Biochemical characterization of G protein coupling to calcitonin gene – related peptide and adrenomedullin receptors using a native PAGE assay

Calcitonin gene-related peptide (CGRP), adrenomedullin (AM), and adrenomedullin 2/intermedin (AM2/IMD) have overlapping and unique functions in the nervous and circulatory systems including vasodilation, cardioprotection, and pain transmission. Their actions are mediated by the class B calcito-nin-like G protein – coupled receptor (CLR), which heterodimer-izes with three receptor activity – modifying proteins (RAMP1 – 3) that determine its peptide ligand selectivity. How the three agonists and RAMPs modulate CLR binding to transducer proteins remains poorly understood. Here, we biochemically characterized agonist-promoted G protein coupling to each CLR · RAMP complex. We adapted a native PAGE method to assess the formation and thermostabilities of detergent-solubilized fluorescent protein – tagged CLR · RAMP complexes expressed in mammalian cells. Addition of agonist and the purified G s protein surrogate mini-G s (mG s ) yielded a mobility-shifted agonist · CLR · RAMP · mG s quaternary complex gel band that was sensitive to antagonists. Measuring the apparent affinities of the agonists for the mG s -coupled receptors and of mG s for the agonist-occupied receptors revealed that both ligand and RAMP control mG s coupling and defined how agonist engagement of the CLR extracellular and transmembrane domains affects transducer recruitment. Using mini-G sq and -G si chimeras, we observed a coupling rank order of mG s > mG sq > mG si for each receptor. Last, we demonstrated the phys-iological relevance of the native gel assays by showing that they can predict the cAMP-signaling potencies of AM and AM2/ IMD chimeras. These results highlight the power of the native PAGE assay for membrane protein biochemistry and provide a biochemical foundation for understanding the molecular basis of shared and distinct signaling properties of CGRP, AM, and AM2/IMD.

disulfide-linked loop structure near the agonist N terminus is required for activation, and truncated peptides that lack this element are competitive antagonists. The RAMPs have an ECD that interacts with the CLR ECD and a single transmembrane helix that contacts the CLR TMD. Soluble RAMP ECD-CLR ECD fusion proteins in which the two ECDs were tethered by an engineered linker bound the aCGRP, AM, and AM2/IMD peptides with selectivity profiles that were similar, although not identical to the intact receptors (20,21). These ECD complexes had reduced binding affinities as compared with the intact receptors, presumably because of the lack of peptide-TMD contacts.
Crystal structures of C-terminal fragments of the three peptides bound to RAMP-CLR ECD complexes (21,22), and cryo-EM structures of the full-length receptors in complex with the agonists and G s heterotrimer (23,24) showed how the peptides occupy the CLR ECD and TMD and revealed minimal RAMPpeptide contacts that were limited to the ECD complexes. The structures and peptide variant studies revealed a peptide-binding selectivity mechanism involving the RAMP-peptide contacts and allosteric modulation of CLR (23,25,26). In contrast to our growing understanding of RAMP-modulated peptide binding, our knowledge of how the agonists and RAMPs affect CLR transducer interactions is limited. The cryo-EM structures showed how the active state receptors bind G s , but how the agonists compare in their abilities to promote G s recruitment and how the RAMPs affect this remain unclear. In addition, the contributions of agonist engagement of each of the two CLR domains to transducer recruitment remain poorly defined. Moreover, the CGRP and AM receptors exhibit pleiotropic Gprotein coupling with the ability to also signal through G q and G i (27), but how the agonists and RAMPs compare in terms of promoting signaling through each G protein remains unresolved. Two groups used cell-based signaling assays in HEK293 or COS-7 cells to examine this issue and found evidence for either dramatic or subtler ligand-and RAMP-dependent G protein signaling bias (28,29). Differences in these reports may be due to the different cell lines used and/or the inherent challenge of studying the signaling bias of these complex heterodimeric receptors in cell-based assays.
Here, we used a biochemical approach to characterize Gprotein coupling to the CGRP and AM receptors. We developed a time-and cost-efficient native PAGE assay to assess the formation and thermostabilities of detergent-solubilized, fluorescent protein-tagged CLR·RAMP heterodimers expressed in mammalian cells. Receptor coupling to G proteins was assessed using purified G protein surrogate mini-G (mG) proteins that were developed for structural studies of active state GPCRs (30,31). The G s a subunit was engineered with several deletions and amino acid substitutions to stabilize it and uncouple receptor binding from nucleotide exchange to create mG s (30), which stabilizes GPCRs in a conformation that recapitulates the features observed for active-state GPCRs bound to nucleotide-free G s heterotrimer (32). Chimeric mG sq and mG si and mG 12 proteins were developed to extend the mG toolkit to all four families of a subunits (31). The mG proteins were shown to be powerful probes for active state GPCRs in living cells (33), and their development and applications were recently reviewed (34). We found that agonist-promoted CGRP and AM receptor coupling to mG s was visible as a mobility-shifted quaternary complex gel band in the native PAGE assay, and we used the assay to measure the apparent binding affinities of each agonist for the mG s -coupled receptors and of mG s for each receptor occupied with each agonist. We also characterized receptor coupling to the mG sq and mG si chimeras. Finally, the value of the native PAGE assay for defining physiologically relevant receptor biochemistry was demonstrated by its ability to predict the cAMP-signaling potencies of novel agonist peptide chimeras. Our results provide important insights into the two-domain agonist-binding and transducer-recruitment mechanism for the CGRP and AM receptors and define how the three agonist peptides and three RAMP accessory proteins control CLR G protein coupling. This work provides a biochemical foundation for understanding commonalities and differences in CGRP, AM, and AM2/IMD signaling through their shared receptors and will aid future structural studies and the development of therapeutic peptide analogs.

Fluorescent protein-tagged CLR and RAMP constructs form detergent-stable complexes visible by high-resolution clear native electrophoresis (hrCNE)
We designed CLR and RAMP expression constructs with Cterminal EGFP and mCitrine tags, respectively (Fig. 1A). N-terminal maltose-binding protein (MBP) tags were originally included on both subunits for purification purposes, but these tags also seemed to promote more defined bands in native PAGE. We used polyethyleneimine (PEI)-mediated transient transfection to express the MBP-CLR-EGFP and MBP-RAMP-mCitrine constructs in HEK293S GnT1 2 cells (35). We reasoned that the homogenous N-glycosylation provided by this cell line would facilitate sharp bands on the native gels. After receptor expression in adherent cultures in 48-well plates, the cells were solubilized with lauryl maltose neopentyl glycol (LMNG)/cholesteryl hemisuccinate (CHS), the lysates were centrifuged at maximum speed in a bench-top microcentrifuge, and the supernatants were analyzed by hrCNE (36) with visualization of the detergent-solubilized receptors by ingel fluorescence (Fig. 1B). Co-expression of the tagged CLR and RAMPs resulted in the appearance of intense sharp bands that migrated slower than the CLR and RAMP alone consistent with formation of the heterodimeric complexes. Analysis of the supernatants by denaturing SDS-PAGE revealed fluorescent bands of the expected molecular masses for the CLR and RAMP constructs (Fig. S1A). A control experiment with each of the RAMPs co-expressed with another class B GPCR, the parathyroid hormone receptor (PTH1R), showed no evidence for detergent-stable PTH1R·RAMP1, -2, or -3 complexes ( Fig.  1C and Fig. S1B), demonstrating specificity. To prove that the starred bands in the native gel (Fig. 1B) were CLR·RAMP complexes, we excised these bands and placed the gel slices in the wells of a denaturing SDS-PAGE gel followed by electrophoresis. Fluorescent imaging revealed the presence of two bands of the correct sizes for the tagged CLR and RAMP constructs (Fig. 1D).
