disruption with glass beads. Unbroken cells were removed by centrifuging twice at 330 ϫ g for 5 min. 1.1 ml of cleared lysate was mixed with an equal volume of Renografin-76 (to give 38% Renografin). A gradient of Renografin was formed by layering successively 8 ml each of 34, 30, 26, and 22% Renografin solutions (prepared by diluting Renografin-76 with 50 mM Tris-Cl, pH 7.5, 1 mM EDTA). Gradients were centrifuged in an SW28 rotor at 27,000 rpm for 20 h at 4°C. Fractions (2.6 ml) were collected into tubes containing 1 ϫ protease inhibitors and stored at Ϫ70°C. Pooled fractions containing plasma membrane marker, Pma1p, were identified by using immunoblotting methods.
Glycerol Gradient Analysis-Purified plasma membranes from strain JH59 were collected by centrifugation (2.5 h at 250,000 ϫ g) and washed with either buffer A (20 mM Tris acetate, pH 7.6, 1 mM magnesium acetate, 1 mM dithiothreitol, and 0.5 ϫ protease inhibitors) or with buffer B (buffer A containing 250 mM NaCl). Washed membranes were resuspended in 180 l of the same buffer containing 0.2 ϫ protease inhibitors, 2 mg/ml dodecyl-␤-D-maltoside, and 0.4 mg/ml cholesterol hemisuccinate. After 2.5 h on ice, the insoluble material was removed by centrifuging for 3 min in a Beckman airfuge. Bovine serum albumin (3 g), mouse gamma globulin (5 g), and bacterial ATCase (1 g) were added as sedimentation markers. 150 l was loaded onto each of two 8 -30% glycerol gradients containing buffer A and buffer B, respectively, and centrifuged in an SW50.1 rotor at 48,000 rpm for 15 h at 4°C . Twenty-three 220-l fractions were collected into tubes containing protease inhibitors (1 ϫ final concentration); fraction 23 contained the resuspended pellet. Proteins were assayed by using immunoblotting methods.
Immunoprecipitation-Fractions 10, 11, and 12 from the glycerol gradient in Fig. 3A were pooled and centrifuged (16,000 ϫ g) 10 min at 4°C to remove insoluble material. 30 l of supernatant was mixed with 5 l of protein A-Sepharose beads that had been coated with either preimmune or immune serum. After incubating overnight at 4°C, the mixture was centrifuged briefly. Supernatants were removed, and pellets were washed in the same buffer. Supernatants and washed pellets were mixed with sample buffer (47), and Ste18p was detected by immunoblotting methods.
Ste4p Turnover Assay-Cells were cultured at 30°C in YM-1 medium to 10 7 cells/ml. Cycloheximide was added to 10 g/ml. At the various time points, a 25-ml sample was removed to a chilled flask containing NaN 3 and KF (final concentration, 10 mM). Cells were collected by centrifugation, washed with 1 ml of membrane buffer, and resuspended in 0.25 ml of ice-cold membrane buffer. Cleared lysates were prepared as for membrane fractionation. Protein concentrations were determined by using the bicinchoninic acid assay (Pierce), and samples were diluted to 2 mg of protein/ml. Two volumes of 3 ϫ sample buffer (47) was added, and the samples were boiled for 10 min. Proteins were resolved by using SDS-PAGE and detected by using immunoblotting methods. Linearity of the detection was verified by analyzing 2-fold dilutions of the zero time point.

