Apelin receptor homodimer-oligomers revealed by single-molecule imaging and novel G protein-dependent signaling

The apelin receptor (APJ) belongs to family A of the G protein-coupled receptors (GPCRs) and is a potential pharmacotherapeutic target for heart failure, hypertension, and other cardiovascular diseases. There is evidence APJ heterodimerizes with other GPCRs; however, the existence of APJ homodimers and oligomers remains to be investigated. Here, we measured APJ monomer-homodimer-oligomer interconversion by monitoring APJ dynamically on cells and compared their proportions, spatial arrangement, and mobility using total internal reflection fluorescence microscopy, resonance energy transfer, and proximity biotinylation. In cells with <0.3 receptor particles/μm2, approximately 60% of APJ molecules were present as dimers or oligomers. APJ dimers were present on the cell surface in a dynamic equilibrium with constant formation and dissociation of receptor complexes. Furthermore, we applied interference peptides and MALDI-TOF mass spectrometry to confirm APJ homo-dimer and explore the dimer-interfaces. Peptides corresponding to transmembrane domain (TMD)1, 2, 3, and 4, but not TMD5, 6, and 7, disrupted APJ dimerization. APJ mutants in TMD1 and TMD2 also decreased bioluminescence resonance energy transfer of APJ dimer. APJ dimerization resulted in novel functional characteristics, such as a distinct G-protein binding profile and cell responses after agonist stimulation. Thus, dimerization may serve as a unique mechanism for fine-tuning APJ-mediated functions.


Formation of APJ homodimers and oligomers is agonist independent.
We further assessed APJ interactions by co-immunoprecipitation (Co-IP), BRET 1 , and Förster resonance energy transfer (FRET). Co-IP analysis of hemagglutinin (HA)-tagged APJ and Myc-APJ in co-transfected CHO cells revealed that immunoreactive bands against the anti-HA and anti-Myc antibodies were evident only when both receptors were co-expressed ( Fig. 2A). BRET studies were performed to confirm APJ homodimerization. In BRET saturation assays, CHO cells co-transfected with APJ-Rluc and APJ-EGFP resulted in a saturation curve, indicating APJ self-interaction. By contrast, the negative controls, mOX2α R-Rluc and mOX2α R-EGFP 13 , induced BRET signals increased with increasing concentrations of mOX2α R-EGFP in a quasi-linear manner (Fig. 2B,C), indicating a non-specific interaction. Furthermore, when cells co-expressing APJ-Rluc and APJ-EGFP (1:1) were treated with an antagonist ( [Ala13] apelin-13) or agonist (apelin- 12,13,15,17,36), BRET signals were not significantly different to those of unstimulated cells (Fig. 2D), indicating that the formation of APJ homodimers is antagonist and agonist independent.
FRET sensitized emission (SE) was also performed to provide further evidence for APJ homodimerization in living cells. Correction factors β and γ showed the highest crosstalk values, whereas α and δ were rather small; α indicated 18% crosstalk for the acceptor channel into the donor channel; β indicated 60% crosstalk for the donor channel into the FRET channel; γ indicated 52% crosstalk for the acceptor channel into the FRET channel; and δ indicated 35% crosstalk for the donor channel into the acceptor channel (Table S1). No FRET signals were detectable in cells transfected with APJ-CFP or APJ-Venus alone. However, a notable FRET signal was detected in CHO cells co-transfected with APJ-CFP and APJ-Venus (Fig. 3A) and FRET efficiency was 52% calculated by equation (3) (Fig. 3B), indicating that APJ exists as a homodimer. Notably, as a result of membrane fluidity, the FRET signal was variable, which was consistent with monomer-dimer-oligomer dynamics on cell membranes (Movie 1).