We were also able to use the native PAGE assay in a thermostability format to test the receptor stabilities in different detergents. Aliquots of the solubilized supernatants were incubated at various temperatures for 30 min, followed by centrifugation and analysis of the supernatant by hrCNE. Of the three detergent systems tested, the CLR·RAMP complexes were least stable in n-dodecyl-b-D-maltopyranoside (DDM)/CHS with the RAMP1 complex falling apart at 20-28°C, the RAMP2 complex being unstable at all temperatures tested, and the RAMP3 complex breaking down above 28°C (Fig. S2A). The three receptor complexes had similar thermostabilities in LMNG/CHS and LMNG/glyco-diosgenin (GDN)/CHS, with each complex being stable up to at least 37°C (Fig. S2, B and C). All experiments hereafter used LMNG/CHS because it solubilized well and gave similar stabilities for the three CLR·RAMP complexes.
Tagged receptor constructs used for hrCNE gel assays exhibit normal cAMP signaling We sought to use the native PAGE assay to characterize peptide ligand and mini-G protein binding to the CLR·RAMP complexes. Toward this end we generated new MBP-RAMP constructs with their native C termini to eliminate the possibility of steric hindrance of mG coupling from the mCitrine tag ( Fig. 2A). Co-expression of these constructs with MBP-CLR-EGFP in HEK293S GnT1 2 cells, solubilization, and hrCNE analysis revealed formation of detergent-stable complexes visible by in-gel fluorescence from the EGFP (Fig. 2B). To ensure that these complexes maintained cAMP signaling properties comparable with the untagged receptors, we co-expressed each of the three tagged complexes in COS-7 cells and measured cAMP accumulation in response to the aCGRP, AM, and AM2/IMD agonists (Fig. 2, C-E). All three complexes exhibited agonist selectivity profiles comparable with those observed for the WT receptors (Fig. 2F), and AM2/IMD was a partial agonist at the RAMP2 complex as reported (37). Table S1 summarizes the potencies obtained in the signaling assays. The agonist potencies were slightly reduced at the tagged constructs as compared with WT receptors, but this effect was minor. The tagged constructs were thus good surrogates for use in experiments to define the ligand and mG-binding properties of the complexes using the hrCNE gel assay.
Agonist-dependent coupling of CLR·RAMP complexes to mG s monitored by hrCNE gel mobility shift To measure peptide ligand and mG binding to the receptor complexes, we turned to the use of membrane preparations to enable uniform receptor addition across multiple reactions. Large-scale transfections of HEK293S GnT1 2 cells were performed for each of the three MBP-CLR-EGFP·MBP-RAMP complexes, the cells were harvested, and crude membranes were prepared. The heterodimer concentration in each preparation was estimated by LMNG/CHS solubilization followed by analysis on SDS-PAGE with comparison to known amounts of purified MBP-EGFP (Fig. S3A). The membrane preparations were each estimated to contain ;100 nM heterodimer enabling their use as 103 stocks to give ;10 nM receptor heterodimer final concentrations in the reactions. All hrCNE experiments hereafter used the membrane preparations. We expressed the mG s , mG sq , and mG si proteins as N-terminally His 6 -tagged SUMO-mG fusion proteins in Escherichia coli and purified them by immobilized metal affinity chromatography and gelfiltration chromatography (Fig. S3B). We originally intended to remove SUMO from the fusion proteins, but we found that the fusions worked better in the native PAGE assay, presumably because of their larger size (see below).
In the course of studies to assess peptide and mG binding to the receptor complexes by hrCNE assay, we noticed the tendency of the purified mG proteins to form disulfide-linked oligomers. The presence of multiple disulfide bonds in the CGRP and AM receptors and a disulfide bond in the peptide agonists prevented the use of DTT or TCEP to solve this problem. Fortunately, we found that a GSH/GSSG redox buffer could maintain the purified mG proteins in a largely reduced monomeric state without damaging the receptors and agonists. For the hrCNE experiments going forward, the purified mG proteins were preincubated in a GSH/GSSG redox buffer before addition to the binding reactions as described under "Experimental procedures." Exogenous synthetic peptide ligands (10 mM) and purified SUMO-mG s (50 mM) were added to the membrane preparations either alone or in combination in the presence of apyrase and incubated on ice for 30 min followed by solubilization with LMNG/CHS for 2 h at 4°C. The solubilized reactions were centrifuged, and the supernatants were analyzed by hrCNE to look for mobility shifts indicative of interactions (Fig. 3A). These experiments used three types of peptide ligands: short single-site ECD-binding high-affinity antagonist variants (21,25,26), traditional dual site ECD/TMDbinding antagonists that lack the N-terminal disulfide-bonded loop, and the disulfide loop-containing agonists that bind both domains. aCGRP peptides were used for the CLR·RAMP1 complex, AM peptides for the CLR·RAMP2 complex, and AM2/IMD peptides for the CLR·RAMP3 complex as described in the Fig. 3 and Fig. S4 legends. No substantial change in mobility of the receptor heterodimer bands was observed in the presence of peptide antagonists, agonists, or SUMO-mG s alone or with the antagonists together with SUMO-mG s . In contrast, agonist and SUMO-mG s together yielded a prominent mobility shift for each of the three receptors consistent with formation of a stable agonist·CLR·RAMP·mG s quaternary complex (Fig. 3A).
To ensure that the shifted bands contained SUMO-mG s , SUMO protease was added to cleave the fusion protein, which should result in a smaller shift for the putative quaternary complex band. This was the case for the RAMP1 complex, whereas for the RAMP2/3 complexes, the protease treatment elimi-nated the shift altogether (Fig. 3A). Analysis of the reactions by denaturing SDS-PAGE with Coomassie Blue staining indicated that the SUMO-mG s fusion protein was properly cut by the protease (Fig. S4A). It is unclear whether the failure of the CLR·RAMP2/3 heterodimer bands to shift in the presence of SUMO protease-liberated mG s was because it did not couple to these receptors or because it bound but did not cause a mobility shift. Nonetheless, these data provided strong evidence for agonist-dependent coupling of the SUMO-mG s fusion protein to each receptor. We also performed the Fig. 3A experiments in the absence of apyrase and obtained similar results (Fig. S4B); however, we elected to include apyrase in all hrCNE assay experiments hereafter.
Next, we assessed the abilities of small molecule and peptide antagonists to antagonize the agonist-and SUMO-mG s -dependent mobility shift of the heterodimer bands. The small molecule antagonist telcagepant was developed as a potential migraine drug. It is specific for the CLR·RAMP1 complex in which it binds a pocket in the ECD complex to block anchoring of the peptide agonist C terminus (22,38). High-affinity singlesite ECD-binding (25,26) and traditional dual site ECD/TMDbinding peptide antagonists appropriate for each receptor complex were used as described in the Fig. 3 legend. Each of the peptide antagonists prevented formation of the shifted bands, and telcagepant selectively prevented formation of the shifted band for the CLR·RAMP1 complex as expected (Fig.  3B). These results provided further evidence that the shifted bands were the quaternary complexes and demonstrated that the detergent-solubilized receptors responded appropriately to antagonists.  Table S1 for a summary of the pEC 50 values with error and statistical analyses.
Quantitation of peptide agonist and mG s binding to CLR·RAMP complexes by hrCNE gel mobility shift assays The agonist apparent binding affinities for the CLR·RAMP complexes in the presence of an excess of SUMO-mG s were determined using the hrCNE gel mobility shift assay. SUMO-mG s was held constant at a high concentration (25 or 50 mM), and the peptide agonists were varied from ; 1 nM to 10 mM (Fig. 4A). The appearance of the quaternary complex band (Fig.  4, B-D) and the disappearance of the heterodimer band (Fig. S5) were quantified by densitometry. Fig. 4 (E and F) shows scatter plots of the pEC 50 values resulting from these analyses applied to three independent replicates, and Table S2 summarizes the values. The two methods of quantitation gave similar results, but hereafter we focus on the values derived from the quaternary complex band appearance. At the CLR·RAMP1 complex, the strongest apparent affinity (EC 50 ) was observed for aCGRP (36 nM) followed by AM2/IMD (68 nM) and then AM with weaker binding (350 nM) (Fig. 4, B and E, and Table S2). At the CLR·RAMP2 complex, similar apparent affinities were observed for AM (45 nM) and AM2/IMD (70 nM), and aCGRP exhibited very weak binding with an unmeasurable apparent affinity greater than 1 mM (Fig. 4, C and E, and Table S2). At the CLR·RAMP3 complex, AM and AM2/IMD had similar apparent affinities of ;35 nM, and aCGRP binding was much weaker with an apparent affinity of 460 nM (Fig. 4, D and E, and Table S2).
