Expression of Dopamine D3 Receptor Dimers and Tetramers in Brain and in Transfected Cells*

The expression and characteristics of the dopamine D3 receptor protein were studied in brain and in stably transfected GH3 cells. Monoclonal antibodies were used for immunoprecipitation and immunoblot experiments. Immunoprecipitates obtained from primate and rodent brain tissues contain a low molecular weight D3 protein and one or two larger protein species whose molecular mass are integral multiples of the low molecular weight protein and thus appear to have resulted from dimerization and tetramerization of a D3 monomer. Whereas D3receptor multimers were found to be abundantly expressed in brain, the major D3 immunoreactivity expressed in stable D3-expressing rat GH3 cells was found to be a monomer. However, multimeric D3 receptor species with electrophoretic mobilities similar to those expressed in brain were also seen in D3-expressing GH3 cells when a truncated D3-like protein (named D3nf) was co-expressed in these cells. Furthermore, results from immunoprecipitation experiments with D3- and D3nf-specific antibodies show that the higher-order D3 proteins extracted from brain and D3/D3nf double transfectants also contain D3nf immunoreactivity, and immunocytochemical studies show that the expression of D3 and D3nfimmunoreactivities overlaps substantially in monkey and rat cortical neurons. Altogether, these data show oligomeric D3 receptor protein expression in vivo and they suggest that at least some of these oligomers are heteroligomeric protein complexes containing D3 and the truncated D3nfprotein.

D 3 receptors belong to the D 2 -class of dopamine receptors known to couple to inhibitory subsets of heterotrimeric G proteins. The members of the superfamily of G-protein-coupled receptors have a predicted membrane topology of seven hydrophobic transmembrane domains connected by alternating extra-and intracellular loops. Although it is generally thought that functional G-protein-coupled receptors contain a single receptor molecule, several recent observations suggest that these receptors also exists as oligomers. For example, muscarinic M2 receptor proteins purified from porcine atria contained, in addition to the monomer, multiples of the monomeric receptor with electrophoretic mobilities suggesting trimeric and tetrameric homoligomers (1). Furthermore, the cooperative ligand binding profile of purified M2 receptors described by Wreggett and Wells (1) appears to fit best a model that assumes a tetrameric configuration of this receptor. Recently, Hebert et al. (2) showed that ␤ 2 -adrenergic receptors form homodimers in transfected cells. These homodimers were resistant to SDS and reducing agents. Most interestingly, this dimerization was found to be essential for ␤ 2 -receptor-mediated stimulation of adenylyl cyclase, and agonist stimulation stabilized the dimeric receptor configuration, whereas inverse agonists favored the monomeric state (2).
Previous immunoblot and immunoprecipitation analyses of a variety of different G-protein-coupled receptors (expressed in insect or mammalian cells) also suggested that these receptors are capable of forming homoligomers. For example, dopamine D 2 receptor immunoprecipitates contained two D 2 -immunoreactive protein species, the larger (ϳ93 kDa) being twice the size of the smaller (ϳ44 kDa) (3). Similarly, dopamine D 1 receptors were found to migrate at ϳ50, 100, and 200 kDa (4) on SDS-PAGE. 1 In addition, oligomers of the m5-HT 1b receptor (5), the mGluR1 receptor (6), the substance P receptor (7), the human C5a anaphylatoxin receptor (8), and the platelet activating receptor (9) were resolved under reducing conditions on SDS-PAGE. Although these results in cell lines were often interpreted as nonspecific aggregation of incompletely folded intermediates, in each case the higher molecular weight species appeared to comprise multiples of a monomer.
Earlier studies on chimeric receptors already suggested the possibility that functional G-protein-coupled receptors can be comprised of multiple receptor molecules. For example, Maggio et al. (10) made chimeric ␣ 2 -adrenergic and M3 muscarinic receptors by replacing transmembrane domains VI and VII of one receptor with the corresponding domain of the other receptor. Although neither chimera was functional when expressed alone, functional receptors were formed when the two chimera were co-expressed. These results demonstrate that intermolecular interactions can occur between G-protein-coupled receptors and that the resultant oligomeric receptor complexes are functional.
Whether G-protein-coupled receptors, in cells in which they are natively expressed, exist in monomeric or oligomeric configurations is still unresolved. However, the demonstration of functional differences between dimers and monomers reported by Wreggett and Wells (1) and Hebert et al. (2) points to the importance of understanding the molecular composition of such receptors in vivo. The present study shows an oligomeric ex-pression of the dopamine D 3 receptor protein in brain which, at least in part, resulted from heteroligomerization of the D 3 protein and the truncated D 3 -like protein D 3nf (see Refs. 11 and 12).

EXPERIMENTAL PROCEDURES
Protein Extraction, Immunoprecipitations, Treatment of Proteins with N-Glycosidase, and Immunoblotting-Our protocols for the extraction of proteins from transfected cells, SDS-PAGE, and Western blotting were previously described (12,13). In addition, membranes were prepared from human brain by homogenizing 0.5 g of postmortem tissue in an ice-cold buffer containing 25 mM NaCl, 2 mM EDTA, and 50 mM Tris (pH 7.6). The homogenate was first centrifuged at 4°C for 10 min at 500 ϫ g to pellet nuclei, and the resultant supernatant was then centrifuged at 100,000 ϫ g for 1 h at 4°C to obtain a pellet containing membranes. This pellet was resuspended in the same buffer, incubated at 37°C for 10 min, and centrifuged again at 100,000 ϫ g for 30 min at 4°C.
