Extracellular loops 2 and 3 of the calcitonin receptor selectively modify agonist binding and efficacy

Graphical abstract Figure. No Caption available. Abstract Class B peptide hormone GPCRs are targets for the treatment of major chronic disease. Peptide ligands of these receptors display biased agonism and this may provide future therapeutic advantage. Recent active structures of the calcitonin (CT) and glucagon‐like peptide‐1 (GLP‐1) receptors reveal distinct engagement of peptides with extracellular loops (ECLs) 2 and 3, and mutagenesis of the GLP‐1R has implicated these loops in dynamics of receptor activation. In the current study, we have mutated ECLs 2 and 3 of the human CT receptor (CTR), to interrogate receptor expression, peptide affinity and efficacy. Integration of these data with insights from the CTR and GLP‐1R active structures, revealed marked diversity in mechanisms of peptide engagement and receptor activation between the CTR and GLP‐1R. While the CTR ECL2 played a key role in conformational propagation linked to Gs/cAMP signalling this was mechanistically distinct from that of GLP‐1R ECL2. Moreover, ECL3 was a hotspot for distinct ligand‐ and pathway‐specific effects, and this has implications for the future design of biased agonists of class B GPCRs.


Introduction
Class B1 G protein-coupled receptors (GPCRs) are the targets for peptide hormones that play major roles in the development and maintenance of lymphatic and cardiovascular function, bone homeostasis, metabolic regulation, migraine, stress and anxiety [1]. Consequently, these receptors are important therapeutic targets.
The calcitonin (CT), Class B1 GPCRs (CTRs), are highly expressed on osteoclasts and have been exploited therapeutically for treatment of bone disorders, including Paget's disease, hypercalcemia of malignancy and osteoporosis [2][3][4][5]. The receptors are also expressed in numerous other cells and tissues including leucocytes and their precursors, the central nervous system, kidney, lung, gastrointestinal tract and reproductive tissues [6], thereby influencing pain perception, feeding and reproduction, and ion secretion [2,7], though these actions are not well understood. Furthermore, CTRs can interact with the receptor activity modifying protein (RAMP) family to form high affinity receptors for amylin (Amy) and calcitonin gene-related peptide (CGRP) [8].
CT peptides from different species have been identified and can be classified into 3 major subgroups based on evolution and sequence conservation: human/rodent, artiodactyl (e.g., porcine) and teleost (salmon/eel)/chicken. Both human and salmon CT have been approved for treatment of bone disorders including Paget's disease and osteoporosis, however, they have distinct binding kinetics, affinity and efficacy [9][10][11] that impact on G protein recruitment and activation [11], suggesting different modes of interaction with CTRs.
Orthosteric peptide ligands of Class B1 GPCRs are proposed to interact with their cognate receptors via a two-domain mechanism, with an initial engagement of the C-terminus of the peptide with the Nterminal extracellular domain (ECD) of the receptor that allows the peptide N-terminus to bind to the transmembrane (TM) spanning receptor core comprising the 7 TM helices and 3 interconnecting extracellular loops (ECLs), leading to receptor activation [12]. This mode of binding is supported by recent full-length, active, Gs-complexed structures of the CTR [13] (bound to sCT) and glucagon-like peptide-1 (GLP-1) receptor (GLP-1R) (bound to GLP-1 [14], or exendin-P5 [15]). Nonetheless, there were marked differences in the orientation of the receptor ECD relative to the receptor core and correlative changes in presentation of the peptides to the receptor core, linked to differences in the degree of secondary structure of the peptides [13][14][15]. Moreover, the CT-family peptides have a cyclic, cysteine-disulfide linked N-terminus between amino acids 1 and 7 (2 and 7 for Amy and CGRP) that contrasts with the extended helix of GLP-1, and alters the relative interaction of the peptide N-termini with the ECLs and proximal TM helix segments.
Mutagenesis and crosslinking studies have shown that the ECLs of Class B1 GPCRs are critical for both peptide binding and propagation of conformational change associated with receptor activation [14][15][16][17][18][19][20]. For the GLP-1R, alanine-scanning mutagenesis revealed that the ECLs, particularly ECL2 and ECL3, were also important in the biased agonism of peptides, but had distinct contribution to pathway specific signalling [16,21]. For this receptor, both ECL2 and ECL3 played a critical role in cAMP formation, and intracellular calcium ( i Ca 2+ ) mobilisation, while effects on ERK phosphorylation (pERK) were principally confined to residues within ECL3.
