Polar transmembrane interactions drive formation of ligand-specific and signal pathway-biased family B G protein-coupled receptor conformations

Abstract

Recently, the concept of ligand-directed signaling—the ability of different ligands of an individual receptor to promote distinct patterns of cellular response—has gained much traction in the field of drug discovery, with the potential to sculpt biological response to favor therapeutically beneficial signaling pathways over those leading to harmful effects. However, there is limited understanding of the mechanistic basis underlying biased signaling. The glucagon-like peptide-1 receptor is a major target for treatment of type-2 diabetes and is subject to ligand-directed signaling. Here, we demonstrate the importance of polar transmembrane residues conserved within family B G protein-coupled receptors, not only for protein folding and expression, but also in controlling activation transition, ligand-biased, and pathway-biased signaling. Distinct clusters of polar residues were important for receptor activation and signal preference, globally changing the profile of receptor response to distinct peptide ligands, including endogenous ligands glucagon-like peptide-1, oxyntomodulin, and the clinically used mimetic exendin-4.

G protein-coupled receptors (GPCRs) are the largest group of cell-surface proteins and mediate signal transduction across cell membranes by recognizing a wide range of extracellular stimuli (1). They signal through heterotrimeric G proteins, as well as various G protein-independent mechanisms (2). These receptors exist in a dynamic equilibrium between different conformational states and activation occurs through a number of intermediate conformations (3, 4). This equilibrium is not only controlled by the binding of specific receptor ligands and effector proteins but is also supplemented by clusters of residues within the receptor that act to stabilize subsets of receptor conformations (5).

Polar transmembrane (TM) residues are rarely found within the core of the membrane bilayer because their insertion in a hydrophobic environment is energetically unfavorable (6). Therefore, most polar residues in TM helices are buried within the interior of the protein, often lining internal water-filled cavities and forming hydrogen-bond interactions with buried water molecules and other polar residues (7, 8). Consequently, they play essential roles in the function of α-helical membrane proteins by mediating and stabilizing their helical interactions (9), in addition to playing key roles in transmission of signals across membranes through forming interactions with ligands and establishing interaction networks required for protein conformational changes (10).

A number of highly conserved polar residues are present in the family A subclass of GPCRs, and there is a wealth of information confirming their functional role. In these receptors, key conformational changes associated with activation occur through local changes in structural constraints that involve reorganization of hydrogen bonds between the polar residues and buried waters (11–14). Recent studies also revealed that distinct ligands interacting at the same receptor can stabilize different subsets of conformational states at the expense of others, which in turn can lead to the engagement of different intracellular effectors (15–17). It is these phenomena that can provide the mechanistic basis for biased agonism.

Family B GPCRs are an important class of physiological and therapeutic targets that are pleiotropically coupled, and there is evidence of ligand-directed stimulus bias for both natural and synthetic ligands of these receptors (18–20). The glucagon-like peptide-1 receptor (GLP-1R) is a prototypical member of this family that is activated by a range of endogenous and exogenous peptides that are used for the treatment of type-2 diabetes, and these ligands can elicit stimulus bias (18). These attributes make the GLP-1R an ideal candidate to study the molecular basis of receptor activation that may lead to both pathway-biased and ligand-dependent signaling occurring in family B GPCRs.

Although family B GPCRs do not share the conserved polar residues that are essential for family A GPCR function, they possess their own unique set of highly conserved intramembranous polar residues that have the potential to serve an role analogous to those in family A. Using the GLP-1R as a model, we have combined mutagenesis with molecular modeling to assess the role of these conserved polar residues (Fig. S1). This study demonstrates the importance of these residues for protein folding and expression, peptide binding, and in controlling activation transition, ligand-biased and pathway-biased signaling.

Results

Universal Numbering System for Residues in Family B GPCRs.

Residues were numbered using a system similar to the nomenclature used for family A GPCRs (21). The most conserved residue in each family B GPCR TM domain (Fig. S1) was assigned the locant of .50, and this number is preceded by the TM number. Each residue is numbered according to its relative position to the residue at .50 in each helix and its absolute residue number is shown in superscript.

Conserved Polar Residues May Form Functionally Important Hydrogen Bonding Networks.

To aid in understanding of mutational data, a model was generated of the GLP-1R TM bundle (Fig. S1). Predicted interactions formed by conserved polar side chains are listed in Table S1. Inspection of the model revealed two extensive hydrogen-bond networks formed between polar residues in TMs 2, 3, 6, and 7 and buried waters (Fig. S1). Three small polar residues, S1.50¹⁵⁵, S2.56¹⁸⁶, and S7.47³⁹², reside outside of these networks and are predicted to facilitate packing between individual TMs.

