Guanine Nucleotide-binding Protein (Gα) Endocytosis by a Cascade of Ubiquitin Binding Domain Proteins Is Required for Sustained Morphogenesis and Proper Mating in Yeast*

Background: The yeast Gα protein contains a unique domain that is monoubiquitinated, leading to vacuolar degradation. Results: A gene deletion screen reveals ubiquitin binding domain proteins necessary for Gα trafficking. Loss of the ubiquitination domain impedes cellular morphogenesis and mating. Conclusion: Proper endocytosis of Gα is required for sustained morphogenesis and efficient mating. Significance: Gα endocytosis promotes signaling. Heterotrimeric G proteins are well known to transmit signals from cell surface receptors to intracellular effector proteins. There is growing appreciation that G proteins are also present at endomembrane compartments, where they can potentially interact with a distinct set of signaling proteins. Here, we examine the cellular trafficking function of the G protein α subunit in yeast, Gpa1. Gpa1 contains a unique 109-amino acid insert within the α-helical domain that undergoes a variety of posttranslational modifications. Among these is monoubiquitination, catalyzed by the NEDD4 family ubiquitin ligase Rsp5. Using a newly optimized method for G protein purification together with biophysical measures of structure and function, we show that the ubiquitination domain does not influence enzyme activity. By screening a panel of 39 gene deletion mutants, each lacking a different ubiquitin binding domain protein, we identify seven that are necessary to deliver Gpa1 to the vacuole compartment including four proteins (Ede1, Bul1, Ddi1, and Rup1) previously not known to be involved in this process. Finally, we show that proper endocytosis of the G protein is needed for sustained cellular morphogenesis and mating in response to pheromone stimulation. We conclude that a cascade of ubiquitin-binding proteins serves to deliver the G protein to its final destination within the cell. In this instance and in contrast to the previously characterized visual system, endocytosis from the plasma membrane is needed for proper signal transduction rather than for signal desensitization.

To fully understand the consequences of G protein trafficking, we must first understand how such trafficking events are regulated. Much of the spatial control of G protein-coupled receptors and G proteins is dependent upon post-translational modification by monoubiquitination. For example, the yeast G␤ (Ste4) and G protein-coupled receptor (Ste2) are monoubiquitinated after stimulation with the mating pheromone ␣-factor (16 -18). Both proteins are subsequently removed from the plasma membrane and delivered to the vacuole (17)(18)(19). For Ste2 this process is mediated in part by endocytic adaptor proteins containing a ubiquitin binding domain (UBD) (21,22). Although structurally diverse, UBDs share the ability to bind non-covalently to ubiquitin-conjugated substrates and serve to transport monoubiquitinated proteins through the various stages of endocytosis (23). Although the general process of monoubiquitination-mediated endocytosis is well understood, questions remain concerning the specific protein components that are important for trafficking of the G protein.
The yeast model system, in which many ubiquitination and cellular trafficking components were originally identified, can facilitate understanding of the interconnections between G protein signaling, monoubiquitination, and trafficking (24). Pheromone binds to Ste2, which is coupled to Gpa1 (G␣) and Ste4/Ste18 (G␤␥). When G␣ is activated, G␤␥ dissociates and stimulates a kinase cascade leading to activation of the yeast MAPK Fus3. A second branch of the pathway leads to activation of a Rho family GTPase, Cdc42 (24). Together these processes result in cell cycle arrest, new gene transcription, and morphological changes that facilitate mating (24). Pheromone pathway signaling is attenuated by the pheromone-dependent monoubiquitination and endocytosis of the G protein-coupled receptor Ste2 (16). As with the receptor, monoubiquitination leads to the endocytosis and eventual vacuolar degradation of Gpa1 (25). In contrast to Ste2, monoubiquitination of Gpa1 is not dependent on pheromone stimulation. Both proteins are monoubiquitinated by the same ubiquitin ligase, Rsp5 (26,27).
Although the Ras-like and ␣-helical domains are highly conserved across species, Gpa1 possesses a unique 109-amino acid insert (ubiquitination domain (UD)) within the ␣-helical domain. The UD contains the known sites of phosphorylation as well as the primary residue for both monoubiquitination and polyubiquitination (25, 26, 28 -31). Given that the ␣-helical domain modulates the activity of some G proteins, we considered whether the UD could influence the activity as well as the cellular distribution of Gpa1. Here we show that the UD does not influence G protein catalytic activity or downstream MAPK signaling but is needed for proper trafficking of Gpa1 to the vacuole. Our screen of 39 gene deletion mutations revealed seven UBD-containing proteins that are required for Gpa1 trafficking. Four of these proteins are required for trafficking of Gpa1 but not Ste2, thus demonstrating that constitutive endocytosis of these proteins occurs by distinct mechanisms. Finally, we show that G␣ endocytosis from the plasma membrane is required for sustained cellular morphogenesis and efficient mating.