Next, we performed the reverse experiment to determine the apparent binding affinities of SUMO-mG s for each of the receptor complexes occupied by each of the peptide agonists. The agonist was held constant at a high concentration (10 mM), and SUMO-mG s was varied from ;11 nM to 75 mM (Fig. 5A). Densitometry was used to quantify the appearance of the quaternary complex band (Fig. 5, B-D) and the disappearance of the heterodimer band ( Fig. S6) with increasing SUMO-mG s . Fig.  5 (E and F) shows scatter plots of the pEC 50 values resulting from these analyses applied to three independent replicates, and the values are summarized in Table S3. At the CLR·RAMP1 complex, aCGRP elicited an apparent mG s affinity (EC 50 ) of ;200 nM, whereas the AM and AM2/IMD peptides equally promoted mG s binding with apparent affinities of ;5 mM ( Fig. 5, B and E, and Table S3). At the CLR·RAMP2 complex, the three peptides elicited weak apparent mG s affinities estimated to be ;70 mM for AM or AM2/IMD and ;120 mM for CGRP (Fig. 5, C and E, and Table S3). AM and AM2/IMD elicited equal apparent mG s affinities of ;10 mM at the CLR·RAMP3 complex, whereas aCGRP promoted mG s binding with an apparent affinity of ;40 mM (  Table S3).

Coupling of mG sq and mG si chimeras to the CLR·RAMP complexes by hrCNE assay
Coupling of purified SUMO-mG sq and SUMO-mG si chimeras to each of the receptor complexes occupied by each of the three agonists was assessed in experiments like those presented in Fig. 5, but we did not quantitate these experiments by densitometry because weaker binding of these mG proteins was observed (Fig. 6). The most notable coupling detected was that elicited by aCGRP at the CLR·RAMP1 complex where binding of the mG sq and mG si proteins was observed with micromolar concentrations and the mG sq exhibited greater apparent affinity than the mG si (Fig. 6, A and B). For the two AM peptides and the two AM receptors, there was evidence for weak mG sq coupling, whereas coupling to mG si appeared to be limited. These results taken together with those in Fig. 5 were consistent with a coupling preference rank order for each receptor complex of mG s . mG sq . mG si , and there did not appear to be significant agonist-dependent alterations of this ranking. For completeness we also assessed the coupling of the three Figure 3. Agonist-dependent coupling of detergent-solubilized CLR· RAMP complexes to mG s and inhibition of quaternary complex formation by small molecule and peptide antagonists assessed by the hrCNE gel shift assay. Fluorescently imaged 8% hrCNE gels using membrane preparations of MBP-CLR-EGFP and MBP-RAMP complexes (CLR:RAMP1, CLR: RAMP2, and CLR:RAMP3) that were solubilized in LMNG/CHS in the presence/ absence of ligands and SUMO-mG s . A, quaternary complex formation. 10 mM agonist or antagonist, 50 mM SUMO-mG s , and 5 mM SUMO protease were added to the membrane preparations as indicated prior to solubilization. Shifted bands signifying formation of a quaternary complex are starred. For CLR·RAMP1, the agonist was aCGRP(1-37), the ECD antagonist was aCGRP (27-37) N31D/S34P/K35W/A36S, and the ECD/TMD antagonist was aCGRP . For CLR·RAMP2 the agonist was AM(13-52), the ECD antagonist was AM (37-52) S48G/Q50W, and the ECD/TMD antagonist was AM . For CLR·RAMP3 the agonist was AM2/IMD(1-47), the ECD antagonist was AM2/ IMD(32-47) H45W, and the ECD/TMD antagonist was AM2/IMD . B, antagonism of quaternary complex formation with simultaneous addition of 300 nM agonists, 10 mM competitive antagonists, and SUMO-mG s (50 mM for CLR·RAMP1 and CLR·RAMP2, and 25 mM for CLR·RAMP3) to the CLR·RAMP membrane preparations before LMNG/CHS solubilization. The agonists were aCGRP(1-37) for CLR·RAMP1, AM(13-52) for CLR·RAMP2, and AM2/IMD(1-47) for CLR·RAMP3. Antagonists included the CLR·RAMP1-selective small molecule telcagepant and peptide antagonists. For CLR·RAMP1 the ECD antagonist was aCGRP(27-37) N31D/S34P/K35W/A36S, and the ECD/TMD antagonist was aCGRP . For CLR·RAMP2 the ECD antagonist was AM (37-52) S48G/Q50W, and the ECD/TMD antagonist was AM . For CLR·RAMP3 the ECD antagonist was AM(37-52) S45R/K46L/S48G/Q50W, and the ECD/TMD antagonist was AM2/IMD . For CLR·RAMP3 the reactions were solubilized overnight (13 h) because 2 h was insufficient to reach equilibrium with the ECD antagonist. A and B show representative images from two independent experiments. SUMO-mG proteins to each of the receptors occupied with the bCGRP agonist, which differs from aCGRP at only three positions. The assays with bCGRP were very similar to those with aCGRP (Fig. S7).
The utility of hrCNE gel mobility shift assays to predict signaling behavior in cells: the ECD complex-binding segments of AM and AM2/IMD drive their cAMP-signaling potency differences at the CGRP receptor Comparing the cAMP signaling results (Fig. 2) with the hrCNE gel shift assay results for agonist binding to SUMO-mG s -coupled receptors (Fig. 4) and SUMO-mG s binding to the agonist-occupied receptors (Fig. 5) raised the question of how the hrCNE gel assay results related to the signaling outcomes observed in cells. We hypothesized that the apparent affinities of the agonists for the SUMO-mG s -coupled receptors were determined by their dual site interactions with the CLR·RAMP ECD complex and the CLR TMD, whereas the apparent affinities of SUMO-mG s for the agonist-occupied receptors were largely driven by the single-site interaction of the N-terminal half of the agonist with the CLR TMD. To test this we focused on the AM and AM2/IMD agonists at the CLR·RAMP1 complex. These agonists exhibited an ;8-fold difference in cAMP-signaling potencies (Fig. 2, C and F) that appeared to be reflected in their different apparent affinities for the SUMO-mG s -coupled receptor (Fig. 4, B and E). In contrast, these two agonists exhibited identical abilities to recruit SUMO-mG s to the agonist-occupied receptor (Fig. 5, B and E).
We reasoned that the different AM and AM2/IMD signaling potencies and apparent affinities for the SUMO-mG s -coupled receptor were largely due to the different affinities of these two peptides for the CLR·RAMP1 ECD complex that we previously reported (21). Thus, we predicted that chimeras of the AM and AM2/IMD agonists with their N-terminal TMD-binding and C-terminal ECD complex-binding segments swapped (Fig. 7, A and B) should exhibit swapped cAMP-signaling potencies with the stronger signaling potency arising from the stronger ECD complex-binding affinity observed for AM2/IMD at the CLR·RAMP1 ECD complex (Fig. 7, B and C). Indeed, as predicted  Table S2 for a summary of the pEC 50 values with error and statistical analyses.
the AM(13-33)-AM2/IMD(28-47) and AM2/IMD(8-27)-AM (34-52) agonist chimeras exhibited cAMP-signaling potencies at the untagged, WT CLR·RAMP1 complex expressed in COS-7 cells that were essentially equivalent to those of the WT agonists that corresponded to their C-terminal segment identity (Fig. 7, D and E, and Table S4). These results indicated that the hrCNE gel shift assays with detergent-solubilized tagged receptors and SUMO-mG s reported on physiologically relevant biochemistry of the CGRP receptor.