Experimental Analysis of GLP-1R Constructs.

WT and mutant GLP-1Rs were isogenically integrated into FlpIn-Chinese hamster ovary (FlpInCHO) cells by recombination. Antibody detection of the N-terminal c-myc epitope and whole-cell binding using [¹²⁵I]exendin (9–39) were used to study cell-surface expression (Fig. S2 and Table 1). Ligand affinities were calculated by whole-cell competition equilibrium binding studies using [¹²⁵I]exendin (9–39) with the agonists GLP-1(7–36)NH₂ (GLP-1), oxyntomodulin, and exendin-4 and an antagonist exendin (9–39) (Fig. S2 and Table 1). To assess the role of chosen residues in strength of coupling to signaling pathways, all were assessed in three pathways, each of which has been physiologically linked to GLP-1R–mediated insulin release [cAMP, phosphorylation of ERK1/2 (pERK1/2), and intracellular calcium (iCa²⁺) mobilization]. Agonist concentration response curves were generated using the three peptide agonists (Figs. S3–S5) and pEC₅₀ and the maximum response (E_max) values determined. Most mutants dramatically altered potencies and/or E_max in signaling through at least one pathway, and all mutations that altered cell-surface expression and/or binding affinity resulted in altered EC₅₀ and/or E_max values. To delineate effects on affinity and efficacy, concentration response curves were analyzed by an operational model of agonism to determine relative signaling efficacy estimates (logτ values). To account for the potential confounding effect of different expression levels, these τ values were normalized to what they would be if the mutant receptor were expressed at the same level as WT (logτ_c values, Table S2). Therefore, where possible, all comparisons between WT and mutant receptors were performed on logτ_c values.

Table 1.

Ligand affinities and cell-surface expression of GLP-1R polar TM mutations

	Ligand binding affinity, pKi				Cell-surface expression, % wt
Receptor	GLP-1(7–36)NH2	Oxyntomodulin	Exendin-4	Exendin (9–39)	ELISA	Bmax
WT	8.67 ± 0.05	7.26 ± 0.04	8.97 ± 0.04	8.11 ± 0.04	100 ± 1	100 ± 2
S1.50155A	8.29 ± 0.10*	7.28 ± 0.13	8.43 ± 0.14*	8.08 ± 0.07	55 ± 2*	48 ± 3*
H2.50180A	ND	ND	ND	ND	18 ± 1*	ND*
S2.56186A	8.60 ± 0.06	7.26 ± 0.05	9.02 ± 0.05	7.94 ± 0.07	100 ± 2	114 ± 4
R2.60190A	7.37 ± 0.09*	7.60 ± 0.13	7.30 ± 0.08*	7.58 ± 0.08*	53 ± 3*	44 ± 2*
N3.43240A	8.19 ± 0.07*	7.43 ± 0.05	8.64 ± 0.06	8.29 ± 0.08	87 ± 3	92 ± 2
E3.50247A	ND	ND	ND	ND	18 ± 4*	ND*
N5.50320A	7.42 ± 0.11*	6.36 ± 0.05*	7.44 ± 0.07*	8.38 ± 0.07	95 ± 4	114 ± 8
T6.42353A	ND	ND	ND	ND	30 ± 2*	ND*
H6.52363A	7.31 ± 0.08*	6.46 ± 0.11*	7.54 ± 0.11*	7.41 ± 0.09*	59 ± 4*	53 ± 2*
S7.47392A	8.42 ± 0.06	7.09 ± 0.07	8.72 ± 0.07	8.45 ± 0.06*	98 ± 3	92 ± 1
Q7.49394A	8.56 ± 0.08	7.26 ± 0.06	8.87 ± 0.07	8.18 ± 0.06	103 ± 3	111 ± 2
Y7.57402A	ND	ND	ND	ND	21 ± 6*	ND*
N7.61406A	8.61 ± 0.06	7.29 ± 0.04	8.77 ± 0.21	8.08 ± 0.05	90 ± 3	119 ± 2

Ligand affinity (K_i) and B_max estimates were derived from competition binding studies. Cell-surface expression was determined by antibody detection of the N-terminal c-Myc epitope tag. All values are expressed as means ± SEM of four to six independent experiments, conducted in duplicate. Data were analyzed with one-way analysis of variance and Dunnett’s posttest (*P < 0.05). ND, data not experimentally defined because no specific radioligand binding could be detected above background in either whole cells or crude membrane preparations.