EXPERIMENTAL PROCEDURES
Strains, Plasmids, and Growth of Cultures-Standard procedures were followed for the growth, maintenance, and transformation of yeast and bacteria. Proteins for biochemical studies were expressed in Escherichia coli BL21 (DE3) RIPL cells (Stratagene, La Jolla, CA) grown at 18°C overnight after induction with isopropyl ␤-D-1-thiogalactopyranoside. The yeast (S. cerevisiae) strain used in this study was BY4741 (MATa leu2⌬ met15⌬ his3-1 ura3⌬) and its derivatives. Cells were grown at 30°C in yeast extract peptone medium or synthetic complete medium containing 2% (w/v) dextrose. The gpa1 ⌬UD mutant strain was generated by transformation of AflII-digested pRS406-GPA1_trunc ⌬UD in wild type cells. In each case transformed cells were grown in synthetic complete medium that lacked the appropriate nutrient or yeast extract peptone medium with the appropriate drug.
Plasmid Construction-Both the native (scGpa1) and codonoptimized Gpa1 (coGpa1) were cloned into the bacterial expression vector pQlinkH (Addgene), which contains an N-terminal tobacco etch virus (TEV) protease-cleavable His 6 tag, using the BamHI and NotI restriction sites. Efficient TEV protease cleavage required the use of Gpa1 ⌬N lacking the first 38 amino acids, which are predicted to be unstructured based on alignment with G␣ i (32). pQlinkH-GPA1 ⌬N⌬UD was constructed by deleting the coding sequence corresponding to amino acids 128 -236 from the pQlinkH-GPA1 ⌬N plasmid. The pRS406-STE2-GFP integrating vector was constructed by PCR amplification of ϳ900 base pairs from the 3Ј end of STE2 inframe with GFP from the yeast GFP strain collection (Invitrogen). Genomic DNA from the above strain was amplified with flanking KpnI and SacI sites for introduction to pRS406. The pRS406-GPA1_trunc ⌬UD integrating vector was constructed by PCR amplification of GPA1 ⌬UD from pRS406-GPA1 ⌬UD -GFP minus the first 81 base pairs of GPA1. The PCR product was flanked by XbaI and SacI sites for introduction to pRS406. Site-directed mutagenesis was used to introduce a silent mutation at bp 181 to generate an AflII site for integration. Construction of the pRS406-GPA1-GFP, pRS406-GPA1 ⌬UD -GFP, pRS423-FUS1-lacZ, and pADM4-GPA1 vectors has been described previously (25,33,34).
Bacterial Protein Expression and Purification-Generating the quantities of pure Gpa1 necessary to complete biophysical studies has been a barrier to progress both because of low initial protein expression and impurities, necessitating additional purification steps. We optimized the process using small-scale batch purification and TEV protease-cleavage to release protein from the nickel beads (Table 1). E. coli were lysed by homogenization (NanoDeBee) at 1000 p.s.i. in 25 mM potassium phosphate buffer, pH 7.0, with 300 mM KCl and 250 M tris(2-carboxyethyl)phosphine. After clarification by centrifugation, the lysate from 500 ml of cell culture was bound to 1 ml of nickel-nitrilotriacetic acid-agarose bead slurry (Qiagen) preequilibrated in phosphate buffer, pH 7.0, for 20 min at 4°C. The beads were washed 3 times at 4°C with 25 mM phosphate buffer, pH 7.0, with 300 mM KCl, then 3 times with Gpa1 buffer (25 mM phosphate buffer, pH 7.0, and 100 mM KCl) with 100 M GDP and 500 M tris(2-carboxyethyl)phosphine. Gpa1 was cleaved from the beads by incubation at 4°C overnight with 500 g of TEV protease. The final product was judged Ͼ95% pure by SDS-PAGE. Protein was stored at 4°C (never frozen) and used within 2 days. Although the use of TEV protease cleavage to release Gpa1 decreased the yield from the first purification step (about half of the Gpa1 remains on the beads), this loss in efficiency is mitigated by the increase in purity (does not require further dialysis or purification steps). Furthermore, we found that 46% of the codons in Gpa1 are rarely used in E. coli, which can significantly reduce the efficiency of expression (35). A codon-optimized Gpa1 yielded 18 mg/liter, an increase of 11-fold over the original method.
Circular Dichroism-Gpa1 was diluted to 5 M in Gpa1 buffer with 50 M GDP, 50 M MgCl 2 , and 550 M tris(2-carboxyethyl)phosphine. Spectra were recorded from 190 to 260 nm on a Chirascan plus CD spectrometer at 25°C using a 1-mm quartz cell. Buffer background was subtracted from the spectra.