To confirm the MS results that revealed the importance of TMD1, 2, 3, and 4 in the formation of APJ dimer interfaces, point mutations were constructed to explore the residues mediating APJ interactions according to previous reports 19,20 . Briefly, we generated 13 mutant receptors including "outward-facing" residues APJM36 1. 40 21 . Plasmids encoding Rluc-tagged mutant APJ receptors and EGFP-tagged wild-type (wt) APJ were transfected into CHO cells, followed by BRET 1 detection as described above (Fig. 6A). Since the correct membrane location is a prerequisite for BRET, to exclude the possibility of incorrect APJ localization leading to a decreased BRET ratio, all mutants were observed by TIRFM before BRET measurement (Fig. 6B). Only mutants with the correct membrane localization were used for BRET measurements. Notably, the APJM36 1.40 A, APJL40 1.44 A, and APJF54 1.58 A mutants in TMD1, and the APJV73 2.48 A, APJV80 2.55 A, APJP83 2.58 A mutants in TMD2 exhibited significantly lower BRET signals than wt APJ, also highlighting the importance of TMD1 and 2 in the formation of the APJ dimer interface. Notably, the mutations in TMD3 and 4 had little effect on the BRET signals, indicating that these mutations may not induce significant conformational changes in TMD3 and 4, or that TMD3 and 4 may be less important for APJ homodimerization in vitro (Fig. 6A). Moreover, the distribution of the intensity of 1189 APJL40 1.44 A-GFP (TMD1) particles was analyzed and fitted to a Gaussian function. Only one peak was observed for APJL40 1.44 A (TMD1) (Fig. 6C), also suggesting that the Leu40 1.44 (TMD1) mutation disrupted the formation of APJ homodimers. Furthermore, the effects of the TMDs on APJ dimers were explored with BRET 1 . CHO cells co-transfected with APJ-Rluc and APJ-EGFP (1:3) were incubated with TMD1, 2, 3, 4, 5, 6, or 7 peptides at 37 °C (4 μ M) and detected by BRET 1 . As shown in Fig. 7, TMD1, 2, 3, and 4 significantly reduced APJ dimer BRET signals (by 55-75%), while TMD5, 6, and 7 had no significant effect. Thus, peptides corresponding to TMD1, 2, 3, and 4 can disrupt the formation of APJ homodimers, further suggesting the involvement of these TMDs in the APJ homodimer interface (Fig. 7).

Measurement of APJ receptor homodimer dynamics on cell membranes. Proximity biotinylation
was also used to examine APJ homodimer dynamics. Since biotin and streptavidin form a tight interaction with a slow off-rate (of days), biotinylated APJ would remain biotinylated even if the dimers dissociated. Here, we detected biotinylation using immunofluorescence imaging and found that acceptor peptide (AP)-APJ can be biotinylated, indicating the formation of APJ dimers (Fig. S4). Subsequently, we observed that the biotinylation increased linearly within ~15 min and reached saturation after 15 minutes, which likely reflects the continuous association and dissociation of APJ in a monomer-dimer dynamic equilibrium (Fig. 8A).
Proteins, including APJ, present on the membrane of living cells are typically dynamic (Movie 2); therefore, APJ could interact by coincidence and be falsely identified as an interaction partner, which represents one of the pitfalls of single-molecule co-tracking. Thus, we detected single-molecule co-tracking of APJ dimers by TIRFM and the co-tracking method was restrained to particle densities of < 1 particle/μ m 2 . The images indicated APJ exists in a dimeric form with short dimer lifetimes (Fig. 8B).
Novel G protein exchange induced by APJ dimers. GPCR dimers can exhibit novel signaling, different from that of their monomers. To explore the novel signaling of APJ dimers, we examined the signaling of APJ dimers using BiFC-BRET. First, we detected concentration for 50% of maximal effect (EC50) with xCELLigence  (Fig. 9B). For G protein dependent signaling, Gα -Rluc8, APJ-VN173, and APJ-VC155 were transfected into CHO cells and BRET signals measured after treatment with 1 μ M apelin-13. Cells expressing Gα -Rluc8 and APJ-Venus served as a positive control, and two negative controls (Gα -Rluc8 and APJ-VN173, Gα -Rluc8 and APJ-VC155A) were also introduced. These tags have been characterized and do not significantly alter the function of the wild type APJ 22 (Fig. S5). To express a similar level, the mount of plasmid in positive controls and negative controls were  transfected, same as that in experiment group. In addition, we have measured the expression levels of donor 23 and acceptor before experiment (Fig. 10). With the positive controls, the increase in BRET signal was rapid, reaching a maximum ~5 min after apelin-13 addition for Gα i1, Gα i2, Gα i3, Gα q, and Gα o, while with cells expressing APJ tagged with VN173 and VC155, an increase in BRET was observed only for Gα i3 and Gα q. No changes in BRET were observed with the negative controls (Fig. 11). The results indicate that APJ monomers activate Gα i1, Gα i2, Gα i3, Gα q, Gα 13, and Gα o, but not Gα s and Gα 12, while APJ dimers activate only Gα i3 and Gα q, but not Gα i1, Gα i2, Gα s, Gα 12, Gα 13, and Gα o.