In the current study, we have performed alanine-scanning mutagenesis of amino acids in ECLs 2 and 3 of the hCTR and interrogated mutant receptors for their effects on cell surface receptor expression, peptide affinity and efficacy for cAMP and IP1 accumulation, as well as pERK in response to calcitonin (sCT, hCT, pCT) and related family (Amy, CGRP) peptides. This work revealed both differences in how the receptor engages with and is activated by the different CT-family peptides, and in the role of ECLs 2 and 3 between the CTR and GLP-1R.  1. Alanine mutation of ECL2 and ECL3 of the hCTR selectively alters expression of the receptor. (A) Alignment of CT and related peptide sequences was performed using Biology WorkBench (workbench.sdsc.edu). Identical residues are highlighted in green, conservative substitutions are coloured blue, and semi-conservative substitutions are in orange. Black text indicates the non-conserved. (B) Snake diagram of the hCTR: highlighted in blue are the residues that constitute the signal peptide of the receptor, in orange the c-Myc tag, and in green the residues that have been mutated to alanine. (C) Active state model of the hCTR (pale blue ribbon), with position of mutated residues displayed as grey surface map. sCT is shown as dark red, with side chains in proximity to the ECLs displayed in x-stick, and residues 1-7 that are critical for receptor activation displayed in transparent cpk. (D) Expression of mutant receptors determined by FACS analysis of antibody binding to the c-Myc epitope. (E) Top view of the ECLs with mutants that significantly altered receptor expression displayed in colour according to the magnitude of effect; grey indicates no significant effect and black mutants where expression could not be measured. The receptor ECD and C-terminal peptide residues are omitted for clarity (F). There was a strong correlation between cell surface receptor expression by FACS and homologous competition radioligand binding. *  and incubated overnight. Initially, pERK1/2 time-course experiments were performed over 30 min to identify the time point when the pERK1/2 response is maximal (6-8 min). Subsequently, this time point was selected to generate concentration response for different agonists with ligand addition performed after overnight serum starvation. pERK1/2 was detected using an Alphascreen assay as previously described [22]. Data were normalized to the maximal response elicited by 10% FBS determined at 6 min.

Data analysis
IC 50 and Bmax values were estimated from competitive inhibition of 125 I-sCT  binding using a 3-parameter logistic equation (log(inhibitor versus response)) in Prism (v7; Graphpad). Data were corrected for radioligand occupancy using the Cheng-Prusoff equation in Prism; as such data are reported as pK i . Emax and EC 50 were estimated from concentration-response curves using with a 3-parameter logistic equation in Prism (v7). These values are a composite of functional affinity, efficacy and stimulus response coupling. The Black and Leff operational model of agonism [23] was applied to separate effects on pathwayspecific efficacy (τ) from those that modify ligand functional affinity (pK A ). Derived τ values were normalized to experimentally determined levels of cell surface expression to provide a measure of efficacy (τ c ) that is independent of affinity and altered cell surface receptor expression [16]. pKi, pK A and log τc values for mutant receptors were statistically compared to those of the WT receptor using a one-way analysis of variance (ANOVA) and Dunnett's post-test. Significance was accepted at P < 0.05.

System equilibration and classic MD run settings
All the following MD simulation stages were performed by using Acemd [35]. Equilibration of the four systems was achieved in isothermal-isobaric conditions (NPT) using the Berendsen barostat [36] (target pressure 1 atm) and the Langevin thermostat [37] (target temperature 300 K) with a low damping of 1 ps −1 . A three-stage procedure with an integration time step of 2 fs was performed: in the first stage, 2000 conjugate-gradient minimization steps were applied to reduce the clashes between protein and lipids. Then, a 10 ns long MD simulation was performed in the NPT ensemble, with a positional constraint of 1 kcal mol −1 Å −2 on protein and lipid phosphorus atoms. During the second stage, 30 ns of MD simulation in the NPT ensemble were performed, constraining all protein atoms but leaving the POPC residues free to diffuse in the bilayer. In the last equilibration stage, positional constraints were reduced by one half and applied only to the protein backbone alpha carbons for a further 10 ns of MD simulation.
For each intermolecular complex, a total of 2 μs of unbiased MD was performed, divided in one replica 1 μs long and two replicas 500 ns long. After equilibration, production MD trajectories were computed with an integration time step of 4 fs in the canonical ensemble (NVT) at 300 K, using a thermostat damping of 0.1 ps −1 and the M-SHAKE Alanine mutation of ECL2 and ECL3 of the hCTR alters CT peptide binding pK i and functional affinity (pK A ) in a peptide-and pathway-specific manner. The effect of mutation on peptide affinity in competition for the antagonist radioligand 125 I-sCT  is displayed as ΔpK i in the upper panels, with functional affinities derived from operation fitting of concentration-response curves in cAMP accumulation, pERK and IP1 accumulation displayed as ΔpK A in the mid and lower panels, respectively. Illustrated is a top view of the active, sCTbound, hCTR model with ECL2 and ECL3 shown in surface representation. Mutations that significantly alter peptide affinity are coloured according the magnitude of effect (from Tables 1  and 5), with mutated amino acids without significant alteration to affinity coloured grey. sCT is shown as dark red, with side chains in proximity to the ECLs displayed in x-stick, and residues 1-7 that are critical for receptor activation displayed in transparent cpk. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)   binding by CT peptides for wild-type (WT) and each of the hCTR mutants stably expressed in CV1-FlpIn cells. Whole cell radioligand binding was performed for each receptor mutant in presence of 125 I-sCT  and competing peptide ranging in concentration between 1 μM and 1 pM. Non-specific binding was determined in the presence of 1 μM of sCT  and was used to calculate% of specific binding. Data were fit with a three-parameter logistic equation. All values are mean + S.E.M. of 3-12 (WT [16][17][18] independent experiments, conducted in duplicate.  algorithm [38] to constrain the bond lengths involving hydrogen atoms. The cut off distance for electrostatic interactions was set at 9 Å, with a switching function applied beyond 7.5 Å. Long range Coulomb interactions were handled using the particle mesh Ewald summation method (PME) [39] by setting the mesh spacing to 1.0 Å

Analysis
Contacts and hydrogen bonds were quantified using VMD [40]. A contact between two residues was considered productive if at least two atoms were detected at distances less than 3.5 Å. A distance between acceptor and donor atoms of 3 Å and an angle value of 20°were set as the geometrical cut-off for hydrogen bonds. Equilibrated coordinates and parameters are available from the following doi: https://doi.org// 10.5526/ERDR-00000075.