Characterization of the Central Interaction Network.

R2.60¹⁹⁰, N3.43²⁴⁰A, N5.50³²⁰, H6.52³⁶³, Q7.49³⁹⁴, and Y7.57⁴⁰² were predicted to reside in a central interaction network (Fig. 1 and Fig. S1). R2.60¹⁹⁰A, H6.52³⁶³A, and Y7.57⁴⁰²A significantly impaired cell-surface expression compared with WT; however, expression of N3.43²⁴⁰A, N5.50³²⁰A, and Q7.49³⁹⁴A were similar to WT (Fig. S2 and Table 1).

Fig. 1.

Ligand-dependent and pathway-dependent effects upon mutation of residues composing the predicted central interaction network. (A) Differences in the coupling efficiency (logτ_c) of GLP-1, exendin-4, and oxyntomodulin to three signaling pathways [cAMP (Top), pERK1/2 (Middle), and iCa²⁺ mobilization (Bottom)] at individual mutants compared with the WT receptor. Statistical significance of changes in coupling efficacy in comparison with WT was determined by one-way analysis of variance and Dunnett's posttest, and values are indicated with an asterisk (*P < 0.05). All values are logτ_c ± SEM of four to six independent experiments, conducted in duplicate. (B) The model of the TM domain of the GLP-1R from the side (Left) and top (Right) highlighting the location of each of the six residues mutated. (C) Interactions in the model between R2.60¹⁹⁰, N3.43²⁴⁰, Q7.49³⁹⁴, and Y7.57⁴⁰². (D) View of the TM domain from the top showing the location of H6.52³⁵⁶³ in relation to R2.60¹⁹⁰, N3.43²⁴⁰, Q7.49³⁹⁴, and Y7.57⁴⁰². (E) N5.50³²⁰ forms an interaction with E6.53³⁶⁴, located next to H6.52³⁵⁶³ in the GLP-1R model.

R2.60¹⁹⁰A displayed reduced affinity for GLP-1 and exendin-4, but not oxyntomodulin. The antagonist affinity was also significantly decreased; however, the loss in agonist affinity was much greater (20- to 47-fold compared with 3-fold) (Fig. S2 and Table 1). Both N3.43²⁴⁰A and Q7.49³⁹⁴A had similar affinity to WT for all ligands with the exception of GLP-1 at N3.43²⁴⁰A, which was significantly reduced (Table 1). H6.52³⁵³A and N5.50³²⁰A reduced the affinity for all three agonists and H6.52³⁵³A also reduced the affinity of antagonist, exendin (9–39), but to a lesser extent [23- to 26-fold (agonists) and 5-fold (antagonist)]. Interestingly, for N5.50³²⁰A, greater reductions were observed for the higher-affinity agonists GLP-1 and exendin-4 (38- to 59-fold) than for oxyntomodulin (9-fold). Due to very low expression of Y7.57⁴⁰²A, no radioligand binding could be detected (Fig. S2 and Table 1).

Mutation to any of these six centrally located residues resulted in impaired signaling, but the effect varied depending on the particular mutation, the activating ligand, and the assay being assessed (Fig. 1). When activated by GLP-1 all six mutations significantly reduced coupling to cAMP, but effects were greater for H6.52³⁵³A and Y7.57⁴⁰²A. A similar trend was observed for iCa²⁺ mobilization; however, R2.60¹⁹⁰A did not reach statistical significance and the impact of H6.52³⁶³A was smaller. H6.52³⁶³A, Q7.49³⁹⁴A, and Y7.57⁴⁰²A also displayed impaired pERK1/2 responses. In contrast, no statistically significant effects on GLP-1-mediated pERK1/2 signaling were observed at R2.60¹⁹⁰A, N2.43²⁴⁰A, or N5.50³²⁰A compared with WT (Fig. 1 and Table S2).

Exendin-4–mediated responses were reduced to an extent similar to GLP-1 in all three pathways for R2.60¹⁹⁰A, N5.50³²⁰A, H6.52³⁶³A, and Y7.57⁴⁰²A. However, exendin-4 displayed a different signaling profile at N3.43²⁴⁰A and Q7.49³⁹⁴A (Fig. 1 and Table S2). N3.43²⁴⁰A did not affect cAMP signaling and showed a small significant increase in pERK1/2 coupling. Neither cAMP nor pERK1/2 reponses were significantly altered at Q7.49³⁹⁴A. However, in iCa²⁺ mobilization both mutations had reductions in efficacy equivalent to those mediated by GLP-1.

