Using the novel HiBiT tag to label cell surface relaxin receptors for BRET proximity analysis

Abstract

Relaxin family peptide 1 (RXFP1) is the receptor for relaxin a peptide hormone with important therapeutic potential. Like many G protein‐coupled receptors (GPCRs), RXFP1 has been reported to form homodimers. Given the complex activation mechanism of RXFP1 by relaxin, we wondered whether homodimerization may be explicitly required for receptor activation, and therefore sought to determine if there is any relaxin‐dependent change in RXFP1 proximity at the cell surface. Bioluminescence resonance energy transfer (BRET) between recombinantly tagged receptors is often used in GPCR proximity studies. RXFP1 targets poorly to the cell surface when overexpressed in cell lines, with the majority of the receptor proteins sequestered within the cell. Thus, any relaxin‐induced changes in RXFP1 proximity at the cell surface may be obscured by BRET signal originating from intracellular compartments. We therefore, utilized the newly developed split luciferase system called HiBiT to specifically label the extracellular terminus of cell surface RXFP1 receptors in combination with mCitrine‐tagged receptors, using the GABA_B heterodimer as a positive control. This demonstrated that the BRET signal detected from RXFP1‐RXFP1 proximity at the cell surface does not appear to be due to stable physical interactions. The fact that there is also no relaxin‐mediated change in RXFP1‐RXFP1 proximity at the cell surface further supports these conclusions. This work provides a basis by which cell surface GPCR proximity and expression levels can be specifically studied using a facile and homogeneous labeling technique such as HiBiT.

Abbreviations

• BRET
• bioluminescnece resonance energy transfer
• FRET
• fӧrster resonance energy transfer
• GABA
• gamma‐aminobutyric acid
• GPCR
• G protein‐coupled receptor
• HEK
• Human embryonic kidney
• LDLa
• Low‐density lipoprotein class A
• LRR
• leucine‐rich repeat
• NTS₁
• Neurotensin receptor 1
• RXFP
• Relaxin family peptide
• SD
• Standard deviation

INTRODUCTION

Relaxin family peptide 1 (RXFP1) receptor is a Class A G protein‐coupled receptor (GPCR) which is of considerable interest as a drug target due to the therapeutic potential of its cognate peptide ligand, relaxin. The large extracellular domain of RXFP1 is unusual for a Class A GPCR (most of which do not contain significant extracellular domains), containing an N‐terminal low‐density lipoprotein class A (LDLa) module preceding the so‐called “linker,” which connects the LDLa module to a 10 leucine‐rich repeat (LRR)‐containing domain (Figure A). High‐affinity relaxin binding is coordinated between two sites, one in the LRR and one in the linker. RXFP1 and RXFP2, which is the receptor for insulin‐like peptide 3, are the only mammalian GPCRs to contain LDLa modules and in both receptors the LDLa module is essential for receptor activation. There is evidence that the LDLa module is a tethered agonist that interacts with and activates the transmembrane domain of RXFP1, and there is also evidence that RXFP1 forms dimers/oligomers in the cell membrane. It has therefore been previously postulated that RXFP1 may be activated as a homodimer, with the LDLa module of one receptor subunit activating the transmembrane domain of the other receptor subunit via a trans‐activation mechanism (Figure B). However, there is weak evidence that RXFP1 forms stable homodimers at the cell surface, so a mechanism involving relaxin activating a homodimer of RXFP1 requires further investigation.

A routinely used method of determining the existence of GPCR dimers is Bioluminescence Resonance Energy Transfer (BRET) in which different receptors are tagged with a luminescent “donor” protein or a fluorescent “acceptor” protein, and are then recombinantly expressed in a model cell line such as HEK293T. The distance dependence of resonance energy transfer (usually described to only occur within distances of about 10 nm for most donor/acceptor pairings) means that the proximity of a population of labeled receptors can easily be measured using bandpass filter‐based light emission measurements. In saturation BRET assays the stoichiometry of acceptor: donor expression (A:D) is varied, keeping a constant amount of donor (linked to GPCR) with increasing amounts of acceptor (linked to the same or a different GPCR). A nonlinear, or hyperbolic, relationship between BRET signal and A:D is generally considered to be evidence of a specific interaction between the two partners. This type of experiment has been applied to RXFP1, indicating “constitutive” homodimerization of the receptor which is not affected by relaxin stimulation. Currently, however, this is the main evidence that RXFP1 forms homodimers.