Discussion
The hrCNE method is a powerful technique for assessing the formation of membrane protein complexes in detergents (36,39). Our results with the CLR·RAMP complexes using the HEK293S GnT1 2 cell line to eliminate N-glycan heterogeneity showed its capability to provide sharp defined bands even for dynamic membrane proteins such as GPCRs. The MBP tags and stabilizing effects of the RAMPs on CLR may also have contributed to the sharp bands we obtained. The CLR·RAMP complexes were specific because we saw no evidence for detergent-stable PTH1R·RAMP complexes (Fig. 1C). This was a nice control, but it was also a bit surprising because PTH1R has been reported to interact with RAMPs. Christopoulos et al. (40) showed an interaction of RAMP2, but not RAMP1/3, with PTH1R using an assay based on trafficking to the cell surface in HEK293 cells. More recently, using a bead-based immunoassay with DDM-solubilized HEK293 cells expressing epitope-tagged receptors, Lorenzen et al. (17) reported interaction of all three RAMPs with PTH1R. Interestingly, they also detected interactions of all three RAMPs with CLR in DDM. This appears to conflict with our observation that the CLR·RAMP2 complex was unstable in DDM/CHS (Fig. S2A). These discrepancies may reflect a requirement for fairly stable complexes to survive electrophoresis in the hrCNE method, or alternatively they may Figure 5. Measuring SUMO-mG s apparent binding affinities for the agonist-occupied, detergent-solubilized CLR·RAMP complexes by hrCNE gel shift assay. A, 8% hrCNE gels showing fluorescent bands corresponding to the MBP-CLR-EGFP and MBP-RAMP heterodimers (CLR:RAMP1, CLR:RAMP2, and CLR:RAMP3) or quaternary complexes formed in the absence or presence of 10 mM of the indicated agonist peptides and increasing concentrations of SUMO-mG s added to the membrane preparations before LMNG/CHS solubilization. Each gel is a representative from three independent experiments. B-D, quantitation of the appearance of the quaternary complex bands from A by densitometry. E, scatter plot showing the replicate pEC 50 values for appearance of the quaternary complex band quantified by densitometry. The CLR·RAMP complexes is denoted as R1, R2, and R3. F, replicate pEC 50 values for the disappearance of the heterodimer band quantified by densitometry. See Table S3 for a summary of the pEC 50 values with error and statistical analyses.
be due to differences in methodological details. Importantly, here we not only demonstrated GPCR·RAMP complex formation but also functionality by coupling to peptide ligands and mG proteins. Given the recent reports of RAMP interactions with numerous other GPCRs (17,18), most of which have not been further validated by other methods or tested for functional consequences, the hrCNE assay may be a useful method with which to examine some of these putative RAMP·GPCR complexes.
The hrCNE methods as applied here can provide an alternative or complement to the fluorescence-detection size-exclusion chromatography (FSEC) and thermofluor stability assays widely used for screening membrane proteins for structure/ function studies. Unlike the thermofluor assays (41), which require purified protein, the hrCNE thermostability assays work with raw unpurified lysates. The FSEC technique also works with unpurified material (42) and can be used for thermostability assays (43), but it requires time-consuming serial gel filtration runs and FPLC fluorescence detection equipment not standard in most biochemistry laboratories. The hrCNE methods applied here use simple equipment and allow analysis of multiple samples in parallel on one or multiple gels. The hrCNE thermostability assay might be of use in campaigns to develop thermostabilized membrane proteins as pioneered by the Tate laboratory (44). Although the hrCNE methods may not work in all cases, they provide another option for consideration alongside the FSEC and thermofluor assays. The hrCNE method can be very powerful for screening and characterizing membrane protein interactions with other proteins. Although not shown here, the agonist-dependent coupling of CLR·RAMP complexes to mG s could be observed using solubilized adherent cultures in 48-well plates. This facilitates rapid screening of interactions without the need for membrane preparations. The native PAGE assay may be of use for future studies of CLR·RAMP interactions with GPCR kinases and b-arrestins.
The tagged CLR and RAMP constructs used for the hrCNE assays with the membrane preparations were quite large (;180 kDa for the heterodimers), so it is not surprising that we did not observe mobility shifts in the presence of the much smaller peptide antagonists or agonists (2-5 kDa) alone, even though these were almost certainly bound to the receptors (Fig. 3A). In future work it should be possible to use peptide ligands labeled with various fluorophores to characterize their binding to the uncoupled receptors. We saw no evidence for agonist-promoted coupling of the tagged receptors to endogenous G proteins in the HEK293 cells, presumably because their levels were low relative to the overexpressed receptors. There was no binding of SUMO-mG s to the receptors in the absence of agonist at most of the concentrations tested, but we cannot rule out some very weak "precoupling" occurring with 75 mM SUMO-mG s as evidenced by the decreased intensities of the heterodimer bands observed at this concentration in the absence of agonist (Fig. 5A). Both agonist and SUMO-mG s were required to obtain a mobilityshifted quaternary complex band, which is consistent with the known behavior of GPCRs. Evidence that the shifted bands were the quaternary complexes was provided by their decreased shifting or disappearance upon SUMO protease treatment (Fig. 3A) and their appropriate sensitivities to small molecule and peptide antagonists (Fig. 3B). It is unclear why the SUMO protease-liberated mG s failed to generate a mobility shift with the agonist-occupied RAMP2 and -3 complexes. It may have bound without causing a mobility shift because of compensating conformational and/or charge changes, or alternatively SUMO may have provided nonspecific interactions with the detergent micelle that increased the fusion protein affinity for the detergent-solubilized receptors. Even if the latter were the case, it is unlikely to have altered the mG interactions with the receptors that are relevant to this study. mG proteins N-terminally tagged with fluorescent proteins have been used in live-cell studies with no apparent detriment to their function (33).
The binding experiments in Figs. 4 and 5 demonstrated that the hrCNE gel assays could provide quantitative information for the binding of peptide agonists to the mG s -coupled receptors and for mG s binding to the agonist-occupied receptors. The two binding events are cooperative and saturating the receptor with mG s in the absence of agonist was not feasible. In addition, we could not rule out the possibility of some perturbation of the equilibrium occurring during the electrophoresis. For these reasons we chose to conservatively fit pEC 50 values and described these as apparent binding affinities. Moreover, for some of the experiments where the agonist was varied, we were approaching ligand-depletion conditions because we had to use ;10 nM receptor to obtain a good fluorescence signal. This was reflected in the increased steepness of these curves, which were better fit with a four-parameter variable slope model (Fig. 4) than the fixed-slope three-parameter model that was used for the experiments in which SUMO-mG s was varied (Fig. 5). Nonetheless, the agonist apparent affinity values we obtained were reasonable considering their known pharmacology (16), and the apparent affinities determined for SUMO-mG s were in line with a prior study. Using a different binding assay, Nehmé et al. Not surprisingly, the Fig. 4 and 5 experiments indicated agonist-dependent control of mG s coupling. More importantly, Figure 7. Cell-based cAMP signaling for the untagged WT CLR·RAMP1 complex in response to chimeric agonist peptides of AM and AM2/IMD designed based on hrCNE assay results. A, amino acid sequence alignment of AM and AM2/IMD. The two segments that were swapped to create the chimeras are indicated. B, agonist chimeras depicted using the AM agonist structure from the active-state AM-and G s -bound CLR·RAMP2 cryo-EM structure (Protein Data Bank code 6UUD) and AM2/IMD agonist structure from the active-state AM2/IMD-and G s -bound CLR·RAMP3 cryo-EM structure (Protein Data Bank code 6UVA). C, scatter plot summarizing the pK I binding affinities of the indicated peptide antagonist fragments for the three purified RAMP-CLR ECD complexes measured in a fluorescence polarization competitive binding assay. Data are taken from the work of Roehrkasse et al. (21). N.D., no binding detected. D, representative concentration-response curves for cAMP accumulation mediated by untagged CLR·RAMP1 transiently expressed in COS-7 cells. Errors are shown as S.D. for technical replicates. E, scatter plot summarizing cAMP potencies for the WT and chimeric peptides plotted as means 6 S.E. from three independent replicates. See Table S4 for a summary of the pEC 50 values with error and statistical analyses. ns, not significant. however, the Fig. 5 experiments also provided evidence that the RAMPs modulate CLR G protein coupling in addition to their well-known function of modulating agonist binding. This is most evident from the data for the AM2/IMD agonist. It exhibited similar double-digit nanomolar apparent affinities for each SUMO-mG s -coupled receptor (Fig. 4), but the apparent affinities of SUMO-mG s for the AM2/IMD-occupied receptors varied, with RAMP1 promoting the strongest apparent affinity (;5 mM), followed by RAMP3 (;10 mM) and then RAMP2 (;70 mM) (Fig. 5). These mG s affinity differences support differential modulation of CLR G protein coupling by the three RAMPs. Stronger RAMP1-mediated coupling was also evident in the experiments with the mG sq chimera (Fig. 6A). The recent cryo-EM studies of active state CLR·RAMP complexes provide a possible molecular basis for these affinity differences. Analysis of dynamic information present in the cryo-EM data found evidence for coordinated motions between the ECD complexes and the G protein that appeared to be RAMP-dependent, with the ECD complex motion being most restricted by RAMP1, followed by RAMP3 and then RAMP2 (23).
