Human CGRP and CRF were purchased from Bachem and human glucagon was purchased from Alta Biosciences. All peptides were made to 1 mM stocks in water containing 0.1% bovine serum albumin (BSA).

The CRFR1a and CRFR1b constructs were gifts from Dr. Simon Dowell (GSK, Stevenage, UK). ICL1 mutants of CLR were generated in a pIRES-RAMP1-SNAP-CLR construct in-house by Sosei Heptares and confirmed through Sanger sequencing. ICL1 mutants of the GCGR were generated as for CLR except using a base vector of pcDNA3.1-GCGR-GFP. H8 mutants for CLR and GCGR were generated by using the QuikChange Lightening Site-Directed Mutagenesis Kit (Agilent Technologies) in accordance with the manufacturer’s instructions and sequenced in house at Cambridge University. All constructs were of human receptors and RAMP.

HEK 293T cells (a gift from Professor Colin Taylor) were cultured in DMEM/F12 GlutaMAX supplemented with 10% fetal bovine serum (FBS), and incubated at 37% in humidified 95% air, 5% CO₂. Plasmids were transfected into HEK 293T cells using FuGENE HD according to the manufacturer’s instructions using a 1:3 w:v ratio of DNA : FuGENE and cultured for 48 hours prior to assaying.

Cell surface expression of WT and mutant GPCRs was determined in HEK 293 cells via flow cytometry as previously described (36). Briefly, after 48 hours, cells were washed three times in fluorescence activated cell sorting (FACS) buffer (PBS supplemented with 1% bovine serum albumin (BSA) and 0.03% sodium azide) before and after 1 hour incubation at room temperature in the dark with appropriate primary antibody; rabbit anti-GCGR (AGR-024, diluted 1:50 (Alomone Labs)) or rabbit anti-SNAP (CAB4255, diluted 1:100, Invitrogen). Cells were then incubated for a further hour at room temperature in the dark in secondary antibody (goat APC-conjugated anti-rabbit IgG polyclonal antibody, diluted 1:150 (ThermoFisher)). Samples were analysed using a BD Accuri C6 flow cytometer (BD Biosciences) Ex. λ 633 nm and Em. λ 660 nm. Mean APC intensity indicated the plasma membrane expression of each GPCR. Data were normalized to the mean APC intensity of cells transfected with WT GPCR as 100% and pcDNA3.1(-) as 0%.

HEK 293 cells expressing WT or mutant GPCR were assayed for cAMP accumulation as previously described (33, 37). cAMP accumulation was measured after 30 minutes stimulation using LANCE^® cAMP Detection Kit (Perkin Elmer Life Sciences) on a Mithras LB 940 multimode microplate reader (Berthold Technologies). Data were normalized to the maximal level of cAMP accumulation from cells in response to 100 μM Forskolin (Sigma) stimulation.

Intracellular calcium mobilisation was measured in transfected HEK 293 cells as previously described using Fluo-4/AM and a Mithras LB 940 multimode microplate reader (33). Data were normalized to the maximal intracellular calcium release in response to 10 μM ionomycin.

HEK 293 cells expressing WT or mutant GPCR were serum starved overnight prior to assaying. Cells were then washed and resuspended in Ca²⁺ free HBSS and seeded at a density of 35000 cells per well in 384-well white Optiplates (Perkin Elmer). Cells were stimulated with ligand for 5 min, before lysis using the supplied lysis buffer and assayed for ERK1/2 phosphorylation using the phospho-ERK (Thr202/Tyr204) Cellular Assay Kit (Cisbio). Plates were read using a Mithras LB 940 multimode microplate reader and data normalized to the maximal response to 100 μM phorbol 12-myristate 13-acetate (PMA, Sigma) (34).