In this study, traditional saturation BRET experiments indicated proximity of RXFP1 receptors across the whole cell with no evidence of relaxin‐induced changes in BRET, consistent with previous reports. However, RXFP1 was observed to target poorly to the cell surface when overexpressed, and so the BRET signal may be resultant from receptor accumulation in intracellular compartments rather than explicitly being involved in the activation mechanism at the cell surface. To circumvent this problem, we applied the recently developed split Nanoluc luciferase system called HiBiT (Promega) to label cell surface RXFP1 receptors with a Nanoluc donor in combination with mCitrine‐tagged RXFP1 to provide a BRET measurement of receptor proximity on the cell surface. The HiBiT tag gave a convenient homogeneous measure of receptor expression; however, BRET experiments in which HiBiT‐RXFP1 was co‐expressed with mCitrine‐RXFP1 indicated that RXFP1 may not predominantly exist as a homodimer at the cell surface. Thus, we believe that relaxin‐mediated activation of RXFP1 does not require receptor homodimerization. This work provides a basis by which only cell surface‐expressed GPCR expression and proximity can be investigated using the facile and homogeneous HiBiT labeling technique.

MATERIALS AND METHODS

Cell culture

Human embryonic kidney (HEK) 293T cells used to express receptors were maintained in DMEM (Life Technology) supplemented with 10% fetal bovine serum (FBS), 2 mmol/L l‐glutamine, and 500 U/mL penicillin/streptomycin. Cells were cultured in 175 cm² flasks in incubators maintained at 37°C, with 5% CO₂ and 85% humidity.

Receptor constructs

All RXFP1 receptor constructs were based on the previously published RXFP1 mammalian expression vector which was cloned into pcDNA3.1/Zeo, containing an initial bovine prolactin signal peptide (BPLSP) followed by a FLAG epitope tag and then the RXFP1 receptor sequence (with the exception of the HiBiT tagged receptor which did not contain a FLAG tag). N‐terminal Nanoluc and mCitrine fusions were added via the insertion of an EcoRI site between FLAG and LDLa module. RXFP1‐Rluc8 and RXFP1‐Venus constructs were constructed by insertion of Rluc8/Venus fragments to the C‐terminal end of the receptor between XhoI/NotI sites. For the N‐terminal HiBiT tagged receptor, a pcDNA3.1/Zeo vector containing BPLSP‐HiBiT was synthesized (Genscript) such that RXFP1 could be inserted C‐terminally to the HiBiT tag via BamHI/NheI sites, and including a 12 amino acid linker (GGGSGGGSGGSG) between HiBiT tag and the start of RXFP1. The pcDNA3.1/Zeo BPLSP‐HiBiT vector was also used for insertion of GABA_B1 (Genscript ORF clone OHu03752C) between BamHI/NheI sites. The GABA_B2 (Genscript ORF clone OHu26227C) construct was synthesized and inserted into a custom‐made pcDNA3.1 BPLSP‐HA vector, which was then further modified by insertion of an mCitrine fusion tag between an EcoRI site. All plasmids were sequenced through the entirety of the ORF to ensure correct sequences, and full amino acid sequences of the ORF for all receptor constructs used are presented in the supplementary information.

Venus/Rluc8 Saturation BRET experiments

For saturation BRET style experiments using RLuc8/Venus‐tagged receptors, HEK 293T cells (15 000 cells/well) were seeded into poly‐L‐lysine coated white, opaque 96‐well microplates (Perkin Elmer). Transient transfections using a constant amount of donor (RXFP1‐Rluc8; 5 ng/well) and increasing amount of acceptor (RXFP1‐Venus; 0‐245 ng/well) were performed the following day using LipofectAMINE 2000 (Invitrogen). Forty‐eight hours after transfection, BRET measurements were performed. In brief, cells were treated with 5 μmol/L coelenterazine h (Promega) in phenol red‐free DMEM containing 10% FBS and 25 mmol/L HEPES buffer (henceforth referred to as PRF‐DMEM). The BRET ratio was defined as the emission intensity at 520‐550 nm divided by the emission intensity at 460‐490 nm. BRET unit was defined as the BRET ratio minus that obtained in cells expressing only RXFP1‐Rluc8. Following BRET measurements, cells were washed with PBS and Venus fluorescence emission was measured at 520‐550 nm after excitation at 479‐491 nm. Background fluorescence from cells expressing only RXFP1‐Rluc8 was subtracted. All measurements were performed using a Polarstar Omega platereader (BMG Labtech) at 37°C. The Venus/Rluc8 expression ratio for each well was defined as RXFP1‐Venus fluorescence signal (a.u.) divided by RXFP1‐Rluc8 luminescence signal (a.u.). All saturation experiments were plotted using GraphPad PRISM and curves fitted using a nonlinear regression one‐site binding curve.