The signaling profiles of these six mutations were distinct when activated by oxyntomodulin (Fig. 1). Coupling to cAMP was impaired at N5.50³²⁰A, H6.52³⁶³A, Q7.49³⁹⁴A, and Y7.57⁴⁰²A and, like the other ligands, these effects were greater at H6.52³⁶³A and Y7.57⁴⁰²A. However, of these, only Q7.49³⁹⁴A and Y7.57⁴⁰²A showed heavily impaired iCa²⁺ mobilization. Additionally, N3.43²⁴⁰A had no effect on functional responses mediated by oxyntomodulin, and R2.60¹⁹⁰A significantly enhanced coupling to cAMP and iCa²⁺. Interestingly, none of these mutations altered oxyntomodulin signaling to pERK1/2 (Fig. 1).

Because the model suggested that both N3.43²⁴⁰ and Q7.49³⁹⁴ interact with R2.60¹⁹⁰, a double mutation (N3.43²⁴⁰A/Q7.49³⁹⁴A) was generated. This receptor displayed impaired coupling to all three pathways when activated by GLP-1_. Unlike the single mutations, which had little impact on exendin-4–mediated cAMP and pERK1/2, this elicited impaired responses for both pathways, indicating that both these interactions are important for exendin-4–mediated signal transmission (Fig. S3). However, oxyntomodulin favors a different mechanism of activation because the double mutation had little effect in pERK1/2 signaling and no greater effect than Q7.49³⁹⁴A alone on coupling to cAMP and iCa²⁺.

Calculation of bias factors between signaling pathways showed that relative to WT, R2.60¹⁹⁰A was significantly biased toward iCa²⁺ mobilization when activated by exendin-4, R2.60¹⁹⁰ and N3.43²⁴⁰ are biased toward pERK1/2 when activated by GLP-1 (and N3.43²⁴⁰ for oxyntomodulin), and significant bias was observed for Q7.49³⁹⁴A toward cAMP and pERK1/2 relative to iCa²⁺ for all ligands (Fig. S6 and Table S3). At H6.52³⁵³A, GLP-1 was significantly biased toward iCa²⁺ mobilization and to some extent pERK1/2 over cAMP, whereas oxyntomodulin and exendin-4 were biased toward pERK1/2. All three ligands were heavily biased toward pERK1/2 at N5.50³²⁰A, but little bias was observed at Y7.57A⁴⁰² (Fig. S6 and Table S3). Collectively, these data suggest that residues forming this central interaction network are crucial for fine-tuning ligand and pathway-specific receptor responses.

Characterization of Mutants in a Hydrogen Bonding Network Near the Cytoplasmic Face.

The predicted cytoplasmic-face hydrogen-bond network involved the amino acid side chains of H2.50¹⁸⁰ (TM2), E3.50²⁴⁷ (TM3), and T6.42³⁵³ (TM6). Ala substitution resulted in receptors that were very poorly expressed (18 ± 1%, 18 ± 4%, and 30 ± 2% of WT for H2.50¹⁸⁰A, E3.50²⁴⁷A, and T6.42³⁵³A, respectively) (Fig. 2, Fig. S2, and Table 1). [¹²⁵I]exendin (9–39) binding at these mutants was not detectable, so agonist affinity could not be calculated. Weak or no signaling was detectable for these receptors in any measured pathway with any of the three agonists. Mutation of all three residues together resulted in a complete loss of receptor expression and no detectable function (Fig. 2 and Fig. S4).

Fig. 2.

Effects on mutation of residues located in the hydrogen-bonding network located at the cytoplasmic face. (A) Cell-surface expression relative to the WT receptor (by antibody detection of the N-terminal cMyc epitope tag) of H2.50¹⁸⁰A, E3.50²⁴⁷A, T6.42³⁵³A, and the triple mutant (H2.50¹⁸⁰A/E3.50²⁴⁷A/T6.42³⁵³A). (B) Interactions formed between H2.50¹⁸⁰A, E3.50²⁴⁷A, T6.42³⁵³A, and waters in the GLP-1R model. (C) Dose–response curves generated in response to GLP-1 in cAMP accumulation (Top), pERK1/2 (Middle), and iCa²⁺ mobilization (Bottom) assays.