Data on the effect of ECL2 and ECL3 mutation on GLP-1R-mediated efficacy were mapped onto the high resolution 3.3A structure of exendin-P5 in complex with the hGLP-1R and dominant negative Gs heterotrimer (PDB = 6B3J [15]). Mapping and visualization of the effect of mutation on receptor structure was performed using ICM (Molsoft).

Results
To assess the importance of ECL2 and ECL3 in CTR function, we performed alanine-scanning mutagenesis of residues within these loops as well as adjacent TMs, and assessed effects on cell surface receptor expression, competitive binding affinity and cAMP accumulation, IP1 accumulation and ERK1/2 phosphorylation (pERK). These pathways are important for physiological, CTR-mediated, signalling [2,7]. Responses were evaluated for representatives of the major structural/ evolutionary clades of CT peptides, specifically, human CT (hCT) and salmon CT (sCT) that are both used clinically, as well as porcine CT (pCT) (Fig. 1A). We also assessed responses to the related peptides Amy and CGRP that bind to and activate the CTR with low affinity/potency, but are potent agonists of CTR/RAMP complexed receptors. Global affinity (pK i ) was determined from competition with the radiolabelled antagonist peptide 125 I-sCT , while functional affinity (pK A ) and efficacy were determined by quantification of concentration-response data with the operational model of Black and Leff [23] that provides independent measures of pK A and efficacy (τ). All experiments were performed in CV-1 FlpIn cells that lack functional CTR and RAMP expression, with receptors stably expressed following isogenic recombination. Data are mapped onto the active hCTR structure (5UZ7), following modelling of missing side chains and sampling by short timescale MD.

Receptor expression
Residues for mutation were selected based on the recent structure of sCT/hCTR/Gs [13] and comprised amino acids I279-I300 (ECL2) and F356-M376 (ECL3) (Fig. 1B,C). Cell surface expression was determined by anti-c-Myc antibody binding to the c-Myc tag inserted after the receptor signal peptide in the N-terminal ECD (Fig. 1B), and quantified by FACS.
Most mutants demonstrated equivalent cell surface expression to that of the WT receptor ( Fig. 1D; Table 1), however, a subset of mutants, particularly within ECL2, had altered expression. Only K370A in ECL3 had no detectable expression. D287A, C289A, W290A, and I300A in ECL2, and I371 in ECL3, also had markedly diminished cell surface expression (< 20% of WT). R281A, N286A, T295A, L297A, Y299A in ECL2 and F356A in ECL3 were expressed at levels between 20% and ∼40% of WT. A smaller but significant attenuation of expression was observed for Y284A, R362A and Y374A. In contrast, T280A significantly increased expression and P360A strongly augmented cell surface receptor expression ( Fig. 1D; Table 1). Within ECL2, those residues with strongly attenuated expression formed a continuum within the central portion of the loop, suggesting that these residues participate in a network that helps to stabilize the apo receptor (Fig. 1E).

ECL2 and ECL3 mutants differentially modulate peptide-specific affinity
Global affinity was determined by competition of 125 I-sCT  binding in whole cells by each of the peptides. Homologous competition with sCT(8-32) revealed a pK i for the WT receptor of 9.70 ± 0.05, and a Bmax of 22,900 ± 2500 sites/cell (Table 1). Overall, there was a good correlation, for the mutant receptors relative to WT, between For each receptor and ligand, data were fit to a three-parameter logistic equation to derive pEC 50 (negative logarithm of the concentration of ligand that produces half the maximal response) and Emax (maximal response, as % of WT). All values are mean ± S.E.M. (independent "n" values are indicated within parentheses). For each ligand, significance of changes in pEC50 and Emax were determined by comparison of mutants to the WT receptor in a one-way analysis of variance followed by Dunnett's post-test (p < 0.05 represented by *). (N.D.) pEC 50 or Emax not determined.
