The Role of Receptor Oligomerization in Modulating the Expression and Function of Leukocyte Adhesion-G Protein-coupled Receptors*

The human leukocyte adhesion-G protein-coupled receptors (GPCRs), the epidermal growth factor (EGF)-TM7 proteins, are shown here to function as homo- and hetero-oligomers. Using cell surface cross-linking, co-immunoprecipitation, and fluorescence resonance energy transfer analysis of EMR2, an EGF-TM7 receptor predominantly expressed in myeloid cells, we demonstrate that it forms dimers in a reaction mediated exclusively by the TM7 moiety. We have also identified a naturally occurring but structurally unstable EMR2 splice variant that acts as a dominant negative modulator by dimerizing with the wild type receptor and down-regulating its expression. Additionally, heterodimerization between closely related EGF-TM7 members is shown to result in the modulation of expression and ligand binding properties of the receptors. These findings suggest that receptor homo- and hetero-oligomerization play a regulatory role in modulating the expression and function of leukocyte adhesion-GPCRs.

In recent years, we and others have identified a group of novel heptatransmembrane (7TM) 4 receptors, termed LNB-TM7, which contain a class B GPCR-related TM7 moiety and an unusually large N-terminal extracellular domain (1). Apart from the TM7 domain that might transduce intracellular sig-nals, the extracellular domain of LNB-TM7 molecules contains multiple repeats of protein modules, such as epidermal growth factor (EGF)-like, lectin-like, Ig-like, and cadherin-like motifs that are known to mediate protein-protein interaction (1,2). Hence, the LNB-TM7 receptors have also been classified recently as adhesion-GPCRs (3,4). Furthermore, members of the LNB-TM7/adhesion-GPCR family all have restricted expression patterns limited to certain cell types or tissues, such as leukocytes, smooth muscle, epididymal epithelium, or brain (1), implicating unique physiological roles. Finally, the majority of the adhesion-GPCRs are known to undergo a post-translational autocatalytic proteolytic modification at the GPCR proteolytic site (GPS). The GPS autoproteolysis cleaves the receptor into two subunits that associate noncovalently as a cell surface heterodimer (5)(6)(7). It has been shown that the GPS autoproteolytic modification is important for the maturation and trafficking of certain adhesion-GPCRs (6). Point mutations predicted to affect the GPS autoproteolytic process have been identified in receptor molecules that are responsible for human hereditary diseases, such as bilateral frontoparietal polymicrogyria and autosomal dominant polycystic kidney disease (8,9).
In view of their unique structural organization and expression patterns, the LNB-TM7/adhesion-GPCR proteins are thought to play important roles in cell type/tissue-specific functions through cellular adhesion mediated by the extracellular domain, followed by signal transduction via the TM7 domain. Indeed, the EGF-like modules of several EGF-TM7 molecules have recently been shown to mediate specific protein-protein interaction (10 -13), and the ␣-latrotoxin receptor, latrophilin, has been known to interact with G o proteins (14). Moreover, a number of the LNB-TM7 receptors have recently been implicated in novel biological functions, including cellular polarity (the flamingo gene in Drosophila (15,16) and Celsr1 in mice (17)), the regulation of dendritic maintenance and growth (mouse Celsr2) (18), epididymal fluid regulation and male fertility (HE6) (19), and the development of human frontal cortex (GPR56) (8). The mechanisms mediating the biological functions of the LNB-TM7/adhesion-GPCR proteins, however, remain unknown.
The self-association of proteins to form dimers and higher order oligomers is a key factor in the regulation of proteins, such as enzymes, ion channels, receptors, and transcription factors (20). In recent years, the concept that GPCRs exist and function as dimers/oilgomers has become more widely accepted through a combination of pharmacological, biochemical, and biophysical approaches (21)(22)(23). These include pharmacological profiling experiments (24), analysis of receptor size (25), co-immunoprecipitation (21), functional complementation experiments (26), and resonance energy transfer techniques (27). More recently, the definitive evidence for GPCR dimerization in vivo has been shown by the identification of a pharmacological ligand that could selectively target a tissuespecific GPCR (opioid) heterodimer (28). In particular, among all of the methods employed, experimental evidence for GPCR dimerization obtained using resonance energy transfer techniques has allowed the observation and analysis of the dimeric receptors under native, physiological conditions (27).