Fractionation of Membranes Containing the Yeast G␤ and G␥
Subunit Homologs-Heterotrimeric G proteins are peripheral membrane proteins that are stably associated with membranes. We wished to determine whether the yeast G␤ and G␥ homologs (Ste4p and Ste18p, respectively) share this property. Membranes, as well as ribosomes and other large macromolecular aggregates, are found in the pellet fraction (particulate fraction) when cleared cell lysates are subjected to centrifuga-tion at 100,000 ϫ g. The solvent conditions that promote dissociation of a specific protein from this particulate fraction provide a criterion that is commonly used to define a peripheral or integral membrane protein. We used immunoblotting procedures to test for the presence of yeast G␤ and G␥ homologs in the soluble and particulate fractions of cleared yeast lysates. Table II shows that Ste4p and Ste18p were extracted from the particulate fraction under similar conditions. When the lysates were prepared in membrane buffer alone, nearly all of Ste4p and Ste18p was associated with the particulate fraction (Table  II). Greater than 75% of the Ste4p and Ste18p was tightly associated with the particulate fraction, that is this portion was resistant to extraction with either 1.0 M NaCl or 4 M urea. Both proteins were efficiently extracted from the particulate fraction by exposing the lysates either to harsh alkaline conditions (Na 2 CO 3 , pH 11) or to a combination of 0.25 M NaCl and the nonionic detergent, dodecyl-␤-D-maltoside. We also found that the G␣ homolog, Gpa1p, was released efficiently with this combination of detergent and NaCl (data not shown). These results suggest that up to 75% of the yeast G␤ and G␥ homologs form a tight peripheral association with membranes in that they were released from the particulate fraction with Na 2 CO 3 , pH 11, but not with 1.0 M NaCl. However, with this technique, it is not possible to distinguish whether the Ste4p and Ste18p in the particulate fraction are associated with membranes or with other non-membranous particles.
We tested whether plasma membranes contain tightly bound G protein subunits by using density gradient centrifugation to fractionate the crude cellular membranes. A gradient of Renografin was layered over a cleared cell lysate that had been adjusted to 38% Renografin (ionic strength, roughly equivalent to 0.5 M NaCl), and the membranes were then allowed to float in a centrifugal field until they reached their buoyant density. We had previously shown that this technique separates plasma membranes from non-membrane proteins and from the more buoyant internal membrane species, i.e. endoplasmic reticulum, Golgi complex, and vacuole (45). The advantages of this fractionation technique are 2-fold. First, by lysing cells rapidly with glass beads and analyzing the crude extracts directly, we minimized potential problems with protein degradation and redistribution of proteins among the cellular membranes. Second, the high ionic strength of Renografin apparently strips the loosely associated proteins from the membranes, maximizing the difference in density between plasma membranes and the internal membranes and thus limiting our analysis to the G a Detergent was a mixture of 2 mg/ml dodecyl-␤-D-maltoside and 0.4 mg/ml cholesterol hemisuccinate (68).
b Although a Ste18p band was evident on the immunoblot, the amount of protein was below the linear range of the assay.
protein subunits that are tightly associated with membranes.
As shown in Fig. 1A, the Ste4p was contained in three different portions of the Renografin gradient. Approximately 40% of the Ste4p (Fig. 1, A and B) cofractionated with plasma membrane ATPase (Pma1p) and with the bulk of the ␣-factor receptor (Fig. 1C). Roughly 30% was found among the more buoyant membranes; the remaining 30% was in the denser fractions that contain none of our membrane markers (Fig. 1B). From a total of eight independent analyses representing three different wild-type strains, we found that a range of 20 -40% of the Ste4p was present among the buoyant membranes, 30 -50% was in the plasma membrane fraction, and 20 -45% in the denser fraction containing nonmembrane proteins. The amount of Ste4p associated with the combined membrane fractions (70%) was somewhat less than the amount of Ste4p found in the particulate fraction (85%) at comparable ionic strength (Table II); thus, it is likely that the particulate fraction contains Ste4p that is not associated with membranes. Consistent with this interpretation, two-thirds of the Ste4p in non-membrane fractions from the Renografin gradient contained Ste4p that was particulate after the fractions had been dialyzed against buffer containing 0.25 M NaCl (not shown). The association of Ste4p with at least two different membrane species (i.e. plasma membranes and the more buoyant internal membranes) apparently does not reflect redistribution of the Ste4p during cell lysis or sample preparation. Each pooled membrane fraction was added to cells containing a deletion of the STE4 gene and then processed as in Fig. 1. In both cases, the membranes migrated to their original position in the second Renografin gradient (not shown).