PDB structures for the inactive and active conformations of CLR complexed with RAMP1 (PDB codes 7KNT and 6E3Y, respectively) and the GCGR (PDB codes 5EE7 and 6WHC, respectively) were obtained from the PDB. Missing atoms were identified from the PDB header and are available in the original PDB file. They were replaced using MODELLER (38) prior to refinement and scoring using Rosetta (39). To reduce errors introduced by loop modelling, ICL2 was excluded from the MODELLER step. Furthermore, no attempt was made to refine existing regions of secondary structure as are as per the original crystallographic file. The best scoring inactive and active conformations of each receptor type were used for an essential dynamics simulation (40). Each protein was embedded in an equilibrated solvated membrane consisting of 280 POPC lipids. NaCl was added at a concentration of 150 mM, with extra Cl^- ions added to the solvent to neutralize the system. Protonation states of charged residues were determined using ProPka (41) prior to the simulation start. Initial equilibration simulations (100 ns) were performed using Gromacs (42) at 310 K for both the active and inactive receptor states for each receptor family. The equilibrated structures were then used for an essential dynamics simulation. During the essential dynamics simulations, each inactive conformation had a fixed potential applied to the first eight non rotational/translations eigenvectors that increased in fixed increments per step to drive the system from the inactive to the active state. Simulations were performed with fixed increments of 1.2 × 10⁻⁶ nm per each simulation step (2 fs). Each simulation was performed 10 times. The amber-ILDN forcefield was used in all simulations. The simulations were combined into an inactive to active trajectory using the best scoring snapshot, using the Rosetta scoring function at each timestep. The snapshots were not chosen, but are equally spaced conformations that describe the trajectory from inactive to active state. The RMSD between snapshots was approximately 0.01 Angstroms.

GPCR sequences were obtained for 289 human GPCRs in Class A and Class B1 from GPCRdb (https://gpcrdb.org) (43). Sequences were aligned for their ICL1 and H8 using GPCRDB outputs. Weblogo’s were then created using free software from] https://weblogo.berkeley.edu using the frequency method where height of each letter represents the frequency of each amino acid.

All pharmacological data was analyzed in GraphPad Prism v9.0 (GraphPad Software, San Diego). Data were fitted to obtain concentration–response curves using either the three-parameter logistic equation to obtain values of E_max and pEC₅₀ or the operational model of agonism (44) as described previously (33, 34), to obtain efficacy (τ) and equilibrium disassociation constant (K_A) values. Statistical differences were analysed using one-way ANOVA followed by Dunnett’s post-hoc (for comparisons amongst more than two groups). Data was normalized to either activity of Wild Type receptor or relative to a system parameter control; 100 μM forskolin stimulation for cAMP accumulation, 10 μM ionomycin for _iCa²⁺ mobilizations assays or 10 μM phorbol 12-myristate 13-acetate (PMA) for ERK1/2 activation. The means of individual experiments were combined to generate the concentration-response curves displayed in the figures. Having obtain values for τ and K_A these were then used to quantify signaling bias as the change in Log(τ/K_A) relative to WT and a reference signaling pathway (ERK1/2) as described previously (34).

There is a high level of conservation in both ICL1 and H8 in all GPCRs ( Figures 1A, B ) and especially Class B1 GPCRs ( Figures 1C, D ). In Class B1 GPCRs only, ICL1 contains an R^12.48KLRCxR ^2.46b motif that extends into TM2 (using the Wootten Class B1 numbering, 46); the C and R in bold are absolutely conserved. In H8 there is an N^7.61bGE ^8.49b-V XxxX(R/K)xW motif where X is hydrophilic, x is any amino acid and the E and V in bold also absolutely conserved ( Figure 1 and Figure S1 ). The sequences of ICL1 and H8 motifs are more precise than our initial reports (32); here we only consider human class B1 GPCRs, rather than the full class B sequences (adhesion + secretin families).