Quantitative Mass Spectrometry of ABD-labeled Cysteines-Gpa1 ⌬N was diluted to 2 M in Gpa1 buffer with 50 M GDP or GTP␥S. Protein sample was mixed with 4-fluoro-7-aminosulfonylbenzoflurazan (ABD-F, Anaspec) on ice (final concentration, 2 mM). 20-l aliquots of the sample were reacted for 3 min at either 42°C or 70°C and then transferred to ice to quench ABD-F labeling. Samples were prepared for mass spectrometry and analyzed as described by Isom et al. (36). Data collected on a Nano-Acquity HPLC solvent delivery system (Waters Corp.) connected through an electrospray ionization source interfaced to an LTQ Orbitrap Velos ion trap mass spectrometer (Thermo Electron Corp.).
Thermal Stability of Gpa1-The fast quantitative cysteine reactivity method (37) was employed to measure Gpa1 thermal stability. Briefly, 2 M protein was incubated with 1 mM ABD-F in Gpa1 buffer in the presence of 20 M GDP or GTP␥S and 2 mM MgCl 2 at the desired temperature for 3 min. The reaction was quenched with 20 mM HCl (final concentration), and ABD-F fluorescence was measured on a PHERAstar plate reader (BMG Labtech, excitation at 400 nm and emission at 500 nm). The data were normalized and fit using GraphPad Prism (GraphPad Software; San Diego, CA) to determine the temperature at which half the protein was unfolded, representing the melting temperature (T m ).
Nucleotide Dissociation Assay-Purified Gpa1 was exchanged into Gpa1 buffer with 50 M MgCl 2 and 100 M GDP. To initiate association, 1 M 2Ј-O-(N-methylanthraniloyl)guanosine 5Ј-diphosphate (MANT-GDP) was added to 1 M protein at room temperature. Gpa1 was determined to be fully loaded when the fluorescence intensity reached a maximum (ϳ250 s). Association was measured as a change in fluorescence intensity over time (excitation, 360 nm; emission, 440 nm) using a LS50B PerkinElmer Life Sciences Luminescence Spectrometer. MANT-GDP dissociation was initiated by the addition of 500 M unlabeled GDP. Fluorescence data were fit in GraphPad Prism to a one-phase exponential association or decay curve.
Intrinsic GTP Binding and Hydrolysis-Purified Gpa1 (200 nM) was equilibrated to room temperature in Gpa1 buffer with 50 M MgCl 2 . GTP at a final concentration of 200 nM was added, and GTP binding and hydrolysis were monitored by the change in intrinsic fluorescence of Gpa1 that occurs upon rearrangement of the tryptophan near the nucleotide binding region (excitation at 284 nm and emission at 340 nm) using a LS50B PerkinElmer Life Sciences Luminescence Spectrometer. Data were fit to exponential association or dissociation curves using GraphPad Prism.
Microscopy Screen for Gpa1 and Ste2 Trafficking-BY4741derived mutants lacking specific ubiquitin binding domaincontaining proteins (disrupted using the KanMX G418 resistance marker, from Research Genetics, Huntsville, AL) were used for the screen. PEP4 was disrupted in the deletion mutant strains by single-step gene replacement with pep4::HIS3 using PCR-mediated gene disruption (38). GPA1-GFP was introduced in each strain by transformation of pRS406-GPA1-GFP after digestion with HindIII (25). Separately, STE2-GFP was introduced in the deletion mutants by transformation of pRS406-STE2-GFP (this study) after digestion with AfeI. Screen hits (deletion mutants that mislocalized Gpa1-GFP) were re-made in BY4741 and BY4741-derived strains by gene replacement with KanMX4 using PCR and pFA6KanMX as the template (39). All deletions were verified by PCR and nucleotide sequence analysis and analyzed for Gpa1-GFP localization.
Transcriptional Reporter Assays-Pheromone sensitivity was measured by a transcriptional reporter assay (33) as described earlier. Briefly, cells transformed with pRS423 FUS1-LacZ were stimulated with different concentrations of pheromone for 90 min. ␤-Galactosidase activity was measured spectrophotometrically (750 nm) using fluorescein di-␤-D-galactopyranoside as the substrate.
Time-lapse Imaging in Microfluidics Chambers-A microfluidics device similar to the one described earlier was constructed (42). Early-mid log phase yeast cells were loaded in the chamber and stimulated with the indicated concentration of ␣-factor pheromone. Differential interference contrast images were acquired every 5 min for 8 h using a 60ϫ PlanApo objective under oil immersion. Images were captured using an Olympus Spinning disc confocal microscope equipped with a motorized XYZ stage and photometrics EM CCD camera. Image processing and analysis was done using ImageJ software.