In a previous study, Weston et al. (29) examined G protein coupling to CLR·RAMP complexes expressed in yeast in which the endogenous G protein GPA1 was modified to contain the C-terminal five amino acids of eleven different human Ga subunits. They observed coupling of each CLR·RAMP complex to the chimeras with G s , G q , and G i and reported dramatic agonist-and RAMP-dependent signaling bias. For each receptor complex, agonist potency rank order changes were observed with the three different G protein chimeras. Some of these findings appeared to translate to signaling studies in HEK293 cells. In the mG sq and mG si chimeras we used here, all residues predicted to contact the GPCR within the larger 19 amino acid segment nearest the C terminus were exchanged. Our Fig. 5 and 6 data were consistent with a coupling rank order of mG s . mG sq . mG si for each receptor, and the different agonists did not appear to elicit substantial changes to this pattern. Thus, dramatic ligand-or RAMP-dependent biasing of mG coupling preferences was not evident in our assays, although the weak mG sq and mG si binding we observed could have prevented detection of bias. The native PAGE assay is likely less sensitive than the yeast signaling assay used by Weston et al. (29). Nonetheless, our data appear to be consistent with the study of Garelja et al. (28) in which activation of several signaling pathways downstream of the CLR·RAMP complexes was measured in COS-7 cells, and more balanced signaling was generally observed. In this study, agonist potency rank order changes were largely limited to experiments measuring ERK activation via the CLR·RAMP1 complex. Notably, they reported that only the CGRP agonist-CLR·RAMP1 pairing was able to elicit a measurable IP1 response (considered to be downstream of G q ), and this is consistent with our findings that mG sq coupling was strongest for CGRP at CLR·RAMP1, and the two AM peptides more weakly promoted mG sq recruitment to the three receptors. Ultimately, obtaining a thorough understanding of signaling bias at the CLR·RAMP complexes will require additional studies with various methods.
Our results provide important insights into the two-domain ligand-binding mechanism and how agonist engagement of the CLR ECD and TMD contributes to G protein recruitment and ultimately the signaling response. The selectivity profiles obtained for agonist binding to the SUMO-mG s -coupled receptors (Fig. 4, E and F) were similar to the agonist cAMPsignaling potency selectivity profiles (Fig. 2F). They were also similar to the peptide ECD complex-binding selectivity profiles previously reported (Fig. 7C) (21), with some deviations likely caused by additional selectivity-altering contacts of the agonists with the TMD. In contrast, the selectivity patterns obtained for SUMO-mG s binding to the agonist-occupied receptors (Fig. 5, E and F) were distinct from the cAMP-signaling potency and agonist-binding patterns. The two AM peptides equally recruited SUMO-mG s to the CGRP receptor despite having different signaling potencies and affinities for the SUMO-mG s -coupled receptor. Also, CGRP was almost as good as the two AM peptides at recruiting SUMO-mG s to the two AM receptors despite its signaling potency and affinity for the SUMO-mG s -coupled receptors being substantially weaker. Comparing the results in Figs. 2, 4, 5, and 7C strongly suggested that the affinities of the agonists for the SUMO-mG s -coupled receptors were determined by their dual-site interactions with the CLR·RAMP ECD complex and the CLR TMD, whereas the recruitment of SUMO-mG s to the agonist-occupied receptors was largely driven by the single-site interaction of the N-terminal half of the agonist with the CLR TMD.
SUMO-mG s recruitment to the agonist-occupied receptors reflects agonist efficacy in part. mG s mimics the nucleotidefree heterotrimer, so it can be thought of as a surrogate for measuring G protein recruitment and GDP release. However, because GTP binding and G protein dissociation are not incorporated, mG s recruitment does not provide a full measure of efficacy. Indeed, different agonists can have different efficacies for a single G s -coupled GPCR because they promote altered Gprotein conformations with different sensitivities to GTP binding and ternary complex disruption (39). Nonetheless, for most of the agonist-receptor pairings studied here, our data appear to provide a reasonable explanation for how the agonist affinity and "efficacy," as reflected by its SUMO-mG s recruitment ability, dictate its cAMP-signaling potency. The AM and AM2/ IMD agonist chimera experiments showed that their different cAMP-signaling potencies at CLR·RAMP1 were determined by their C-terminal ECD-binding segments, which have different ECD complex affinities (Fig. 7). These results were consistent with the N-terminal TMD-binding segment of the two AM peptides having equal efficacies for signaling through G s at this receptor as suggested by the SUMO-mG s binding assay (Fig. 5,  B and E).
Our data provide a plausible explanation for why CGRP signaling potency at CLR·RAMP1 and its binding affinity for the SUMO-mG s -coupled receptor were stronger than AM2/IMD (Figs. 2, C and F, and 4, B and E), even though its ECD complex-binding affinity was weaker (Fig. 7C). This likely results from the much stronger SUMO-mG s recruitment promoted by the CGRP N-terminal TMD-binding segment as compared with AM2/IMD (Fig. 5, B and E), which may be due in part to CGRP residue Ala 5 in the disulfide-linked loop, which is a Gly at the equivalent positions in AM2/IMD (Gly 13 ) and AM (Gly 19 ). It was recently reported that AM2/IMD G13A and AM G19A exhibited increased cAMP-signaling potencies and CGRP A5G had decreased cAMP-signaling potencies at all three receptors (28,37). Last, our data suggest that the differences in cAMP-signaling potencies of the agonists at the CLR·RAMP2 and CLR·RAMP3 complexes (Fig. 2, D-F) are determined in large part by their different ECD complex-binding affinities (Fig. 7C) and to a lesser extent by their abilities to recruit SUMO-mG s , which were similar (Fig. 5, C-E). One signaling aspect for which our native PAGE data did not provide an obvious explanation is the partial agonism of AM2/IMD at the CLR·RAMP2 complex.
In conclusion, our results revealed a role for RAMPs in modulating CLR G protein coupling and provided a framework for understanding how agonist engagement of the two domains of CLR affects transducer recruitment and cAMP signaling. Our data support a model in which the binding of the C-terminal half of the agonists to the ECD complexes is a primary determinant of receptor selectivity, and in several cases this also largely explains their differences in cAMP-signaling potencies. The binding of the N-terminal half of the agonists to the CLR TMD determines the strength of G protein recruitment, and for the CGRP agonist at CLR·RAMP1, this component is a significant contributor to its stronger cAMP-signaling potency. This information provides a valuable biochemical foundation for understanding the shared and distinct signaling properties of CGRP, AM, and AM2/IMD in human physiology and disease and guiding the development of therapeutic peptide analogs with desired signaling properties.