For each receptor and ligand, data were fit to a three-parameter logistic equation to derive pEC 50  measured Bmax and cell surface expression data from antibody binding (Fig. 1F), although there was a high error in Bmax estimates, consistent with expectations of cold saturation experiments with only two radioligand concentrations. sCT, hCT and pCT had pKi values for the WT receptor of 9.87 ± 0.03, 6.72 ± 0.06 and 8.27 ± 0.07, respectively (Table 1), consistent with those reported in previous studies [9,10]. Mutants with very low expression, N286A, D287A, C289A, W290A, I300A, K370A and I371A, displayed little or no detectable 125 I-sCT(8-32) binding, thereby preventing assessment of global affinity, whereas all other mutant receptors exhibited a sufficient radioligand binding window to determine peptide affinity. Of these, only R281A, L291A and P360A reduced sCT(8-32) affinity, whereas T295A increased affinity of this peptide (Table 1, Fig. 2; Fig. 3). Similarly, there was only limited effect on sCT affinity, with reduced affinity for the R281A, P360A and D373A mutants, and increased affinity for S292A. Intriguingly, the L291A mutant that had reduced affinity for the antagonist and both hCT and pCT, did not alter sCT affinity. Both hCT and pCT were more broadly sensitive to mutation, with those affecting pCT common with hCT, including R281A, L291A, T295A, L298A, F359A-P363A and Y372A that principally resided either within 5 Å of sCT in the CTR model, or were involved in the network of amino acids in the core of ECL2 that was linked to receptor stability/expression (Table 1, Figs. 1  and 2). There was a significant reduction in pCT affinity for the Y299A mutant that also trended lower for hCT. Moreover, there was selective, significant attenuation of hCT affinity for L297A, L298A, V357A, V358A, D373A and M376A (Table 1, Fig. 2).
CGRP and Amy displayed no detectable competition with the radioligand at the wildtype or mutant receptors within the concentration range assessed (up to 10 μM), confirming previous findings that these have low affinity for the CTR.

ECL2 and ECL3 mutants alter functional affinity in a ligand and pathway dependent manner
Concentration-response data, for cAMP accumulation, in response to each peptide was established for WT and mutant receptors (Fig. 4, Table 2), IP1 accumulation ( Fig. 5; Table 3), and ERK1/2 phosphorylation ( Fig. 6; Table 4). Functional affinity for each of the pathways was determined by operational fitting of the concentration response data. The effect of mutation on pK A for cAMP formation was broadly similar to the derived pK i values (Figs. 6 and 7), clustering to either residues in proximity to the peptide or in the central segment of ECL2 that was important for receptor stability and expression (Fig. 2, Fig. 8). Unlike the competition binding assay, estimates of pK A could be derived for at least one of the peptides for all mutants, except K370A that was not expressed at the cell surface (Fig. 2, Fig. 8, Table 5). Of these,  Fig. 7. Correlation between changes in global peptide affinity (pK i ) derived from heterologous competition binding assays, and functional pK A for cAMP (upper panels), IP1 (middle panels) or pERK (bottom panels) signalling, for hCT (left hand panels), sCT (middle panels) or pCT (right hand panels). For all peptides, the highest correlation was seen between pK i and pK A derived from operational analysis of cAMP response data. Significant but weaker correlations were also observed between pK i and functional affinities for IP1 and pERK signalling for hCT, and for pERK signalling alone for pCT. No correlation was observed for sCT pK A values from IP1 or pERK assays or pCT IP1 assays.
N286A had no impact on affinity, while D287A, C289A and W290A caused a marked decrease in pK A for cAMP formation for all peptides, and these were the only mutations to alter sCT cAMP pK A values. The weak IP1 responses for these mutants made interpretation of effect difficult, however they had clear differential impact on pERK pK A values. While none of the mutants altered sCT pERK pK A , D287A decreased both hCT and pCT functional affinities, and W290A decreased that for pCT (hCT was not detectable), and there was a selective loss of pERK functional affinity for hCT at the C289A mutant (Fig. 2, Fig. 8, Table 5). There were distinct patterns in the effect of mutation on pK A values both across pathways and between peptides. sCT functional affinities were the least sensitive to mutation, and in particular, mutations in ECL3 had limited impact. This contrasts with hCT and pCT where ECL3 mutations had widespread effect with greatest impact on pK A values derived from IP1 and cAMP signalling (Figs. 2, 8 and 9, Table 5).
The F359A and P360A mutants induced a global decrease in functional affinity for pERK and IP1 pathways, but had differential effect on cAMP responses. Neither affected sCT, while P360A attenuated pK A values for both hCT and pCT, with a selective loss of affinity for hCT observed for the F359A mutant (Figs. 2, 8 and 9, Table 5).