In the present study, we examine the structural organization of human EMR2, a myeloid-restricted EGF-TM7 molecule (29 -31), using multiple complementary approaches, including surface receptor cross-linking, co-immunoprecipitation, fluorescence resonance energy transfer (FRET), and functional modulation of co-expressed receptors. We demonstrate that EMR2 is expressed constitutively as a dimer, and the receptor dimerization is mediated exclusively by the 7TM moiety. EMR2-⌬S, an alternatively spliced EMR2 variant with an 11amino acid truncation in the stalk region, was identified to be structurally unstable and poorly expressed. EMR2-⌬S acts as a dominant negative modulator and down-regulates the cell surface expression of the full-length EMR2 through receptor dimerization and subsequent degradation. Finally, heterodimerization between closely related EGF-TM7 receptors is also demonstrated, and the functional implication in EGF-TM7 receptor biology is discussed.
Construction of Expression Vectors-All expression vectors were constructed using pcDNA3.1(ϩ) or pcDNA3.1/Myc-His vectors (Invitrogen) unless otherwise specified. EMR2-mFc expression constructs have been described previously (32). In some cases, the c-Myc epitope tag was replaced with the HA epitope (YPYDVPDYA). Specific cDNA sequences encoding EMR2 and CD97 full-length (FL) as well as site-directed mutant proteins described previously were used as templates. The 7TM/Myc and 7TM/HA constructs were made by cloning the EMR2 7TM region into pSecTag2A via the BamHI and XhoI sites. To make EMR2-pIRES-EGFP, the entire EMR2 coding sequence containing either the FL or the alternatively spliced stalk (⌬S) was inserted into SalI and SacII sites of the pIRES2-EGFP vector (BD-Clontech). The CD97(125)-pIRES-EGFP and rat thyrotropin-releasing hormone receptor (rTRHR)-pIRES-EGFP were made in a similar fashion. An N-terminal HA-tagged rTRHR (HA-rTRHR/pcDNA3) was a gift from Dr. A. Chakera of the Sir William Dunn School of Pathology. For FRET analysis, cDNAs of proteins of interest were cloned in frame immediately in front of two GFP variants, ECFP and EYFP, so that the enhanced GFP variants were tagged at the C terminus of the proteins. All constructs were confirmed by sequencing. The following controls were used throughout the FRET experiments: a negative control, ECFP ϩ EYFP, was a kind gift from Dr. J. McIlhinney (Department of Neuropharmacology, Oxford University). Tandem EYFP-ECFP constructs linked by 15 (CY-15) and 24 (CY-24) amino acids were kindly provided by Dr. Yuechueng Liu (Department of Pathology, University of Oklahoma Health Sciences Center) (33). EYFP-Ki6.7-ECFP was a gift from Dr. J. Lipiat (University Laboratory of Physiology, Oxford University) (34). The last three were used as a positive control.
Receptor Cross-linking Experiments-Receptor cross-linking by cell-impermeable cross-linker BS 3 (Pierce) was carried out according to the manufacturer's protocol. Briefly, transfected CHO-K1 cells were washed twice with PBS, followed by 30 min of incubation with 1 mM BS 3 in PBS at room temperature. The reaction was stopped by adding PBS/Tris-HCl, pH 7.5, buffer to a final concentration of 50 mM Tris-HCl and incubated for a further 15 min. Cells were then washed with cold PBS three times before collecting total cell lysate in Nonidet P-40 cell lysis buffer (20 mM Tris-HCl, pH 7.4, 1% Nonidet P-40, 5 mM MgCl 2 , 100 mM NaCl) supplemented with 1 mM sodium orthovanadate, 1 mM 4-(2-aminoethyl)-benzenesulfonyl fluoride, 5 mM Levamisole, and 1ϫ Complete TM protease inhibitor (Roche Applied Science) as described previously (5). Proteins were quantified using a bicinchoninic acid (BCA) protein assay kit (Pierce) according to the manufacturer's instructions.

Immunoprecipitation, Western Blotting, and Other Protein
Analysis-Transiently expressed EMR2 proteins were immunoprecipitated from cell lysate or conditioned medium (CM) using anti-c-Myc-conjugated agarose beads (A7470; Sigma) as described previously (5,7). Briefly, cell lysate (200 g) or CM (1 ml) was precleared with protein A-Sepharose (A7786; Sigma), followed by incubation with anti-c-Myc-conjugated agarose beads for at least 2 h at 4°C. After extensive washes with 1% bovine serum albumin/PBS, the immunocaptured proteins were subjected to Western immunoblotting analysis. For Western blotting, proteins were denatured and separated in 8 or 10% SDS-polyacrylamide gels, transferred to polyvinylidene difluoride membranes (Immobilon-P; Millipore), and probed with anti-Myc or anti-HA Abs. The blots were extensively washed and then incubated with appropriate horseradish peroxidaseconjugated secondary mAb for ECL detection (Amersham Biosciences).