Previous workers (49) have resolved electrophoretically distinct forms of Ste4p that differ by their state of phosphorylation. We found that about 10% of Ste4p migrates more slowly during SDS electrophoretic analysis, although we have no direct evidence that this species is in fact phosphorylated. In our Renografin density gradient analysis (Fig. 1A), we found that nearly all of this slower-migrating form was associated with the non-membrane fractions (fractions 13-15). Moreover, it was also largely excluded from the particulate fraction in the analysis described in Table II (not shown). A lesser amount of the slower-migrating species was found among the internal membranes (Fig. 1A, fractions 1-6), whereas it was undetectable in the plasma membranes (fractions 7-11). In the presence of ␣-factor, slower migrating species were found in all three regions of the gradient (data not shown).
The fractions from the Renografin gradient depicted in Fig.  1A were assayed for the presence of Ste18p (Fig. 2A). In order to distinguish Ste18p from other cross-reacting proteins, we included control strains that either contained a deletion of the STE18 gene (Fig. 2B) or contained an over-producing plasmid (Fig. 2C). Like Ste4p, a portion of Ste18p was found to be tightly associated with plasma membranes (fractions 7-10) in addition to its association with internal membranes (fractions 2-6) and to the non-membrane fractions (fractions 13 and 14). We also examined Renografin gradients for Gpa1p; it was present in both internal membrane and plasma membrane fractions (not shown). We could not determine whether Gpa1p was in the non-membrane fraction due to the presence of crossreacting proteins.
Solubilized Plasma Membranes Contain a Complex of Ste4p and Ste18p-To obtain biochemical evidence for a Ste4p⅐Ste18p complex, we extracted these proteins from purified plasma membranes with detergents and then examined their sedimentation properties under two different solvent conditions. To maximize detection of the proteins in this experiment, we used a diploid strain containing plasmids that express STE4, STE18, and GPA1 from strong promoters. Renografin gradient analysis of Ste4p and Ste18p from this strain showed a similar profile as observed for the wild-type haploid strain (see Figs. 1 and 2). Plasma membranes were purified by using Renografin gradient centrifugation and then solubilized in buffer containing dodecyl-␤-D-maltoside and cholesterol hemisuccinate in the presence and in the absence of 250 mM NaCl. The detergentextracted proteins were resolved on 8 -30% glycerol gradients that contained the same detergent and salt concentrations. Under both conditions, Ste4p and Ste18p cosedimented (Fig. 3), despite the large difference in their molecular masses (51 and 15 kDa, respectively (5)). When compared with marker proteins, the apparent sedimentation coefficient was 6 S at 250 mM NaCl (Fig. 3A) and 8 S under low salt conditions (Fig. 3B). The complexes depicted in Fig. 3 did not appear to be a consequence of protein overproduction or expression in diploid cells since identical sedimentation coefficients under both low and high salt conditions were obtained for the Ste4p that had been extracted from the plasma membranes of wild-type haploid cells (not shown); however, in these experiments, the concentration Ste18p was below the limit of detection. It is presently unclear whether the larger complex (Fig. 3B) results from dimerization of G␤␥, from increased association of detergent, or from the binding of additional protein factor(s). Although the Gpa1p peak overlapped with Ste4p and Ste18p in 250 mM NaCl (Fig. 3A), the sedimentation coefficient of Gpa1p was not influenced by salt (Fig. 3B). Hence, under the conditions of our assay, we could detect no association of the G␣ subunit homolog, Gpa1p, with the G␤␥ complex.
To verify that the Ste4p and Ste18p on the glycerol gradient (Fig. 3A) were contained in the same complex, we tested for the ability of anti-Ste4p antiserum to precipitate Ste18p. As shown in Fig. 4, Ste18p was precipitated specifically with either anti-Ste4p (compare lanes 1 and 2) or with anti-Ste18p antiserum (compare lanes 3 and 4). Ste4p was not analyzed as it comigrates with immunoglobulin heavy chains on SDS-PAGE.
