The Extracellular N-terminal Domain and Transmembrane Domains 1 and 2 Mediate Oligomerization of a Yeast G Protein-coupled Receptor*

G protein-coupled receptors (GPCRs) can form homodimers/oligomers and/or heterodimers/oligomers. The mechanisms used to form specific GPCR oligomers are poorly understood because the domains that mediate such interactions and the step(s) in the secretory pathway where oligomerization occurs have not been well characterized. Here we have used subcellular fractionation and fluorescence resonance energy transfer (FRET) experiments to show that oligomerization of a GPCR (α-factor receptor; STE2 gene product) of the yeast Saccharomyces cerevisiae occurs in the endoplasmic reticulum. To identify domains of this receptor that mediate oligomerization, we used FRET and endocytosis assays of oligomerization in vivo to analyze receptor deletion mutants. A mutant lacking the N-terminal extracellular domain and transmembrane (TM) domain 1 was expressed at the cell surface but did not self-associate. In contrast, a receptor fragment containing only the N-terminal extracellular domain and TM1 could self-associate and heterodimerize with wild type receptors. Analysis of other mutants suggested that oligomerization is facilitated by the N-terminal extracellular domain and TM2. Therefore, the N-terminal extracellular domain, TM1, and TM2 appear to stabilize α-factor receptor oligomers. These domains may form an interface in contact or domain-swapped oligomers. Similar domains may mediate dimerization of certain mammalian GPCRs.

G protein-coupled receptors (GPCRs) 1 constitute the largest class of transmembrane receptors in multicellular organisms. GPCRs mediate the actions of an enormous array of peptides, hormones, bioactive lipids, neurotransmitters, and sensory stimuli, and they are the targets of many drugs used in clinical medicine. Members of this class of receptor have a similar structural architecture consisting of an N-terminal extracellular domain, followed by a bundle of seven transmembrane ␣-helices connected by intracellular and extracellular loops, and terminated with a cytoplasmic tail. Therefore, many aspects of GPCR structure, function, signaling, and regulation are conserved.
Defining the mechanisms that direct the formation of GPCR dimers/oligomers is required to understand how GPCRs activate G proteins and to suggest which GPCRs preferentially form homodimers/oligomers versus those that also interact as heterodimers/oligomers with other GPCRs. Studies of chimeric receptors and computational methods (evolutionary trace and correlated mutations) have suggested that dimerization/oligomerization may occur by exchange of one or more complementary TM domains between receptor subunits (e.g. TM1 and -2 of one receptor are reciprocally exchanged for their counterparts in the second receptor (19 -23)). Alternatively, GPCR dimer/oligomerization may occur by formation of simple contacts that do not involve exchange of TM domains between receptor subunits (24).
Several types of molecular interactions occur in GPCR dimer/oligomers. Intermolecular disulfide bonds occur within the N-terminal extracellular domains of calcium-sensing (25)(26)(27) and metabotropic glutamate (28,29) receptors and within the second extracellular loop of m3 muscarinic receptors (30). However, these disulfide bonds are dispensable for dimer/oligomerization of these receptors, indicating that non-covalent interactions are sufficient (31)(32)(33). Indeed, the N-terminal extracellular ligand-binding domain of mGluR1 forms a high affinity homodimeric complex (34,35), which is likely to promote dimer formation by this receptor. Likewise, a C-terminal cytoplasmic coiled-coil motif contributes to the formation of heterodimers between GABA B -R1 and -R2 in the endoplasmic reticulum (36 -40). However, this coiled-coil motif is not the sole determinant of dimerization because C-terminally truncated GABA B receptors lacking this motif still interact (39,40), and the extracellular N-terminal domain of GABA B -R1 heterodimerizes with that of GABA B -R2 (41). Whether GPCRs generally dimerize in the endoplasmic reticulum is unknown, although this would seem likely.
We have used the ␣-factor receptor (STE2 gene product) of the yeast Saccharomyces cerevisiae as a model system to investigate the mechanisms of GPCR oligomerization. We have shown previously that this receptor constitutively forms homooligomers, as detected in living yeast cells by performing fluorescence resonance energy transfer (FRET) experiments between CFP-and YFP-tagged receptors (42). ␣-Factor receptor oligomers can also be detected by showing that GFP-tagged endocytosis-defective receptors interact with untagged wild type receptors to be recruited into the endocytic pathway (42,43). Analysis of dominant-negative mutants of the ␣-factor receptor suggests that oligomerization may be important for signal transduction (42).
To elucidate mechanisms of ␣-factor receptor oligomerization, we have characterized intracellular compartments where oligomerization occurs and identified receptor domains that mediate oligomerization in living cells. The results implicate which domains of ␣-factor receptors are involved in oligomerization, address whether ␣-factor receptors interact by monomer-monomer contact or by exchange of complementary TM domains (domain swapping), and suggest whether ␣-factor receptors and certain mammalian GPCRs oligomerize by employing similar mechanisms.