Quantitative Mating Efficiency Assay-The mating assay was performed as previously described (43). Equal amounts of earlylog phase MATa cells (BY4741) and MAT␣ cells (BY4742, leu2⌬ his3⌬ ura3⌬ lys2⌬) were mixed, sterile-filtered onto nitrocellulose membranes, and overlaid onto yeast extract peptone medium plates at 30°C for 4 h. Cells were diluted and plated onto synthetic complete medium and synthetic dextrose medium with only His, Ura, and Leu to select for diploids. Mating efficiency was calculated by dividing the number of diploid colonies by the total number of colonies on the synthetic complete medium plate.

Structural Contributions of the Ubiquitination Domain to
Gpa1-Gpa1 is distinct from other G␣ proteins because of the UD. A comparison of available genome sequences shows that the UD is found among Saccharomycotina but is not present in other eukaryotes ( Fig. 1A and supplemental Fig. 1). Based on sequence alignment with G␣ proteins, the UD is at the end of the A/B helix in the ␣-helical domain (Fig. 1B). Previous analysis has shown that the UD is the site of post-translational modifications including phosphorylation, monoubiquitination, and polyubiquitination (25, 26, 28 -31).
To be modified, amino acids must be accessible to the modifying enzyme. The ability to modify a given amino acid requires that the residue be present at the surface of the protein or in a region lacking secondary structure. Indeed, phosphorylation and ubiquitination often occur on intrinsically disordered regions of proteins (44). Therefore, we postulated that the UD is without substantial secondary structure. To test our hypothesis we employed a series of biophysical measurements of protein conformation. First, we used circular dichroism to determine the contribution of the UD to the secondary structure content of Gpa1. This analysis revealed no difference between Gpa1 with or without the UD (Fig. 1C). Precise removal of the UD resulted in expression comparable with that of the full-length protein, whereas removing just one additional amino acid abolished expression entirely. We infer that the UD does not contribute to the structure of the adjoining helical domain.
To further assess whether the UD contains unfolded regions, we asked whether resident cysteines were especially accessible to labeling with a small modifier ABD-F (36). Gpa1 is an ideal candidate for this method because it contains cysteines dispersed throughout the protein. There are two in the Ras-like domain (Cys-333 and Cys-443), four in the helical domain (Cys-105, Cys-117, Cys-258, and Cys-288), and one in the UD (Cys-208). Using this method, we detected ABD-F labeling for five of the seven cysteines. As seen in Fig. 1D, cysteines within the well folded helical domain are highly protected, whereas the cysteine in the UD is highly exposed (value close to one). These results are consistent with the circular dichroism data, suggesting that the UD lacks secondary structure. Taken together, our results are consistent with the hypothesis that the UD is an evolutionarily unique, structurally distinct, and largely unstructured domain.
Functional Contributions of the Ubiquitination Domain to Gpa1 Activity-The ␣-helical domain is known to influence G protein nucleotide exchange activity (45). The UD is adjacent to the A/B helix of the helical domain, a region whose dynamics were previously shown to promote the rapid, receptor-independent, nucleotide exchange activity of Arabidopsis thaliana G␣ (45). However, it is not known whether the UD regulates the enzymatic activity of S. cerevisiae Gpa1. To address this question we used the cysteine reactivity method to measure the thermal stability of Gpa1 in the presence and absence of the UD. As shown in Fig. 2A, removing the UD did not alter Gpa1 thermal stability despite the loss of 109 amino acids. We then measured the ability of guanine nucleotides to bind Gpa1 in the presence and absence of the UD. For these experiments we used the fluorescent nucleotide analog MANT-GDP. The rate of nucleotide exchange was measured as a loss of fluorescence over time as Gpa1 released MANT-GDP and bound unlabeled GDP, added in excess. As shown in Fig. 2B, the rate of nucleotide dissociation was not altered by the absence of the UD. Finally, we measured the rate of GTP hydrolysis using a method that monitors a nucleotide-dependent change in the intrinsic fluorescence of Gpa1 (46). Using this assay, we observed similar rates of GTP hydrolysis in the presence and absence of the UD (Fig. 2C). Together, these data suggest that the UD does not alter the structure or enzymatic activity of Gpa1.
A Cascade of Ubiquitin Binding Domain Proteins Transport Gpa1 to the Vacuole-The data presented above indicate that the UD does not directly contribute to the structure and enzymatic function of Gpa1. However, the UD functions as a site of phosphorylation, polyubiquitination (for targeting to the proteasome), and monoubiquitination (for trafficking to the vacuole) (25, 26, 28 -31). Gpa1 is monoubiquitinated by the Rsp5 ubiquitin ligase (26). However, the proteins involved in Gpa1 endocytosis and specifically those that recognize monoubiquitin within the UD are not known. Accordingly, we embarked on a search for proteins that recognize the monoubiquitinated form of Gpa1.