For immunoprecipitation experiments, proteins from human, monkey, and rat brain tissue homogenates or membrane preparations and from transfected rat GH3 cells were solubilized in a buffer containing 0.1 M Tris (pH 8), 0.15 M NaCl, 0.001 M EDTA, 0.5% Nonidet P-40, 1% Triton X-100 supplemented with the protease inhibitors aprotinin (2 g/ml) and phenylmethylsulfonyl fluoride (1 mM). Solubilized protein samples were precleared with either protein G or protein A-agarose (40 l; Boehringer Mannheim) and subsequently incubated with primary antibody (a D 3 -specific monoclonal antibody IgG or a D 3nf -specific polyclonal antibody; see below) at 4°C for 15 h. After adding 40 l of protein G-or protein A-agarose slurry, the incubation was continued for at least 2 h. The immunoprecipitate was washed 4 times in a buffer containing 50 mM Tris pH 7, 150 mM NaCl, 5 mM EDTA, and 1% Triton X-100, followed by 2 additional washes in the same buffer without Triton X-100. The protein A-or G-antibody-antigen complexes were boiled for 5 min in 1 ϫ Laemmli buffer (containing 5% ␤-mercaptoethanol) and analyzed by Western blotting (probed with the polyclonal D 3nf -specific antibody and the D 3 -specific monoclonal antibody (IgM; see below)). Bound antigen was visualized using the appropriate peroxidase-conjugated secondary antibodies (goat anti-rabbit IgG, goat anti-mouse IgG, or goat anti-mouse Ig(GϩM); Kirkegaard & Perry Laboratories, Gaithersburg, MD) in conjunction with enhanced chemiluminescence (Pierce, Rockfort, IL).
Immunoprecipitated proteins subjected to treatment with N-glycosidase were separated from the agarose beads and antibody by boiling for 5 min in 1 ϫ Laemmli buffer containing 2% SDS. After boiling, samples were centrifuged for 10 min to collect the supernatant. Equal volumes of a buffer containing 100 mM sodium phosphate (pH 7.2), 20 mM EDTA, and 1% Nonidet P-40 was added to the supernatant. Samples were boiled again for 2 min and allowed to cool down to room temperature. N-Glycosidase F (0.6 to 1 units; Boehringer Mannheim) was then added and the samples were incubated at 37°C for 16 h.
Antibodies-The rabbit polyclonal antibody raised against the unique 60-amino acid residue-long carboxyl-terminal of the human D 3nf protein was previously characterized (12). Two monoclonal D 3 antibodies (IgG and IgM) were raised against a fusion protein incorporating amino acid residues 252-284 of the putative third cytoplasmic loop of the human D 3 receptor (14) fused in-frame to the carrier protein glutathione S-transferase. The fusion protein was expressed in bacteria and purified by affinity chromatography using glutathione according to the protocol provided by the manufacturer (Pharmacia Biotech, Piscataway, NJ). BALB/c mice were immunized with 50 g of purified fusion protein. The animal with the highest titer for binding to the fusion protein was boosted with 24 g of fusion protein administered intravenously. Single-cell suspensions of spleen cells were mixed with SP2/0 myeloma cells in serum-free media and fused using polyethylene glycol 4000. Cells were deposited in Iscove's medium (10% fetal calf serum, G-418, sodium pyruvate, L-glutamine, hypoxanthine, and thymidine). After selecting for fused cells with azaserine, supernatants from wells exhibiting growth were tested for the presence of antibodies which bound to the fusion protein, but not to glutathione S-transferase alone. Positive cell lines were cloned by limiting dilution. Ascites were then prepared by injecting 10 6 hybridoma cells into the peritoneal cavity of pristane-primed BALB/c mice.
For immunoblotting experiments, the D 3nf -specific antiserum was used at a dilution of 1:2000, the D 3 monoclonal antibody (IgG, supernatant) at a dilution of 1:100, whereas the D 3 monoclonal antibody (IgM, ascites) was used at a dilution of 1:500. For immunoprecipitation experiments, 20 l of the IgG supernatant or 10 l of the D 3nf -specific antiserum were added to 1 ml of protein (7-9.5 mg/ml) solubilized in immunoprecipitation buffer.
Generation of Stably Transfected GH3 Cell Lines-Experiments were performed on stably transfected rat GH3 cells in which the expression of D 3 , D 3nf , or both is under the transcriptional control of a tetracyclineresponsive promotor. The generation and characterization of the D 3expressing rat GH3 cell line used in this study was described previously (13). D 3nf -expressing GH3 cells were generated in a similar manner by transfection via lipofection of the plasmid pTA-N (encoding a tetracycline-responsive transactivator and the neomycin-resistance gene) and pTET-Spl-D 3nf (see Ref. 13). Double transfectants were generated by transfecting the plasmid pTET-Spl-Hyg-D 3nf (encoding D 3nf and the hygromycin-resistance gene) into the stably D 3 -expressing and G-418resistant GH3 cell line. Selection for double transfectants was obtained by supplementing the media (50% Ham's F-10, 50% Dulbecco's modified Eagle's medium, 15% horse serum) with G-418 (0.3 mg/ml) and hygromycin B (0.5 mg/ml). The expression of D 3 and D 3nf mRNA and protein was verified by Northern blotting (D 3 mRNA), S1 nuclease protection assays (using D 3nf -specific splice-junctional oligoriboprobes), and immunoblots (D 3 , D 3nf ).
Immunocytochemistry in Macaque Monkey and Rat Brain-Four macaque monkeys (2 Macaca fascicularis and 2 Macaca mulatta), and four rats were used for the immunocytochemical studies. No qualitative differences in staining were observed between the two monkey species. All experimental protocols were conducted according to NIH guidelines for animal research and were approved by the Institutional Animal Care and Use Committee (IACUC) at the Mount Sinai School of Medicine. The animals were perfused transcardially as described previously (15,16). Briefly, the monkeys were deeply anesthetized with ketamine hydrochloride (25 mg/kg) and sodium pentobarbital (20 -35 mg/kg intravenously), intubated, and mechanically ventilated. The chest was then opened, the heart exposed, and 1.5 ml of 0.1% sodium nitrite was injected into the left ventricle. The descending aorta was clamped and the animals were perfused transcardially with cold 1% paraformaldehyde in PBS and then for 8 -9 min with cold 4% paraformaldehyde in PBS. Following perfusion, the brain was removed from the skull, cut into 1 cm-thick blocks and postfixed in 4% paraformaldehyde for 6 h at 4°C. It was then immersed in PBS until processing. Rats were anesthetized with xylazine and ketamine (10 and 30 mg/kg, respectively, intramuscularly) and then perfused transcardially with 1% paraformaldehyde (1 min) and 4% paraformaldehyde (8 min).