Due to the low affinity of amylin and CGRP for CTR, and poor coupling to IP1 signalling (data not shown), no competitive binding data or IP1-derived functional affinity data could be obtained for these peptides. pK A values from operational analysis of the cAMP signalling could be derived for most mutants, revealing only limited impact on either amylin or CGRP functional affinity. This was particularly true for amylin that mirrored the observations for sCT, with ∼10-fold loss of affinity for the C289A and W290A mutants. No quantifiable response for the D287A and Y372A mutants (that had weak reductions in sCT functional affinity) was observed, while all other mutants failed to significantly alter affinity ( Fig. 8; Table 5, Fig. 6). Although the magnitude of effect was also limited for CGRP in this pathway, there were broader effects of mutations, including loss of detectable response for D287A, C289A, and W290A, and reduced affinity for L291A, T295A, F356A, W361A and M376A. These residues were also important for hCT functional affinity at this pathway (Figs. 3 and 10; Table 5).
Both amylin and CGRP are only weakly coupled to pERK, and alanine mutation either had no effect or attenuated responses such that they there were not quantifiable. Nonetheless, interesting differences were observed between the effect on amylin and CGRP signalling. Residues with very low expression, including D287A, C289A, W290A and I371A had no detectable response for either peptide. Loss of response for both peptides was also seen for the P360A, Y372A and D373A mutants, despite similar or greater (P360A) cell surface expression of the receptor. The L291A, and Y299A mutants selectively attenuated, and the N286A mutant selectively enhanced, CGRP functional affinity. Functional CGRP affinity for T295A was not significantly altered, and not determined for amylin due to large error in parameter estimates, despite a measurable response (Figs. 8 and 10; Table 5).

Calcitonin peptides
Interestingly, despite detrimental effects on functional affinity, only enhancement of cAMP efficacy was observed for any of the CT peptides (Figs. 11A and 12A; Table 6). For sCT and hCT, the effect of mutation was similar, and confined to ECL2, with the exception of F356A that resides at the base of the peptide binding pocket. Enhanced efficacy was seen for both sCT and hCT for R281A, D287A, C289A, W290A, T295A and I300A (Fig. 11A and 12A; Table 6). These mutants also had reduced cell surface expression (Fig. 1), indicative of destabilization of the receptor in a manner that lowers the barrier to Gs coupling. N286A caused a selective enhancement of hCT efficacy, while pCT had both overlapping and distinct effects following receptor mutation. In ECL2, the effect of mutation was conserved with the other peptides except that Fig. 10. Alanine mutation of ECL2 and ECL3 of the hCTR alters amylin and CGRP functional affinity (pK A ) in a peptide-and pathway-specific manner. Functional affinities derived from operation fitting of concentration-response curves in cAMP accumulation (upper panel) and pERK (lower panel) are displayed as ΔpK A . Illustrated is a top view of the active, sCT-bound, hCTR model with ECL2 and ECL3 shown in surface representation. Mutations that significantly alter peptide affinity are coloured according the magnitude of effect (from Table 5), with mutated amino acids without significant alteration to affinity coloured grey. sCT is shown as dark red, with side chains in proximity to the ECLs displayed in x-stick, and residues 1-7 that are critical for receptor activation displayed in transparent cpk. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) there was no change in pCT efficacy with the D287A mutant, and a selective enhancement of efficacy at the L297A mutant. Strikingly, ECL3 residues were also important for pCT cAMP efficacy, with P360A, R362A, P363A and D373A causing selective enhancement of efficacy ( Fig. 11A and 12A; Table 6).
There was only limited impact of mutation on CT-mediated IP1 efficacy (Fig. 11B, Fig. 12B, Table 6). Within ECL2, no quantifiable response was seen with W290A for any of the peptides, with C289A for hCT and pCT (but no effect on sCT), and with D287A for sCT and hCT, but an enhanced efficacy for pCT with this mutation. The Y299A mutant attenuated efficacy for all peptides, while I300A abolished responses to sCT and pCT but not hCT (Table 6). Within ECL3, there was no pCT response with the F356A mutant, but this did not alter efficacy for sCT or hCT. The I371A mutant significantly enhanced efficacy for all peptides. The P360A mutant selectively reduced hCT efficacy, while the Y372A and M376A mutants selectively attenuated efficacy for pCT. No other mutants altered peptide-mediated IP1 efficacy (Fig. 11B, Fig. 12B, Table 6).

Amylin and CGRP
Within ECL2 there was no quantifiable response for D287A for either amylin or CGRP, and no measurable response to CGRP at the C289A and W290A mutants, while the C289A mutant, along with N286A, had enhanced amylin efficacy (Fig. 11A, Fig. 12A, Table 6). Both amylin and CGRP efficacy were enhanced at the I300A mutant but there were no other significant effects for either peptide. Within ECL3, the F356A and I371A mutants enhanced efficacy for both peptides, while the Y372A mutant attenuated CGRP efficacy and abolished the response to amylin, but no other mutants impacted on amylin efficacy. For CGRP, loss of efficacy also occurred at the F359A, P360A, D373A and M376A mutants (Fig. 11A, Fig. 12A, Table 6).