Monoubiquitin-mediated trafficking pathways are often composed of ubiquitin binding domain (UBD)-containing proteins (47). In most cases, multiple UBD-containing proteins are involved in the passage of a monoubiquitinated protein to its final destination (23). Moreover, the route of endocytosis appears to be dictated in part by the identity of the monoubiquitinated protein (23). Numerous UBDs have been previously defined (48) and are now included in the protein descriptions found in the Saccharomyces Genome Database (SGD; www. yeastgenome.org). Using UBD proteins known at the start of this project, we screened a comprehensive set of gene deletion strains for endocytosis and trafficking defects affecting the delivery of Gpa1 from the plasma membrane to the vacuole. Because Gpa1 and Ste2 are targeted for internalization by the same ubiquitin ligase, we also examined trafficking of Ste2, with the expectation that both proteins share the same trafficking machinery components. Accordingly, we monitored the localization of GFP-tagged Gpa1 (Gpa1-GFP) and Ste2 (Ste2-GFP) in a total of 39 UBD deletion strains (49,50). Gpa1-GFP was introduced in a strain lacking the master vacuolar protease (pep4⌬) so as to preserve the GFP signal after delivery to the vacuole (25). Ste2-GFP was expressed from the native STE2 locus.
As a proof of concept we first monitored GFP localization in the absence of Vps23, Vps27, Vps9, or Rpn10. Major trafficking defects were observed for both Gpa1-GFP and Ste2-GFP in the absence of Vps23 or Vps27, two known components of the ESCRT (Endosomal Sorting Complexes Required for Transport) machinery (Fig. 3) (51, 52). Trafficking defects were also observed for Ste2-GFP after deletion of Vps9, a protein required for efficient endocytic trafficking of proteins (53). Finally, deletion of the polyubiquitin binding proteasomal protein Rpn10 did not affect trafficking of the receptor or G protein (Fig. 3), indicating that the screen specifically monitors the fate of monoubiquitinated proteins.
Having validated the screening approach, we conducted a more comprehensive screen of the known UBD-containing proteins in yeast (Fig. 4A). Our general strategy was to look for UBD mutants that prevent accumulation of GFP in the vacuole. As shown in Fig. 4B, vacuoles were visualized by pulse-staining with an amphiphilic styryl dye (FM4-64, red), which initially incorporates into the plasma membrane and gradually accumulates in the lumen of the vacuole (54). In wild type cells Gpa1-GFP was present at the plasma membrane but not in vacuoles, consistent with our previous findings (25) (Fig. 4B, left ). In pep4⌬ cells Gpa1-GFP accumulated in the vacuole, indicative of monoubiquitin-dependent vacuolar translocation (Fig. 4B,  middle). Of the 39 UBD deletion strains tested, seven exhibited Gpa1-GFP mislocalized to puncta within the cytoplasm (Fig.  4B, right). Major trafficking defects were observed in the absence of Vps36, a component of the ESCRT machinery (51). Additionally, defects in Gpa1 localization were seen upon deletion of the following: Bul1, the ubiquitin binding component of the Rsp5 ubiquitin ligase (55); Ede1, a component of the early endocytic machinery (56); Ddi1, a DNA damage inducible v-SNARE-binding protein regulating exocytosis (57); Rup1, a regulator of Rsp5 (58) (Fig. 4C). Defects in Ste2-GFP trafficking were likewise observed in the absence of Vps36. However, four of the UBD deletions that affected Gpa1 trafficking had no effect on Ste2 endocytosis (Fig. 4D). In total, we identified seven UBD proteins necessary for proper vacuolar delivery of Gpa1 (Figs. 3 and 4). Deletion of four of these proteins disrupted trafficking of Gpa1 but not Ste2, indicating that their trafficking pathways differ. Alternatively, there could be a high degree of functional redundancy among UBDs in regulating Ste2, but not Gpa1, trafficking.
Gpa1 Is Delivered to the Plasma Membrane in the Absence of UBD Proteins Involved in Endocytosis-Having identified mutants that alter the cellular distribution of Gpa1, we next considered whether the defect was due to removal from, or delivery to, the plasma membrane. That is, Gpa1 mislocalization could be caused by impaired synthesis, maturation, or delivery to the plasma membrane. To rule out this possibility we compared the localization of Gpa1 with that of Gpa1 ⌬UD , which in wild type cells is delivered to the plasma membrane but not the vacuole (25). For all the UBD mutants tested, we found that Gpa1 ⌬UD -GFP was localized normally to the plasma membrane (Fig. 5). These data indicate that the UBD-containing proteins identified in our screen are specifically required for trafficking of monoubiquitinated Gpa1 from the plasma membrane to the vacuole.