To understand the function of ICL1 and H8, we first performed an alanine scan centred on the R^12.48KLRCxR and E^8.49VXxxX motifs in CLR. The ICL1-H8 motifs (underlined) are represented in the sequences Y¹⁶⁵FK^167/12.48bSLSCQR ^173/2.46b and N^388/7.61bGE^390/8.49bVQAIL ^395/8.54b. Almost all mutants were expressed at the cell surface at a third or greater than that of WT ( Tables S1, S2 – except I394A), a level which we have previously established that causes little change in agonist potency (47). It should however be noted that we had not checked for effects on efficacy/E_max. L169A expression was reduced by 72% and there were also sizeable decreases in expression for E390A, I394A, I397Aand L398A and R399A; the expression of L395A and K167A were reduced by 50%.

When cAMP production was measured, mutation of F166, L169, C171 and Q172 ( Table S3A and Figure 2 ) all reduced CGRP potency or E_max while S170A slightly increased it. For H8 and the base of TM7 there were significant reductions in pEC₅₀ for N388A, E390A (also decreasing E_max), V391A and W399A ( Table S3A ).

To provide a comparison of how the ICL1 mutants might influence downstream signaling activity, we renormalised the cAMP accumulations data shown in Figure 2 to now account for the maximal level of cAMP accumulation achievable for HEK 293T cells (stimulations with the non-selective adenylyl cyclase activator forskolin), ( Figure 3A and Table S3B ). Beyond cAMP accumulation, CLR can also couple to Gα_q/11 to increase intracellular calcium (33, 48). When this response is measured, the ICL1 mutants show a somewhat different pattern of activity; whilst L169A and C171A again reduce potency, with S168A there was a decrease in apparent affinity calculated from the operational model ( Figure 3B and Table S4 ). No mutants enhanced coupling. Alanine mutation of the ICL1 appeared to show little overall effect upon CGRP-induced ERK1/2 activation [which was largely independent of Protein Kinase A activation, (34)], with only K167A showing any significant changes; an increase in potency but decrease in E_max ( Figure 3C and Table S5 ).

While the changes in potency for cAMP and _iCa²⁺ mobilisation appear to suggest changes in signaling bias profiles, to formally confirm this, and to remove potential confounding issues of system bias, we refitted the data in Figure 3 using the operational model of receptor agonism (44). Since the ERK1/2 activation seemed to tolerate mutation at any positions in ICL1, all data was normalised to this parameter ( Figure 3D and Tables S3B, S4, S5 ). This analysis reconfirmed that mutations of residues L169 to Q172 all resulted in a signaling response that was biased away from _iCa²⁺ mobilisation and towards cAMP. Moreover, mutations of C171 and Q172 also displayed bias away from cAMP accumulation to a lesser extent than for mobilisation of _iCa²⁺.

We next performed the same array of assays as described for the ICL1 mutants on the CLR H8 mutants. Normalisation of the cAMP accumulation data showed that N388A, E390A, V391A, L395A and W399A all displayed reduced potency ( Figure 4A and Table S3B ). For _iCa²⁺ mobilisation, only E390A showed a dramatic reduction in potency ( Figure 4B and Table S4 ). Unlike the ICL1 mutants, mutations at every position within the H8 motif changed either potency (with increases for E390A, V391A, and Q392A) or E_max when ERK1/2 activation was quantified ( Figure 4C and Table S5 ). Analysis using operational parameters confirmed these observations ( Figure 4D and Tables S3B, S4, S5 ). I394A displayed reduced cell surface expression and failed to generate responses for _iCa²⁺ mobilisation or ERK1/2 activation so was excluded from any analysis although it did show a WT cAMP response, for reasons that are not obvious. Thus, the effects of mutation on ICL1 and H8 differ depending on which coupling pathway is examined and ERK1/2 activation is much more sensitive to mutations in H8 than cAMP or _iCa²⁺ elevation. These differences are highlighted in the bias plots for ICL1 and H8.