Ste4p Stability Requires Ste18p-As a test for interactions among the three G protein subunits and the receptor in vivo, we tested whether mutations in the genes encoding these components affect steady-state accumulation of Ste4p and Ste18p. The levels of Ste4p and Ste18p observed for the gpa1 and ste2 mutants were essentially the same as those from the wild-type control strain (Table III); however, the steady-state level of Ste4p in a ste18 mutant was reduced 50%. To ascertain whether this decrease was due to an increased rate of Ste4p turnover or to a reduced rate of synthesis, we examined the rate at which Ste4p disappeared from cultures that had been blocked for protein synthesis. Cycloheximide was added to growing cultures of STE ϩ and ste18 strains, and samples were assayed for the amount of Ste4p that remained in the culture over a 3-h time course (Fig. 5). The kinetics of Ste4p decay in the ste18 mutant were consistent with a first-order reaction and a half-life of 90 min; whereas in the wild-type control, the level of Ste4p (Fig. 5) and Ste18p (not shown) remained undiminished for the duration of the time course. The rate of decay in the ste18 mutant is expected to produce a 2-fold reduction in the steady-state level of Ste4p in growing cultures since Ste4p is also diluted as a result of cell division (90 min doubling time). Ste18p was barely detectable in the ste4 mutant, consistent with rapid decay (Table III)  to the Plasma Membrane-As a second criterion for the presence of specific protein complexes in vivo, we examined whether G␤ requires the receptor or the other G protein subunits for its tight association with plasma membranes. Lysates from ste2, ste18, and gpa1 mutant cells were subjected to Renografin gradient analysis, and the fractions were assayed for Ste4p. In the ste18 mutant, greater than 90% of Ste4p was found in non-membrane fractions (Fig. 6A), indicating that the G␥ homolog is required for stable association of G␤ with both the plasma membrane and the internal membranes. In the gpa1 mutant, Ste4p was present only in internal membrane and non-membrane fractions (Fig. 6B), suggesting that G␣ is required for stable accumulation of G␤ at the plasma membrane. In contrast to ste18 and gpa1, the ste2 mutation did not significantly affect distribution of Ste4p (Fig. 6C), indicating that the receptor is not required for G␤ to accumulate at either membrane location.
Effect of Ste18p on the Aggregation State of Ste4p-As an additional test for interaction between Ste4p and Ste18p, we examined the sedimentation properties of Ste4p in the ste18 mutant. When lysates of the ste18 mutant were analyzed as described in Table II, 84% of the Ste4p protein was found in the particulate fraction, and 65% of this particulate fraction was solubilized by extracting it with a combination of detergent and 0.25 M NaCl. When the solubilized extract was analyzed by using glycerol gradient centrifugation (as described in Fig. 3A), we found a value of 4 S for the sedimentation coefficient of Ste4p from the ste18 mutant (consistent with a 54-kDa Ste4p monomer), whereas we obtained a value of 8 S for the STE18 ϩ control strain under these conditions. Thus, Ste18p is necessary for assembly of Ste4p into a more rapidly sedimenting complex. This complex is likely to contain additional protein components since Ste18p (15 kDa) alone is too small to have an appreciable influence on sedimentation of Ste4p. The observation that the sedimentation coefficient for the wild-type control is larger than the value obtained for purified plasma membranes from diploid cells (Fig. 3A) or from haploid cells (not shown) implies that protein factors associated with G␤␥ are removed during plasma membrane purification.
G␤ Is Not Subject to Receptor-mediated Endocytosis-Previous studies (45,50) have demonstrated that, when haploid a or ␣ cells are exposed to the pheromone from the opposite cell type, the receptors exit the plasma membrane within 20 min and accumulate in the vacuole. If G proteins remain bound to these receptors during the endocytic process, they should also be eliminated from the plasma membrane. We tested this possibility by performing Renografin gradient analysis on cells that had been exposed to ␣-factor in the absence of new protein synthesis. As shown previously (45), the ␣-factor receptor was quantitatively eliminated from the plasma membrane after 20 min of pheromone exposure (Fig. 7). In contrast, pheromone treatment resulted in no major change in the distribution of Ste4p (Fig. 7). Therefore, very little, if any, G␤ is released from the plasma membrane as a consequence of receptor endocytosis or G protein activation.