Disruption of Gpa1 Trafficking Promotes Accumulation of Ubiquitinated Gpa1-Post-translational modification by monoubiquitination and polyubiquitination are used to regulate the quantity of Gpa1 present at the plasma membrane. Whereas monoubiquitinated Gpa1 is delivered to the vacuole, where it is eventually degraded, polyubiquitinated Gpa1 is recruited to the proteasome, where it is likewise degraded. We reasoned that disruption of Gpa1 trafficking might allow the  monoubiquitinated protein to accumulate and perhaps undergo additional rounds of ubiquitination. To test our hypothesis we monitored the level of Gpa1 ubiquitination in the four UBD deletion mutants that exhibited trafficking defects for Gpa1 (but not Ste2). To visualize the rare monoubiquitinated species we overexpressed Gpa1 using a multicopy plasmid (28). As expected we observed an increase of monoubiquitinated Gpa1 in three of the four UBD deletions (ede1⌬,  ddi1⌬, and rup1⌬). The increase was similar to that seen upon deletion of PEP4, consistent with a defect in vacuolar degradation of Gpa1 (Fig. 6). Additionally, we observed an increase in polyubiquitinated Gpa1, particularly upon deletion of EDE1 (Fig. 6). The increase in monoubiquitinated Gpa1 was not due to an overall increase in the amount of loaded substrate protein, as evident from shorter exposures of the blot. These data suggest that disruption of Gpa1 trafficking slows the clearance of monoubiquitinated Gpa1 and drives the system toward polyubiquitination (26). ; UEV, ubiquitin-conjugating enzyme E2 variant. B, general scheme of the genetic screen to identify binding partners of Gpa1-ubiquitin. A total of 39 deletion mutants, each lacking a specific UBD-containing protein, were separately transformed with plasmids encoding either Gpa1-GFP (pRS406-GPA1-GFP) or Ste2-GFP (pRS406-STE2-GFP). Gpa1 and Ste2 localization was monitored by differential interference contrast and fluorescence microscopy in the presence of a vacuolar dye (red, FM4-64). In the absence of endocytosis the proteins were retained at the plasma membrane (PM; green, left). After delivery to the vacuole (Vac), the proteins colocalized with FM4-64 (yellow, middle). If trafficking was disrupted at any point, the proteins were mislocalized (Misloc) to puncta in the cytoplasm (right). Microscopy of mutants that mislocalized Gpa1-GFP (C) but not Ste2-GFP (D) is shown. Mutant vps36⌬ is shown as an example of a UBD that affects trafficking of both Gpa1 and Ste2. Scale bar, 5 m.  absence (bul1⌬, ede1⌬, ddi1⌬, vps36⌬, vps23⌬) of the indicated UBD proteins. Vacuoles are stained with FM4-64. Scale bar, 5 m. Despite repeated attempts, we were unable to delete RUP1 in the Gpa1 ⌬UD background.

Retention of Gpa1 ⌬UD at the Plasma Membrane Inhibits
Proper Morphogenesis and Mating-Previous studies revealed that persistent GTP activation leads to an accumulation of Gpa1 at endosomes (12). Conversely, removal of the UD leads to an accumulation of Gpa1 at the plasma membrane (25). With the availability of the Gpa1 ⌬UD mutant, as well as UBD mutants that disrupt translocation of the wild type protein, we have the ability to study the functional consequences of Gpa1 trafficking without the confounding effects of altered GTPase activity. Accordingly, we compared the pheromone response in wild type cells and mutants defective in Gpa1 trafficking using MAPK activation and transcription reporter assays. In the case of Gpa1 ⌬UD , we replaced the wild type gene with the UD mutant via PCR-mediated integration. This was done at the endogenous GPA1 locus so as to maintain proper G protein expression and subunit stoichiometry. As shown in Fig. 7A, there was no significant change in MAPK activation in the UBD mutants (bul1⌬, ede1⌬, ddi1⌬, and rup1⌬) when compared with wild type cells. Furthermore, we observed no significant differences when comparing the Gpa1 ⌬UD mutant and wild type protein. Similarly, we observed only modest differences in the pheromone transcription response (Fig. 7B). Taken together, these data indicate that the MAPK branch of the pheromone pathway is largely unaffected by the trafficking-deficient mutants.
In addition to the MAPK and transcription response, the pheromone pathway has a second branch that activates the small G protein Cdc42. Active Cdc42 promotes polarization and morphogenesis to facilitate mating (59). We, therefore, asked whether defects in Gpa1 trafficking had any effect on cellular morphogenesis (shmoo formation). We monitored each mutant in a microfluidic chamber with constant pheromone stimulation, as previously described (42). At a saturating concentration of pheromone, wild type cells can form up to three mating projections. As shown in Fig. 7C there were no differences between wild type and bul1⌬, ddi1⌬, or rup1⌬ mutants. Cells lacking Ede1 did exhibit an increased number of projections, as compared with wild type, but this same phenotype was evident in an ede1⌬ Gpa1 ⌬UD double mutant, indicating that the ede1 phenotype is not the result of impaired Gpa1 trafficking. In contrast to all of the other strains tested, cells that express Gpa1 ⌬UD were able to form just a single mating projection (Fig. 7C). These cells formed a shmoo and are subsequently elongated. Taken together these results suggest that Gpa1 must be endocytosed for proper morphogenesis to occur. Because the same morphological defect was not observed in the UBD deletion mutants, it is the first step of removing Gpa1 from the plasma membrane that is necessary for the formation of additional mating projections.