Due to weak coupling of amylin and CGRP to IP1 signalling, effects of mutations on efficacy could not be determined.
As noted above for functional affinity data, as coupling of amylin and CGRP to pERK is relatively weak, many mutants had responses that could not be operationally quantified. Although some of these had selective effects on amylin or CGRP, in these cases the effects on affinity versus efficacy could not be separated.
Within ECL2, most mutants either had no quantifiable signalling or did not affect peptide efficacy. Lack of signalling occurred for both peptides at the D287A, C289A and W290A mutants. Signalling was not quantifiable for L291A and Y299A for CGRP, and attenuated for amylin, while amylin signalling was not quantifiable at the T295A mutant, with no effect on CGRP (Fig. 11C, Fig. 12C, Table 6). Within ECL3, no quantifiable signalling for either peptide was observed for the P360A, D373A and I371A mutants. At the Y372A mutant, efficacy was abolished for amylin and attenuated for CGRP. Efficacy of both peptides was reduced at the W361A and M376A mutants. It was reduced for amylin and abolished for CGRP at the P363A mutant. In general, mutations to the membrane proximal segment of TM6/ECL3 had greater impact on CGRP efficacy, with either loss (V357A) or attenuation (V358A, F359A, R362A) of efficacy for CGRP with less pronounced effects amylin efficacy (Fig. 11C, Fig. 12C, Table 6).

Discussion
Recent structural biology breakthroughs for the CTR and GLP-1R have provided new understanding of class B GPCR peptide binding and receptor activation that includes reorganisation of the packing of loop residues, and major, conserved, conformational changes in TM6/ECL3/ TM7 at the extracellular face of the receptor that are linked to outward movement of the intracellular face of TM6 to accommodate G protein binding [13][14][15]. However, these studies also revealed peptide/receptor specific differences in presentation of the peptide N-terminus to the core and their engagement with the receptor surface, in particular for ECLs 2 and 3. For the GLP-1R, these loops play an important role in peptide binding, efficacy and biased agonism [15,16,21]. Intriguingly, the role of ECLs 2 and 3 of the CTR was generally distinct when compared to the GLP-1R. This could in part be attributed to distinct effector coupling profiles exhibited by the two different receptors with the GLP-1R capable of coupling to both G proteins and β-arrestins [16], whereas CTR is unable to recruit the latter when activated by CT peptides [41].  13. Comparison of the effect of alaninescanning mutagenesis of ECL2 and ECL3 on cAMP (A) and pERK (B) efficacy of sCT at the CTR or exendin-4 at the GLP-1R. Top views of active state structures of sCT/hCTR/Gs or Ex-P5/hGLP-1R/ Gs, with the receptor ECD and peptide C-terminus omitted for clarity. Mutations that significantly alter peptide efficacy are coloured according the magnitude of effect, with mutated amino acids without significant alteration to efficacy coloured grey. Efficacy data for exendin-4 (Ex4) are from , and are mapped onto the structure of exendin-P5/hGLP-1R/Gs (PDB: 6B3J). sCT is shown as dark red, with side chains in proximity to the ECLs displayed in x-stick, and residues 1-7 that are critical for receptor activation displayed in transparent cpk. Exendin-P5 is shown as dark blue, with side chains in proximity to the ECLs displayed in x-stick, and residues 1-8 that are important for receptor activation displayed in transparent cpk. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) 4.1. CTR ECL2 plays a key role in conformational propagation linked to Gs/cAMP signalling that is distinct from that of GLP-1R ECL2 CTR stability, as indexed by cell surface expression, was highly sensitive to mutations in the core of ECL2 that formed an interconnected network but were located away from the principal binding site for sCT in the active structure. Moreover, mutation of this ECL2 network enhanced efficacy selectively for cAMP (all peptides) suggesting that the ECL2 destabilized state is linked to lowered barrier for Gs activation, despite decreased affinity of some mutations for peptides (lower pK A ). Indeed, some mutants demonstrated higher Emax than WT receptor, despite low cell surface expression. However, there was limited correlation of the loss of cell surface expression with efficacy in other pathways, indicating that this ECL2 conformation is poorly linked to activation of other pathways for this receptor.