DISCUSSION
Consistent with the proposed function of Ste4p and Ste18p in the pheromone response pathway, we have found that these G␤ and G␥ homologs form a complex associated with the plasma membrane. First, approximately 40% of Ste4p and Ste18p fractionated with the plasma membrane ATPase in Renografin buoyant density gradients, and association of Ste4p with plasma membranes was dependent upon functional Ste18p. Second, the Ste4p and Ste18p that had been extracted from plasma membranes were found to cosediment under two different solvent conditions, and the Ste18p in these fractions was precipitated in the presence of anti-Ste4p antiserum. Finally, Ste4p and Ste18p were interdependent for their stability in vivo. Levels of Ste18p in ste4 mutant cells were barely detectable; in ste18 mutant cells, Ste4p was reduced in amount and exhibited a more rapid turnover rate. These results represent the first direct evidence for a G␤␥ complex at the plasma membrane. Other biochemical evidence supporting a direct interaction between Ste4p and Ste18p was provided by Song et al. (51) who found that Ste4p and Ste18p coprecipitate with Gpa1p in a guanine nucleotide-dependent manner. Other indirect evidence was provided in a G protein-coupling assay, that is the ability of guanine nucleotides to influence ␣-factor binding is disrupted in both ste18 and STE4 Hpl mutants (2).
Under high salt conditions, the protein complex containing Ste4p and Ste18p sedimented more rapidly than predicted for a simple complex of these two proteins, and this discrepancy was greater under low salt conditions. An increase in the sedimentation rate can be explained by the binding of additional molecules of protein and/or detergent, whereas deviations in shape or hydration would both tend to decrease the rate. It is therefore likely that more protein or detergent binds under low salt conditions than under high salt conditions. Although we have not determined the relative contributions of protein and a Strain DJ602-15-1 served as the wild-type control for strains DJ1006-17-2 and DJ1004-11-1. Strain DJ803-11-1 was the control for DJ803-2-1, and strain DJ147-1-2 was the control for DJ240-4-1.
b Background value from the ste18 mutant was subtracted. detergent to the sedimentation behavior, there are a number of proteins predicted to interact with Ste4p. Candidates include Gpa1p, the pheromone receptor, Ste5p (21,22), Ste20p (22)(23)(24)(25), Cdc24p (26), Akr1p (28,29), and Syg1p (27). Ste5p and the receptor were not expressed in the diploid cells analyzed in Fig.  3. Although Gpa1p was expressed in these cells, its sedimentation rate was not influenced by salt; hence, under the conditions of our assay, Gpa1p does not appear to bind the G␤␥ complex. Ste20p, Cdc24p, Akr1p, and Syg1p are expressed in both haploid and diploid cells; hence, they are potentially components of the protein complex that contains Ste4p and Ste18p (Fig. 3). Ste20p encodes a protein kinase (52); analyses of double mutant strains suggest that Ste20p acts in the pheromone response pathway at a step that is executed after Ste4p (23,24) and before Ste5p (25). Cdc24p is required for budding and for generating cell polarity (53); recent evidence also suggests that it is involved in pheromone signaling (26,54). The ankyrin repeat-containing protein Akr1p is required for normal bud and projection formation and appears to have a negative effect on signaling (28,29). Interestingly, in the two-hybrid system, Akr1p shows an interaction with free G␤␥ but not with the G␣␤␥ heterotrimer (28). Consistent with this requirement for binding, we do not detect Gpa1p in the complexes containing Ste4p and Ste18p. Finally, the truncated form of a putative transmembrane protein, designated Syg1p, has been shown to interact with Ste4p by using the two-hybrid system, and it has also been shown to suppress the lethality of a gpa1 deletion (27). Syg1p is proposed to be a transmembrane signaling component that can respond to or transduce signals through G␤␥.