Flow cytometric and immunocytochemical analyses were carried out using standard procedures as described previously (7,35). Briefly, transfected cells were fixed ϳ20 h post-transfection with 4% paraformaldehyde. Cells were blocked in PBS con- taining 4% normal goat serum and 1% bovine serum albumin (blocking buffer) for 30 min and then stained with appropriate mAbs in blocking buffer to detect the expression of the protein of interest. For intracellular staining, cells were permeabilized with 0.1% saponin, 1% bovine serum albumin, PBS before the blocking and staining steps. Cells were collected and analyzed on a FACScan by flow cytometry using CellQuest software (for flow cytometry) and detected by a Bio-Rad laser-scanning confocal microscope (inverted MicroRadiance AG2 or Radiance 2000), using Lasersharp software (for confocal microscopy). Further confocal image processing was performed in Metamorph (version 5.0.4.8) and Adobe Photoshop. In experiments using pIRES-EGFP constructs to quantify the extent of down-modulation of surface EMR2-FL by EMR2-⌬S, the level of surface EMR2 expression as detected by APC-conjugated secondary Ab was normalized to enhanced green fluorescent protein (EGFP) expression by the formula, (APC test/APC control)/(EGFP test/EGFP control) ϫ 100, to give a percentage score in comparison with cells not co-transfected with EMR2-⌬S.
Wide Field Fluorescent FRET Imaging-For FRET imaging studies, transfected cells were cultured on glass bottom microwell dishes (Matek Corp.) and imaged live in bicarbonatebuffered saline (containing 140 mM NaCl, 5 mM KCl, 10 mM glucose, 1 mM CaCl 2 , 23 mM NaHCO 3 , and 10 mM HEPES, pH 7.2) under 5% CO 2 at 37°C. A Zeiss Axiovert S-200 TV microscope was used. The primary objective lens was a Plan-Apochrome ϫ63 oil immersion lens with a numerical aperture of 1.4. The microscope was mounted into a heated incubation chamber to maintain constant temperature and CO 2 level. Images were acquired on a cooled charge-coupled device camera (12bit Coolsnap HQ; Photometrics). To increase signal to noise ratio, which permitted shorter exposure times and therefore reduced fluorophore bleaching, 2 ϫ 2 binning was used. A Lambda LS xenon arc lamp (Sutter Instrument Co.) was used as a light source. Optical filter sets were from Chroma. For CFP, we used s436/20 for excitation and D480/30 for emission. For YFP, we used HQ500/20 for excitation and HQ535/30 for emission. A dual pass dichroic mirror, 51017BS, was used to decrease the shoulder of the CFP emission into the YFP channel during FRET experiments. The optical filters were placed in a Lambda 10-2 high speed filter wheel (Sutter Instrument Co.) and were rotated into position using Metamorph software (Universal Imaging Corp.). To observe transfected cells preimaging, a 51017BS dichroic (beam splitter) with CFP/YFP emitter was manually rotated into place. For demonstration of cell morphology, differential interference contrast was used. A 10% neutral density filter was set in place to reduce photobleaching during imaging. Exposure times were kept constant for each filter set (typically 100 ms) and taken in the order YFP, FRET, CFP to avoid bleaching CFP during imaging.
The average intensities of regions of interest in transfected cells were measured using Metamorph software under FRET, YFP, and CFP filter sets, respectively, and exported to an Excel file, where all subsequent calculations were performed. Regions without cells were used as background intensity and were deduced from the regions of interest measurement to give FRET intensity (I FRET ), YFP intensity (I YFP ), and CFP intensity (I CFP ). Bleed-through corrections were calculated as follows: a is a norm of the percentage of CFP bleed-through, and b is a norm of the percentage of YFP bleed-through under the FRET filter set. There were no bleed-through signals from CFP under YFP filter sets and vice versa. The values for the bleed-through using this equipment were on average 60% (a) and 6% (b), respectively. These were determined by analyzing multiple images of cells of differing intensities expressing only CFP or YFP and quantifying the relative intensity ratio under the FRET/CFP or FRET/YFP filter sets. A "compensated" FRET measurement was obtained using the equation, Fa/D ϭ I FRET Ϫ I YFP ϫ b/I CFP (36). For each FRET experiment, both positive and negative control constructs were included on each occasion.