EXPERIMENTAL PROCEDURES
Yeast Strains and Plasmids-The S. cerevisiae strain used in this study was KBY58 (MATa ura3-52 leu2-3,112 his3⌬-1 trp1 ste2⌬) (44), which carries a complete deletion of the ␣-factor receptor structural gene (STE2) that allows various combinations of receptor mutants to be expressed from plasmids. Unless stated otherwise, all receptor constructs used in this study lacked sequences encoding the C-terminal cytoplasmic domain, which is dispensable for agonist binding and signaling but which is required for receptor internalization and desensitization (45). Use of receptors lacking their C-terminal domain is necessary to detect FRET between CFP-and YFP-tagged ␣-factor receptors (42), because FRET is exquisitely sensitive to interfluorophore distance, orientation, and mobility. Because preliminary experiments indicated that deletions removing various transmembrane domains of the ␣-factor caused significant expression defects, it was necessary to express deletion mutants from the STE2 promoter on high copy plasmids. To generate these overexpression plasmids, we cloned BamHI fragments containing the STE2 promoter and coding region for Ste2⌬tail-YFP and -CFP fusions (42) into the BamHI sites of the high copy plasmids pRS423 and pRS424, yielding pRS423STE2⌬tail-YFP and pRS424STE2⌬tail-CFP. The following method was used to delete coding sequences for various transmembrane domains. First, pRS423STE2⌬tail-YFP and pRS424STE2⌬tail-CFP were subjected to two separate rounds of site-directed mutagenesis (Stratagene QuikChange TM mutagenesis kit) to create a series of constructs with all possible pairwise combinations of in-frame SphI sites that introduced either a 6-bp insertion or substitution encoding an Ala-Cys dipeptide at the following positions within the STE2-coding region ( Fig. 1): bp 4 -9 (substitution of amino acids Ser-2 and Asp-3); bp 137-142 (substitution of Ser-47 and Thr-48 immediately preceding TM1); bp 236 -241 (insertion between Pro-79 and Ile-80 in intracellular loop 1); bp 341-346 (insertion between Thr-114 and Gly-115 in extracellular loop 1); bp 485-490 (insertion between Thr-155 and Glu-156 in intracellular loop 2); bp 593-598 (insertion between Ala-198 and Thr-199 in extracellular loop 2); bp 707-712 (insertion between Leu-236 and Gly-237 intracellular loop 3); bp 818 -823 (insertion between Gly-273 and Thr-274 in extracellular loop 3); bp 897-902 (substitution of Ala-297 and Ala-298 immediately following TM7). Introduction of the Ala-Cys substitution or insertion encoded by the SphI sites at these positions preserved receptor function as indicated by quantitative assays of agonist-induced growth arrest (data not shown). These constructs were then digested with SphI and ligated to generate deletion mutations (Fig. 1), all of which were non-functional, as expected.
Subcellular Fractionation-Subcellular fractionation of yeast cell lysates was performed as described previously by equilibrium density gradient centrifugation (46). Briefly, cells were grown in selective medium to a density of 10 7 cells/ml and were killed by addition of 10 mM NaN 3 and 10 mM KF. Cells were harvested and washed once with 25 ml of sorbitol buffer (10 mM Tris, pH 7.6, 0.8 M sorbitol, 10 mM NaN 3 , 10 mM KF, 1 mM EDTA, pH 8.0), once with 1 ml of sorbitol buffer, once with 1 ml of sucrose buffer (10 mM Tris, pH 7.6, 1 mM EDTA, pH 8.0, 10% (w/v) sucrose), and suspended in 1 ml of sucrose buffer containing protease inhibitors 0.1 mM phenylmethylsulfonyl fluoride, 0.1 mM benzamidine, 20 M tosylphenylalanine chloromethyl ketone, 5 M pepstatin A, 5 M leupeptin, and 1 mM N ␣ -ptosyl-L-arginine methyl ester. Cells were lysed by mechanical disruption and cleared by centrifugation at 300 ϫ g for 5 min. The supernatant fraction was mixed with 0.5 ml of 50% (w/v) sucrose in 10 mM Tris, pH 7.6, 1 mM EDTA, and layered on top of a 4-ml, 35-60% linear sucrose gradient prepared in 10 mM Tris, pH 7.6, 1 mM EDTA. Gradients were centrifuged at 150,000 ϫ g in an SW50.1 rotor at 4°C for 20 h. Fractions (350 l) were collected from the top of the gradient. Aliquots (150 l) were assayed in FRET experiments as described below. Aliquots (50 l) of gradient fractions were diluted 1:2 with 2ϫ Laemmli sample buffer containing 8 M urea for use in immunoblotting analysis with antibodies directed against tagged receptors and markers of various organelles (Vph1 (vacuole), Gda1 (Golgi), Dpm1 (ER), and Pma1 (plasma membrane)).