cAMP activity assays

All RXFP1 receptor constructs used here were tested for their ability to signal in response to relaxin stimulation using a cAMP reporter gene assay as previously described. Briefly, HEK 293T cells (5 × 10⁵ cells/well) were seeded into six‐well plates followed by transfection the following day with 50 ng receptor DNA and 2 µg of β‐galactosidase reporter gene DNA using LipofectAMINE 2000 (Invitrogen). Twenty‐four hours after transfection, cells were lifted and seeded into CELLBIND 96‐well plates (Corning) at a density of 5 × 10⁴ cells/well. The following day, cells were stimulated for 6 hours at 37°C with varying concentrations of relaxin or 5 μmol/L forskolin and subsequently frozen at −80°C overnight. Cells were then lysed and the amount of cAMP driven β‐galactosidase reporter expression was determined. Experiments were performed in triplicate and were normalized to the cAMP response induced by 5 μmol/L Forskolin. A nonlinear regression sigmoidal dose‐response curve was fit using GraphPad PRISM to obtain pEC₅₀ and E_max values.

FLAG receptor expression assays

Cell surface and total cellular expression of FLAG‐tagged RXFP1 receptors were measured using a method described previously. HEK 293T cells were seeded into poly‐L‐lysine coated clear 96‐well plates (for Figure D) or 24‐well plates (for Figure ). Twenty‐four hours later, cells were transfected with increasing amounts of FLAG‐tagged RXFP1 receptor DNA using LipofectAMINE 2000. For all transfections performed, the amount of transfected DNA was kept constant (250 ng/well for 96‐well plate format and 1000 ng/well for 24‐well plate format) between conditions using empty pcDNA3.1 vector DNA. Twenty‐four hours after transfection, cells were washed once in assay buffer (TBS pH 7.4, 2 mmol/L CaCl₂) and fixed for 15 minutes by addition of assay buffer containing 3.7% formaldehyde (for cell surface) or 3.7% formaldehyde/0.25% Triton‐X (for total expression). Cells were then washed twice with assay buffer, blocked for 45 minutes in assay buffer containing 1% BSA, incubated at room temperature with mouse anti‐FLAG M1 monoclonal antibody (Sigma Aldrich), washed once in assay buffer, incubated at RT in goat anti‐mouse Alexa 488 conjugated antibody (Invitrogen), and washed twice in assay buffer. Finally, cells were lysed and transferred to black walled 96‐well optiplates for fluorescence measurement at 520 nm after excitation at 479‐491 nm. Measurements were performed using a Polarstar Omega platereader (BMG Labtech).

HiBiT receptor expression assays

HEK293T cells were seeded into clear 24‐well plates (Costar) at a density of 200 000 cells/well and the following day were transfected with receptor DNA using LipofectAMINE 2000 (Invitrogen). Twenty‐four hours after transfection, cells were resuspended and seeded into white, opaque 96‐well microplates (Perkin Elmer) at a density of 5 × 10⁴ cells/well for assay the following day. On the day of the experiment, cell culture media was aspirated from all wells and replaced with 100 µL of PRF‐DMEM and placed back in the 37°C incubator for 1 hour before further addition of 100 µL HiBiT complementation reagent and subsequent luminescence measurements in a Polarstar Omega platereader (BMG Labtech). HiBiT complementation reagent was prepared according to the manufacturer's instructions using the Nano‐Glo ® HiBiT Extracellular Detection system kit (Promega) and contained a final concentration of 200 nmol/L LgBiT protein. To permeablize cells and label intracellular HiBiT tags for measurement of total expression, digitonin (Sigma) was added to the HiBiT complementation reagent to achieve a final concentration of 0.01%.

Nanoluc/HiBiT BRET assays

HEK293T cells were prepared for assay the same as above (HiBiT receptor expression assays), but co‐transfected with HiBiT/Nanoluc‐ and mCitrine‐tagged receptor DNA. For saturation‐style BRET assays 10 ng per well of HiBit/Nanoluc‐labeled RXFP1, or 5 ng per well of HiBit‐labeled GABA_B1, and increasing amounts (0‐4000 ng) mCitrine‐labeled receptor per well were transfected, with empty pcDNA3.1 vector included to make up equal total DNA amounts. BRET measurements were taken immediately after addition of HiBiT complementation reagent. For experiments where Nanoluc was used, PRF‐DMEM containing a 1:250 dilution of Nano‐Glo® luciferase substrate (Promega) was used instead of HiBiT complementation reagent. The BRET ratio was defined as the filtered light emission intensity at 520‐620 nm divided by the emission intensity at 410‐490 nm and measured on a Polarstar Omega platereader (BMG Labtech) at 37°C. BRET unit was defined as the BRET ratio of wells transfected with both donor and acceptor tagged receptor, minus the BRET ratio obtained in cells transfected with Nanoluc/HiBiT‐tagged receptors and the appropriate untagged receptor.