Identification of the Alternatively Spliced EMR2 Transcript-To define the expression frequency of the EMR2 isoform with an alternatively spliced (⌬S) stalk in the general population, two experimental approaches were used. First, pooled leukocyte cDNA samples (BD-Clontech) were used as a template using the following primers: F1300, 5Ј-CAGAAGCAAGTAGACAG-GAGTGTCACCTTG; R1500, 5Ј-CCGTCCTGCAGCAAGC-CCTGGTGTGTCTCA. The resulting PCR products revealed a 219-bp (73-amino acid) FL fragment and a 186-bp (62-amino acid) ⌬S variant. Second, total RNAs from freshly isolated granulocytes and monocytic cells of six healthy donors were isolated, reverse-transcribed, amplified by PCR, and analyzed. The cDNAs encoding the FL stalk region of EMR2 and the ⌬S variant were amplified using a sense primer in exon 10 (5Ј-CGAT-TCTTCGACAAAGTCCAGGACC-3Ј) and an antisense primer (5Ј-TGGGTCACCAGATTTCTGTGCCTG-3Ј) in exon 12. The amplified cDNA sequences were cloned into the pCR4-TOPO vector (Invitrogen), and the DNA sequence was determined. The frequency of the stalk variants were determined by comparing multiple DNA samples.
Red Blood Cell Binding Assay-Red blood cell binding assays were performed as previously described (37). Briefly, CHO-K1 cells transfected with the indicated expression vectors were subcultured into 12-well dishes (ϳ5.0 ϫ 10 4 cells/well) at day 1 post-transfection. At day 2, transfected cells were washed extensively with PBS. Heparinized human whole blood cells were collected, diluted 1:100 (v/v) in OPTI-MEM I, and added to the transfected cells (1 ml/well) for 30 min at room temperature with occasional mixing. Nonbinding blood cells were removed by gentle but thorough washing with OPTI-MEM I. Rosettes of red blood cells around the transfected cells were observed and imaged by microscope. The extent of red blood cell adhesion was quantified by measuring the peroxidase activity of hemoglobin (E 450 ) of methanol-fixed red blood cells using the TMB substrate (T8665; Sigma).

Biochemical Characterization of EMR2 Homo-oligomerization-To
address whether the EGF-TM7 receptors might form dimers/oligomers, BS 3 cross-linking experiments were performed on EMR2transfected CHO-K1 cells. We found that samples from untreated cells showed a broad band of ϳ70 -75 kDa that is consistent with the size of the glycosylated EMR2 extracellular subunit derived from receptor GPS autoproteolysis (Fig.  1A, lane 3). In contrast, samples from BS 3 -treated cells show additional protein bands of higher order structures (ϳ105 and Ն250 kDa), which match the sizes of the EMR2-FL receptor and its oligomers, respectively (Fig. 1A, lane  4). Mock-transfected cells do not yield any specific signals (Fig. 1A, lanes 1 and 2). Since BS 3 is a membrane-impermeable cross-linker with a spacer arm of 11.4 Å, this result suggests that the EMR2 N-terminal extracellular subunit is very close to the C-terminal 7TM subunit (Յ11.4 Å) and to other EMR2 molecules on the cell surface.
To verify whether the previous result is due to EMR2 homo-oligomerization, co-immunoprecipitation (co-IP) was performed on cells transfected with EMR2 constructs tagged at the C terminus with either a c-Myc or an HA epitope. IP experiments show that the HA-tagged EMR2 TM7 fragment is co-precipitated readily with the Myctagged fragment only when both constructs were co-transfected (Fig. 1B, lane 3). No co-IP is observed in samples from control singly transfected cells or a mixture of cell extracts of singly transfected cells (Fig. 1B, lanes 1, 2, and 4). Because the cell extract mixtures of individually transfected cells do not show similar protein interaction, this indicates that the detection of co-IP EMR2 receptor subunits is specific and rules out protein aggregation resulting from cell lysis as a potential artifact. The same results were obtained in a reciprocal manner using anti-HA-agarose for IP (supplemental Fig. S1). Therefore, we conclude that EMR2 receptors form homo-oligomers in intact cells.
FRET Analysis of EMR2 Homo-oligomerization-To demonstrate that EMR2 oligomerization occurs under native conditions in living cells, a technique based upon distance-dependent energy transfer, FRET, was employed. To this end, we chose to use a wide field fluorescence microscope with three filter sets (yellow, cyan, and FRET channels). Cells were co-transfected with EMR2-YFP and EMR2-CFP constructs, and the fluorescent signal in each of the three channels was recorded and quantified (Fig. 2). In comparison with the data obtained from positive and negative controls, the compensated FRET intensities in cells co-transfected with EMR2-YFP and EMR2-CFP were found to be significantly positive (Fig. 2C). Most importantly, the FRET signal of cells co-transfected with EMR2-CFP and F4/80-YFP, a murine EGF-TM7 receptor, was at the same level as that of the negative controls, further confirming the specificity of the interaction (Fig. 2C). Since FRET only occurs over distances of Ͻ100 Å, the positive FRET signal detected here indicates that the EMR2 receptors are closely associated with each other in living cells.