Fluorescence Resonance Energy Transfer (FRET)-Scanning fluorometry of intact yeast cells or subcellular fractions co-expressing CFP-and YFP-tagged ␣-factor receptors was used to detect FRET between oligomerized receptors in vivo and in vitro as described previously (42). In each FRET experiment, fluorescence emission spectra were recorded from four samples that contained the same number of cells (or membrane protein for in vitro experiments) as follows: (a) control cells that do not express tagged receptors; (b) cells that express only CFP-tagged receptors; (c) cells that express only YFP-tagged receptors; and (d) cells that co-express CFP-and YFP-tagged receptors. Fluorescence emission due to FRET between CFP-and YFP-tagged receptors was detected by employing the following three-step procedure. Briefly, this involved irradiation of cells at the optimized max for excitation of CFP, recording emission spectra, and subtracting the components of the emission spectra due to cell autofluorescence, CFP emission, and YPF emission due to direct excitation; this resulted in the YFP emission spectrum due only to FRET.
Step 1 involved data acquisition and correction for cell autofluorescence. Control cells and cells that co-express CFP-and YFP-tagged receptors were irradiated at 425 nm, near the max for excitation of CFP, which gives reduced direct excitation of YFP. The emission spectrum obtained from control cells (no tagged receptors expressed) was subtracted from that obtained with an equivalent number of cells co-expressing CFP-and YFP-tagged receptors, resulting in the CFP ϩ YFP emission spectrum. This emission spectrum is a composite of CFP emission, YFP emission due to direct excitation, and YFP emission due to FRET. Step 2 involved subtraction of CFP emission from the CFP ϩ YFP emission spectrum. Cells expressing only CFP-tagged receptors were irradiated at the optimized max for excitation of CFP (425 nm). This spectrum was normalized to give a CFP emission peak value identical to the CFP emission peak value of the CFP ϩ YFP spectrum obtained from the preceding step. After normalization, the CFP spectrum was subtracted from the CFP ϩ YPF emission spectrum. This resulted in a YFP emission spectrum composed of a FRET component and a direct excitation of YFP component. Step 3 involved obtaining the YFP emission spectrum due to FRET. To obtain the YFP emission spectrum due only to FRET, it was necessary to subtract the component of the total YFP emission that was due to direct excitation of YFP. To accomplish this, the YFP emission spectra of cells co-expressing CFPand YFP-tagged receptors versus cells expressing only YFP-tagged receptors were normalized for small differences in YFP expression level. This was achieved by irradiating these two types of cells at the max for YFP (510 nm) and recording their respective YFP emission spectra. Because CFP was not excited at this wavelength, these emission spectra quantify only the level of YFP-tagged receptors. The ratio of the YFP emission peak heights of these two spectra was then used as a scaling factor to normalize the emission spectrum obtained when cells expressing YFP-tagged receptors were irradiated at 425 nm, near the max for CFP. This normalized emission spectrum was then subtracted from the total YFP emission spectrum obtained in step 2 from cells co-expressing CFP-and YFP-tagged receptors. The result was a YFP emission spectrum due solely to FRET.
To calculate the apparent efficiency of FRET, we used the following two spectra obtained during the process of generating the FRET emission spectrum described above: the YFP emission spectrum due specifically to FRET, and the YFP emission spectrum obtained by irradiating cells co-expressing CFP-and YFP-tagged receptors at the max for YFP. Apparent FRET efficiency was calculated by dividing the integrated area of the FRET spectrum by the integrated area of the YFP emission spectrum obtained by excitation at the max for YFP.
Fluorescence Imaging-Fluorescence and Nomarski images of cells expressing wild type and various receptor fragments tagged with YFP were captured by using a DAGE cooled CCD camera mounted on an Olympus BH-2 microscope equipped with a DPlanApo100UV 100ϫ objective, as described previously (44). Endocytosis assays were performed as described previously (42). Briefly, cells co-expressing fulllength untagged wild type receptors overexpressed from the PGK1 promoter on the high copy plasmid pPGK (47) and various tailless receptor fragments tagged with YFP expressed from the STE2 promoter on high copy plasmids were treated with agonist (5 M ␣-factor). Images were collected before and at the indicated times after agonist treatment.

Using FRET to Detect Association between ␣-Factor Receptor
Fragments-To identify domains of the yeast ␣-factor receptor involved in oligomerization, we used deletion mutagenesis to identify receptor fragments that could oligomerize as indicated by FRET. Deletion mutants rather than chimeric receptors were used to map oligomerization domains because ␣-factor receptors from sufficiently closely related species of yeast have not yet been cloned. The utility of deletion mutants was suggested by studies of Dumont and colleagues (48), which indicated that a functional receptor can be generated by co-expressing complementary receptor fragments (e.g. TM1 expressed from one plasmid and TM2-7 expressed from another in the same cell). These findings and similar studies of mammalian GPCRs suggested that receptor fragments can be expressed in a relatively native form (34,49,50). Consistent with this hypothesis, synthetic peptides corresponding to certain individual transmembrane segments of the ␣-factor receptor adopt ␣-helical conformations when reconstituted in lipid vesicles (51,52).