RESULTS

Previous studies investigating RXFP1 homodimerization with saturation BRET used RXFP1‐Rluc/ RXFP1‐Venus or RXFP1‐Rluc/RXFP1‐GFP²  pairings for their receptor constructs. We chose to use a Rluc8/Venus pairing as it has been shown to give improved sensitivity in BRET measurements. The functionality of newly cloned RXFP1‐Rluc8 and RXFP1‐Venus receptor constructs were tested using a cAMP reporter gene assay (Table and Figure A), confirming that the C‐terminal fusions did not adversely affect receptor signaling as compared to the untagged receptor. Saturation BRET experiments yielded a hyperbolic curve, indicative of proximity between RXFP1‐Venus and RXFP1‐Rluc8 (Figure B), concordant with that previously published. We also co‐expressed RXFP1‐Rluc8 and RXFP1‐Venus at different levels and treated cells with vehicle or 100 nmol/L relaxin before measuring BRET (Figure C) which showed no changed in BRET ratio due to relaxin activation of receptors, also as previously published. Notably, we have not performed control saturation BRET experiments with RXFP1‐Rluc8 and another Venus‐tagged GPCR as these have been performed previously and this study is focussed on measuring cell surface proximity. These experiments report on the proximity of tagged receptors in all compartments of the cell, not just on those at the cell surface. It was therefore important to investigate the cell localization of RXFP1 receptors to allow interpretation of BRET data. Receptor expression was quantified by virtue of an N‐terminal FLAG tag. It was found that, when overexpressed in HEK 293T cells, only about 5% of RXFP1 was trafficked to the cell surface (Figure D). C‐terminal Venus‐ and Rluc8‐tags appeared to improve the trafficking of RXFP1 to the cell surface when expressed at similar levels; however, cell surface expression remained only around 15% and 10% of the total receptor pool, respectively (Figure D).

Relaxin‐mediated cAMP activity of tagged RXFP1 constructs used in this study

	pEC50	Emax (% of 5 µmol/L forskolin response)	n
aRXFP1	10.80 ± 0.10	114 ± 17	9
aRXFP1‐Venus	10.77 ± 0.10	98 ± 1.9	3
aRXFP1‐Rluc8	11.23 ± 0.20	106 ± 4.9	3
amCitrine‐RXFP1	10.32 ± 0.10	117 ± 3.6	3
aNanoluc‐RXFP1	10.53 ± 0.03	101 ± 7.6	3
HiBiT‐RXFP1	10.43 ± 0.04	113 ± 12	3

Values represent the mean ± SD of n experiments performed in triplicate. ^aReceptor also contain a FLAG epitope tag at the N‐terminus.

Further analysis of RXFP1 expression in HEK 293T cells using increasing DNA transfection levels revealed that, while there was a good linear relationship between the DNA transfection amount and total protein expression (Figure A), there was a saturable limit to the amount of receptor that could be trafficked to the cell surface (Figure B) hence the relative percentage of the total receptor pool which is actually trafficked to the cell surface is highly dependent on the total expression of the receptor (Figure C).

As the majority of RXFP1 appears to be located intracellularly when overexpressed in HEK293T cells, it could be possible that saturation BRET experiments (using Rluc8/Venus fusions) are predominantly reporting interactions occurring in intracellular compartments (especially at the highest expression levels), which could explain the lack of relaxin‐mediated BRET change that may hypothetically be occurring. We therefore tagged RXFP1 at the extracellular N‐terminus, which we reasoned offered the possibility that if RXFP1 is indeed activated as a homodimer, conformational changes in the extracellular domain due to relaxin binding to both receptors might be more readily detectable since the efficiency of resonance energy transfer is influenced by both distance and angular orientation of the donor/acceptor species. Tagging of GPCRs for BRET analysis often uses a variant of Renilla luciferase; however, this has been noted to adversely affect cell surface trafficking when tagged to the N‐terminus. To assess the possibility of BRET transfer between RXFP1 receptors suitably tagged at their N‐terminus, we used RXFP1 with Nanoluc at the N‐terminus and paired that with N‐terminally mCitrine tagged RXFP1 as the acceptor (Figure A). Importantly, both fusions (Nanoluc and mCitrine) were well tolerated and did not perturb relaxin‐mediated signaling from these receptors (Table and Figure B).