EMR2 Homo-oligomerization Is Independent of GPS Autoproteolysis and Is Mediated via the 7TM Subunit-EMR2 undergoes a GPS autoproteolytic reaction during receptor biosynthesis and is expressed as an ␣␤ heterodimeric entity on the cell surface. We have demonstrated previously that the GPS autoproteolysis is mediated by an Ntn-hydrolase-like mechanism. Interestingly, most Ntn-hydrolases, such as aspartylglucosaminidase (38,39), L-asparaginase (40), and ␥-glutamyltranspeptidase (41-43), have been shown to form a ␣␤␤␣ heterotetrameric structure consisting of two copies of the cleaved Ntnhydrolase ␣␤ heterodimer. Indeed, homodimerization has been shown as a prerequisite step for Ntn-hydrolase autoproteolysis (44 -46). In view of the unique dimeric/oligomeric structural arrangement of EMR2 molecule and its unusual post-translational proteolytic modification, it is important to demonstrate the relationship between receptor oligomerization and GPS autocleavage, as well as the protein domains mediating the formation of receptor oligomers.
Next, to examine whether GPS autoproteolysis requires protein dimer formation and to evaluate the role of the extracellular subunit in protein oligomerization, HA-and c-Myc-tagged soluble chimeras consisting of the extracellular domain fused to EGFP were used. Since EGFP is known not to form dimers, these fusion proteins are ideal reagents for the analysis. As expected, immunoblotting of the CM revealed that the chi-meric fusion proteins were processed fully, producing cleaved protein fragments of correct sizes (Fig. 4A, left; data not shown). This is consistent with our previous finding that GPS proteolysis requires a complete stalk region but not the 7TM domain (5). Interestingly, IP of CM using anti-Myc-agarose failed to co-immunoprecipitate the HA-tagged chimeras (Fig. 4A, right). This indicates that the 7TM region but not the extracellular domain is responsible for the formation of receptor oligomers and shows that, unlike the processing of Ntn-hydrolases, dimerization is not a prerequisite for GPS autoproteolysis.
To confirm the role of the 7TM subunit in receptor oligomerization, HA-and Myc-tagged constructs encoding only the 7TM region of EMR2 were expressed. Similar to what was found previously with the EMR2-FL protein, co-IP of the Myctagged 7TM receptor subunit and its HA-tagged 7TM counterpart (Fig. 4B, lane 3) was detected only when the two constructs were co-transfected, suggesting that EMR2 homo-oligomerization is indeed mediated by the 7TM region.
Characterization of a Novel EMR2 Alternative Splice Variant, EMR2-⌬S-As reported previously, the EGF-TM7 receptors are characterized by extensive alternative splicing, especially at the N-terminal EGF-like motif region, which generates multiple splice variants (31). These splice variants contain different numbers of EGF-like motifs and have been shown to interact with different cellular ligands (31,32,37,47). Thus, alternative splicing provides a means for the functional diversity of EGF-TM7 receptors. In addition to the alternative splicing within the EGF-like motif region, we have identified a common splicing event in the mucin-like stalk region of EMR2 (Fig.  5, A and B). Due to the differential usage of a distinct splice acceptor site at exon 12, this splicing event is expected to produce a variant EMR2 isoform, EMR2-⌬S, that is 11 residues shorter than the FL isoform (Fig. 5A). Reverse transcription-PCR and sequencing analysis indicate that the frequency of the different stalk isoforms in the general human population is ϳ4:1 (FL/⌬S), using both a commercially available pooled leukocyte cDNA preparation (Marathon-ready cDNA library; Clontech) and RNA isolated from healthy individuals (n ϭ 6) (Fig. 5B). Moreover, different subsets of myeloid cells (monocytes and neutrophils) are found to have a similar expression frequency (Fig. 5B), and individual donor frequencies were also similar (data not shown).
Interestingly, the EMR2-⌬S variant was poorly expressed and not proteolytically processed. Thus, although both the EMR2 (125)-FL-mFc and EMR2(12345)-FL-mFc molecules showed efficient GPS cleavage, the EMR2-⌬S-mFc proteins were detected as an uncleaved single polypeptide chain and required Ͼ20-fold concentrations for detection comparable with the FL forms ( Fig. 5C and supplemental Fig. S2). These results clearly indicated that alternative splicing in the stalk region, but not the EGF-like motifs, affects both the protein expression and GPS proteolysis. The subcellular localization of EMR2-⌬S was predominantly perinuclear, whereas EMR2-FL is located primarily on the cell surface (Fig. 5D). Similarly, a quantitative analysis of the two receptor isoforms using the pIRES2-EGFP bicistronic expression system demonstrated that no surface expression was detected for EMR2-⌬S, whereas its intracellular expression is significantly weaker than that of EMR2-FL (Fig. 5, E and F). These results suggested that EMR2-⌬S might have an unfolded conformation due to the splicing of the 11 residues in the stalk region. Cell surface glycoproteins that are not folded correctly in the ER are usually degraded by the ER-associated degradation (ERAD) system involving cytosolic proteasomes. To test whether this is the case for EMR2-⌬S, EMR2-⌬S-transfected cells were treated with or without proteasome inhibitors, such as MG-132, lactocystin, and ALLN. As shown in Fig. 5, F and G, the EMR2-⌬S protein expression increased significantly in cells treated with the proteasome inhibitors. In addition, the uncleaved EMR2-⌬S receptor is retained within the ER, where it co-localizes with the ER-specific protein PDI. This indicates that the EMR2-⌬S isoform is a misfolded unstable protein, which is retained within the ER and degraded by ERAD.