Disruption of APJ dimerization results in distinct cell responses after apelin-13 stimulation.
Our previous study demonstrated that APJ monomers and APJ-bradykinin 1 receptor (B1R) heterodimers show different effects on the regulation of eNOS phosphorylation in human umbilical vein endothelial cells (HUVECs) 24 eNOS is involved in the regulation of endothelial cell proliferation and/or migration. Here, we also measured the effects of interference peptides on apelin-13-induced cellular responses in HUVECs, as measured with an xCELLigence Real-Time Cell Analyzer. As shown in Fig. 12, apelin-13 stimulation elicited a further increase in the HUVEC cell index (CI); however, co-stimulation with interference peptides corresponding to  TMD1, which was demonstrated to disrupt APJ homodimers, led to a decreasing trend in the CI compared to that before treatment, suggesting disruption of APJ homodimers induced distinct cell responses after agonist stimulation.

Discussion
GPCR dimerization and oligomerization in vivo and in vitro has been demonstrated by a range of techniques including co-IP, resonance energy transfer, proximity ligation assays, and protein fragment complementation assays 25,26 . Here, we examined APJ homodimerization and oligomerization with BRET, FRET, MS, TIRFM, and traditional Co-IP. Although previous studies demonstrated the existence of APJ heterodimerization with other GPCRs in vitro, no study to date has revealed the characteristics of human APJ homodimers and oligomers. Here, analyses of BRET, BiFC, and traditional Co-IP data provide solid evidence for the formation of functional APJ homodimers and oligomers. Furthermore, using single-molecule fluorescence imaging methods (TIRFM), we revealed monomer-to-dimer interconversion of APJ molecules on the cell membrane. TIRFM provides advantages for monitoring membrane receptors within a thin layer 100 nm from the coverslip and improves the detection of single fluorescent molecules on the membrane [27][28][29] . Therefore, our observations likely reflect the dynamics of APJ on the cell membrane. We also measured the dynamics of the quaternary organization of APJ by observing GFP-labeled APJ as individual, mobile, fluorescent spots that were evenly distributed on the cell surface. In cells with less than 0.3 receptor particles/μ m 2 , the proportion of APJ monomers, dimers, and oligomers was ~40%, ~36%, and ~24%, respectively. Agonist stimulation further increased the density of APJ oligomer, which has also been reported by several other recent studies in which agonist stimulation promoted GPCR dimerization 30 . Notably, even in the absence of extracellular stimulation, many receptors form dimers, including epidermal growth factor receptor 31 and various GPCRs [32][33][34] . Moreover, our observations also indicate that APJ receptors on the cell surface are in a dynamic equilibrium with constant formation and dissociation of new receptor complexes, which was also demonstrated previously 30,35 . Recently, using single-particle tracking experiments in cells expressing GFP-tagged adenosine A1 and mCherry-tagged A2A receptors, a preponderance of cell surface 2:2 receptor heteromers (dimer of dimers) that coupled with two G proteins was identified, suggesting the quaternary structure of the GPCR may be involved in the molecular intricacies of GPCR function 12 .