We generated deletions that removed sequences encoding the N-terminal extracellular domain or various transmembrane domains to produce all possible receptor fragments that allow a CFP or YFP tag appended at the C terminus of the molecule to be located intracellularly (Fig. 1). With the exception of deletions removing TM1 or TM7, deletions of single TM domains could not be analyzed because they would disrupt the topology of the receptor. To analyze receptor mutants by FRET, all of the mutant receptors lacked a C-terminal cytoplasmic domain (which is dispensable for signaling but is required for endocytosis (45)) to bring CFP and YFP within sufficient proximity. Each mutant receptor tagged at the C terminus of its last TM with CFP or YFP was expressed from its normal promoter on a high copy vector, which was necessary to express receptor fragments at levels similar to that of wild type receptors ( Fig.  2 and data not shown). As expected, when expressed alone none of these fragments produced a functional receptor (data not shown).
To determine whether receptor fragments could be useful to identify domains involved in ␣-factor receptor oligomerization, we determined whether FRET could detect association between pairs of complementary receptor fragments that are known to reconstitute a functional receptor (48). For this purpose, we analyzed cells that co-expressed a YFP-tagged receptor fragment containing the N terminus and TM1-5 with a CFP-tagged receptor fragment containing TM6 -7 ( Fig. 1), or that co-expressed a YFP-tagged receptor fragment containing the Nterminal domain and TM1-3 with a CFP-tagged receptor fragment containing TM4 -7 (Fig. 1). Cells were excited at 440 nm (CFP excitation maximum), and fluorescence emission spectra were recorded. After correcting spectra for differences in protein expression levels, we subtracted the emission spectra obtained with control cells (expressing only the CFP-or YFPtagged receptor fragment fusion) from the spectrum obtained upon co-expressing the relevant pair of receptor fragments. This procedure yielded the fluorescence emission spectrum due to FRET (indicated by the solid curve in all figures), as published previously (42) and described under "Experimental Procedures." By using this method, we detected FRET between tagged wild type receptors and between pairs of receptor fragments expected to reconstitute a functional receptor (Fig. 2). Quantitation of apparent FRET efficiency (see "Experimental Procedures") indicated that complementary receptor fragments associated efficiently with one another (TM1-3 ϩ TM4 -7, 15.0 Ϯ 1.0%; TM1-5 ϩ TM6 -7, 16.5 Ϯ 1.4%), validating the utility of the approach.
␣-Factor Receptor Oligomerization in the ER and Plasma Membrane Fractions-Before we could analyze the ability of receptor fragments to self-associate in FRET experiments us-FIG. 1. Schematic of the ␣-factor receptor and deletion mutants analyzed in this study. Each mutant is designated by the identities of the transmembrane domains that were retained. The protein coding region remaining in each deletion mutant is indicated with a solid black line, and the specific amino acid (a.a.) residues present or deleted in each fragment are indicated. Transmembrane domains are indicated with shaded boxes and the CFP or YFP tag is depicted as a black oval.
FIG. 2. Use of FRET to detect oligomerization in vivo of wild type receptors and interaction between co-expressed receptor fragments that reconstitute a functional receptor. The indicated CFP-and YFP-tagged receptor or receptor fragments were co-expressed for FRET experiments. FRET data were collected and analyzed as described under "Experimental Procedures." Each panel in this and subsequent figures shows four emission spectra obtained upon excitation of cells at the max for CFP as follows: one from cells co-expressing the indicated CFP-and YFP-tagged receptors or receptor fragments (dotted plus dashed line); a second from cells expressing only the indicated CFP-tagged receptor or receptor fragments (dotted line); a third from cells expressing only the indicated YFP-tagged receptor or receptor fragment (dashed line); and a fourth that shows the fluorescence emission due specifically to FRET (solid line) from cells co-expressing the indicated CFP-and YFP-tagged receptors or receptor fragments. Emission due to FRET was determined by subtracting the second and third emission curves from the first emission curve, as described under "Experimental Procedures." YFP-and CFP-tagged receptors or receptor fragments were co-expressed from plasmids in cells carrying a deletion of the chromosomal gene encoding the ␣-factor receptor.
ing intact cells, it was necessary to determine whether receptor fragments trafficked normally to the plasma membrane. If receptor fragments are blocked or delayed in the secretory pathway prior to where oligomerization normally occurs, the absence of a FRET signal could not be interpreted as a defect in oligomerization per se. Indeed, as indicated by fluorescence microscopy of cells expressing individual YFP-tagged receptor fragments (Fig. 3), most of the mutant receptors localized to the plasma membrane and the perinuclear endoplasmic reticulum (ER), although ER localization was abnormally prominent in some cases. Furthermore, one receptor mutant possessing the extracellular N-terminal domain and TMs1-4 and 7 was mislocalized nearly exclusively in intracellular vesicles of unknown identity. These vesicles were not endosomes because they were not motile, in contrast to endosomes bearing wild type receptor-GFP fusions (44). 2 As expected, a YFP fusion containing only the hydrophilic N-terminal domain of the receptor was cytoplasmic, albeit with some concentration in the nucleus.