Co-expression of the EMR2-⌬S Variant Isoform Specifically Reduces the Surface Expression of the EMR2-FL Receptor-In
view of the expression of EMR2-⌬S transcripts in primary myeloid cells (monocytes and neutrophils) and the evidence demonstrating that EMR2 could itself form homodimers, it is reasonable to think that both EMR2-FL and EMR2-⌬S could also dimerize. Taking into account the distinct subcellular localization and protein stability, one would predict that should significant dimerization occur between the two isoforms, it would either increase the surface expression of EMR2-⌬S or decrease the surface expression of EMR2-FL. To address this question,  8 (white arrow)). Note that the signals in lanes 4 and 7 are very weak. D, subcellular localization of the EMR2-FL (left) and EMR2-⌬S (right) isoforms were detected using anti-HA and a Cy-3-conjugated secondary antibody. Both confocal images were taken using the same laser settings. The white bar represents 5 m. E and F, fluorescence-activated cell sorting analysis of EMR2-IRES-EGFP transfectants confirms that there is little, if any, surface (E) and intracellular (F) expression of the ⌬S receptor. The plots are representative of three independent experiments. The graph in the right panel shows the mean values of normalized surface EMR2 staining (n ϭ 3). CHO-K1 cells transfected with pIRES-EGFP alone were used as a negative control. Intracellular expression (F) was detected on cells permeabilized by saporin. G, CHO-K1 cells were transfected with EMR2-⌬S-7TM/pcDNA3.1, treated with or without proteasome inhibitors as indicated for 24 h, and subjected to Western blotting. H, confocal immunofluorescent staining reveals that EMR2-⌬S is "trapped" in the ER. CHO-K1 cells were transfected with EMR2-FL-YFP (top panel) or EMR2-⌬S-YFP (middle and lower panels). Cells were treated without (top and middle panels) or with the proteasome inhibitor MG-132 (lower panels). Cells were fixed, permeabilized, and probed using the ER-specific protein PDI and the Golgi-specific marker Golgi-58 (data not shown). Bound primary Ab was detected using a Cy-3-conjugated secondary antibody. White bar, 5 m. expression patterns of co-transfected EMR2-FL and EMR2-⌬S isoforms were examined using a confocal microscope and flow cytometry. It was found that EMR2-FL mostly co-localizes with EMR2-⌬S in the perinuclear region and that the cell surface EMR2-FL was reduced in a dose-dependent manner with increasing amounts of co-transfected EMR2-⌬S (supplemental Figs. S3 and S4).
To demonstrate an inhibitory effect of EMR2-⌬S on the cell surface expression of EMR2-FL, a more stringent quantitative flow cytometric analysis based on the pIRES2-EGFP bicistronic expression system was performed (Fig. 6). The cDNA encoding for the EMR2-FL isoform was cloned into the pIRES2-EGFP vector to generate EMR2-FL-pIRES-EGFP. As expected, the cell surface expression of EMR2-FL isoform was found to be correlated proportionally to that of EGFP (Fig. 6A). The extent of down-modulation of surface EMR2-FL by EMR2-⌬S was quantified by measuring the surface EMR2 expression normalized to EGFP expression (see "Experimental Procedures"). For each transfection, the same amount of EMR2-FL-pIRES-EGFP was used. Increasing amounts of EMR2-⌬S-pcDNA3.1 were then co-transfected, using F4/80-pcDNA3.1 as a "stuffer" plasmid to keep the final total amount of transfected DNA constant (Fig. 6B). The fluorescence-activated cell sorting dot plot profiles (Fig. 6A) indicated that the addition of EMR2-⌬S resulted in a dose-dependent decrease in surface EMR2 expression. As shown in Fig. 6B, the normalized APC staining of surface EMR2 was reduced to 49% of the control when an optimal amount of EMR2-⌬S was co-transfected. Cotransfection of more EMR2-⌬S DNA resulted in the additional inhibition of cellular EGFP production, most likely due to the ER "stress" resulting from accumulated, unfolded protein within this compartment. Other "stuffer" expression constructs, MARCO-pcDNA3.1 and rTRHR-HA-pcDNA3, were also used in independent experiments and showed similar results (data not shown).