Additionally, we examined the interacting interfaces of APJ dimers using interference peptides and MS. The results demonstrated that peptides encoding TMD1, TMD2, TMD3, and TMD4 can bind to APJ, suggesting these TMDs are involved in interface formation. To the best of our knowledge, this is the first time that MS and interference peptides were utilized to demonstrate the GPCR dimer and examine dimer interfaces simultaneously. Given the high sensitivity and accuracy of the mass analyzer, this method could provide extremely useful information for understanding GPCR structure and conformation changes during the formation of dimers. Furthermore, point mutations in TMD1 and TMD2 also disrupted the APJ dimers, which is consistent with previous findings that TMD1 and TMD2 are important for GPCR dimerization [36][37][38] . It is most likely that two of these are the dimer and two drive the tetramer interaction, which needs further investigation. Notably, these finding also highlight the therapeutic potential of interference peptides, as they can disrupt dimerization and influence receptor function. Indeed, Bouvier et al. showed that a peptide derived from TMD6 of β 2AR could disrupt the dimer and decrease receptor function 39 . Moreover, BRET also revealed that peptides corresponding to TMD1, 2, 3, and 4 can effectively disrupt the formation of APJ homodimers, which highlights the potential of these peptide for targeting distinct dimer-specific physiological effects.
Recent studies demonstrate that GPCR dimerization is of physiological relevance. In general, receptor dimerization or oligomerization is important in many signaling pathways, often as the first step for inducing intracellular signals upon ligand binding [40][41][42] . For example, orexin receptor-corticotropin-releasing factor receptor heteromers in the ventral tegmental area serve as targets for cocaine and promote long-term disruption of orexin-A-CRF negative crosstalk 37 . Therefore, GPCR dimer-monomer interconversion may be important for precise control of signal transduction. Here, APJ dimers induced distinct signaling and functional characteristics from those of monomers; APJ dimers activated only Gα i3 and Gα q. These novel characteristics will help develop more selective drugs that have fewer side effects. Recent studies demonstrate that monovalent drugs specific for GPCR heterodimers 43 and bivalent ligands 44,45 are powerful tools for evaluating activation of signaling pathways and cellular responses elicited by GPCR dimerization and oligomerization. Further studies are needed to generate specific agonists for APJ dimers/oligomers to investigate the novel G-protein binding characteristic of APJ dimers.
Our results also revealed distinct cellular responses profiles in interference peptide-treated versus non-interference peptide-treated HUVECs after apelin-13 stimulation. Recent studies used the xCELLigence system to monitor activation of endogenous GPCRs. For example, administration of UK14,304 (0.15 to 333 nM) to CHO-K1 cells expressing human α -adrenergic 2 A receptors (Gα i coupled) induces a dose-dependent increase in CI. Stimulating serum-deprived CHO-K1 cells with calcitonin (5 pM to 5 μ M) also resulted in a transient agonist-induced increase in CI lasting up to a few hours. Here, we also observed a transient increase in HUVEC CI after apelin-13 stimulation for 2-3 h; this observation is consistent with previous findings. Furthermore,  interference peptides abolished these apelin-13-induced changes, suggesting dimerization may be involved in APJ-mediated morphological or signaling alterations in HUVECs.
Taken together, this study advances our knowledge of the structure and function of APJ; however, much needs to be learned about the neuroanatomy, pharmacology, and signaling of APJ oligomers both in vitro and in vivo before new therapeutic drugs that affect their function can be developed.