Because certain receptor fragments localized more prominently to the ER, it was necessary to determine whether oligomerization of wild type receptors occurs in the ER. To address this issue, we took advantage of the fact that ␣-factor receptors lacking their C-terminal domains and tagged with CFP or YFP are present both in the perinuclear ER and the plasma membrane (Fig. 3). This allowed us to use FRET experiments to determine whether oligomerization of these receptors occurs in the ER and plasma membrane fractions prepared by sucrose gradient centrifugation. To carry out FRET experiments using methods similar to those described above, we prepared subcellular fractions from four types of cells: 1) cells co-expressing Ste2⌬tail-CFP and Ste2⌬tail-YFP; 2) cells expressing only Ste2⌬tail-CFP; 3) cell expressing only Ste2⌬tail-YFP; and 4) cells that do not express fluorescently tagged receptors. After correcting emission spectra for autofluorescence and differences in membrane protein content, we subtracted the emission spectra of control gradient fractions (prepared from cells expressing only CFP-or YFPtagged receptors) from the spectrum of the equivalent experimental gradient fraction (prepared from cells co-expressing CFP-and YFP-tagged receptors). This procedure yielded the fluorescence emission spectrum due to FRET, which then was used as described before to obtain an apparent FRET efficiency.
As shown in Fig. 4, the efficiencies of FRET obtained with ER/Golgi-enriched fractions (fractions 7 and 9), a fraction with overlapping ER and PM markers (fraction 11), and a purified PM fraction (fraction 13) were statistically indistinguishable from one another (12-14 Ϯ 2%). Therefore, we conclude that ␣-factor receptors can oligomerize in the ER. Furthermore, because FRET efficiency obtained with wild type receptors was the same in ER and plasma membrane fractions, any differences in FRET efficiency observed with receptor fragments in intact cells are unlikely to be due to differences in subcellular localization and are likely to be due to differences in the intrinsic ability of the fragments to oligomerize.
Using FRET to Identify ␣-Factor Receptor Domains Involved in Oligomerization-Because the intracellular and extracellular loops of the ␣-factor receptor are relatively small (ϳ5-27 amino acids), it was likely that oligomerization is mediated primarily by the extracellular N-terminal domain (49 amino acids) and/or various transmembrane domains. To test this 2 F. Chang and K. Blumer, unpublished data.
FIG. 3. Subcellular localization patterns of YFP-tagged ␣-factor receptor deletion mutants. The indicated wild type and deletion mutants of the ␣-factor receptor tagged at their intracellular C termini with YFP were examined by fluorescence microscopy. YFP-tagged wild type receptors were expressed from their normal promoter on a single copy plasmid, whereas YFP-tagged deletion mutant receptors were expressed from their normal promoter on high copy plasmids to achieve an expression level similar to that of YFP-tagged wild type receptors. All receptors were expressed in cells that carried a deletion of the chromosomal ␣-factor receptor gene. hypothesis, we performed several types of experiments to determine which receptor fragments could self-associate in a manner that is specific and likely to reflect a biologically relevant interaction rather than nonspecific aggregation. First, we used FRET experiments to determine which receptor fragments could self-associate when co-expressed as a pair of CFPand YFP-tagged fusions. Second, we used FRET experiments to determine which receptor fragments could associate with wild type tailless receptors, as would be expected of a receptor fragment that possesses a functional oligomerization domain. Finally, we used FRET experiments to determine whether self-association of a receptor fragment could be inhibited upon overexpression of untagged wild type receptors, as would be expected for a specific, saturable interaction rather than nonspecific aggregation of misfolded receptor fragments.
The results of FRET experiments used to detect self-association of receptor fragments are shown in Figs. 5-7. Control experiments indicated that wild type tailless receptors formed homo-oligomers with an apparent FRET efficiency of 11.5 Ϯ 2.2% ( Fig. 2 and Table I) and that this interaction was inhibited by overexpression of untagged wild type receptors ( Fig. 8 and Table III), as we have shown previously (42). Similarly, receptor fragments that possess the N-terminal domain and TM1-5, TM1-3, or TM1 could also self-associate (Fig. 5). Fragments containing the N-terminal domain and TM1-5 or TM1-3 selfassociated with an efficiency (14.7 Ϯ 2.4% and 11.6 Ϯ 2.9%, respectively; Table I) similar to that of wild type tailless receptors. Furthermore, self-association of these fragments was specific because the FRET signal was reduced upon overexpression of untagged wild type receptors ( Fig. 8 and Table II). A fragment containing the N-terminal domain and TM1-5 also interacted with wild type tailless receptors in FRET assays ( Fig. 9 and Table III in a manner similar to that of wild type receptors. Taken together, these findings suggested that oligomerization does not require TMs 4 -7 and provided an initial suggestion that oligomerization involves TM1-3. A fragment containing only the N-terminal domain and TM1 self-associated but with lower apparent efficiency (7.7 Ϯ 2.5%; Fig. 4 and Table I), whereas the hydrophilic N-terminal domain expressed as an unglycosylated cytoplasmic protein did not self-associate to a detectable degree (Fig. 5 and Table I). However, this does not exclude the possibility that the Nterminal domain can interact with itself when glycosylated and attached to a transmembrane domain. Nevertheless, the results are consistent with the hypothesis that the extracellular N-terminal domain is unlikely be the principle oligomerization domain, whereas such a function may be subserved by TM1.