All tissues were cut at 50 m on a vibratome. Rat brains were cut either coronally or parasagittally. In the monkeys, the following blocks were taken: a parasagittal block through the central sulcus at the level of the arcuate spur containing portions of both the primary motor cortex and the primary somatosensory cortex and a coronal block through the principal sulcus (prefrontal cortex). These neocortical areas were chosen because they are among the areas receiving the richest dopaminergic innervation in the monkey brain (17)(18)(19)(20). Corresponding regions were analyzed in rats.
The free-floating tissue sections were incubated overnight at 4°C in primary antibody (D 3nf , diluted 1:1000 in PBS or 1:500 for fluorescence microscopy; anti-D 3 IgM, ascites: diluted 1:100 to 1:250). Sections were then rinsed with PBS and placed into a biotinylated anti-rabbit IgG or a anti-mouse Ig(GϩM) secondary antibody solution (Vector Laboratories, Burlingame, CA; 1:200) for 1 h. Some sections were double labeled with the anti-D 3 (IgM) antibody and the D 3nf -specific antibody. For brightfield microscopy, the sections were processed with the avidinbiotin method (using a Vectastain ABC kit (Vector Laboratories) and 3,3Ј-diaminobenzidine as a chromogen), rinsed, mounted onto gelatincoated slides, air dried, and immersed in 0.067% OsO 4 for 8 min to intensify the 3,3Ј-diaminobenzidine. Sections analyzed by fluorescence and laser scanning confocal microscopy were placed into a solution containing fluorescein isothiocyanate-avidin D (1:200) for 2 h. For sections that were incubated in both anti-D 3nf and anti-D 3 (IgM) antibodies, rhodamine-conjugated anti-rabbit IgG secondary antibody (Boehringer Mannheim; 1:100) was added to the biotinylated anti-mouse IgG secondary antibody. Prior to visualization, the sections were coverslipped with Vectashield mounting medium. A Zeiss Axiophot microscope equipped with appropriate filters to visualize fluorescein isothiocyanate and rhodamine was used for brightfield and epifluorescence microscopy. For confocal analysis, sections were scanned on a laser scanning confocal microscope (Zeiss LSM 410) using Zeiss NeoFluar ϫ 63 or ϫ 40 objectives (Zeiss, Oberkochen, Germany). Rhodamine was visualized using an ArKr 488/568 laser with a 568-nm excitation filter and a 590-nm long-pass emission filter; fluorescein isothiocyanate was visualized with a 488-nm excitation filter and a 515-540-nm bandpass emission filter.

RESULTS
Polyclonal and monoclonal antibodies that were raised against peptide sequences specific for the human D 3 receptor were used to characterize the properties and distribution of D 3 receptor species expressed in mammalian brain and in stably transfected rat GH3 cells.
Characterization of the Antibodies-The characterization of one of the D 3 receptor antibodies used in this study has been reported previously (13). This antipeptide antibody, which was raised against the amino terminus of the human D 3 receptor (Cambio, Cambridge, UK), detected D 3 immunoreactivity in immunoblot experiments using proteins extracted from rat GH3 cells that stably express the human dopamine D 3 receptor under the transcriptional control of a tetracycline-responsive promotor. The expression of D 3 receptor mRNA and protein in these stably transfected cells is suppressed by including tetracycline in the culture medium and reaches steady-state levels 24 h after induction (13). Two novel monoclonal antibodies (IgM/D3 and IgG/D3), which were raised against peptide sequences constituting amino acid residues 252-284 that are part of the putative third cytoplasmic loop of the human D 3 protein, were then tested for their ability to detect the same approximately 50-kDa D 3 -immunoreactivity recognized by the polyclonal antipeptide antibody raised against the amino terminus of the D 3 protein. As shown in Fig. 1A, all three antibodies recognize the same protein of approximately 50 kDa, which appears on immunoblots of proteins extracted 5-9 h after induction of D 3 expression in stably transfected GH3 cells. This D 3 immunoreactivity reaches steady-state levels 24 h after the induction of expression and is not detected in non-transfected GH3 cells. The expression of this D 3 immunoreactivity is abolished 3 days after the inhibition of D 3 mRNA expression by the addition of tetracycline (2 g/ml) to the culture medium ( Fig. 1A).
In addition, we synthesized a T7 RNA polymerase transcript of the human cDNA encoding the D 3 receptor (cloned into the plasmid vector pRc/CMV; Invitrogen) and translated this mRNA in vitro in rabbit reticulocyte lysates. The resulting proteins were analyzed by immunoblotting using monoclonal antibodies IgM/D3 and IgG/D3 as well as the polyclonal anti-D 3 antiserum. As shown in Fig. 1B, all three antibodies recognized an in vitro translation product that also migrated at approximately 50 kDa. When D 3 mRNA was omitted from the in vitro translation reaction, no D 3 immunoreactivity was detected on immunoblots, as shown in Fig. 1B for the polyclonal anti-D 3 antiserum. It is noted in both experiments on transfected GH3 cells and on in vitro translation products that all three antibodies also recognize smaller protein species of about 37 and 30 kDa. These proteins are likely to be degradation products of the 50-kDa D 3 -immunoreactivity because they are not seen in nontransfected cells and they are also absent from in vitro translation reactions lacking D 3 mRNA. Because these shorter protein species are also detected with the polyclonal antibody that recognizes amino-terminal D 3 peptide sequences it is likely that these proteins represent D 3 degradation products that lack carboxyl-terminal peptide sequences. The appearance of these putative degradation products is not due to insufficient supplementation of the SDS-containing solubilization buffer with protease inhibitors because they were detected in protein samples prepared in the presence of phenylmethylsulfonyl fluoride and aprotinin, and in samples that were additionally supplemented with pefabloc, pepstatin, and leupeptin (13). Altogether, the results shown in Fig. 1 demonstrate that all three antibodies (which were raised against non-overlapping peptide sequences of the human D 3 receptor) recognize the same D 3 immunoreactivity.