Finally, previous reports indicated that multiple projections are necessary for optimal mating efficiency (60). Accordingly, we observed a 2-fold higher mating efficiency in wild type as compared with the Gpa1 ⌬UD mutant strain (Fig. 7D). Thus the cellular morphogenesis defects exhibited by ubiquitin-deficient Gpa1 ⌬UD mutants leads to reduced mating fitness. Because we have shown that removal of the UD does not alter Gpa1 enzymatic activity or MAPK signaling, we conclude that monoubiquitination and removal from the plasma membrane contributes to proper morphogenesis and mating.

DISCUSSION
Here we have used biochemical, biophysical, genetic, and cell biological approaches to demonstrate a positive signaling role for G␣ endocytosis. Taking advantage of the yeast system, we identified seven ubiquitin binding domain-containing proteins required for constitutive internalization of the G protein. Three of these proteins act on Ste2 as well as Gpa1. Four others act on Gpa1 alone, demonstrating that endocytosis of the G protein and receptor are distinct processes regulated by distinct binding partners. We further show that the Gpa1 UD is targeted by the UBD proteins. In the absence of the ubiquitination domain, Gpa1 is retained at the plasma membrane, and cellular morphogenesis is curtailed. A second effector pathway leading to MAPK activation is unaffected.
The proteins specifically required for Gpa1 trafficking (Rup1, Ddi1, Bul1, and Ede1) act at different stages of endocytosis (Fig.  8). Rup1 binds to Rsp5 and stimulates Rsp5 autocatalysis and substrate ubiquitination (58). We suggest that Rup1 may promote Rsp5-mediated ubiquitination of particular membrane proteins. Bul1 is the ubiquitin binding component of the Rsp5 ubiquitin ligase (55). Curiously, we find that Bul1, but not the closely related protein Bul2, promotes Gpa1 endocytosis. Even though Bul1 and Bul2 are homologous, they appear to have distinct functions within the cell. For example, Bul1 is a negative regulator of the tryptophan permease Tat2, but Bul2 is not (61,62). Ddi1 is a DNA damage-inducible v-SNARE-binding protein (57) that was previously shown to be a negative regulator of the late secretory pathway, a function that is independent of the UBA domain of Ddi1 (63). Finally, Ede1 is a member of the epsin family of adapter proteins and a component of the early endocytic machinery (56,61,64,65). Deletion of Ede1 but not other epsin proteins (Ent1, Ent2) leads to Gpa1 mislocaliza- tion. Although Ede1 is implicated in Ste2 trafficking, we did not observe Ste2 mislocalization in the absence of Ede1. Ste2 trafficking has previously been monitored in agonist-stimulated cells. In contrast, we measured Gpa1 and Ste2 trafficking in unstimulated cells. Thus it is possible that Ede1 is important for pheromone-dependent, but not basal, trafficking of Ste2.
It is of note that when endocytosis of Gpa1 is inhibited (rup1⌬, ddi1⌬, ede1⌬), we could detect more of the monoubiquitinated protein and, in some cases, more polyubiquitinated Gpa1 as well. Differences in the levels of Gpa1 monoubiquitination and polyubiquitination likely reflect the roles of the UBD-containing proteins in the removal of monoubiquitinated substrates. When normal trafficking is abrogated, the monoubiquitinated species accumulates and may undergo additional ubiquitination steps resulting in the accumulation of polyubiquitinated substrate. Conversely, we observed no increase in the amount of ubiquitinated Gpa1 in the strain lacking BUL1 despite the trafficking defect in that mutant. Because Bul1 is a ubiquitin binding component of the ubiquitin ligase Rsp5, it is possible that the ligase is unable to act effectively on Gpa1 when Bul1 is absent. Taken together these data suggest that the quantity of Gpa1 at the plasma membrane is finely tuned and that if Gpa1 cannot be removed through monoubiquitination and delivery to the vacuole, it is instead removed through polyubiquitination and delivery to the proteasome.