Plasmids
Plasmids for expressing the tagged human receptors in mammalian cells used the pHLsec vector (45). The GPCRs were tagged at their N terminus with an HPC4 epitope tag and MBP and at their C terminus with monomeric EGFP and a 1D4 epitope tag. 3C protease cleavage sites flanked the receptors. The receptor constructs were as follows (with the receptor amino acid residue numbers indicated): HPC4-MBP-3C-CLR.29-403-3C-EGFP-1D4 (pMW084) and HPC4-MBP-3C-PTH1R.29-478-3C-EGFP-1D4 (pMW113). The RAMPs were tagged at their N termini with MBP followed by a 3C site, and their C termini were either tagged with mCitrine or left native. The RAMP constructs were as follows: MBP-3C-RAMP1.24-148-mCitrine (pMW051), MBP-3C-RAMP2.55-175-mCitrine (pMW070), MBP-3C-RAMP3.25-148-mCitrine (pMW069), MBP-3C-RAMP1.24-148 (pAMR026), MBP-3C-RAMP2.55-175 (pAMR027), and MBP-3C-RAMP3.25-148 (pAMR011). Cloning was performed by PCR/restriction enzyme digest/ligation or Gibson assembly methods. The fusion constructs were cloned between the AgeI and KpnI sites of pHLsec, and EcoRV and NotI sites flanked the receptor/RAMP-encoding fragments. Oligonucleotides were synthesized by Integrated DNA Technologies (Coralville, IA) The plasmid for bacterial expression of MBP-TEV cleavage site-EGFP-His 8 (pMW086) was made by PCR amplification of EGFP-His 8 , restriction digest with BamHI and NotI, and ligation into a pETDuetI vector with MBP-TEV in the first multiple cloning site and DsbC in the second multiple cloning site (46). DNA sequences for expression of the three mini-G proteins: mini-G s (construct 393), mini-Gsq (construct 70), and mini-Gsi (construct 43), were purchased as Gene art strings from Thermo Fisher Scientific using E. coli codon-optimized DNA sequences based on the published amino acid sequences (31). The Gene art strings included BamHI and NotI sites that were digested, and the inserts were ligated into the pSUMOT7Amp vector for bacterial expression (Lifesensors, Malvern, PA) to produce His 6 -SUMO-mG s (pMW101), His 6 -SUMO-mG si (pMW099), and His 6 -SUMO-mG sq (pMW100). The University of Oklahoma Health Sciences Center Laboratory for Molecular Biology and Cytometry Research core facility was used to confirm the coding sequences of all plasmids by DNA sequencing. Amino acid sequences of the tagged CLR, PTH1R, RAMP, and mini-G constructs are listed in Figs. S8-S10. For protein expression in mammalian cells, the plasmids were purified using a Macherey-Nagel midi kit according to the manufacturer's instructions. The pcDNA3.1 plasmids encoding untagged CLR and RAMP1 used for the cAMP signaling assays in Fig. 7 were previously described (26).

Peptides
Synthetic agonist peptides: aCGRP(1-37), bCGRP(1-37), AM(13-52), and AM2/IMD(1-47) were purchased from Bachem (Torrance, CA). Antagonist peptides and agonist chimeras were custom synthesized and HPLC-purified by RS Synthesis (Louisville, KY). The lyophilized powders were dissolved in sterile ultrapure water to 10 mg/ml and stored in aliquots at 280°C. Peptide concentrations were determined by UV absorbance at 280 nm following dilution of the peptides in 10 mM Tris-HCl, 1 mM EDTA, pH 8.0. Extinction coefficients were calculated from the Tyr, Trp, and cystine content of each peptide. Because CGRP peptides lack Tyr and Trp resides, peptide concentration was calculated based on the peptide content reported by Bachem (78.5% w/w). The peptide sequences are listed in Table S5.

Bacterial expression and purification of H 6 -SUMO-mG fusion proteins
The proteins were expressed in E. coli BL21 (DE3) with 6 liters of total culture volume. The cultures were induced by the addition of 0.4 mM isopropyl b-D-thiogalactopyranoside at mid-log phase, and the temperature was reduced to 16°C with expression overnight. The harvested cell pellets were resuspended in 100 ml of 50 mM Tris-HCl, pH 7.5, 10% (v/v) glycerol, 150 mM NaCl, 25 mM imidazole, and stored at 280°C. All purification steps were carried out on ice or at 4°C, and the column steps used an AKTA purifier (GE Healthcare). 5 mM b-mercaptoethanol, 50 mM GDP, and 1 mM MgCl 2 were added to the thawed, resuspended pellets. The cells were lysed by sonication, and the total lysate was spun at 25,000 3 g for 30 min. The supernatant was loaded on a 50 ml of nickel-chelating Sepharose column equilibrated in buffer A (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 10% (v/v) glycerol, 25 mM Imidazole, 5 mM b-mercaptoethanol, 50 mM GDP, and 1 mM MgCl 2 ). The column was washed with 200 ml of buffer, with the imidazole raised to 72.5 mM and then eluted with the buffer containing 262.5 mM imidazole. The protein was concentrated by precipitation with solid ammonium sulfate to 65% saturation, and the precipitated protein was pelleted at 15,000 3 g and resuspended in a minimal volume of 25 mM Tris-HCl, pH 7.5, 10% (v/v) glycerol, 2 mM DTT, 1 mM MgCl 2 , 1 mM GDP. The SUMO-mG fusions were further purified by size-exclusion chromatography using a 320ml Superdex200HR column (GE healthcare) and a buffer of 50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 10% (v/v) glycerol, 2 mM DTT, 1 mM MgCl 2 , 1 mM GDP. The final pooled protein was concentrated using an Amicon Ultra 10,000-Da molecular mass cutoff concentrator (Milipore Sigma) and dialyzed overnight against 1 liter of 25 mM HEPES, pH 7.5, 150 mM NaCl, 50% (v/v) glycerol, 0.5 mM DTT, 1 mM MgCl 2 , 1 mM GDP for storage at 280°C. Protein concentrations were determined by Bradford assay with BSA standard and confirmed by UV absorbance at 280 nm. Final yields for each protein were 30 mg for H 6 -SUMO-mGs393, 47 mg for H 6 -SUMO-mG sq 70, and 25 mg for H 6 -SUMO-mG si 43.

Bacterial expression and purification of MBP-TEV cleavage site-EGFP-H 8
Expression and purification by immobilized metal affinity chromatography were as above for the SUMO-mG fusion proteins except that the buffers lacked b-ME, GDP, and MgCl 2 . Peak fractions from the nickel column were pooled and loaded onto an amylose high flow (NEB) column pre-equilibrated in buffer C (50 mM Tris-HCl, pH 7.5, 5% (v/v) glycerol, 150 mM NaCl). The column was washed with buffer C and then eluted using a linear gradient of buffer C to D (50 mM Tris-HCl, pH 7.5, 5% (v/v) glycerol, 150 mM NaCl, 10 mM maltose), dialyzed to storage buffer (25 mM sodium HEPES, pH 7.4, 50% (v/v) glycerol, 150 mM NaCl), and stored at 280°C. Protein concentration was determined by visible absorbance at 488 nm using the molar absorptivity of EGFP (55,000 M 21 cm 21 (47)) and was further verified by Bradford assay using a BSA standard curve.