An additional polyclonal antibody that specifically recognizes a unique carboxyl-terminal peptide sequence contained only within the truncated D 3 -like protein named D 3nf (11), but not within the D 3 protein, was used in some experiments. The characterization of this antibody has been reported previously (12).
D 3 Receptor Immunoreactivity Expressed in Brain-The monoclonal antibodies described above were used to characterize D 3 immunoreactivity expressed in brain tissue. Because of the low levels of the D 3 protein, immunoprecipitation experiments were performed using the monoclonal antibody (IgG/ D3). The composition of the immunoprecipitate was analyzed on immunoblots using the monoclonal IgM/D3 antibody.
We first analyzed the expression of D 3 immunoreactivity expressed in human motor cortex. Fig. 2A shows the immunoprecipitates obtained with the IgG/D3 antibody. A protein of ϳ50 kDa is clearly detected. However, an additional major D 3 immunoreactivity of more than 180 kDa (ϳ200 kDa), and a smear of minor proteins migrating just below the largest protein, are also detected. The D 3 immunoprecipitate of solubilized total human brain proteins (Fig. 2A, lane 2) was then compared with the immunoprecipitate obtained from a membrane prep- On all three blots, the lanes marked 5, 9, 24, and 48 contain proteins extracted from cells 5, 9, 24 and 48 h, respectively, after induction of D 3 receptor expression by withdrawal of tetracycline from the culture medium. Five g of total cellular protein was loaded onto each lane. B, immunoblot of proteins translated in vitro from D 3 mRNA. An in vitro transcribed full-length human D 3 receptorencoded mRNA was incubated with a rabbit reticulocyte lysate translation mixture. A fraction of the translation mixture was electrophoretically separated on a 10% SDS-PAGE gel and transferred onto Immobilon polyvinylidene difluoride membrane. The blot on the left, marked IgM, was probed with the monoclonal IgM/D3 antibody. The blot in the middle, marked IgG, was probed with the monoclonal IgG/D3 antibody. The blot on the right was probed with the polyclonal antipeptide antibody. The lane marked pc contains a translation mixture that was incubated in the presence of RNA. The lane marked with a minus (Ϫ) contains a translation mixture that was incubated in the absence of RNA. All three antibodies recognize an approximately 50-kDa translation product. Bound antigens were visualized with the appropriate horseradish peroxidase-conjugated secondary antibodies used in conjunction with enhanced chemiluminescence (Pierce). aration of the same tissue ( Fig. 2A, lane 1; see "Experimental Procedures"). Also proteins solubilized from the membrane pellet contained the two major protein species indicating that both D 3 protein species are integral membrane proteins. In lieu of the results shown in Fig. 1, the detection of the high molecular mass D 3 immunoreactivity, which appears to be a multiple of the ϳ50-kDa D 3 protein, was unexpected. It should be noted, however, that even extensive post-translational modifications of the ϳ50-kDa protein in vivo are unlikely to account for the mass of the additional large protein. For example, although dopamine receptors are thought to be extensively N-glycosylated, treatment of the D 3 immunoprecipitate with N-glycosidase F prior to blotting did not abolish the detection of the ϳ200-kDa D 3 protein (Fig. 2A, lane 3). It did, however, abolish the smear of proteins seen directly below this protein species which revealed two sharp protein bands of slightly lower molecular mass. Furthermore, a doublet of proteins of ϳ50 kDa is detected after treatment of N-glycosidase F, indicating that removal of N-linked sugars in a substantial portion of the D 3 proteins results in only a small shift of their electrophoretic mobility.
In further experiments, we tested the ability of the IgG/D3 antibody to immunoprecipitate the D 3 protein expressed in monkey and rat brain. As shown in Fig. 2B, the antibody also immunoprecipitated the D 3 protein expressed in these two species, and in both species the protein composition of the immunoprecipitate is very similar. In addition to a ϳ45-kDa protein, two larger proteins are contained in the immunoprecipitate that migrate at approximately 85 and 180 kDa. D 3immunoreactive signals were not detected when solubilized proteins from spleen or muscle tissue (which do not express D 3 receptors) were incubated with the monoclonal IgG/D3 antibody (Fig. 2C).