In mammals, G␤␥ subunits apparently have a broad range of functions as indicated by the variety of proteins with which they interact. In addition to interacting with G␣, G␤␥ subunits appear to bind receptors (55) and several effector proteins including phospholipase C␤ and certain isozymes of adenylyl cyclase (reviewed in Ref. 56). In other examples, G␤␥ complexes are known to promote attenuation of signaling by binding the ␤-adrenergic receptor kinase (57) and by binding phosducin (58). Pumiglia et al. (59) used the two-hybrid system to show that the protein kinase, Raf-1, binds the G␤ 2 subunit, thereby implicating G␤␥ in regulation of the mitogen-activated protein kinase pathway. The small GTPase ARF, a component required for transport of proteins among Golgi compartments, has also been shown to bind G␤␥ (60), supporting the notion that heterotrimeric G proteins are involved in the control of vesicular protein traffic.
As judged from our Renografin gradient analysis, a substantial portion of G␤␥ was not tightly associated with plasma membranes. Approximately 30% of Ste4p and Ste18p was associated with internal membrane fractions (containing membranes of the endoplasmic reticulum, Golgi, and vacuole), while the remaining 30% was not tightly bound to membranes. The biological significance of this fractionation pattern is as yet unclear. Association with different cellular compartments may reflect independent pools of Ste4p and Ste18p or it may reflect the presence of intermediates in the assembly, turnover, or activation of G␤␥. Some, or all, of the Ste4p that was confined to the non-membrane fractions on Renografin gradients may be associated with membranes in vivo. Roughly 25% of Ste4p and Ste18p was released from the particulate fraction in the pres-FIG. 6. Membrane fractionation and localization of Ste4p in ste18, gpa1, and ste2 mutants. Cultures were subjected to Renografin gradient analysis as depicted in Fig. 1. A, ste18 mutant (DJ1006-17-2, Ⅺ) and wild-type control (DJ602-15-1, f). B, gpa1 mutant (DJ803-2-1, Ⅺ) and wild-type control (DJ803-11-1, f). Both strains contained the ste5-3 mutation and were cultured at 34°C; consequently, the gpa1 mutation did not result in constitutive arrest of the cell division. C, ste2 mutant cells (DJ240-4-1, Ⅺ) and wild-type control (DJ147-1-2, f). Amount of Ste4 protein (solid lines) is the percentage of Ste4p present in each fraction; for the ste18 mutant (A), this value was normalized to reflect the reduced level of Ste4p described in Table III. The plasma membrane marker, Pma1p, is indicated by the dashed line for each mutant strain and by the dotted line for each wild-type control. ence of 1.0 M NaCl (Table II); thus, the G␤␥ that was weakly associated with membranes may be released in the presence of 38% Renografin (ionic strength roughly equivalent to 0.5 M NaCl). This material appears, at least in part, to represent a chemically distinct form of G␤␥ since a Ste4p species that migrated more slowly on SDS-polyacrylamide gels was primarily limited to the non-membrane fractions of the Renografin gradient ( Fig. 1) and since it was extracted from the particulate fraction with 1.0 M NaCl (not shown). Interestingly, the ␥ 1 subunit of transducin (G t ) is modified with the C 15 farnesyl group (61), as is Ste18p, and transducin does not require detergent for extraction from the membrane (62). In contrast, the ␥ 2 subunit of G s , G i , and G o contains the C 20 geranylgeranyl lipid, and these G proteins require detergent for extraction (63,64).
Cole and Reed (49) found that phosphorylated forms of Ste4p migrate more slowly on SDS-polyacrylamide gels and that the abundance of these species increases upon exposure to ␣-factor. Furthermore, these modifications are likely to regulate the Ste4p activity since ste4 mutants that block phosphorylation become hypersensitive to pheromones. In our experiments, when a cell cultures were treated with ␣-factor and the membranes resolved on Renografin gradients (not shown), Ste4p was converted to slower-migrating species in plasma membrane, internal membrane, and non-membrane fractions. Assuming that these species represent phosphorylated forms, then the phosphorylated Ste4p that was tightly associated with membranes may represent a transient intermediate that subsequently is either dephosphorylated, degraded, or reduced in its affinity for membranes. Two models are consistent with the presence of phosphorylated forms of Ste4p in both membrane fractions. Protein kinases that become activated upon ␣-factor binding may modify Ste4p molecules both at the plasma membrane and at internal membrane sites; alternatively, Ste4p may be modified at one site and then transported to the other site.