This segment of ECL2 contains a number of residues that are very highly conserved across the CTR and GLP-1R (and indeed all class B GPCRs), including R281/K288 (CTR/GLP-1R amino acids and residue number, respectively), C289/C296, W290/W297 and an polar/acidic motif between these residues N286,D287,N288 (CTR) and E292,D293,E294 (GLP-1R) [13][14][15]. Despite this, these residues are differentially important in receptor activation between those receptors. With minor exception, GLP-1R expression/stability was not markedly affected by mutation for any of the ECLs, including ECL2, however, ECL2 was broadly required for both Gs-(cAMP) and Gq-(iCa2+) mediated signalling with mutation decreasing peptide efficacy [16,21]. This contrasts with both the enhancement of cAMP efficacy for CT peptides, and the very limited importance of ECL2 in CT efficacy for IP1 signalling (Figs. 11, 12 and 13A). Comparison of the Gs complexed structures of the two receptors provides some potential insight into why these differences may occur, in particular, there are marked differences in positioning of W290/297 and the packing interactions of conserved residues around this residue. In the GLP-1R, W297 is completely flipped and buried within the core of the loop and this conformation is stabilized by K288, with the acidic/polar residues forming additional interactions that stabilize this conformation (Fig. 14). In contrast, the aromatic functional group of W290 in the CTR remains oriented towards the receptor core with extensive interactions observed with sCT and hCT that are stable in MD simulations (Fig. 15A, Table 7); D287 packs tightly with W290 and C289, while R281 forms alternate interactions to stabilize the loop conformation (Fig. 14B). The two receptors have very distinct preferred orientations of the N-terminal ECD and the peptide ligands enter the receptor core at different angles, with GLP-1/ ExP5 closer to ECL2 such that their entry may require the major reorientation of W297 [15]. In this vein, it is interesting to note that while alanine mutations of C296 and W297 dramatically diminished GLP-1 and exendin-4 binding, they did not alter oxyntomodulin affinity [21], and it is possible that this peptide engages the receptor core in a manner more similar to that observed for sCT. For ECL2, CGRP and amylin were generally less affected by mutation and efficacy effects were either unmeasurable or only found in a subset of those with altered efficacy for CT peptides.
The CTR shares greatest homology with the calcitonin receptor-like receptor (CLR), including strong conservation of residues within ECL2. Unlike CTR, CLR requires RAMP interaction for functional cell surface expression and to form CGRP (CGRP 1 , CLR/RAMP1) or adrenomedullin (AM 1 , CLR/RAMP2; AM 2 , CLR/RAMP3) receptors. In contrast to the CTR, CLR cell surface expression was not greatly impacted by alanine mutation of ECL2 residues [42,43]. However, there were similarities in the impact of mutation of conserved residues on cAMP pK A (CTR) or cAMP potency (CLR/RAMP receptors). This included reductions in CGRP and adrenomedullin potency, for their respective receptors, for R274A (R281A, in CTR), D280A (D287A), C282A (C289A), W283A (W297A), and I284A (L291A) [42] that paralleled the losses in functional affinity seen with hCT and pCT, although this could be RAMPdependent for the adrenomedullin receptors [43].
Within ECL2, amino acids proximal to the peptide in the sCT/CTR/ Gs structure (W290-T295; L297,L298) tended to display peptide and/or pathway selective effects. These likely form dynamic and differential interactions with key polar residues of the peptides (S 2,sCT/pCT /G 2,hCT , N 3 , T 6 , Q 14,sCT/hCT ,R 14,pCT , H 17,sCT /N 17,hCT/pCT ) to influence peptide binding and signalling ( Fig. 15A; Table 7; Supplementary Movie 1 Movie 1. Interactions between hCT (left hand panel) or sCT (right hand panel) with the CTR during 1 μs MD simulation. Interactions between hCT and the CTR are less stable than those of sCT and the receptor. As a consequence, ECL2 in the hCT-bound CTR undergoes more dynamic conformational sampling. ). Of note, T288A, L290A, and residues in red x-stick and the highly conserved C296, W297 and K288 residues displayed in transparent cpk representation. Shown in orange are GLP-1R residues in the conserved polar network E292, D293, and E294. (B) CTR displaying loop residues in green x-stick and the highly conserved C289, W290 and R281 residues displayed in transparent cpk representation. Shown in light green are the CTR residues in the equivalent conserved polar network N286, D287, and N288. Ribbon representations of the proximal amino acids of exendin-P5 (blue) or sCT (dark red) are also displayed. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) L291A of CLR, in a RAMP-dependent manner, also attenuated adrenomedullin (T288A) or CGRP (T288A,L290A,L291A) cAMP potency [42,43]. Intriguingly, comparison of 1 μs MD simulations of hCT and sCT bound to the CTR indicated that ECL2 was more conformationally dynamic when the receptor was bound to hCT (Supplementary Movie 1), and this may also contribute to the differential effects of mutation between CT peptides.