Materials. Chinese hamster ovary cells and human umbilical vein endothelial cells were obtained from
American Type Culture Collection (ATCC). All restriction enzymes were from NewEngland BioLabs. [Ala13] apelin-13, apelin-12, 13, 15, 17, 36 were purchased from Phoenix Pharmaceuticals. Lipofectamine 2000, streptavidin-phycoerythrin (SA-PE) and Opti-MEM I were obtained from Invitrogen Life Technologies. Isopropyl-beta-D-thiogalactopyranoside (IPTG) and Coelenterazine h were obtained from Promega. HEPES-buffered phenol red− free medium, Dulbecco's modified eagle medium and Incomplete RPMI-1640 culture medium were purchased from Gibco. Anti-HA-agarose was obtained from Pierce Chemical Co. Polyclonal horseradish peroxidase-conjugated goat anti-rabbit immuno-globulins/HRP was obtained from Zhong Shan Gold Bridge Biology Corporation (China). Anti-Myc antibody and anti-HA antibody were purchased from Cell Signaling Technology Plasmid construction. APJ-VN173 and APJ-VC155 were generated by fusing either the N-terminal fragment of Venus (Venus N: amino acids 1-172) or the C-terminal fragment of Venus (Venus C: amino acids 156-239) to the C-terminus of the wild-type (wt) full-length apelin receptor. Gα i1-Rluc8, Gα i2-Rluc8, Gα i3-Rluc8, Gα q-Rluc8, Gα oa-Rluc8, and Gα s-Rluc8, pET-21a(+ )-GFP, FLAG-BirA-APJ, and HA-AP-APJ were constructed as described previously 9,46 . Cell culture and cDNA transfection. CHO  Western blotting. Transfected cells were lysed in lysis buffer. Cell lysates was separated by 10% SDS-PAGE followed by transfer to PVDF membranes. Proteins of interest were probed with primary and secondary antibodies as described above. Enhanced chemiluminescence (ECL) kits were used to visualize protein bands. Films were scanned and bands analyzed using a ChemiDoc MP Imaging System (Bio-Rad).

Total internal reflection fluorescence microscopy (TIRFM). A commercial TIRF system (Leica
Microsystems) equipped with a 100 × 1.47 NA oil-immersion objective, electron multiplying charge-coupled device camera, and a heatable mounting frame (M-H) was used Illumination was restricted within a single focal plane, with relatively short exposure times (three frames per second), and only 8% laser power for imaging events at the cell surface, thereby avoiding photobleaching during image acquisition. The penetration depth of the evanescent field was ~150 nm. Additionally, the following factors were taken into consideration: Gaussian fitting. In our experiments, only CHO cells with < 0.3 receptor particles/μ m 2 were analyzed. The single-particle tracking algorithm used to identify and track individual APJ receptors has been described previously. APJ monomers, dimers, and oligomers were distinguished according to the smallest normalized intensity and using equation (1): where y is the normalized intensity, w and xc are the standard deviation and the mean, respectively. Receptor particle intensities were analyzed by Gaussian fitting (see equation (2)). Briefly, the Gaussian fitting was performed with Origin 8.0 software and the following equation (2): where y is the frequency of the particles and w and xc are the standard deviation and mean, respectively. trypsinized and plated on a 96 well microplate for 24 h in HEPES-buffered phenol red-free medium (Invitrogen, Life Technologies). Coelenterazine h (5 μ M; Promega) was added for BRET measurements using a Tristar LB941 plate reader (Berthold) with an Rluc filter (400-475 nm) and a EGFP filter (500-550 nm). To monitor the induced interactions of APJ, BRET signals were measured as described above, with slight modifications. Briefly, cells were transiently transfected with APJ-Rluc and/or APJ-EGFP and stimulated with an antagonist ([Ala 13 ]apelin-13) or agonist (apelin- 12,13,15,17,36). To measure the effects of interference peptides on APJ dimers, CHO cells were co-transfected with APJ-Rluc and APJ-EGFP (1:3) and incubated with interference peptides corresponding to TMD1, TMD2, TMD3, TMD4, TMD5, TMD6, and TMD7 at 37 °C (4 μ M), and BRET 1 detected as described above.