The size of the 45-kDa D 3 immunoreactivity obtained from monkey and rat brain tissue corresponds to the calculated molecular mass of the D 3 core protein. Similar to the composition of the D 3 immunoprecipitate obtained from human brain tissue, the two larger protein species appear to be multiples of the 45-kDa protein. The results, therefore, suggest that the D 3 protein may be expressed as a monomer, a dimer, or a tetramer in rodent and primate brain. Interestingly, in all three species examined the tetrameric configuration of the D 3 protein is the most abundant protein species present in the immunoprecipitate. In fact, only the tetrameric, but not the dimeric, D 3 protein configuration was detected in human brain ( Fig. 2A). In all three species, these higher-order structures of the D 3 receptor were shown to be resistant to reducing agents (thus suggesting that monomers are not linked to each other via disulfide bonds) and were maintained after exposure to SDS. D 3 Receptor Immunoreactivity Expressed in Stably Transfected rat GH3 Cells-The results shown in Fig. 2 motivated us to examine in more detail whether higher-order structures of the D 3 protein can also be detected in a single type of cell in which the level of D 3 receptor expression can be controlled. We therefore re-examined the GH3 cells that stably and inducibly express D 3 mRNA and protein. Fig. 3A (middle panel) shows the results of immunoblot experiments from three independent GH3 cell clones in which the expression of the D 3 protein was maximally induced. In each of these clones, however, the major D 3 immunoreactivity (detected by the monoclonal IgG/D3 antibody) is that of the approximately 50-kDa immunoreactivity also seen in Fig. 1. Some higher-order structures of the D 3 protein are also expressed albeit at very low levels. For example, low expression of an ϳ90-kDa D 3 immunoreactivity is detected and the largest band, although barely visible, migrates at about 180 kDa. (These faint bands are best seen in Fig. 3B, last two lanes.) Thus, in contrast to the results we obtained with proteins extracted from brain tissue, the major D 3 protein species detected in these cells appears to be a monomer. It should be noted that D 3 protein expression has been characterized in a larger number of independent clones during various levels of induction, steady-state expression, or inhibition of expression. In all these clones we have never observed significant levels of expression of higher-order structures of the D 3 protein. Interestingly, however, high levels of D 3 immunoreactivity with an electrophoretic mobility of approximately 180 kDa was detected on immunoblots of proteins from cells that co-express the D 3 receptor and a truncated D 3 -like protein (named D 3nf ) which differs from the D 3 protein only in its carboxyl-terminal peptide sequence and therefore is unable to form transmembrane spanning domains VI and VII (11,12). An immunoblot analysis of D 3 protein expression in 5 randomly picked G-418-and hygromycin-resistant double-transfected GH3 cell clones (see "Experimental Procedures") is shown in Fig. 3A (right panel). Three of the 5 clones analyzed express, in addition to the 50-kDa D 3 immunoreactivity also seen in single FIG. 2. D 3 -immunoreactivity expressed in human, monkey, and rat brain. A, D 3 immunoprecipitate of human motor cortical tissue. Lane 1, D 3 immunoprecipitate obtained from 10 mg/ml protein of motor cortical membrane pellets. Lane 2, D 3 immunoprecipitate obtained from 6.5 mg/ml protein of total motor cortical tissue homogenate. Lane 3, the D 3 immunoprecipitate of motor cortical tissue protein was treated with 1 unit of glycosidase F prior to gel electrophoresis. B, D 3 immunoprecipitate obtained from monkey and rat brain tissues. Lane 1, D 3 immunoprecipitate of proteins extracted from monkey basal ganglia. Lane 2, D 3 immunoprecipitate obtained from proteins extracted from rat prefrontal cortex. Proteins were immunoprecipitated with the monoclonal IgG antibody. C, D 3 immunoprecipitates of different rat tissues. Lane 1, prefrontal cortex; lane 2, rat muscle; lane 3, rat spleen. All proteins were immunoprecipitated with the monoclonal IgG/D3 antibody. The immunoblots of these precipitates were probed with the monoclonal IgM antibody. Bound antigens are visualized as described in the legend to Fig. 1. D 3 transfectants, a major D 3 immunoreactivity migrating at approximately 180 kDa. Just as this result is different from the one obtained with single D 3 transfectants, it is also different from the results obtained with single D 3nf transfectants in which the major D 3nf immunoreactivity, recognized by a D 3nfspecific polyclonal antiserum (12), migrates at approximately 45 kDa (Fig. 3A, left panel).
The expression of the 180-kDa D 3 immunoreactivity does not appear to be the result of co-transfection per se, because it was not detected in D 3 -expressing GH3 cells that also express high mRNA levels of the (homologous) human D 2 receptor (Fig. 3B,  third lane). Furthermore, we observed the 180-kDa D 3 immunoreactivity in a GH3 cell clone in which both D 3 and D 3nf are expressed at low levels (Fig. 3B, first lane) as well as in clones in which both proteins are expressed at very high levels (Fig.  3B, second lane). However, in clones with high levels of D 3 , but very low levels of D 3nf , the expression of the 180-kDa protein was not detected. This is, for example, the situation found for protein samples of the two G-418 and hygromycin-resistant GH3 clones loaded onto first and fifth lanes of Fig. 3A (right  panel). (Note that the gel shown in Fig. 3B is a 7% SDS-PAGE while the onces shown in Fig. 3A are 10 and 12% gels. This results in differences in the banding thickness of the 50-kDa protein.) In total, these results suggest that in GH3 cells the formation of the 180-kDa D 3 immunoreactivity is specifically promoted by expression of the D 3nf protein. The results also suggest that it is the molar ratio of the two proteins, and not their absolute level of expression, that determines the appearance of the higher-order D 3 species; and the results do not therefore support the conclusion that the 180-kDa D 3 immunoreactivity results from nonspecific protein aggregation due to overexpression of the transfected proteins ("molecular crowding").
In further experiments, proteins from the three double transfectants that express the 180-kDa D 3 immunoreactivity were immunoprecipitated with the monoclonal IgG/D3 antibody and the protein composition was analyzed on immunoblots probed with the monoclonal IgM/D3 antibody (Fig. 3C). As shown in Fig. 3C, in each of the 3 clones, strong immunoreactive signals of approximately 180, 90, and 50 kDa were detected with monoclonal antibody IgM/D3. Thus, the pattern of D 3 immunoreactivity contained in the immunoprecipitate of proteins from these D 3 /D 3nf double transfectants is very similar to that seen in immunoprecipitates of proteins extracted from brain tissues (Fig. 2). It is noted, however, that the dimeric D 3 immunoreactivity is only detected in immunoprecipitation, but not immunoblotting, experiments on proteins of the GH3 double transfectants and that it is also not detected in immunoprecipitates of proteins extracted from human brain (Figs. 2 and 3). Although the reason for this discrepancy remains to be resolved, it is likely that the pellets of the immunoprecipitates contained mostly the 180-kDa protein complex which, during the preparation for SDS-PAGE analysis, could then be dissociated into its various protein constituents and that the detection of the dimeric D 3 protein complex is therefore only possible when large amounts of the 180-kDa complex (i.e. those found in transfected cells, but not in brain tissue) are expressed.