Our second major finding is that the UD promotes Gpa1 trafficking without altering Gpa1 enzymatic activity in vitro or MAPK activation in vivo. These results are particularly striking given that (a) the ubiquitination domain is located near a key dynamic region of the ␣-helical domain, (b) the ubiquitination domain is essential for transport of Gpa1 to the vacuole, and (c) there are a number of Gpa1 binding partners that specifically target the ubiquitination domain. These binding partners include enzymes responsible for monoubiquitination (Rsp5) FIGURE 7. Proper endocytosis of Gpa1 is required for sustained morphogenesis and efficient mating. A, MAPK activation profiles. Wild type (WT) cells, bul1⌬, ede1⌬, ddi1⌬, rup1⌬, and gpa1 ⌬UD were either left untreated or treated with 3 M ␣-factor (␣F) for 30 min. MAPK activation was determined by immunoblotting with phospho-MAPK p44/42 (P-Fus3, P-Kss1), Fus3 (Total Fus3) and G6PDH (Load control) antibodies. The image is representative of three independent experiments. Bottom, densitometry of P-Fus3 bands normalized to total Fus3. Results show the mean Ϯ S.E. for three individual experiments. B, transcriptional activation (␤-galactosidase activity) in response to ␣-factor treatment was measured spectrofluorometrically in wild type, bul1⌬, ede1⌬, ddi1⌬, rup1⌬, and gpa1 ⌬UD cells. Strains were transformed with a FUS1-lacZ reporter and treated with the indicated concentrations of ␣-factor for 90 min. Data for the individual mutants were normalized to WT in each case. Results show the mean Ϯ S.E. for three independent experiments, each performed in quadruplicate. C, loss of ubiquitination domain of Gpa1 results in morphological abnormalities under prolonged pheromone stimulation. Wild type bul1⌬, ede1⌬, ddi1⌬, rup1⌬, and gpa1 ⌬UD cells were stimulated with a constant saturating dose (150 nM) of ␣-factor in a microfluidic chamber. Differential interference contrast images were taken every 5 min for 8 h to monitor changes in morphology. Arrows point to mating projections in wild type cells. Scale bar, 5 m. D, mating efficiency assay. Separate cultures of wild type mating-type ␣ cells (BY4742) and wild type or gpa1 ⌬UD mating-type a cells (BY4741) were used. Cells from each culture were mixed and incubated for 4 h. Results are the mean Ϯ S.E. from four replicates. (26), polyubiquitination (SCF/Cdc4) (29), deubiquitination (Ubp12) (25), phosphorylation (Elm1, Tos1, Sak3) (30,31,66), and dephosphorylation (Reg1) (66) as well as the seven UBD proteins identified here. We conclude that the ubiquitination domain likely evolved to serve a unique trafficking function and that this function is wholly separate from the regulation of G protein catalytic activity.
Although the UD is unique to yeast G␣ proteins, G␣ trafficking is not. Trafficking of G␣ has long been established in the visual system, where a redistribution of the protein serves to decrease signaling, allowing adaptation to bright light (67). Whereas the visual G␣ acts directly to stimulate an effector enzyme, yeast Gpa1 acts indirectly to sequester G␤␥ and prevent downstream signaling. Therefore, the final outcome of G␣ endocytosis will be different in the visual and yeast signaling systems. G␣ endocytosis attenuates signaling in the visual system but promotes signaling in the yeast mating pathway. On the other hand, G␣ proteins have been shown to be activated at the endosomal membrane compartment both in yeast and in mammalian cells (12,14). Before our analysis, however, little was known about the proteins necessary for proper G␣ trafficking.
Finally, our analysis of the UD may lead to insights regarding the function of unique inserts in other signaling proteins. For example, there are families of small GTPases that are known to contain inserts within the highly conserved Ras-like domains. Members of the Rho family of small GTPases have a unique insert, called the Rho insert, that is not present in other small GTPases (68). The presence or absence of the Rho insert does not alter the intrinsic activity of these small GTPases (69). However, when the insert is absent Rho can bind, but no longer activate, its downstream effector Rho kinase (69). The Rho insert in the small GTPase Rac1 was recently shown to be monoubiquitinated (70). Although no function has yet been assigned to monoubiquitination of Rac1, it is possible that this modification is involved in the mechanism by which Rho interacts with downstream effectors. Monoubiquitination of mammalian Ras proteins can lead to cell transformation (71) and does so by promoting nucleotide exchange or by impeding the binding of the GTPase activating protein (72,73).
In summary we have identified four UBD-containing proteins as specific regulators of G protein trafficking. When trafficking is abrogated, the morphogenesis branch of the pathway is attenuated. None of these UBD proteins affect the MAPK branch of the signaling pathway, and none affect trafficking of the receptor. Looking forward, our integrated approach should be broadly applicable as more ubiquitination substrates, ubiquitination domains, and UBD-containing proteins are identified. Recently, dysregulation of NEDD4/Rsp5-mediated trafficking was shown to promote neurodegenerative disease (20). In view of these findings, any components responsible for monoubiquitination and protein trafficking represent potential targets for future drug development efforts.