While loss of the receptor did not affect localization of Ste4p, loss of Gpa1p prevented stable association of Ste4p with the plasma membrane. Gpa1p did not influence association of Ste4p with internal membranes. These results suggest that G␣ exerts a physical effect on G␤␥ in vivo; Gpa1p may help to anchor G␤␥ to the plasma membrane. Lipid modifications of G␣ and G␥ are thought to mediate (at least in part) the membrane attachment of G proteins (see Ref. 37). It has been proposed (37) that the G␣ subunit facilitates binding of G␤␥ to membranes by supplementing the low binding energy provided by the G␥ isoprenyl group. Ste18p appears to be farnesylated since its sequence contains a consensus site for farnesylation and since mutants affecting that site or the farnesyl transferase enzyme affect signal transduction (15,20,65,66). Gpa1p is both myristoylated (14) and palmitoylated. 2 In a recent related paper focusing on N-myristoylation of Gpa1p, Song et al. (51) report on a number of observations that pertain to subcellular localization of G␤␥. Some of their results differ from ours and provide possible insights for properties of G␤␥ and limitations of the two approaches. Using sucrose gradient fractionation, Song et al. (51) could detect Ste4p only in fractions containing plasma membranes. Lack of Ste4p in the non-membrane fractions is likely to reflect the lower salt conditions used in their assay. In Table II, we found that a portion of Ste4p was associated with the particulate fraction only under low salt conditions. Perhaps, this species is associated weakly with the plasma membrane and suggests the presence of at least two populations of G␤␥ species at this location. The inability of Song et al. (51) to detect Ste4p in other membrane fractions may have resulted from strain differences, from a failure to resolve these membrane species from the plasma membrane, or from degradation of Ste4p during spheroplast formation. In gpa1 mutants, Song et al. (51) found that only a portion of the Ste4p was resolved from plasma membranecontaining fractions, whereas in the present study, we found that essentially all of the tightly associated Ste4p was resolved from the plasma membrane. These results may reflect either the presence of a Ste4p species that is weakly associated with plasma membranes or insufficient resolution of the sucrose gradient technique. When we examined the sedimentation behavior of detergent-solubilized G␤␥ on glycerol gradients, we found no evidence for association with Gpa1. In contrast, Song et al. (51) found that at least a portion of Ste4p and Ste18p from whole cell extracts coprecipitates with the Gpa1p that had been 2  tagged with glutathione S-transferase. The apparent absence of such a complex in our study may reflect the differences in the assay conditions or the fact that we limited our analysis to purified plasma membranes.
Although a number of proteins have been proposed to interact with Ste4p, the mechanism by which G␤␥ activates subsequent events in the pheromone response pathway remains unknown. The activity of G␤␥ is thought to be stimulated upon its dissociation from G␣, that is when occupied receptors stimulate guanine nucleotide exchange or when G␣ is inactivated by a mutation in the GPA1 gene. When cells were treated with a saturating concentration of ␣-factor for 20 min, we found that cellular distribution of Ste4p was not altered discernibly. As this is sufficient time to provide maximal response to ␣-factor, it appears unlikely that activation of the response pathway requires movement of G␤␥ from the plasma membrane to another cellular compartment or that removal of G␤␥ from the plasma membrane regulates its activity. However, we cannot rule out the possibility that the exit of a small quantity of G␤␥ in response to ␣-factor is sufficient for maximal activation. In contrast, gpa1 mutant cells failed to accumulate G␤␥ that was tightly bound to the plasma membrane, thus cells that lack Gpa1p are not in the same physiological state as wild-type cells treated with pheromone. G␣ may retain some contacts with G␤␥ during ␣-factor stimulation; G␣ may facilitate the assembly of G␤␥ into a structure that remains tightly bound to the plasma membrane when activated by ␣-factor, or chronic stimulation of G␤␥ in the gpa1 mutant may disrupt its stable contacts with the plasma membrane. Even though G␤␥ is no longer tightly associated with the plasma membrane in the gpa1 mutant, the requisite events for signal transduction may be mediated by G␤␥ molecules that are only weakly associated with the plasma membrane.