Protein Composition of the Higher-order D 3 Immunoreactivity in Brain and in D 3 /D 3nf -expressing GH3 Cells-One possible explanation for the appearance of the ϳ90and 180-kDa protein species in co-transfected cells is that D 3 and D 3nf form heteroligomers. Both proteins have identical amino acid sequences that extend into the amino-terminal sequence of the putative third cytoplasmic domain, and thus include transmembrane spanning domains I through V. Although the D 3nf protein is unlikely to have the additional transmembrane spanning domains VI to VII found in the D 3 receptor protein, previous studies have shown molecular interactions between split G-protein-coupled receptor proteins (21) or chimeric G-proteincoupled receptors (10). The following experiments therefore tested whether the ϳ85/90and 180-kDa D 3 immunoreactivities seen in immunoprecipitates of proteins extracted from D 3 /D 3nf -expressing GH3 cells and, most importantly, from brain tissue also contain D 3nf immunoreactivity.
The results shown in Fig. 4 were obtained with proteins extracted from rat prefrontal cortex (lane 1) and from GH3 D 3 /D 3nf double transfectants (lane 2) that were immunoprecipitated with the monoclonal IgG/D 3 antibody. The immunoblot of these precipitates was probed with the D 3nf -specific polyclonal antibody. Both the 180-and the ϳ85/90-kDa protein species, but not the 50-kDa D 3 monomer, gave D 3nf -immunoreactive signals. A small fraction of the pellets of these immunoprecipitates (approximately 10%) was analyzed on immunoblots probed with the monoclonal IgM/D3 antibody. The results (not shown) are identical to the results shown in Figs. 2 and 3C, e.g. all three D 3 -immunoreactive protein species, including the ϳ50-kDa monomer, were detected. In addition, proteins extracted from the rat brain were immunoprecipitated with the D 3nf -specific polyclonal antiserum and the immunoblot of this precipitate was probed with the monoclonal IgM/D3 antibody. These co-expressing cells were obtained after co-transfecting D 3nf and D 2 , respectively, into the parent D 3 -expressing clone. Left blot, the first two lanes, marked D 3 /D 3nf , contain proteins extracted from two independent clones with low (first lane) and high (second lane) levels of D 3 and D 3nf expression. The lanes marked D 3 /D 2 and D 3 contain proteins extracted from cells that co-express D 3 and D 2 receptors and from cells that express only D 3 receptors, respectively. Five g of total cellular proteins were separated by SDS-PAGE and the blot was probed with the monoclonal IgG antibody. All cells are in the fully induced state. C, immunoprecipitation of proteins extracted from D 3 /D 3nf double transfectants. Proteins extracted from the three double transfectants that express the 180-kDa D 3 immunoreactivity were immunoprecipitated with the monoclonal IgG antibody. The immunoprecipitate was electrophoresed on SDS-PAGE. The blot was probed with the monoclonal IgM antibody (left). Bound antigens are visualized as described in the legend to Fig. 1. Fig. 4 (lane 3), both the higher-order D 3 species, but not the 45-kDa D 3nf immunoreactivity seen in single D 3nf transfectants (see Fig. 3A), are also recognized by the D 3specific antibody. (In general, however, we found the D 3nf antiserum less suitable for immunoprecipitation experiments because such an immunoprecipitate also contains additional bands that are weakly recognized by the monoclonal IgM/D3 antibody. Whether these bands result from a partial degradation or partial dissociation of the immunoprecipitated 180-kDa protein complex remains to be elucidated.)

As shown in
Immunocytochemical Analysis of D 3 and D 3nf Protein Expression-The results shown above suggest a heteroligomerization of the D 3 and D 3nf proteins. A prerequisite for such an intermolecular interaction between D 3 and D 3nf in vivo must therefore be that both proteins co-localize in the same neuron. The following series of immunocytochemical experiments sought to address whether, and to what extent, D 3 and D 3nf immunoreactivities overlap in monkey and rodent neurons.
Light microscopic studies revealed that both D 3 -and D 3nfimmunoreactive structures are abundantly found in rat and monkey neocortex. No significant differences were observed in the overall pattern of immunoreactivity for the two species. D 3 immunoreactivity was present in all neocortical areas studied, and it was most pronounced in neuronal somata (Figs. 5, A and C; 6, A and C). Dendritic immunostaining was also evident, which varied on a laminar and regional basis, with the greatest density in areas of most dense dopaminergic innervation, such as the primary motor and prefrontal cortex (data not shown). D 3nf immunoreactivity was found in every cortical area that contained D 3 immunoreactivity. However, as shown in Figs. 5 and 6, its cellular distribution differed. As noted, D 3 immunoreactivity was found most prominently in neuronal perikarya, whereas D 3nf -like immunoreactivity (while present in perikarya) was most intense more distally in pyramidal neuron apical dendrites. In neocortex, areas in which intense D 3 receptor somatic labeling was evident also contained thick D 3nfimmunoreactive dendrites, as well as fine (presumably dendritic) processes throughout the neuropil. D 3nf -immunoreactive dendrites were frequently found in bundles. These dendrites were especially prominent in layer V, and they were also seen coursing through layers IV and III (Figs. 5D and 6, B and D).