Co‐expression of Nanoluc‐RXFP1 and mCitrine‐RXFP1 produced a BRET signal which was stable for at least 20 minutes after addition of furimazine (the coelenterazine analogue developed specifically for Nanoluc), indicating close proximity of Nanoluc‐ and mCitrine‐tagged RXFP1 receptors across the whole cell, and again there was no effect of relaxin treatment (Figure C). Additionally, the specificity of the BRET signal between Nanoluc‐RXFP1 and mCitrine‐RXFP1 was tested (Figure F and Figure S1) by co‐expressing mCitrine‐RXFP1 with other related and unrelated Nanoluc‐tagged GPCRs – RXFP2, RXFP3, α₁‐adrenoceptors (α_1A, α_1B, α_1D)_, and the neurotensin receptor 1 (NTS₁). Notably only Nanoluc‐RXFP2 demonstrated a specific BRET signal with mCitrine‐RXFP1 to a similar level (~75% of the RXFP1 BRET signal) that has previously been shown using C‐terminally tagged receptors. Importantly, all Nanoluc‐labeled receptors were expressed as demonstrated by luminescent measurements and none of the paired receptors disrupted mCitrine‐RXFP1 expression measured as fluorescence (Figure S1).

Furthermore, saturation BRET style analyses demonstrated a hyperbolic curve for Nluc‐RXFP1/mCitrine‐RXFP1 co‐transfections (Figure E) and a linear relationship for Nluc‐RXFP1/mCitrine‐GABA_B2 co‐transfections (Figure F) which were used as a negative control.

To separate BRET signal originating from intracellular compartments from that at the cell surface, we next took advantage of the recently described split Nanoluc complementation system called HiBiT both as a homogeneous assay of receptor expression and as a labeling technique for cell surface receptors. The FLAG tag present in our RXFP1 expression construct was replaced with the 11 amino acid HiBiT tag (VSGWRLFKKIS) which did not adversely affect relaxin‐mediated signaling (Table : Figure A). Addition of the HiBiT complementation reagent (containing furimazine and LgBiT, the protein which binds the HiBiT tag to form the active Nanoluc luciferase) to HEK293T cells transiently expressing HiBiT‐RXFP1 resulted in a luminescence signal indicative of RXFP1 expression at the cell surface, with negligible luminescence from untransfected cells (Figure B). The luminescence signal from complemented HiBiT‐RXFP1 rose slowly after addition of the HiBiT complementation reagent and generally reached a maximum after 20 minutes (Figure B), hence we used the luminescence values at this timepoint to indicate the receptor cell‐surface expression level. To further demonstrate that cell surface localized HiBiT‐RXFP1 receptors were successfully being labeled, we also used fluorescently labeled relaxin (TamRLX); used as a fluorescent BRET acceptor with Nanoluc‐RXFP1 in recent studies concerning relaxin binding kinetics. Preincubation of HiBiT‐RXFP1 expressing cells with 10 nmol/L of TamRLX for 30 minutes before addition of the HiBiT complementation reagent produced a BRET signal between bound TamRLX and labeled HiBiT‐RXFP1 which was stable for 60 minutes and the signal was fully attenuated by co‐incubation with a large excess of non‐fluorescent relaxin as a competitor (Figure C).

Our previous experiments using a FLAG tag to determine total receptor expression used 0.25% Triton‐X (to permeabilize cell membranes and allow labeling of the whole receptor pool); however, we found that Triton‐X strongly inhibited Nanoluc luminescence (Figure S2). Instead, by co‐addition of 0.01% digitonin with the HiBiT complementation reagent we were able to obtain an estimate of the total expression of HiBiT‐RXFP1 (Figure D). Similar to our experiments using a FLAG‐tagged receptor, there was a saturable limit to the amount of HiBiT‐RXFP1 receptor that could be trafficked to the cell surface (Figure E) and a good linear relationship between the DNA transfection amount and total protein expression (Figure D). Again, it was clear that there was an appreciable intracellular pool of HiBiT‐RXFP1 (Figure F).

Having developed a suitable system to assess BRET between tagged RXFP1 receptors at the cell surface, we then co‐expressed a 1:1 ratio of HiBiT‐ and mCitrine‐tagged RXFP1 in live HEK293T cells (Figure A), at a range of transfection levels which should correlate to a varied cell surface expression based on the results from Figure E. This 1:1 DNA ratio was chosen in an attempt to express equimolar amounts of HiBiT/mCitrine‐tagged receptors in cells, which seems likely given that all FLAG‐tagged RXFP1 constructs used in this study demonstrated similar molar expression levels (Figure S5).

We observed a slow increase in the BRET signal over time, with a magnitude correlating to the amount of receptor cell surface expression (Figure B). This is partly unexpected since BRET is a ratiometric measurement and thus should produce a relatively stable BRET signal over time even as the luminescence signal changes. The increase in overall BRET signal as a result of increased surface expression suggests that the BRET signal results from proximity (either random collisions from high receptor density or due to proximity “bystander BRET”) of receptors rather than strictly a specific dimeric physical interaction. Co‐addition of digitonin with the HiBiT complementation reagent, to label intracellularly expressed receptors, showed a sharper increase in the BRET signal, which plateaued at a higher level consistent with additional BRET signal from the now exposed intracellular receptors (Figure C). We then treated non‐permeabilized cells with relaxin and saw no changes in BRET signal (Figure D), clearly demonstrating that relaxin binding does not influence the apparent proximity of cell surface RXFP1 protomers. These experiments thus confirm that there is no relaxin‐mediated change in proximity of RXFP1 at the cell surface.