Cell-based cAMP accumulation assays cAMP accumulation assays were performed as previously described (21), except that stimulation time was reduced to 15 min. In brief, COS-7 cells were transiently transfected with the expression constructs using PEI, and the cells were stimulated with agonist 48 h after transfection. The cells were lysed, and the cAMP content was determined using the LANCE cAMP kit (PerkinElmer) according to the manufacturer's directions and a PolarSTAR Omega Plate Reader with an advanced optic head (BMG Labtech, Ortenberg, Germany). DNA (43 mg/dish; 1:1 ratio of the CLR and RAMP constructs) was added to 4.3 ml of DMEM followed by the addition of PEI (64.5 mg/dish). The transfection mixture was incubated at room temperature for 10 min, and then valproic acid was added to 35 mM. Growth medium was removed from each of the dishes by aspiration and replaced with 26 ml of DMEM with 2% FBS, 13 NEAA, and 50 units/ml penicillin, 50 mg/ml streptomycin/dish. 4.3 ml of the transfection mixture was added to each plate (final concentration, 5 mM valproic acid), and the dishes were incubated at 30°C and 5% CO 2 for 3 days.

Crude membrane preparations
The medium was aspirated, the cells were washed with 8 ml of PBS/plate and harvested by adding 8 ml of PBS with 5 mM EDTA and scraping with a cell scraper. The cells were divided into four 50-ml conical tubes and pelleted by centrifugation at 1000 3 g for 5 min at 20°C . The supernatants were discarded, and cell pellets were resuspended in 24 ml/conical ice-cold hypotonic buffer (25 mM HEPES, pH 7.5, 2 mM MgCl 2 , 1 mM EDTA, with 13 EDTA-free Pierce protease inhibitor tablet (PI)) and homogenized using an Ultra Turrax T25 (IKA Works, Wilmington, NC) for 30 s at 10,000 rpm with the samples cooled in an ice-water bath. The samples were incubated for 10 min and then homogenized a second time followed by a lowspeed spin of 800 3 g at 4°C for 10 min to pellet cell debris. The supernatants from the low-speed spin were transferred to ultracentrifuge tubes and spun at 100,000 3 g for 1 h at 4°C. The pelleted membranes from four tubes were combined in 6 ml of storage buffer (25 mM HEPES, pH 7.5, 25 mM NaCl, 2 mM MgCl 2 , 10% (v/v) glycerol, 13 PI) and homogenized three times at 5,000 rpm for 30 s with cooling in between. Single-use aliquots were flash-frozen in liquid nitrogen and stored at 280°C.
To estimate the concentration of the CLR·RAMP heterodimer in the preparation, 80 ml was diluted to 400 ml in solubilizing buffer (25 mM HEPES, pH 7.5, 140 mM NaCl, 10 mM KCl, 1 mM MgCl 2 , 2 mM CaCl 2 , 13 PI, 0.5% (w/v) LMNG, and 0.05% (w/v) CHS) and incubated 2 h on a tube tumbler at 4°C. The lysate was spun at 16,100 3 g at 4°C for 10 min. 10 ml of the supernatant was combined with 5 ml of purified MBP-TEV-EGFP at known concentrations, run on a nonreducing 12% SDS-PAGE gel, and imaged by in-gel fluorescence using a Chemidoc MP (Bio-Rad). The protein concentration for the MBP-3C-CLR-3C-EGFP-1D4 was estimated by comparing the band intensity with that of the MBP-EGFP. We estimated ;100 nM heterodimer in each of the membrane preparations, and this allowed their use as 103 stocks for the native PAGE assays. Total protein content was determined for the membrane preparations using the DC protein assay (Bio-Rad), yielding concentrations for the MBP-CLR-EGFP·MBP-RAMP1, -2, and -3 preparations of 4.56, 4.11, and 6.36 mg/ml, respectively.
Adherent cell expression of GPCR·RAMP heterodimers and detergent solubilization HEK293S GnT1 2 cells were seeded at 120,000 cells/well into 48-well plates and grown to ;90% confluency (overnight) in 250 ml growth media/well. The cells were transfected with 0.3 mg of total DNA and 0.45 mg of PEI/well. For co-transfections, 0.15 mg of each receptor component was used per well. The DNA and PEI were combined in 30 ml of DMEM, incubated at room temperature for 10 min, and then valproic acid was added (46.7 mM). The growth medium was removed by aspiration and replaced with 250 ml of DMEM with 2% FBS, 13 NEAA, 50 units/ml penicillin, 50 mg/ml streptomycin, and 30 ml of the transfection mixture. The plates were incubated for 3 days at 30°C, 5% CO 2 . For solubilization, the medium was removed by aspiration, the cells were washed with 250 ml of PBS/well, the PBS was aspirated, and the plates were placed on ice. 100 ml of solubilization buffer (25 mM HEPES, pH 7.5, 140 mM NaCl, 10 mM KCl, 1 mM MgCl 2 , 2 mM CaCl 2 , 3% (v/v) glycerol, 0.6253 PI, 0.5% (w/v) LMNG, and 0.05% (w/v) CHS) was added to each well, and the plate was incubated at 4°C with rocking for 2 h. The lysates were transferred to prechilled microcentrifuge tubes and centrifuged at 16,100 3 g at 4°C for 10 min in a benchtop microcentrifuge, and the supernatants were used for hrCNE analysis.
Thermostability assay for CLR·RAMP complexes expressed in adherent cells HEK293S GnT1 2 cells were seeded at 500,000 cells/well into 12-well plates and grown to ;90% confluency (24 h). The cells were transfected with 1.2 mg of total DNA/well (1:1 ratio of CLR and RAMP constructs) using PEI as described above and a 120-ml volume transfection mixture/well. The growth medium was removed from each well by aspiration and replaced with 1 ml of DMEM with 2% FBS, 13 NEAA, 50 units/ml penicillin, 50 mg/ml streptomycin, and 120 ml of the transfection mixture, and the plate was incubated for 3 days at 30°C and 5% CO 2 . The medium was aspirated, the cells were washed with 1 ml PBS/well, the PBS was aspirated, and the plates were placed on ice. 200 ml of solubilization buffer was added to each well, and the plate was incubated at 4°C rocking for 2 h. The solubilization buffers were 25 mM HEPES, pH 7.5, 140 mM NaCl, 10 mM KCl, 1 mM MgCl 2 , 2 mM CaCl 2 , 3% (v/v) glycerol, 0.6253 PI, and one of three detergent systems: 1.5% (w/v) DDM, 0.15% (w/v) CHS; 0.5% (w/v) LMNG, 0.05% CHS; or 0.25% (w/v) LMNG, 0.25% (w/v) GDN, 0.05% (w/v) CHS. The lysates were transferred to cold microcentrifuge tubes and spun at 16,100 3 g, 4°C for 10 min. The supernatant was aliquoted (22 ml) into cold microcentrifuge tubes, and each tube was incubated at the indicated temperature for 30 min. The tubes were cooled on ice for 5 min and spun at 16,100 3 g at 4°C for 10 min, and the supernatants were used for hrCNE analysis.
Binding and solubilization reactions using membrane preparations with exogenous addition of ligands and SUMO-mG proteins These reactions were assembled in microcentrifuge tubes with 30-50 ml of total assay volume. For all assay formats (Figs. 3-6 and Figs. S4 and S7), the final solubilized reactions contained ;10 nM receptor heterodimer from the membrane preparation and 0.05 unit/ml apyrase, ligands, and purified SUMO-mG proteins as indicated in binding buffer (25 mM HEPES, pH 7.5, 140 mM NaCl, 10 mM KCl, 1 mM MgCl 2 , 2 mM CaCl 2 ) supplemented with 0.075 mg/ml FAF-BSA, 1-2% (v/v) glycerol, 0.5383 PI, 0.5% (w/v) LMNG, and 0.05% (w/v) CHS for solubilization. Where SUMO-mG proteins were used, they were preincubated with the redox pair GSH/GSSG in binding buffer supplemented with 0.1 mg/ml FAF-BSA and 0.253 PI on ice for 30 min, because the purified mG proteins tended to form intermolecular disulfide bonds. The GSH/GSSG was present at a fixed molar ratio with the SUMO-mG such that the GSH/ GSSG was diluted out proportionally to the SUMO-mG. The GSH and GSSG were kept at a 5:1 molar ratio, and unless otherwise noted, the GSH was 3-fold higher molarity than SUMO-mG, e.g. for 50 mM SUMO-mG s , 150 mM GSH and 30 mM GSSG were added. Prior to solubilization, the binding reactions contained the membrane preparation and other components (apyrase, ligands, and SUMO-mG as indicated) at 1.33 of the final concentrations in binding buffer supplemented with 0.1 mg/ml FAF-BSA, 0.253 PI. The binding reactions were incubated for 30 min on ice; then detergent buffer with the detergent at 43 (binding buffer supplemented with 13 PI, 2% (w/v) LMNG, 0.2% (w/v) CHS) was added; and the reactions were tumbled at 4°C for 2 h to solubilize followed by centrifugation at 16,100 3 g at 4°C for 10 min and analysis of the supernatants by hrCNE. The preincubation and set up of binding reactions differed slightly based on the assay format as described below.