FRET SE. The donor plasmid APJ-CFP and receptor plasmid APJ-Venus were co-transfected into CHO cells as the FRET channel. In addition, the donor and acceptor channel are used to eliminate crosstalk into the FRET channel. FRET sample preparations must therefore include references of the donor in the absence of the acceptor (donor only control) and acceptor in the absence of the donor (acceptor only control). APJ-CFP and APJ-Venus were transfected into CHO cells as donor and acceptor channels, respectively, to obtain calibration coefficients to correct for excitation and emission crosstalk. After 12-24 h, FRET signals were detected with a FRET Kit for a Leica AM TIRF MC system (Leica Microsystems). Calculation of FRET efficiency. EA is the apparent FRET efficiency. A, B, C correspond to the intensities of the three signals (donor, FRET, acceptor) and α , β , γ , and δ are the calibration factors generated by the acceptor only and donor only references (equation (3)): Design and synthesis of TMD peptides. Peptides derived from human APJ TMDs were custom synthesized, and their primary sequences are shown in Table 1. The identity of the TMD peptide sequences was analyzed with liquid chromatography (LC)-MS. HIV TAT (YGRKKRRQRRR) was fused to the N-terminus of even-numbered TMDs and to the C-terminus of odd-numbered TMDs to obtain the correct orientation for the inserted peptide, because HIV TAT binds to phosphatidylinositol-(4,5)-bisphosphate found on the inner surface of the membrane 47 . After custom synthesis, the identity of the TMD peptide sequences was analyzed using a LC-MS system (Shimadzu2020 and Water1010). The molecular weight of TMD1, 2, 3, 4, 5, 6, and 7 was 4067.95, 4054.84, 4186.01, 3824.75, 4085.9, 4228.27, and 4261.1 Da, respectively. As shown in Fig. S3, all TMDs were synthesized correctly.
Mass spectrometry. MS was performed to identity APJ dimer interfaces in samples treated with TMD peptides. Cells were transfected with APJ and, 48 h later, treated with or without the indicated HIV TAT-TMD fused peptides (4 μ M) for 60 min at 37 °C. Extracted proteins were immunoprecipitated using an anti-APJ antibody. Protein A/G PLUS-agarose beads were incubated with proteins for 2 h and washed four times with lysis buffer. The APJ complex was eluted from beads as described previously 48 . Then, the endogenous APJ complex was analyzed with an AXIMA matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) system (Shimazdu) in linear positive mode. The mass range was 4-10 kDa and the matrix was Sinapic Acid (SA), 10 mg/ml (50% acetonitrile (ACN), 50% H 2 O, 0.1% Sodium trifluoroacetate (TFA)). The results were calibrated with a Shimadzu MALDI calibration kit (Laser BioLabs). Proximity biotinylation. Escherichia coli biotin ligase (BirA) and an AP were fused to APJ. In the presence of biotin, BirA can site-specifically biotinylate AP. Cells transfected with FLAG-BirA-APJ and HA-AP-APJ were incubated for a designated time (5, 10, 15, and 30 min) with biotinylation media (DMEM, 100 mM biotin, 5 mM MgCl 2 , and 1 mM ATP) at 37 °C. Cells were then rinsed with PBS, detached with pancreatin, pelleted, and resuspended in 1% bovine serum albumin (BSA)/PBS for 15 min. Next, cells were labeled with streptavidin-phycoerythrin (SA-PE) in 1% BSA/PBS. Flow cytometry was performed with a BD FACSCalibur flow cytometer (BD Biosciences). PE was excited by 488 nm laser and emission examined using a 575/24 filter.
Real-time cell analyzer. An xCELLigence Real-Time Cell Analyzer (RTCA) DP system (ACEA Biosciences) was used to measure the effects of interference peptides on the apelin-13-induced cellular responses as reported via a cell index. A background step (step 1) was performed before other experimental steps. The background step was performed with each well of the E-Plate 16 containing only culture media (100 μ l/well) for 30 min at 37 °C, prior to the addition of 10000 HUVECs per well (100 μ l/well). Different concentrations of apelin-13 (1 nM, 0.1 μ M, and 10 μ M) with or without TMD1 (4 μ M) were then added during step 3. The schedule settings were as follows (RTCA Software 2.0): Step Sweep Status Sweeps Interval Unit Total Time Step 1 2 2 1 minute 0:01:26 Step 2 145 241 10 minute 24:51:56 Step 3  181  181  1  minute  28:25:24 Statistical analysis. Results were analyzed using GraphPad Prism 5.0 and Origin 8.0 software, and are shown as the mean ± SEM. Student's paired t-test was used to assess the statistical significance of differences. One-way ANOVA was adopted for multiple group comparisons. P < 0.05 was considered statistically significant.