Additionally, saturation BRET style analyses were attempted using HiBiT‐RXFP1/mCitrine‐RXFP1 co‐transfections (Figure E), or HiBiT‐RXFP1/mCitrine‐GABA_B2 as a negative control (Figure F). The BRET signal for HiBiT‐RXFP1/mCitrine‐RXFP1 appeared pseudo‐hyperbolic which may indicate RXFP1 homodimers at the cell surface. However, there were clear qualitative differences to that obtained using Nanoluc‐RXFP1/mCitrine‐RXFP1 (Figure D) suggesting that these interactions are different at the cell surface compared to the whole cell context.

Given these results for RXFP1, we applied this HiBiT/mCitrine BRET approach to the GABA_B receptor, a bona fide stable GPCR heteromer composed of GABA_B1 and a GABA_B2 subunits, as a positive control (Figure A). The GABA_B1 subunit contains an ER retention motif in the C‐terminal tail that inhibits its trafficking to the cell surface when expressed alone. Co‐expression of GABA_B1 with GABA_B2 masks the ER retention motif of GABA_B1, allowing both to traffic to the cell surface where they exist as a stable di‐sulphide linked heteromer. As expected, HiBiT‐GABA_B1 was poorly expressed alone (though intracellular expression could be detected with the addition of digitonin) and co‐expression of GABA_B2 greatly enhanced the expression of HiBiT‐GABA_B1 at the cell surface as measured by HiBit luminescence (Figure B). Co‐expression of a 1:1 ratio of HiBiT‐GABA_B1/mCitrine‐GABA_B2 in HEK293T yielded results that were consistent with a specific, stable heterodimeric interaction. The BRET signal from labeled GABA_B heteromers was stable over time and was not influenced by different expression levels of the receptor at the cell surface (Figure D‐E). Additionally, a saturation BRET style analysis using a titration of mCitrine‐GABA_B2 with a constant amount of HiBiT‐GABA_B1 showed a clear saturable curve indicative of heterodimerization (Figure F). Comparison of the HiBiT/mCitrine BRET results for the GABA_B heteromer with that of RXFP1 suggested that RXFP1 does not exist as a stable homodimer at the cell surface.

DISCUSSION AND CONCLUSIONS

Hetero‐ and homodimerization of GPCRs has been a topic of great interest in the drug discovery field for several decades now. However, the functional implications of such interactions are difficult to resolve. The receptor for relaxin, RXFP1, has a unique mode of activation which lends itself to the possibility that the functional unit may be a homodimer (Figure ). However, there were unresolved questions about whether RXFP1 is necessarily a homodimer at the cell surface and what role, if any, homodimerization of RXFP1 plays on its mechanism of activation.

A common method for determining GPCR proximity is the use of saturation BRET experiments, and these have been published for RXFP1 showing “constitutive” homodimerization, which appear to be unaffected by relaxin binding. Our primary aim was to further investigate RXFP1 homodimerization in order to assess whether it is indeed a necessary requirement for relaxin‐mediated activation of RXFP1, thus we sought appropriate tools to investigate RXFP1 proximity in live cells. Saturation BRET experiments inherently give a readout of the proximity of receptors across the whole cell (not simply at the cell surface) and involve a titration of receptor expression by increasing DNA transfection amounts. We produced our own saturation BRET experiments, using slightly different fusion proteins (Rluc8/Venus) to previously published reports. This yielded a BRET saturation curve indicative of a close proximity between RXFP1‐Venus and RXFP1‐Rluc8 receptors across the whole cell. We could not, however, detect any change in the BRET ratio upon stimulation with relaxin using this technique. Only a small proportion of RXFP1 (with or without BRET tags) was reaching the cell surface, thus indicating that a large proportion of the BRET signal was coming from receptors expressed in intracellular compartments, which may obscure any potential relaxin‐induced change in BRET signal at the cell surface. As our main goal was to understand what is happening at the cell surface, we looked at the localization of FLAG‐tagged RXFP1 when overexpressed in HEK 293T cells over a range of DNA transfection amounts. While receptor DNA transfection amounts corresponded linearly with the total amount of receptor expressed in the cell, the level of receptor expressed at the cell surface had a saturable limit. Thus, increasing the amount of RXFP1 DNA transfected (as in saturation BRET experiments) leads to accumulation of receptor within intracellular compartments. It has previously been published that RXFP1 targets poorly to the cell surface when transiently expressed in HEK 293 cells, however, this involved confocal imaging of permeabilized vs non‐permeabilized transfected cells, and no quantification of the percentage of receptor at the cell surface was presented. Kern et al also neatly showed co‐localization of RXFP1 receptor with RXFP1 splice variants that were retained in the endoplasmic reticulum. Co‐expression of these splice variants with full length RXFP1 resulted in a decrease in the cell surface targeting of the receptor which, taken together, supports the view that RXFP1 homodimerization in the endoplasmic reticulum is involved in receptor maturation and subsequent targeting to the cell surface. This is not unique to RXFP1, however, as homo‐ and heterodimerization of receptors in the endoplasmic reticulum is believed to be common across the GPCR family to allow appropriate trafficking to the cell surface.