After preliminary experiments using saturation mutagenesis of ICL1 on cAMP production, the effect of glutamate, arginine, histidine, glycine and isoleucine were explored in detail, to examine the effect of charge and size at each position. This showed that substitutions at every position can alter receptor activation and/or receptor expression ( Figure 5 ). Several interesting features appear from this extended data. Substitution by R within the motif is well tolerated apart from at L169; S168R increases CGRP potency by almost 100-fold and there was a small increase in Emax with Q172R. Substitution by H was best tolerated at C171 and replacement by E gives particularly large decreases in activity at every position apart from F165. Replacement of L169 by any residue apart from I reduces expression by at least ²/₃ compared to WT. Glycine substitution, generally deleterious, increases potency at L169 (albeit with a reduced E_max) and C171.

In the GCGR, the sequence of ICL1 is G^165/1.61bL^12.47b SKLHCTR and the corresponding H8 sequence is NE^406/8.49bVQSEL (underlined residues are the motifs, as shown for CLR in section 3.1 and 3.2). This is a more typical H8 sequence as 8.53b is E in 12/15 human class B GPCRs. To probe the differences/conservation of regions of activity, we focussed upon mutations of the ICL1 region of the GCGR. C171A failed to express and so was without activity in all signaling assays; K168A, L169A and H170A showed a 50% reduction in expression. For the other mutants, cell-surface expression was not changed ( Figure 6A ). Alanine mutation of L169^12.50 and T172^12.52 reduced the extent of cAMP production; the other mutants that expressed were without effect on this pathway ( Figure 6B ). By contrast, both elevation of _iCa²⁺⁺ mobilisation and (in contrast to the CLR) ERK1/2 signaling were far more sensitive to alanine substitution throughout ICL1 ( Figures 6B–E and Table S6 ).

Immediately outside ICL1, R173^2.46bA reduced the potency of glucagon, almost 100-fold, on cAMP accumulation (pEC₅₀ WT, 9.09 ± 0.1; pEC₅₀ R173A, 7.14 ± 0.1, p>0.01, n= 5). A limited characterisation of H8 was carried out, focussing on the two charged residues E406^8.49b and E410^8.543b. The only effect on cAMP production was a very small increase in basal activity seen with E406A (WT 7.6 ± 1.1% of response to forskolin, E406A, 11.1 ± 1.5%, p>0.05, n=5). With the double mutant E406AE410A, the basal was further increased to 19.6 ± 1.4% (p>0.01), again with no change in potency. However, for E406A, expression was 53.8 ± 5.8% of WT and this was further reduced to 21.5 ± 1.3% with E406AE410A, suggesting an increase in the effectiveness of coupling to Gs given the increase in basal and the maintenance of potency.

Our data suggested that ICL1 might have a role in determining G protein preference of Class B1 GPCRs. The CRFR1 exists in two splice forms (1a and 1b) which differ due to the presence of a 29 amino acid insert in the RKLR motif which does not influence cell surface expression (49). Analysis of the signaling profiles suggests that the CRF1a receptor, which lacks the insert, is biased towards Ca²⁺ signaling, which is consistent with the motif being important for G protein specificity ( Figure 7 and Table S7 ).

A comparison between the inactive and active CGRP and GluR structures (the latter in complex with both Gs and Gi) indicated that ICL1 only underwent subtle changes during the process of receptor activation. To further understand why mutations throughout this region can change activation, a molecular dynamic simulation was used to study the inactive-active transition in the CGRP receptor ( Figure 8 and Videos S1A, B ). Early during the simulation, the proximal region of ICL1 underwent a flexing movement centred on S168^12.49. This subtly changes the orientation of ICL1 with respect to both ICL2 and H8. This was followed by movements of ICL2. Around the middle of the simulation, N388^7.61b and E390^8.49 in ICL4 move towards L169^12.50. There is then a bending of the distal part of TM1 but ICL1 with H8 and ICL2 move largely as a rigid body as a consequence of this. At the very end of the simulation there are further movements of N388 and E390, coupled eventually to a change in the side-chain orientation of R173. At this stage, the G protein binding pocket is now fully open. The distance between L169 in ICL1 and E390 in H8 reflects these changes ( Figure 8B ). Broadly similar movements were seen during a molecular dynamics simulation of the glucagon receptor, with both early and late movements in ICL1 ( Video S2 ).