For the assays investigating peptide ligand and SUMO-mG s interactions with the CLR·RAMP complexes ( Fig. 3A and Fig.  S4), the binding reactions were assembled by first combining buffer components as a mastermix (binding buffer, with or without apyrase for Fig. 3A versus S4B, FAF-BSA, and PI) followed by addition of SUMO-mG s and GSH/GSSG to indicated reaction tubes for the preincubation. Then the ligands, membrane preparation, and SUMO protease were added to the reactions as indicated.
For experiments with competing agonists and antagonists (Fig. 3B), SUMO-mG s and the GSH/GSSG redox pair were diluted to 2.673 of the final concentrations for the 30-min preincubation. Ligands, apyrase, and the membrane preparation were combined as indicated at 2.673 of final concentrations (15 ml volume) in binding buffer with 0.1 mg/ml FAF-BSA and 0.253 PI. The membrane preparation was added after the ligands for simultaneous exposure to agonists and antagonists. 15 ml of the SUMO-mG s with GSH/GSSG or binding buffer supplemented with 0.1 mg/ml FAF-BSA and 0.253 PI for wells without SUMO-mG s was then added to the reaction tubes containing ligands, apyrase, and the membrane preparation.
To measure agonist binding to CLR·RAMP complexes in the presence of excess SUMO-mG s (Fig. 4), the preincubation was assembled with apyrase, SUMO-mG s , and GSH/GSSG at 2.673 of the final concentration in binding buffer with 0.1 mg/ ml FAF-BSA and 0.253 PI and incubated for 30 min on ice. The membrane preparation was then added to the mastermix. For the CLR·RAMP3 reactions, the redox pair concentration was 3-fold lower (133.3 mM SUMO-mG s , 133.3 mM GSH, 26.67 mM GSSG) because this receptor complex seemed to be more sensitive to the reducing agent in this assay format. The agonist was diluted to 2.673 of the final highest concentration in binding buffer with 0.1 mg/ml FAF-BSA and 0.253 PI and then 3fold serially diluted in the same buffer. The binding reactions were assembled by combining equal volumes of the master mix and agonist dilutions (or binding buffer with 0.1 mg/ml FAF-BSA and 0.253 PI for the reaction lacking agonist).
For measuring SUMO-mG binding to agonist-occupied CLR·RAMP complexes (Figs. 5 and 6 and Fig. S7), the SUMO-mG and GSH/GSSG were diluted to 2.673 of the highest final concentration (200 mM SUMO-mG with 600 mM GSH and 120 mM GSSG except in the case of SUMO-mG si , for which GSH/ GSSG was 3-fold higher because it has more cysteines) in binding buffer with 0.1 mg/ml FAF-BSA and 0.253 PI for the preincubation and then 3-fold serially diluted in binding buffer with 0.1 mg/ml FAF-BSA and 0.253 PI. Agonists, apyrase, and the membrane preparation were combined in a mastermix at 2.673 of final concentrations in binding buffer with 0.1 mg/ml FAF-BSA and 0.253 PI, and the binding reactions were assembled by combining equal volumes of the master mix with the SUMO-mG dilution or with binding buffer supplemented with 0.1 mg/ml FAF-BSA and 0.253 PI for the reaction lacking SUMO-mG.

hrCNE and SDS-PAGE
The native gels were set up with cold buffers and run at 4°C. The hrCNE protocols, gel, and buffer recipes were based on those described (36,39). 8% resolving polyacrylamide hrCNE gels were used with cathode buffer: (;175 ml) 50 mM Tricine, 7.5 mM imidazole, pH 7.0, supplemented with the detergent used for solubilization at the following concentrations: 0.01% (w/v) LMNG, 0.001% (w/v) CHS; 0.02% (w/v) DDM, 0.002% (w/v) CHS; or 0.005% (w/v) LMNG, 0.005% (w/v) GDN, 0.001% (w/v) CHS; anode buffer (1 liter): 25 mM imidazole, pH 7.0. The gels were prerun at 100 V for 20 min, 20 ml of solubilized lysate supernatant was loaded, and the gels were run at 200 V for 2.5 h in experiments examining heterodimer formation or thermostability and for 3.5 h in experiments examining peptide and SUMO-mG interactions. The gels were rinsed in ddH 2 O and imaged by in-gel fluorescence using the ProQEmerald488 preset on a Chemidoc MP (Bio-Rad). This channel detects both EGFP and mCitrine.
For SDS-PAGE gels, 15 ml of the solubilized lysate supernatants was combined with 5 ml of 43 nonreducing SDS loading dye, run on 12% polyacrylamide gels, and visualized by in-gel fluorescence as above. To excise the hrCNE gel heterodimer bands for analysis by SDS-PAGE, the fluorescent images of the gels were printed at actual size, the hrCNE gel was placed over the printed image, and a gel fragment corresponding to the dimer band was excised with a razor blade. Excised fragments were placed in the wells of a 12% SDS-PAGE gel and covered with 13 nonreducing SDS loading dye. The SDS-PAGE gels were run at 50V until the loading dye passed through the stacking gel and then at 200 V for 1.5 h. 5 ml of the WesternC (Bio-Rad) protein standard was loaded in lane 1, and fluorescent marker bands were imaged using the Cy3 and Cy5.5 channels on the Chemidoc MP.

Densitometry analysis
All gels that were used for quantifying dimer or quaternary complex bands were imaged at similar exposure times (;80 s), ensuring that there were no saturated pixels. Densitometry analysis was conducted using the ImageLab software (Bio-Rad). Using the Lane Profile tool, the distance from the top of the gel to the boundaries of the fluorescent bands was determined. Lane 1 was used to determine the boundaries of the dimer band, and lane 10 was used to determine the boundaries of the quaternary complex band. These measurements were used to manually determine the quaternary complex and dimer bands in the other lanes. The adjusted volume (Intensity) of the selected bands was plotted against agonist or SUMO-mG concentration in GraphPad Prism. The binding curves for SUMO-mG apparent affinities for the agonist-occupied receptors were fit to a three-parameter (fixed slope) dose-response model to determine pEC 50 values. The binding curves for agonist apparent affinities for the mG scoupled receptors were fit to a four-parameter (variable slope) dose-response model to determine pEC 50 values.

Data analysis and statistics
All experiments that involved quantifying results were done as three independent experiments on different days. Statistical analysis for the SUMO-mG s and agonist affinities for the CLR·RAMP complexes quantified by densitometry was done by comparison of pEC 50 values for appearance of the quaternary complex band or disappearance of the dimer band using an ordinary one-way analysis of variance with Tukey's multiple comparisons test. For the cell-based assays measuring cAMP potencies at the tagged constructs ( Fig. 2 and Table S1) and comparing WT to chimera peptides at the WT untagged receptors (Fig. 7 and Table  S4), pEC 50 values were compared using an ordinary one-way analysis of variance with Tukey's multiple comparisons test. Statistical significance was defined as p , 0.05.

Data availability
The data described in this article are available from the corresponding author (augen-pioszak@ouhsc.edu) upon reasonable request.