Due to the significant intracellular accumulation of RXFP1 receptor, we therefore aimed to develop a system where the BRET signal from intracellular compartments could be excluded, by only detecting signal from cell surface receptors. This is not a new idea – previous successful approaches to investigate GPCR oligomerization have used fluorescent antibodies directed against N‐terminal epitope tags or used specific labeling proteins such as the SNAP tag to perform time resolved FRET experiments. More recently, surface labeled SNAP tag fused receptors were used in combination with Nanoluc‐tagged VEGR2 (a receptor tyrosine kinase) to investigate the possibility of interactions between VEGR2 and the β₂‐adrenergic receptor.

We chose a new approach in using the Nanoluc split luciferase system, HiBiT, which was developed by Promega and has recently entered use by academic labs. This involved fusion of the HiBiT tag to the N‐terminus of RXFP1, such that cell surface expression could be measured by exogenous addition of the complementary Nanoluc fragment LgBiT (an 18 kDa protein which does not cross the cell membrane). This has many advantages – no requirement for removal of unbound labeling reagent, and no requirement for laser excitation of the donor as is necessary for FRET approaches. Indeed, the HiBiT tag proved to be an excellent method for detection of RXFP1 cell‐surface and total expression (as well as for the GABA_B heteromer), comparable to using a FLAG epitope tag but far less labor‐intensive and applicable for use on live cells at 37°C. In future, HiBiT labeling of receptors could be combined with previous strategies (ie, SNAP surface labeling) to further refine BRET experiments investigating cell surface receptor interactions.

In order to quickly and fully label all cell surface HiBiT‐tagged RXFP1 receptors, the concentration of LgBiT used in our assay conditions was around 100 nmol/L – well above the reported dissociation constant (K_D) for the LgBiT:HiBiT interaction (K_D = 700 pM). It is generally found that the on‐rate (k_on) for purely diffusion limited protein association is in the 10⁵‐10⁶ M/sec range which, through some basic simulations (Figure S3) shows it reasonable that equilibrium should be attained within a few minutes at most, on the assumption of a simple reversible one‐step interaction occurring according to the law of mass action.

The ability to detect the real‐time presentation of membrane‐bound, extracellularly expressed HiBiT tag in live cells at 37°C is unique and may potentially provide information about the dynamics of trafficking of membrane receptors to and from the cell surface. We note that the increase in luminescence upon addition of HiBiT complementation reagent to HiBiT‐RXFP1 expressing cells was unusually slow (Figure B), contrasting with the profile of HiBiT‐GABA_B1 labeling which showed the expected rapid rise in luminescence followed by the expected slow signal decay (Figure C). In our early experiments using HiBiT for labeling of RXFP1, we used a short four residue glycine/serine linking sequence between HiBiT and RXFP1 and considered that this slow increase in luminescence may be due to poor accessibility of the HiBiT tag due to being too close to the N‐terminus of RXFP1. However, the slow rise in luminescence remained even in the 12 residue linker which we subsequently used for these studies (Figure S4), indicating that steric inaccessibility of the HiBiT tag is not an issue. An alternative explanation for the slow rise in luminescence for HiBiT‐RXFP1 complementation is that the receptor is constitutively being recycled between the cell surface and endosomal compartments during the experimental time course. RXFP1 has been demonstrated to undergo constitutive endocytosis in HEK293T cells, therefore it follows that there must also be a constitutive recycling of receptors back to the cell surface in order to maintain a dynamic equilibrium of cell surface expression. Hence, the luminescence signal of labeled HiBiT‐RXFP1 is likely to be a summation of all receptors which have resided at the cell membrane over the experimental time course, including those that have subsequently been endocytosed but continue to emit luminescence. This explanation would also resolve the slow rise in BRET between HiBiT/mCitrine tagged RXFP1 receptors (Figure B), which contrasts with the temporal stability of the BRET signal for the obligate GABA_B heteromer (Figure E), meaning that RXFP1 receptors may only come into proximity upon constitutive internalization into endosomes. These possibilities can easily be investigated in future using pharmacological inhibitors of endocytic machinery, such as pitstop 1/2 or dynasore. Additionally, live cell imaging techniques could also be applied to investigate the dynamics of RXFP1 subcellular localization and trafficking in future.