Because of the noted difference in the most pronounced cellular distribution of D 3 and D 3nf immunoreactivity (as illustrated in Figs. 5 and 6), we examined with confocal microscopy whether both proteins show regions of co-expression in the same neuron. Indeed, confocal analysis of double labeled material confirmed the presence of both proteins in the same pyramidal-like neurons. In agreement with the results obtained with light microscopic analysis of immunolabeled tissue, D 3 immunoreactivity was concentrated in the somata, and D 3nf immunoreactivity was concentrated in the dendritic profiles. However, as shown in Figs. 7, C and E, the expression of both proteins also overlaps. Whereas D 3 immunoreactivity tapers off in intensity with distance from the soma, D 3nf immunoreactivity gradually increases, with the result that the region of greatest co-localization of both proteins is the proximal portion of the apical dendrite. DISCUSSION The present study shows that dopamine D 3 receptors expressed in brain exist as oligomers. Furthermore, results obtained with transfected GH3 cells suggested that the detection of D 3 receptor oligomers is significantly enhanced when the truncated D 3 -like protein D 3nf is co-expressed in these cells. Although the enhancing effect of D 3nf expression on the detection of D 3 oligomers may be unique to GH3 cells, results from these studies have pointed to the possibility of a D 3 /D 3nf heteroligomerization. Indeed, D 3 and D 3nf immunoreactivities expressed in brain and in D 3 /D 3nf -expressing GH3 cells co-precipitate in immunoprecipitation experiments, and both immunoreactivities co-migrate at approximately 85/90 and 180 kDa. The heteroligomerization of D 3 and D 3nf proteins in vivo is further supported by results of immunocytochemical studies demonstrating that D 3 and D 3nf proteins co-localize in distinct regions of cortical neurons.
Two novel monoclonal antibodies (raised against a peptide sequence of the human D 3 protein that constitutes a region of the unique third cytoplasmic domain of the receptor) were used in this study. The ability of these antibodies to detect the expression of recombinant human D 3 receptor proteins (translated either in vitro or stably expressed in GH3 cells with active tetracycline-responsive promotors), as well as the similarity of the results obtained with the monoclonal antibodies and with an antipeptide D 3 receptor antibody raised against a completely different (amino-terminal) epitope, suggest that these antibodies react specifically with the D 3 receptor protein. Furthermore, these antibodies do not cross-react with the truncated D 3 -like protein D 3nf (Figs. 3A and 4), and they fail to recognize the homologous D 2 receptor protein (Fig. 3B). It is worth noting, however, that these antibodies immunoprecipitate not only the D 3 protein expressed in human brain. As shown in Fig. 2B, they also immunoprecipitate the D 3 protein expressed in rodent and monkey brain tissue, and they produce similar immunocytochemical staining of monkey and rat brain tissue sections (Figs. 5 and 6). Because of the considerable divergence of the peptide sequence of the third cytoplasmic domain of the receptor protein between primates and rodents, it is likely that the monoclonal antibodies described here recognize either a distinct structural motif of the D 3 receptor protein (rather than a linear peptide sequence) that is common to both rodent and primates or that they recognize a very small number of amino acid residues. Neither scenario would be unusual for monoclonal antibodies.
The availability of D 3 -specific monoclonal antibodies and a polyclonal D 3nf -specific antibody that recognizes peptide sequences that are not present in the D 3 protein enabled us to use these antibodies sequentially in immunoprecipitation and Western blot experiments. Results from these studies showed that the smallest D 3 protein has a molecular mass ofϳ 50 kDa in human motor cortex, and ϳ45 kDa in rodent and primate brain. In addition to these low molecular mass proteins, however, two larger D 3 protein species of ϳ85 and ϳ180 kDa were immunoprecipitated from rodent and primate brain, and one protein species of ϳ200 kDa was detected in immunoprecipi- FIG. 4. Immunoprecipitations of proteins extracted from rat brain and D 3 /D 3nf -transfected GH3 cells. Lanes 1 and 2, proteins extracted from rat prefrontal cortical tissues (lane 1) and from D 3 /D 3nf double transfectants (lane 2) were immunoprecipitated with the monoclonal IgG/D3 antibody. The composition of the immunoprecipitate was analyzed on Western blots probed with the D 3nf -specific antibody. Lane 3, proteins from rat prefrontal cortex were immunoprecipitated with the D 3nf -specific antibody and the immunoprecipitate was analyzed on a Western blot probed with the monoclonal IgM/D3 antibody. Bound antigens are visualized as described in the legend to Fig. 1. tates of human brain tissues. These results suggest for the first time the existence of D 3 receptor dimers and tetramers in brain tissue. Surprisingly, similarly abundant oligomeric D 3 receptor species were not observed in GH3 cells singly transfected with cDNA encoding the D 3 receptor. However, D 3 oligomers were found to be abundantly expressed in GH3 cells that co-express the D 3 and D 3nf proteins. Furthermore, only in double transfectants in which the relative expression levels of the D 3 and D 3nf proteins were similar, a substantial proportion of the D 3 immunoreactivity is expressed in the tetrameric form (see Fig.  3, A and B). Whereas these results already suggested the possibility of a heteroligomerization of the D 3 protein, the dem-onstration that D 3 and D 3nf proteins co-precipitate in immunoprecipitation experiments provided further evidence for a physical interaction between D 3 and D 3nf molecules both in brain and in transfected cells (Fig. 4).
A presently unresolved issue is the relative contribution of the D 3 and D 3nf proteins to the oligomeric states of the D 3 protein. It is noted that a small amount of the D 3 protein expressed in single GH3 transfectants also appears to form higher-order structures suggesting that, in addition to the heteroligomerization described here, a proportion of the D 3 protein can also form homoligomers. Furthermore, because of the different nature of the different antibodies used in this study, we cannot determine the relative contri-FIG. 5. Low-magnification photomicrographs of dopamine D 3 receptor and D 3nf immunoreactivity in rat and monkey neocortex. In both rat (A and B) and macaque monkey (C and D) neocortex, D 3 immunoreactivity (A and C) is most pronounced in neuronal somata, and to a lesser extent in dendrites. D 3nf immunoreactivity (B and D) is also present in somata, but is much more intense in dendritic profiles, which often appear in bundles (arrows). Scale bar ϭ 100 m.