Because H2.50¹⁸⁰ was protonated in the GLP-1R model, this side chain is predicted to form a hydrogen bond with N7.61⁴⁰⁶ in TM7 (Fig. S7). N7.61 is absolutely conserved in family B GPCRs; however, no dramatic alterations in signaling were observed for N7.61⁴⁰⁶A, although subtle changes to receptor bias occurred for all ligands, with a selective increase in iCa²⁺ signaling, no change in cAMP, and a small reduction in pERK1/2, albeit none of these changes reached statistical significance (Figs. S4 and S6 and Tables S2 and S3).

Conserved Small Polar Residues Fine-Tune Activation Transition, Resulting in Controlled Activation of Different Intracellular Signaling Pathways.

The three remaining conserved polar TM residues are all serines in the GLP-1R, two of which are small residues in all family B GPCRs (1.50 and 2.56) (Fig. S2). Mutation of S2.56¹⁸⁶ and S7.47³⁹² had no effect on the cell-surface expression but S1.50¹⁵⁵A expression was significantly impaired (Fig. S2). S2.56¹⁸⁶A and S7.47³⁹²A did not alter the affinity for any agonist, whereas S1.50¹⁵⁵A displayed small reductions in affinity for GLP-1 and exendin-4 but not oxyntomodulin (Fig. S2 and Table 1).

Interestingly, all three mutants had global effects on stimulus bias regardless of the activating ligand. S1.50¹⁵⁵A had reduced coupling efficiency for cAMP and iCa²⁺ mobilization (where no response was detectable) yet enhanced efficacy for pERK1/2. In contrast, S2.56¹⁸⁶A and S7.47³⁹²A showed efficacy similar to WT for cAMP, reduced efficacy for pERK1/2, and an enhanced ability to mobilize iCa²⁺ (Fig. 3, Fig. S5, and Table S2). The changes in global bias induced by these mutations can be clearly seen in bias plots (Fig. 3) that provide a visual representation of relative pathway response. Calculation of bias factors confirmed S1.50¹⁵⁵A was prejudiced toward pERK1/2 relative to cAMP and iCa²⁺, whereas S2.46¹⁸⁶A and S7.47³⁹²A exhibited bias toward iCa²⁺ signaling relative to cAMP and pERK1/2, although these bias factors did not reach statistical significance for all ligands (Table S3).

Fig. 3.

Stimulus bias effects upon mutation of small polar residues in TMs 1, 2, and 7. (A) The coupling efficiency relative to WT (Δlogτ_c) of GLP-1, exendin-4, and oxyntomodulin to three signaling pathways [cAMP (Top), pERK1/2 (Middle), and iCa²⁺ mobilization (Bottom)] at S1.50¹⁵⁵A, S2.56¹⁸⁶A, and S7.47³⁹²A. Statistical significance was determined by one-way analysis of variance and Dunnett's posttest (*P < 0.05). (B) The relative location of S1.50¹⁵⁵, S2.56¹⁸⁶, and S7.47³⁹² in the GLP-1R model. (C) Predicted packing interactions between TMs 1 and 7 formed by S1.50¹⁵⁵ and S7.47³⁹². (D) Predicted packing interactions formed by S2.50¹⁸⁶ between TMs 2 and 3. (E) Bias plots of WT, S1.50¹⁵⁵A, S2.50¹⁸⁶A, and S7.47³⁹²A for cAMP vs. pERK1/2 (Top), pERK1/2 vs. iCa²⁺ mobilization (Middle) and cAMP vs. iCa²⁺ mobilization (Bottom) for GLP-1. Data for each pathway are normalized to the maximal response elicited by peptide at the WT GLP-1R and analyzed with a three-parameter logistic equation as defined in Eq. S1, with 150 points defining the curve. A similar bias was observed when cells expressing receptors were stimulated with exendin-4 and oxyntomodulin as evidenced by bias factor calculations (Fig. S6 and Table S3).