Plate‐based assays which use BRET to investigate GPCR:GPCR interactions provide only limited evidence concerning GPCR dimerization. Indeed, the great difficulty and caveat of resonance energy transfer methods to investigate protein‐protein interactions is that it is inherently a readout of proximity only, rather than black‐and‐white evidence of a molecular interaction. On the other hand, single‐molecule microscopy methods have previously been employed to determine the 2D interaction kinetics of a few Class A GPCRs, demonstrating that these interactions are quite transient. Such methods are far more labor intensive; however, and so higher throughput plate‐based methodologies such as saturation BRET and the experiments described here are still important tools when weighed against other available data.

As previously mentioned, the evidence that RXFP1 is indeed a functional homodimer at the cell surface is weak. Alongside previously published saturation BRET experiments, relaxin has been reported to bind RXFP1 with negative cooperativity, which seemingly strengthens a theory RXFP1 may be a functional homodimer. A co‐operative binding interaction implies the existence of more than one binding site, in which the occupancy of one site allosterically decreases (negative cooperativity) or increases (positive cooperativity) the affinity of a second binding site. An RXFP1 homodimer provides a structural basis by which two allosterically coupled relaxin binding sites might exist. Svendsen et al reported negative cooperativity of relaxin binding using a method originally developed for the insulin receptor, in which an undefined concentration of radio‐labeled relaxin was incubated with high expressing HEK293T‐RXFP1 cells, and the amount of remaining bound radio‐labeled relaxin was later quantified after removal of unbound radio‐labeled relaxin (to allow dissociation of bound radio‐labeled relaxin) with or without varying concentrations of unlabeled relaxin. The observation that increasing concentrations of unlabeled relaxin led to a concentration dependent decrease in the amount of remaining bound radio‐labeled relaxin indicated a competitor induced “acceleration” of radiolabeled relaxin dissociation which was taken as evidence of negative cooperativity (even though dissociation rates were not quantified). However, this methodology does not take into account the general phenomena that even strictly isolated single‐site binding processes will necessarily show an apparent competitor‐induced acceleration of dissociation due to the competitors ability to occlude the rapid rebinding of the initially bound ligand. Additionally, the complex mechanism by which relaxin is now understood to bind RXFP1 (co‐ordinated by multiple distinct sites within the ECD) could also explain an apparently cooperative binding interaction when a receptor monomer is assumed. Furthermore, the most recent investigations into the kinetics of relaxin:RXFP1 binding found no evidence of negative binding cooperativity when relaxin dissociation rates were quantified in the presence of varying concentrations of competitor relaxin. Therefore, the most current evidence concerning the mode of relaxin binding to RXFP1 does not support the idea that it is activated as homodimer.

Given the results of these studies, a non‐homodimer mechanism of RXFP1 activation by relaxin currently appears more likely. These studies have shown that the localization of receptors is an important consideration when interpreting the results of plate‐based assays using BRET methodologies. We have demonstrated that the HiBiT tag is an excellent tool for cell labeling both Class A (RXFP1) and Class C (GABA_B) GPCRs with a luminescent tag which can participate in BRET transfer, and that it may even be useful to study the real‐time dynamics of receptor trafficking in future. Importantly, by isolating the BRET signal occurring at the cell surface, we can say with more certainty that activation of RXFP1 by relaxin does not induce homodimerization at the cell surface. Indeed, the outcomes of these studies suggest that further investigations should focus on the dynamics of RXFP1 trafficking to and from the cell surface for which the use of a real‐time labeling strategy such a HiBiT may be informative.

ACKNOWLEDGEMENTS

This research was supported by National Health and Medical Research Council of Australia project grant [1100676] (RADB and DJS) and the Victorian Government Operational Infrastructure Support Program. RADB is supported by an NHMRC Research Fellowship and DJS is supported by an NHMRC Dementia Fellowship. The authors thank Tania Ferraro and Sharon Layfield for technical assistance and Promega for early access to HiBiT complementation reagents.

DISCLOSURES

The authors declare no competing interests.

AUTHOR CONTRIBUTIONS

BH and SB performed the experiments and analyzed the data. RADB, DJS, MK, and BH conceived experiments. B.H, DJS, SB, and RADB wrote the manuscript. All author contributed to the editing of the manuscript.

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the corresponding author upon reasonable request.