To demonstrate that the inhibitory effect of EMR2-⌬S on surface EMR2 expression is specific, an N-terminally HA-tagged rTRHR-pIRES-EGFP was used to replace EMR2-FL-pIRES-EGFP. The surface expression levels of rTRHR, an unrelated GPCR, were found to be unchanged in response to the addition of EMR2-⌬S-pcDNA3.1 plasmids (Fig. 6, C and D), thus confirming the specificity of EMR2-⌬S in modulating the surface expression of EMR2-FL. The hetero-oligomerization of EMR2-FL and EMR2-⌬S isoforms were further confirmed by FRET analysis (Fig. 7). Similar to EMR2 homo-oligomerization, the EMR2-⌬S isoform was found to be closely associated with the EMR2-FL isoform within the ER compartment, generating a positive FRET signal.

Modulation of Expression and Function of EGF-TM7 Molecules by Receptor
Hetero-oligomerization-The 7TM regions of human EGF-TM7 members are extremely homologous. In fact, EMR3 is most closely related to EMR2 at the TM7 region (85% amino acid identity) (12,49). In view of the fact that EMR2 forms dimers/oligomers via its 7TM region, we questioned whether EMR2 would form heterodimers with other related human EGF-TM7 receptors. In order to address this, EMR2-FL-Myc was co-transfected with CD97-HA or EMR3-HA. IP of EMR2-FL-Myc using anti-Myc-agarose beads was found to co-IP both CD97-HA and EMR3-HA, although at a lesser extent than that seen with EMR2-FL-HA (data not shown). This result suggests that related EGF-TM7 receptors could heterodimerize.
As previously demonstrated, EMR2-⌬S can reduce the surface expression of EMR2-FL by receptor dimerization. Therefore, if EMR2 is indeed able to dimerize with either EMR3 or CD97, surface expression of these receptors should also be down-regulated by EMR2-⌬S. This was subsequently tested on CD97 using the pIRES-EGFP system as described previously. The cell surface CD97 expression of CD97-pIRES-EGFP-transfected cells decreases in a dose-dependent manner when EMR2-⌬S but not a control F4/80 expression vector is co-transfected (Fig. 8A). Again, the maximum dominant negative effect of EMR2-⌬S on CD97 expression is less than that seen for EMR2-FL expression (ϳ40% reduction of surface CD97 levels versus ϳ51% reduction of EMR2 levels). To confirm the flow cytometric data, fluorescent microscopy was performed on CHO-K1 cells co-transfected with various relevant receptors. Once again, EMR2-⌬S appears to decrease the surface expression of CD97 but not that of F4/80 (Fig. 8B).
Finally, a ligand-binding functional assay was performed to demonstrate the dominant negative effect of EMR2-⌬S on CD97 receptor function. The CD97(125) isoform interacts with CD55 specifically, which can be shown by the binding of CD55expressing erythrocytes to the CD97(125)-transfected cells (Fig. 8C) (37). Increasing amounts of co-transfected EMR2-⌬S but not F4/80 expression vectors cause a quantitative reduction of erythrocyte binding to the CD97-expressing cells, where the extent and size of erythrocyte rosettes are clearly reduced dosedependently (Fig. 8, C and D). Taken together, these data have shown for the first time the evidence of homo/hetero-oligomerization in human EGF-TM7 receptors. Furthermore, it suggests that receptor oligomerization represents a regulatory mechanism whereby the surface expression level and the function of the leukocyte-restricted EGF-TM7 receptors are modulated.

DISCUSSION
In the present study, the myeloidrestricted adhesion-GPCR, EMR2, has been shown by co-IP, FRET, and quantitative expressional and functional analyses to form homodimers as well as heterodimers with other closely related EGF-TM7 proteins. The dimerization is mediated exclusively by the 7TM region and does not involve the post-translational GPS autoproteolytic modification commonly found in these receptors. In view of the identification of a rapidly degraded novel EMR2-⌬S isoform, we hypothesized that dimerization of the two variants could potentially result in the modulation of EMR2-FL expression. In keeping with this hypothesis, co-expression of EMR2-FL and EMR2-⌬S indeed resulted in a marked decrease of surface expression of the FL isoform.