Although TM1 appeared to be important for oligomerization, the results of an additional experiment suggested that the hydrophilic N-terminal domain of the receptor may facilitate receptor-receptor interaction. A receptor fragment lacking the N terminus but possessing all 7 transmembrane domains (TM1-7⌬N-term) showed reduced efficiency of FRET (Fig. 6).
Although the N-terminal domain may facilitate oligomerization of the receptor, this could not occur via disulfide bond formation because this domain lacks cysteine residues.
To determine whether transmembrane domains other than or in addition to TM1 contribute to the formation of receptor oligomers, we examined the ability of other N-terminally truncated receptor fragments to self-associate. We were unable to detect a significant FRET signal in experiments that examined self-association of TM2-7, TM3-7, TM6 -7, or TM7 (Fig. 6), providing a further suggestion that TM1 (possibly in concert with the N-terminal domain) is necessary for receptor oligomerization.
In contrast, self-association of fragments containing only TM4 -7 and TM5-7 was detected (Fig. 6). However, neither of these fragments interacted significantly with wild type tailless receptors in FRET experiments ( Fig. 9 and Table III), and overexpression of untagged wild type receptors did not prevent self-association of these receptor fragments (data not shown). Therefore, fragments containing only TM4 -7 or TM5-7 may self-associate non-specifically or aggregate because they expose FIG. 7. Homo-oligomerization of ␣-factor receptor internal deletion mutant fragments detected by FRET. CFP-and YFP-tagged forms of the indicated receptor mutants were co-expressed and analyzed for interaction by FRET as described in the legend to Fig. 2.   FIG. 8. Inhibition of homo-oligomerization of wild type ␣-factor receptors or receptor deletion mutant fragments by overexpressed untagged wild type receptors. CFP-and YFP-tagged forms of the indicated receptor mutants were co-expressed wild untagged wild type receptors overexpressed from a high copy plasmid and analyzed for interaction by FRET as described in the legend to Fig. 2.

TABLE I Apparent efficiencies of FRET for homo-oligomerization of wild type tailless ␣-factor receptors and receptor fragments
The ability of wild type tailless receptors and the indicated receptor fragments to self-associate was quantified by calculating the apparent FRET efficiency as described under "Experimental Procedures." The domains present in each mutant are indicated with a ϩ (N, N- residues normally buried or involved in intramolecular contacts in wild type receptors. Indeed, there is evidence of an intramolecular contact occurring between TM3 and TM6 of the ␣-factor receptor (53). Alternatively, these fragments may selfassociate specifically with such high affinity that interaction with wild type tailless receptors is precluded. If so, TM4 -7 may contain a secondary contact site important for oligomerization, although the inability of TM2-7 and TM3-7 fragments to selfassociate is inconsistent with this model. In light of the above results, we further investigated the role of TM1 and/or other transmembrane domains by performing FRET experiments with all internal deletion mutants that would maintain the normal topology of the protein (Fig. 7 and Table I). The results indicated that a fragment containing the N-terminal domain and TM1-4 and -7 could self-associate. However, this result was not readily interpretable because this fragment was severely mislocalized (Fig. 3), and its ability to self-associate was not inhibited by overexpression of untagged wild type receptors ( Fig. 8 and Table II).
Results obtained with other internal deletion fragments were more clear. Receptor fragments containing the N-terminal domain and TM1-2 and -5-7, TM1-3 and -6 -7, or TM1-2 and -7 self-associated with nearly wild type efficiency. In contrast, significant but less efficient self-association was observed with fragments containing only the N terminus and TM1 and -4 -7 or TM1 and -6 -7 ( Fig. 7 and Table I). These interactions appeared to be specific because they were inhibited by overex-pression of untagged wild type receptors (Fig. 8, Table II, and data not shown). These results support the hypothesis that TM2 is part of the oligomerization domain and/or facilitates oligomerization indirectly by positioning or stabilizing TM1. An indirect role for TM2 may be more likely because a fragment containing TM2-7 did not self-associate (Fig. 6). Taken together, the results of the FRET experiments (Table I) suggest that TM1 is necessary and possibly sufficient for oligomerization and that regions flanking TM1 (the hydrophilic N terminus and TM2) facilitate this interaction.