FIG. 6. High-magnification photomicrographs of D 3 and D 3nf immunoreactivity in rat and monkey motor cortex. At this magnification, the difference in localization of these two proteins is evident, with somata intensely immunoreactive for D 3 and more distal portions of the apical dendrite preferentially containing D 3nf (arrows). Letters A-D correspond to those of Fig. 5. Note the presence, in D, of lightly stained large neuronal somata adjacent to intensely labeled dendrites. Scale bar ϭ 25 m. butions of D 3 and D 3nf proteins to the higher-order structures of the D 3 immunoprecipitates.
The D 3 receptor oligomers described here are resistant to SDS and reducing agents. This finding is similar to results obtained for higher-order structures of other G-protein-coupled receptors (see above). Several studies have now shown that oligomers of G-protein-coupled receptors are characteristically of defined size and that they always comprise multiples of the monomer. Although random intermolecular interactions between hydrophobic molecules might be expected to yield similar large protein aggregates, such interactions would also be expected to yield aggregates of varying size and intermediate molecular mass, and if they resulted from random formations of disulfide bonds they would be disrupted by reducing agents. Furthermore, if all studies that identified oligomers of G-protein-coupled receptors suffer from the same artifactual oligomerization of such hydrophobic molecules in solution, one may expect that mixing protein extracts from two single transfectants would also result in apparent oligomerization. However, we have never observed higher-order D 3 species after mixing and boiling extracts of D 3 and D 3nf single transfectants in the presence of SDS (not shown). Finally, if the higher-order structures described in this study are in fact aggregates forming under denaturing conditions prior to electrophoresis (this procedure is identical for all samples) it would be difficult to explain the absence of such aggregates in protein samples obtained single GH3 transfectants. The weight of the evidence therefore suggests that the observed intermolecular interactions between D 3 and D 3nf molecules (and the likely interactions between D 3 molecules) are specific, a conclusion also supported by our finding that D 3nf , but not the homologous D 2 receptor, induces D 3 receptor oligomerization.
It has recently been shown that the heterodimeric assembly of a full-length and a truncated amino-terminal peptide requires only that the truncated peptide still possesses an aminoterminal hydrophobic transmembrane-spanning domain (22). Furthermore, it has been shown that a dimer with a single carboxyl terminus can still be functional (23,24). But what is the functional relevance of oligomerization of G-protein-coupled receptors and, more specifically, what role do D 3 /D 3nf heteroligomers play in vivo? Because G-protein-coupled receptors are integral membrane proteins, a prerequisite for a functional role of receptor oligomers is their presence in such membranes. As shown in Fig. 2A, both D 3 monomers and tetramers could be immunoprecipitated from membranes prepared from human brain tissues, demonstrating that oligomers are indeed inserted into membranes. Furthermore, our immunocytochemical analysis shows that D 3 and D 3nf immunoreactivities colocalize within individual neurons with overlapping, but also different intracellular distributions. The existence of functional heteroligomers could perhaps explain several puzzling characteristics of these proteins and their distribution. For example, the functionally competent form of the D 3 receptor protein is predominantly localized in the soma, whereas the bulk of dopaminergic terminals synapse on dendritic shafts and spines (25)(26)(27). Conversely, the D 3nf protein has an anatomic distribution consistent with that of a dopamine receptor, but it is not expected itself to act as a functional receptor. Thus, one role of the D 3nf protein, which is located more peripherally in the dendritic tree (and in the vicinity of dopaminergic afferents), could be to target a proportion of the D 3 receptor pool to those areas in which D 3nf is expressed. Thus the localization of the D 3nf protein, not the D 3 receptor protein itself, would signal the presence of functional heteroligomers. Interestingly, the ␤ 2adrenergic receptor has been shown to dimerize, and the presence of agonist promotes dimer formation (2). If such a mechanism is operable in the dopamine D 3 receptor, the localization of both D 3nf and D 3 receptor protein might be influenced by the activity of dopaminergic afferents impinging on different parts of the dendritic arbor. Thus, future studies on transfected polarized cells that express D 3 in the presence and absence of D 3nf will have to test whether D 3nf influences the trafficking of the D 3 protein.
Finally, different molecular sizes of oligomers might have differing anatomic distributions and affinities for receptor ligands. For example, if only the most abundant form, the tetrameric D 3 receptor, were recognizable by the D 3 -specific ligands currently in use, this could explain the discrepancies between immunocytochemical (28,29) and ligand-binding studies of receptor distribution (30 -33).
Although the functional consequences of D 3 receptor oligomerization remain to be elucidated, the results reported here suggest that D 3 receptor oligomers represent the predominant form of the receptor in brain. Results from transfected GH3 cells also suggest that co-expression of the D 3nf protein influences the assembly of the D 3 receptor, further suggesting that alterations of D 3nf expression may have important functional consequences for dopaminergic neurotransmission. FIG. 7. Confocal micrographs of D 3 and D 3nf immunoreactivity in rat sensorimotor neocortex. D 3 immunoreactivity is pseudocolored green (A and D), D 3nf immunoreactivity is pseudocolored red (B and E), and the images are overlaid in C and F, where colocalization is pseudocolored yellow. Note the dense D 3nf -immunoreactive meshwork in the neuropil (B and E), and the region of greatest colocalization in the proximal apical dendrite (arrows). The arrowhead indicates a portion of the proximal apical dendrite that is not visible because it is out of the confocal plane. Scale bar ϭ 30 m.