Due to the findings that transcripts of both isoforms are found in human primary leukocytes, we suggest that a potential function of the ⌬S splice variant is to retain the FL isoform in the ER by dimerization, targeting it for degradation and thereby reducing its cell surface expression. The resulting decrease in surface FL receptor levels could potentially modulate its multiple functions, such as ligand binding (to CS for example), signaling (via the TM7 region), and cellular adhesion or migration. Thus, the EMR2-⌬S isoform, although unstable, acts in a dominant negative manner. The fact that this dominant negative effect is specific to the EMR2 molecule but not unrelated 7TM proteins (Fig. 6) further validates the results obtained by co-IP and FRET analysis.
Our result adds EGF-TM7 receptors to the group of GPCRs that form dimers or oligomers. In fact, very little is known about the extent and basis of homointeraction for adhesion-GPCRs (50). To date, there are only a few reports of LNB-TM7 receptors potentially forming oligomers. ETL appears to form dimers after cell lysis that was detergent-sensitive (51). Ig-hepta forms disulfide-linked dimers through intermolecular covalent association in the extracellular domain in a reaction that requires the transmembrane region. Latrophilin was shown to induce receptor clustering after the addition of the exogenous ligand, but no evidence of a direct receptor interaction was demonstrated (52).
Using various biochemical, biophysical, and functional assays, we have established herein that EMR2 forms dimers constitutively, most likely independent of the interaction with its ligand(s). Although the positive FRET signals detected here are relatively small, the specific increase above the multiple negative controls is indicative of receptors in close proximity (less than 100 Å). Using FRET techniques also allowed us to examine the dimerization of EMR2 under native conditions in different compartments of the cell. The fact that FRET occurred both at the cell surface (EMR2-FL) and intracellularly (EMR2-⌬S) also suggests that EMR2 dimerization is indeed ligand-independent and constitutive.
The findings of this study also establish the EGF-TM7 receptors as another example of how heterodimerization of GPCR isoforms or receptor variants are able to modulate cell surface receptor expression and function. What is of particular interest is that EMR2-⌬S is a naturally occurring EMR2 variant that is co-expressed at significant levels in primary myeloid leukocytes with other EGF-TM7 receptors, such as CD97 and EMR3. Therefore, in addition to modulating EMR2 receptor expression by homodimerization, EMR2-⌬S is potentially capable of regulating the expression and function of CD97 and EMR3 by heterodimerization. Although little is known about the function of EMR3, CD97 has been clearly shown to play a role in angiogenesis, granulocyte migration, and homeostasis in vivo as well as in the modulation of CD4 ϩ T cell activation in vitro. The down-regulation of CD97-erythrocyte binding by EMR2-⌬S provides an intriguing insight into the regulation of CD97. How EMR2-⌬S may regulate the function of other EGF-TM7 receptors remains of great interest and requires further investigation.
Although there are a few examples of naturally occurring receptor isoforms producing this type of modulation in the GPCR family, as far as we are aware, this is one of very few examples that would modulate a proposed adhesion function in myeloid cells. The naturally occurring isoform of the CCR5 receptor, CCR5⌬32, which is protective against acquiring HIV may behave in a similar fashion. CCR5⌬32, which like EMR2-⌬S is also retained in the ER, appears to dimerize with wild-type CCR5, thus preventing the surface expression of CCR5 (53). This may explain the lower CCR5 levels and slower progression to disease in patients with HIV harboring this isoform.
Other examples of diseases involving mutant GPCR receptors reported to produce a similar dominant negative effect on cell surface expression include the vasopressin V 2 receptor (V 2 R) isoform, del 62-64 V 2 R, which can cause nephrogenic diabetes insipidus (54), and the dopamine D3 receptor splice form, D3nf, which has been identified in schizophrenic brains (55). Heterodimerization of the rabbit calcitonin receptor isoforms C1a and CTR⌬e13 prevents transport of the receptor to the cell surface (56). A similar situation exists for the rat V 2 R, where the V 2a and V 2b splice variants heterodimerize to reduce surface expression and [H 3 ]arginine vasopressin binding (48).
Our preliminary data suggest that EMR2 (1-5)-FL enhances myeloid cell migration to various chemoattractants and that the 7TM region is required for this function. 5 It seems likely that EMR2-⌬S could influence this isoform-specific migratory function via the 7TM-mediated dimerization. Although we have not yet found the conditions (data not shown), it is highly possible that the expression of the dominant negative EMR2-⌬S transcript in myeloid leukocytes could be modulated  by certain external stimuli, such as cytokines or chemokines, thus affecting the functions of the EMR2-FL and other related EGF-TM7 receptors.
In summary, we have characterized the homo-and heterodimerization of EGF-TM7 receptors and revealed its effect on the expression and function of these leukocyte adhesion-GPCRs. The results presented here might have broader implications in the receptor biology of other LNB-TM7/adhesion-GPCRs, a number of which have been linked to human hereditary diseases.