Mapping Domains Required for Receptor-Receptor Interaction by Monitoring Heterodimerization between Endocytosisdefective Receptor Fragments and Wild Type
Receptors-Loss of a FRET signal does not necessarily indicate a loss of oligomerization activity because of the exquisite sensitivity of FRET to changes in intrafluorophore distance and/or orientation. Therefore, it was important to use an independent means of mapping receptor domains involved in oligomerization. Although crosslinking methods could be used, they have not been established for this receptor, and they require extensive specificity controls to determine whether a specific, stable interaction relevant to in vivo oligomerization is being detected.
As an alternative we used our previously published endocytosis-based assay that detects the ability of GFP-tagged endocytosis-defective receptors to interact with and be rescued by co-expressed untagged wild type receptors (42). Rescue of endocytosis-defective receptors occurs by hetero-oligomerization between endocytosis-deficient and -competent receptors (42,43), rather than by stimulation of bulk internalization of plasma membrane proteins. Therefore, we could determine whether the endocytosis defects of representative deletion derivatives of GFP-tagged tailless receptors could be rescued by co-expression with untagged wild type receptors.
As reported previously (42,43), the endocytosis defect of tailless but otherwise full-length GFP-tagged receptors (Ste2⌬tail-GFP) was rescued upon co-expression with untagged wild type receptors ( Fig. 10; compare Ste2⌬tail-GFP with Ste2⌬tail-GFPϩWT), as indicated: 1) fluorescent labeling of the lysosome-like vacuole (V), and highly motile endosomes (E) whose abundance increased following agonist treatment (the identities of these compartments have been previously established (44)); and 2) time-dependent loss of GFP-tagged tailless receptors from the cell surface following stimulation with agonist. To quantify the results, we counted 1000 cells of each type to determine the proportion that contained labeled, motile endosomes 1 h after agonist treatment. At this time point, 90% of cells co-expressing GFP-tagged tailless receptors and untagged full-length receptors contained labeled endosomes. Similarly, many of the cells expressing GFP-tagged receptor fragments containing the N-terminal domain and   TM1-5 or TM1 displayed labeled endosomes (79 and 43%, respectively) when untagged wild type receptors were co-expressed (Fig. 10). However, neither of these GFP-tagged fragments was removed completely from the plasma membrane even after prolonged (2 h) exposure to agonist, indicating that they do not interact as efficiently with untagged wild type receptors. In contrast, a GFP-tagged receptor fragment lacking TM1 and -2 (TM3-7⌬tail-GFPϩWT) was unable to interact with untagged wild type receptors in order to be recruited into endosomes or the vacuole (no motile, labeled endosomes detected; Fig. 10). Although punctate labeling of intracellular organelles was occasionally observed, these were not endosomes because they were not motile, and their abundance was not agonist-induced. This inability to undergo endocytosis was unlikely to be caused by segregation of the tagged receptor fragment and untagged wild type receptors in different plasma membrane subdomains, because their localization patterns were similar (TM3-7⌬tail-GFP versus wild type tailless receptors tagged with GFP (STE2⌬tail-GFP); Fig. 10). Therefore, the results of these endocytosis experiments provided independent in vivo evidence supporting the hypothesis that ␣-factor receptor oligomerization is mediated by TM1, possibly in concert with the N-terminal domain and TM2. DISCUSSION Here we have presented in vivo biophysical and cell biological analysis of an extensive collection of ␣-factor receptor deletion mutants to provide direct evidence that specific domains of a GPCR are involved in dimer/oligomer formation. We also have shown that oligomerization of ␣-factor receptors occurs with equivalent efficiency in the endoplasmic reticulum and the plasma membrane. Because previous studies (42, 45, 46, 54 -56) have indicated that the overall structural organization and signaling mechanisms of the yeast ␣-factor receptor are similar to those of mammalian GPCRs, the results of our in-vestigation are likely to be applicable to other receptors.
Oligomerization of ␣-factor receptors in the ER is consistent with the general view that membrane proteins are folded and properly assembled in this compartment prior to subsequent trafficking to their final destinations. Indeed, GABA B receptor heterodimerization occurs in the ER, which masks an ER retention signal in the C-terminal domain of the GABA B -R1 subunit and allows the heterodimeric receptor complex to exit the ER and transit to the plasma membrane (39,40). It would be interesting to determine whether GPCR dimer/oligomer assembly in the ER is spontaneous or involves chaperones or other factors.
The second major conclusion of the present investigation is that oligomerization of the ␣-factor receptor is mediated significantly by TM1, possibly in concert with the N-terminal extracellular domain and TM2. This conclusion has been reached by analyzing a collection of receptor deletion mutants for the ability to self-associate and to interact with wild type receptors in FRET and endocytosis assays using living cells. Several factors indicate that the results obtained from analysis of deletion mutants and the conclusions derived therefrom are likely to be relevant to the mechanisms that mediate oligomerization of intact ␣-factor receptors. First, several controls have established the specificity of the interactions observed with receptor deletion mutants, indicating that most of the deletion mutants do not aggregate because they are grossly misfolded. Second, several of the receptor deletion mutants can be co-expressed with one another to reconstitute a functional receptor (48), indicating each fragment is not misfolded irreversibly. Third, individual TM domains of the ␣-factor receptor adopt highly ␣-helical structures when reconstituted into lipid vesicles (51,52), which is thought to resemble their structures in the context of the full-length receptor. Fourth, although certain receptor fragments localize somewhat less evenly in the plasma membrane than wild type receptors, they retain the ability to self-associate and interact with wild type receptors in FRET and endocytosis experiments.