Discussion

Polar residues in membrane proteins are under evolutionary pressure for conservation and hence maintain common functions with essential roles in stability, activation, and interhelical association through the formation of hydrogen bonds (8). Based on our GLP-1R model, two main hydrogen-bond networks involving conserved polar side chains located between TMs 2, 3, 6, and 7 and waters are evident. Six conserved residues lie within a centrally located network including R2.60¹⁹⁰, N3.43²⁴⁰, N5.50³²⁰, H6.52³⁶³, Q7.49³⁹⁴, and Y7.57⁴⁰². Mutations in this central core often [although not in all cases (e.g., N3.43²⁴⁰A)] had similar effects on cAMP and iCa²⁺ mobilization, but effects on pERK1/2 signaling were less pronounced. This suggests that fine control of GLP-1R signaling is linked to changes in interactions formed by these buried polar residues. In addition, although in the majority of cases activation by GLP-1 and exendin-4 was similar for the six different mutations across three signaling pathways, oxyntomodulin had a strikingly different pattern of behavior, and only the mutation of Y7.57⁴⁰² (and to some extent Q7.49³⁹⁴) had a trend consistent with that of the other two peptides. Therefore, oxyntomodulin’s interaction with the GLP-1R and/or the precise mechanism by which it activates the receptor is different from that of the other peptides. This is in agreement with previous studies on bias at the WT receptor (18). In addition, Ala mutations to residues in extracellular loop 2 differentially altered ligand interactions and functional responses for oxyntomodulin compared with GLP-1 and exendin-4 (22).

R2.60¹⁹⁰A significantly lowered the affinity and efficacy of GLP-1 and exendin-4. In other receptors belonging to the family B subclass [vasoactive intestinal polypeptide type-1 and type-2 receptors (VPAC1R and VPAC2R) and secretin receptor (SecR)] the equivalent residue has been argued to interact with an Asp at position 3 of the agonist ligand (23, 24). GLP-1 and exendin-4 both contain an equivalent Glu at position 3 and therefore R2.60¹⁹⁰ could serve a similar role. Interestingly oxyntomodulin contains a Gln at position 3 and the affinity of this peptide was unaltered at R2.60¹⁹⁰A, supporting a differential mode of binding. However, the antagonist exendin (9–39) does not contain the first eight amino acids, and yet its binding was also reduced, albeit to a lesser extent.

Although less common, there were also significant differences in effect of mutation between pathways activated by GLP-1 and exendin-4 (Fig. 1 and Fig. S6). This indicates there are differing mechanisms of receptor activation for these two ligands. Thus, all three of the related peptides activate the receptor via subtly different mechanisms, which is particularly relevant becaus GLP-1 and oxyntomodulin both act endogenously and exendin-4 is used clinically to mimic the physiological functions of GLP-1.

All GPCRs are able to activate common G proteins, suggesting a conserved mechanism for G protein activation. For transition from inactive to active conformations, family A GPCRs undergo a global rearrangement of the helix bundle that shifts the cytoplasmic end of TM6 (and to a smaller extent TM5) away from the receptor core by a rotation in TM6, aided by a bend in the helix caused by a highly conserved proline (P6.50) (25–27). Early studies performed using Zn²⁺ binding suggest similar helical movements of TM6 occur in the β₂AR (a well characterized family A GPCR) and the parathyroid hormone (PTH) receptor (a family B GPCR) and, like family A GPCRs, family B receptors contain a highly conserved proline in TM6 that is important for coupling to G proteins (28–30). These large global changes are nonetheless mediated by small conformational changes within or near the receptor binding pocket that are propagated through the receptor core (26, 31). From this study, we propose that binding of different peptides to the GLP-1R results in differential (perhaps even minor) changes around the binding pocket that are linked either directly through binding or indirectly via other interactions to a hydrogen-bond network involving the conserved residues R2.60¹⁹⁰, N3.43²⁴⁰, Q7.49³⁹⁴, structural waters and other nonconserved residues. This is likely to be an early event in receptor activation because the different ligands had different effects between mutations and across signaling pathways and no mutant displayed the same profile in all pathways for the same ligand or for all ligands in the same pathway. A role for these three conserved residues (2.60, 3.43, and 7.49) was also proposed for the early stages of activation of the VPAC1R (32).

Although the magnitude of effect upon mutation varied, the effects on function observed upon mutation of N5.50³²⁰, H6.52³⁶³, and Y7.57⁴⁰² were more consistent between pathways and ligands and thus may assist the formation of larger-scale conformational changes within TMs 5 and 6, transmitting conformational movements from the early stages of activation that lead to coupling of effectors at the cytoplasmic face. Potentially, these residues may play roles in activation analogous to W6.48 (H6.52³⁶³), Y5.58 (N5.50³²⁰), and Y7.53 (Y7.57⁴⁰²) in family A GPCRs (25).

H6.52³⁶³A resulted in a marked reduction in affinity of all ligands and all functional responses, independent of the decreased affinity (with the exception of oxyntomodulin-induced pERK1/2 and iCa²⁺). In water pockets, histidines can be readily protonated and deprotonated depending on the local environment (33). In the model, H6.52³⁶³ is only singly protonated, so it is uncharged, but changes in the protonation state owing to the exchange of bulk waters could arise owing to an opening up of the bundle upon peptide binding. If this does occur, TM6 would rotate away from TM2 and TM3 because of the proximity of the positively charged R2.60¹⁹⁰, consistent with helical movements that occur in family A GPCRs (26, 34).