ECL3 is a gateway for ligand, receptor and pathway specific modulation of class B GPCR function
Across the available active CTR and GLP-1R structures, the largest difference in the receptor core was the angle of tilt of TM6, and to a lesser extent TM7, and the interconnecting conformation of ECL3, with the CTR exhibiting the greatest outward movement of this domain [13][14][15]. Nonetheless, this was also the region of greatest divergence between the structures of GLP-1/GLP-1R/Gs and the G protein-biased analogue complex, ExP5/GLP-1R/Gs, indicating that peptide interactions with ECL3 play a critical role in differential modes of receptor activation [15]. Indeed, comparison of the effect of mutation in ECL3 across multiple peptides and pathways, and between the CTR and GLP-1Rs, revealed significant diversity in how ECL3 was engaged and contributed to peptide binding and propagation of conformational change linked to efficacy (Fig. 13). Interestingly, mutations that altered GLP-1R mediated Ca 2+ (Gq) and pERK primarily clustered within ECL2 and ECL3 respectively, whereas CTR mutations that altered IP1 (Gq) and pERK displayed similar clustering, predominantly within ECL3 that distinct from those required for cAMP. pERK can be activated downstream of multiple effector proteins and is often a composite of many divergent signalling pathways [16,44]. While pERK1/2 mediated by the GLP-1R is a composite of both G protein and β-arrestin signalling [16], the CTR is unable to recruit β-arrestins, suggesting that the pERK response is likely to be predominantly G protein mediated [41]. While the CTR can couple to multiple different G proteins, similar clustering in our of residues important for IP1 and pERK in our mutational analysis suggests that CTR mediated pERK, at least in part, may be downstream of Gq coupling, although further experimental data will be required to confirm this.
Despite the diversity in how CTR and GLP-1Rs engage ECL3 to promote signalling, there were clear patterns with respect to clustering of residues that were functionally important, particularly around the TM6 and TM7 proximal segments of ECL3 that were located with 5 Å of the sCT ligand, and the network of residues within the loop that stabilized these interactions. For hCT and pCT, the sequence of residues at the apex of the TM extension (K366-G369) was important, in a peptide specific manner, for IP1 or pERK signalling, indicating that secondary structure in this segment of ECL3 contributes to propagation of conformation linked to these pathways for the less well coupled peptides. Of particular note were the clear distinctions in the patterns of important residues for pCT versus other CT peptides, and CGRP across all peptides suggesting that these peptides have different modes of ligand engagement with the receptor relative to the other peptides.
Uniquely among class B GPCR peptide ligands, the CT-family peptides contain an N-terminal disulphide bridge between residues 1 and 7 (2 and 7 for CGRP and adrenomedullin), with a consequent bulky loop structure that contrasts to the linearly extended GLP-1R peptides observed in the active structures. This loop is oriented toward ECL3 and the larger outward movement of this domain is required to accommodate the peptide N-terminus. However, the peptides are predicted to make relatively weak (non-polar) and transient interactions with ECL3 ( Fig. 15A; Table 7), and this is consistent with the high mobility of TM6/ECL3 in the sCT/CTR/Gs structure that could not be resolved at high resolution [13]. Linear analogues of sCT maintain high affinity and potency in cAMP signalling, whereas equivalent analogues of hCT have attenuated potency [9,45,46]. sCT has higher helical secondary structure propensity in the mid-region of the peptide, compared to hCT The CT residues least engaged by the receptor (0% contact) are coloured blue, while residues most engaged by the receptor (100% contact) are coloured red. (A) hCT (magenta), (B) sCT (magenta). (C) Upper panel; secondary structure of hCT during the 1 μs MD simulation of hCT bound to CTR, as determined using VMD. Lower panel; secondary structure of sCT during the 1 μs MD simulation of sCT bound to CTR, as determined using VMD. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) [9,47] that likely constrains the location of the N-terminus to maintain interactions, whereas the additional constraints imposed by the disulphide bridge are required to facilitate interactions for hCT. Greater secondary structure of sCT versus hCT is seen in MD simulations of bound peptides (Fig. 15C) and this contributes to predicted differences in peptide-receptor interactions for these two peptides ( Fig. 15B; Table 7, Supplementary Movie 1).
Intriguingly, comparison of the ExP5, and the GLP-1, receptor complexes, revealed distinct positioning of peptides (including minor difference in the relative orientation of the ECD) that likely contributes to engagement with ECL3/TM6 and TM7, and this had implications for ligand-dependent G protein conformations and cAMP signalling efficacy [14,15]. Amongst the cryo-EM structures, the GLP-1R complexes exhibited a single major conformation of the ECD relative to the receptor core, while the sCT/CTR/Gs complex contained multiple conformations of the ECD that were discernible at lower resolution [13]. CT family peptides have a relatively unstructured, and more flexible Cterminus than GLP-1 and related peptides [13,48], and it is likely that there would be greater potential for different CT peptides to have altered orientation within the receptor core, relative to sCT, as would be predicted from MD simulations (Fig. 15C). This would be consistent with the differential impact of ECL3 mutation on sCT versus hCT and pCT, and between hCT and pCT (eg. for pK A in IP1 and pERK, and efficacy for cAMP and pERK).
Though it is difficult to draw direct comparisons, it is intriguing that select mutation of residues in ECL3 differentially affected potency/ functional affinity of GLP-1 relative to ExP5 [15], and that this was also seen for sCT versus hCT. Like hCT and sCT that have altered Gs-mediated efficacy linked to different ligand-induced conformations of Gs [11], higher efficacy was observed for ExP5 (relative to affinity)