Accordingly, we suggest that the N-terminal domain, TM1, and TM2 provide a homophilic interaction surface that stabilizes ␣-factor receptor dimers/oligomers. TM1 may be a major contact site because this TM is sufficient to interact with itself or wild type receptors, and because point mutations affecting TM1 impair oligomerization of intact receptors. 3 However, on its own TM1 self-associates with reduced efficiency. This effect is probably not due to steric hindrance between CFP and YFP tags used for FRET, because these tags allow efficient homodimerization of the single transmembrane domain of the receptor tyrosine phosphatase-␣ (57). Therefore, TM1 may selfassociate with reduced efficiency because other domains of the 3 M. C. Overton and K. Blumer, manuscript in preparation.
FIG. 11. Models of ␣-factor receptor dimerization suggested by this study. A contact dimer between receptor monomers and a domainswapped dimer involving reciprocal exchange of TM1 and TM2 between receptor subunits are shown.
FIG. 10. Interaction between wild type receptors and receptor deletion mutant fragments detected during endocytosis. Fluorescence microscopy was used to localize the indicated GFP-tagged endocytosis-defective (tailless) receptors or receptor mutants expressed alone or with untagged wild type receptors. Images were acquired before (0 min) or at the indicated times after the addition of agonist (␣-factor, 5 M). Endosomes (E), the lysosome-like vacuole (V), and the endoplasmic reticulum (R) are indicated as documented previously (44).
␣-factor receptor position or stabilize TM1 or provide additional contacts that stabilize the receptor complex. Such auxiliary functions may be provided by the N-terminal extracellular domain or TM2. Although neither the N terminus nor TM2 appears to be sufficient for self-association, receptor fragments lacking either of these domains (e.g. TM1 and -4 -7, TM1 and -6 -7, and TM1-7 lacking the N terminus) displayed reduced FRET efficiency. In contrast, transmembrane domains 3-7 do not appear to have a significant auxiliary role, because the TM1-2 and -5-7, TM1-3, and TM1-5 fragments self-associate as efficiently as wild type receptors. Therefore, the results to date support the conclusion that TM1 is a primary oligomerization contact site and that the extracellular N-terminal domain and TM2 may stabilize receptor-receptor association directly or indirectly. TM1 in concert with TM2 and the N terminus could form an interface in the context of contact dimers/oligomers in which there is no exchange of TM domains between receptor subunits, or they could be regions that participate in a domain swapping mechanism of dimer/oligomer formation (Fig. 11). We currently favor the simple contact model because small receptor fragments, including one consisting only of the N-terminal extracellular domain, and TM1 can self-associate, unlike several receptor fragments that lack TM1.
Our findings and recent studies of mGluR1, B2 bradykinin, and GABA B receptors suggest that similar domains are used for dimer/oligomerization of certain GPCRs. Studies (7, 28, 29, 34, 37, 39, 58 -60) of these mammalian receptors indicate that dimer/oligomer formation involves several domains (the N-terminal cytoplasmic domain, TM1, TM7, and/or the membraneproximal portion of the C-terminal cytoplasmic domain) that are likely to be contiguous on one lateral face of the receptor. Although certain mammalian GPCRs and the ␣-factor receptor may use similar domains to dimer/oligomerize, there is no significant sequence homology between the extracellular N-terminal domain, TM1, and TM2 of the ␣-factor receptor and mammalian GPCRs (data not shown). Presumably, different sequence motifs in these domains allow these receptors to form specific dimeric-oligomeric complexes with themselves and/or a restricted number of other GPCRs.
Other domains of GPCRs also may mediate dimer/oligomerization. Studies of chimeric mammalian GPCRs (19), inhibition ␤ 2 -adrenergic receptor dimerization in vitro by TM6 peptides (12), evolutionary trace analysis, and correlated mutations among related mammalian GPCRs (22) have provided indirect evidence suggesting that TM5 and/or -6 are involved in dimer formation, either as contact or domain-swapped dimers. Assuming various GPCRs indeed use different domains for dimer/ oligomer formation, this may provide an additional mechanism whereby GPCRs form specific homo-and/or heterodimeric complexes, thereby restricting the combinations of GPCRs that can be used to generate functionally novel classes of receptors. Detailed analysis of the structural and sequence motifs in dimer/oligomerization domains will be necessary to understand more fully the mechanisms that direct the formation of specific GPCR dimers/oligomers.