N5.50³²⁰A reduced affinity of GLP-1 and exendin-4 and their ability to activate both cAMP and iCa²⁺ mobilization; however, there was no change in coupling to pERK1/2. There was a similar effect on oxyntomodulin affinity and coupling to cAMP; however, in this case pERK1/2 and iCa²⁺ mobilization were unaltered. Our model suggests an interaction of N5.50³²⁰ with E6.53³⁶⁴, located next to H6.52³⁶³, that is also crucial in activation of the GLP-1R leading to cAMP signaling. Mechanistically this is interesting, because a rotation in TM6 upon activation would result in a movement of TM5 that would be aided by the proposed interaction between N5.50³²⁰ with E6.53³⁶⁴, opening up the helical bundle at the cytoplasmic face (Fig. S7). Mutation of N5.50³²⁰ seems to alter the ensemble of conformations that the GLP-1R can sample (or the frequency with which they sample subsets of populations) in response to agonists such that there is a greater propensity to form conformations linked to pERK1/2 than to cAMP and iCa²⁺. Therefore, N5.50³²⁰ may mediate receptor transitions thorough aiding movement of TM5 to open up the bundle allowing G-protein coupling but is less important for transitions enabling G-protein-independent signaling (pERK1/2).

Y7.57⁴⁰²A had a global impact on receptor function regardless of the activating ligand. Y7.57⁴⁰² forms part of a conserved VXXXY motif that may be the family B equivalent of the NPXXY motif, playing a role similar to Y7.53 in family A GPCRs (27). Y7.57⁴⁰² in the GLP-1R forms hydrophobic packing contacts through its bulky aromatic ring and a hydrogen bond interaction with N3.43²⁴⁰A and also waters through its polar moiety. These interactions are likely to be broken upon receptor activation and may aid transmission of signal from the hydrogen-bond network in the core of the protein to the cytoplasmic face, stabilizing conformations that allow effectors to bind.

In addition to a role in aiding conformational rearrangements R2.60¹⁹⁰, H6.52³⁶³, and Y7.57⁴⁰² are also required for efficient cell-surface expression. The reduction in cell-surface-expressed receptors for R2.60¹⁹⁰A, H6.52³⁶³A, and Y7.57⁴⁰²A indicates that these mutations either disrupt trafficking/insertion of the receptor into the bilayer and/or disrupt the stability of the helical bundle. R2.60¹⁹⁰ could perform both of these roles. In our model, His6.52³⁶³ resides in close proximity to R2.60¹⁹⁰ and forms π stacking interactions with F6.56³⁶⁷ and F7.45³⁹⁰, polar interactions with T7.46³⁹¹ in TM7, and hydrophobic packing with L6.49³⁶⁰ and L6.48³⁵⁹, whereas Y7.57⁴⁰² packs between two phenylalanine residues in TM2. These packing interactions are likely to provide structural integrity (Fig. S7). Although we have proposed that effects on signaling following mutation of these residues may be due to their involvement in a central interaction network, these effects could also arise due to indirect effects on folding and structural integrity via other mechanisms.

The predicted bottom hydrogen-bond network involves H2.50¹⁸⁰A, E3.50²⁴⁷, T6.42³⁵³, and multiple waters. Mutation of these residues individually resulted in heavily impaired cell-surface expression and subsequently impaired functional responses to all ligands in all pathways. One proposal is that H2.50¹⁸⁰ and E3.50²⁴⁷ play a role in family B GPCRs similar to the highly conserved D(E)RY motif in family A GPCRs. Family A GPCRs also have a proposed lock between TMs 3 and 6 that stabilize these receptors in their inactive conformation (35, 36). The extended network in our model revealed water-connected hydrogen-bond interactions between E3.50²⁴⁷ and T6.42³⁵³ that may play a similar role, locking the receptor in an inactive conformation. Often in family A GPCRs, mutation of either the D(E)RY motif or the interacting residues in TM6 (E6.30 and T6.34) results in constitutive activity (36). However, sometimes this manifests as a reduction in cell-surface expression, due to a destabilization of the receptor structure and/or constitutive internalization (36). The inability to detect binding in crude membrane fractions and whole cells upon mutation of these residues supports a crucial role of this bottom network for receptor folding and stability. It is also interesting to note that mutation of both H2.50 and T6.42 in the PTH and SecR receptors and of H2.50 in the VPAC1R leads to constitutive activation, supporting the evolutionarily conserved role of these residues in ground-state stabilization (37–40).

In addition to large polar and charged amino acids, mutation to three small polar residues promotes receptor conformations that altered the equilibrium between the different signaling cascades; however, these mutants had global impacts on signaling regardless of the activating ligand. Relative to WT, S1.50¹⁵⁵A displayed significant bias toward pERK1/2 over calcium and cAMP, whereas S2.50¹⁸⁶A and S7.47³⁹²A were biased toward iCa²⁺ (Fig. 3). Small weakly polar residues (Ser, Thr, and Cys) within the TM domains of membrane proteins are key determinants in helix–helix interactions (41). In our model, S1.50¹⁵⁵ packs up against two small residues in TM7 (G7.50³⁹⁵ and A7.54³⁹⁹), aiding close packing of TMs1 and 7, and S2.50¹⁸⁶ in TM2 packs with C3.39²³⁶ in TM3, allowing tight packing of these helices. S2.50¹⁸⁶ is also in close proximity to A3.42²³⁹, and an interaction with this residue may occur on conformational transition (Fig. 3). It is not clear whether Ala mutation results in more tightly packed helices or creates space in the structure, but clearly subtle changes in these regions have very selective effects on the conformational landscape that the receptor can explore. Interestingly, S7.47³⁹² in TM7 does not cause tight packing but instead borders a solvated pocket, forming a direct interaction with Y1.48¹⁴⁸ in TM1 and through water contact with D2.68¹⁹⁸ (Table S1). This Asp has been highlighted as a potential contact for peptides and when mutated significantly alters GLP-1R function (42). Ala mutation of S7.47³⁹² did not alter affinity of any ligand, but its ability to selectively enhance signaling to iCa²⁺ but not pERK1/2 or cAMP indicated that disruption of these interactions significantly lowers the energy barrier, allowing the receptor to mobilize iCa²⁺. In addition, N7.61⁴⁰⁶A also resulted in an increase in iCa²⁺ for all ligands, although the pathway selectivity was not as pronounced as those mentioned above. Nonetheless, these effects were similar to those of S2.56¹⁸⁶ and S7.47³⁹² (also located in TMs 2 and 7). Taken together, these residues seem to optimally pack TM helices to allow fine-tuning of receptor response to different intracellular effectors.

Collectively, this study indicates that conserved polar residues within the TM domain of the GLP-1R are essential for structural integrity and activation transition, including signaling preferences and ligand-directed stimulus bias. This likely involves reordering of interaction networks between polar side chains and buried waters, providing a mechanism through which receptors and ligands achieve conformational complexes associated with signaling bias. The high degree of conservation of these residues suggests that they may play a similar role in signaling of other family B GPCRs.

Materials and Methods

Receptor Mutagenesis and Cell Culture.

Thirteen conserved polar residues located in the TM domain that are predicted to be buried within the core of the GLP-1R were selected for mutagenesis (Fig. S1). This was achieved using the QuikChange site-directed mutagenesis kit (Stratagene), and stable cell lines were generated using gateway technology (Invitrogen).

Expression, Binding, and Functional Assays.

Cell-surface expression, radioligand binding (22), cAMP accumulation, pERK1/2, and iCa²⁺ mobilization assays (18) were performed as previously described.

Molecular Modeling.

All atom canonical alpha helices for each TM region of the GLP-1R were generated using TINKER (43, 44). Each helix was fitted to the equivalent helix in bovine rhodopsin (1GZM) using lipid facing residue information extracted from the Trans-Membrane helix-LIPid web server. For those helices containing proline residues helices were picked from the pool of helical turns to minimize polar residue exposure to the lipid environment. A coarse-grain representation of the helical bundle was generated according to the MARTINI force field before insertion into a palmitoyl oleoyl phosphatidylcholine bilayer (45). The system was simulated for 1 µs with snapshots taken every 10 ns, then transformed from coarse-grain to all-atom representation using PULCHRA and OPUS_ROTA and scored using OPUS_PSP. The best scoring bundle was subjected to 5e8 steps in a replica exchange Monte Carlo simulation as implemented in hippo using default values. Water positions were predicted using DOWSER.

Data Analysis.

All data were analyzed using Prism 5.04 (GraphPad Software Inc.) as previously described (46).

For more details on experimental methods and data analysis, see SI Materials and Methods.