Scholarly article on topic 'The Extracellular Surface of the GLP-1 Receptor Is a Molecular Trigger for Biased Agonism'

The Extracellular Surface of the GLP-1 Receptor Is a Molecular Trigger for Biased Agonism Academic research paper on "Biological sciences"

CC BY
0
0
Share paper
Academic journal
Cell
OECD Field of science
Keywords
{}

Abstract of research paper on Biological sciences, author of scientific article — Denise Wootten, Christopher A. Reynolds, Kevin J. Smith, Juan C. Mobarec, Cassandra Koole, et al.

Summary Ligand-directed signal bias offers opportunities for sculpting molecular events, with the promise of better, safer therapeutics. Critical to the exploitation of signal bias is an understanding of the molecular events coupling ligand binding to intracellular signaling. Activation of class B G protein-coupled receptors is driven by interaction of the peptide N terminus with the receptor core. To understand how this drives signaling, we have used advanced analytical methods that enable separation of effects on pathway-specific signaling from those that modify agonist affinity and mapped the functional consequence of receptor modification onto three-dimensional models of a receptor-ligand complex. This yields molecular insights into the initiation of receptor activation and the mechanistic basis for biased agonism. Our data reveal that peptide agonists can engage different elements of the receptor extracellular face to achieve effector coupling and biased signaling providing a foundation for rational design of biased agonists.

Academic research paper on topic "The Extracellular Surface of the GLP-1 Receptor Is a Molecular Trigger for Biased Agonism"

Cell

Article

The Extracellular Surface of the GLP-1 Receptor Is a Molecular Trigger for Biased Agonism

Graphical Abstract

Authors

Denise Wootten, Christopher A.

Reynolds, Kevin J. Smith.....Laurence J.

Miller, Arthur Christopoulos, Patrick M. Sexton

Correspondence

denise.wootten@monash.edu (D.W.), patrick.sexton@monash.edu (P.M.S.)

In Brief

Understanding how different ligands engage and activate signaling pathways downstream of the GLP-1 receptor, a key type II diabetes target, suggests strategies for shaping signaling outcomes to favor physiologically beneficial effects.

Highlights

• Effect of mutation on affinity and efficacy of biased ligands mapped onto 3D models

• Biased agonists form distinct interactions with the GLP-1R extracellular surface

• Engagement of unique elements of the extracellular surface promotes biased agonism

• Insights into class B GPCR activation/biased agonism can aid rational drug design

Wootten et al„ 2016, Cell 165, 1632-1643 ciossMark June 16, 2016 © 2016 The Authors. Published by Elsevier Inc. http://dx.d0i.0rg/l 0.1016/j.cell.2016.05.023

CelPress

Article

The Extracellular Surface of the GLP-1 Receptor Is a Molecular Trigger for Biased Agonism

Denise Wootten,1* Christopher A. Reynolds,3 Kevin J. Smith,3 Juan C. Mobarec,3 Cassandra Koole,14 Emilia E. Savage,1 Kavita Pabreja,1 John Simms,5 Rohan Sridhar,1 Sebastian G.B. Furness,1 Mengjie Liu,2 Philip E. Thompson,2 Laurence J. Miller,6 Arthur Christopoulos,1 and Patrick M. Sexton1*

1Drug Discovery Biology 2Medicinal Chemistry

Monash Institute of Pharmaceutical Sciences, Monash University, Parkville, Victoria 3052, Australia 3School of Biological Sciences, University of Essex, Colchester CO4 3SQ, UK

4Laboratory of Chemical Biology and Signal Transduction, The Rockefeller University, 1230 York Avenue, New York, NY 10065, USA 5School of Life and Health Sciences, Aston University, Birmingham B4 7ET, UK

6Department of Molecular Pharmacology and Experimental Therapeutics, Mayo Clinic, Scottsdale, AZ 85259, USA 'Correspondence: denise.wootten@monash.edu (D.W.), patrick.sexton@monash.edu (P.M.S.) http://dx.doi.org/10.10167j.cell.2016.05.023

SUMMARY

Ligand-directed signal bias offers opportunities for sculpting molecular events, with the promise of better, safer therapeutics. Critical to the exploitation of signal bias is an understanding of the molecular events coupling ligand binding to intracellular signaling. Activation of class B G protein-coupled receptors is driven by interaction of the peptide N terminus with the receptor core. To understand how this drives signaling, we have used advanced analytical methods that enable separation of effects on pathway-specific signaling from those that modify agonist affinity and mapped the functional consequence of receptor modification onto three-dimensional models of a receptor-ligand complex. This yields molecular insights into the initiation of receptor activation and the mechanistic basis for biased agonism. Our data reveal that peptide agonists can engage different elements of the receptor extracellular face to achieve effector coupling and biased signaling providing a foundation for rational design of biased agonists.

INTRODUCTION

G protein-coupled receptors (GPCRs) are critical for the transmission of extracellular signals across the cell membrane to initiate intracellular responses (Fredriksson et al., 2003) and are the leading targets of currently marketed therapeutics (Overing-ton et al., 2006). It is therefore vital to understand molecular interactions that govern ligand binding and how these interactions initiate intracellular signaling. Key advances in GPCR structural biology have greatly enhanced our knowledge of ligand interaction with GPCRs and yielded insight into receptor activation (reviewed in Katritch et al., 2013). However, to date, full-length structures have only been solved for a subset of class A GPCRs, mostly in complex with small-molecule ligands and in single

inactive conformations. In contrast, there is limited information addressing the molecular details by which peptide binding at class B GPCRs couples to effector activation.

Class B peptide hormone receptors are a subfamily of GPCRs that are major targets for the treatment of chronic disease, including type 2 diabetes, obesity, and dis-regulated bone metabolism (Couvineau and Laburthe, 2012). They include receptors that bind calcitonin, calcitonin gene-related peptide, vasoactive intestinal polypeptide, pituitary adenylate cyclase-activating polypeptide, corticotropin releasing factor (CRF), gastric inhibitory polypeptide, parathyroid hormone, glucagon, and glucagon-like peptides (GLP-1 and GLP-2). Class B GPCRs share the basic seven transmembrane (TM) topology common to all GPCRs but also possess a large N terminus that forms the major binding site for selective recognition of peptide ligands (Couvineau and Laburthe, 2012). Despite sequence divergence in this region between different receptors, this extracellular domain (ECD) contains key conserved residues, including three disulphide bonds that aid in stability and confer structural similarities between receptors.

Structural data for class B receptors are limited to partial domains, including several NMR and crystal structures of pep-tide-bound N-terminal domains (reviewed in Pal et al., 2012) and, more recently, two inactive structures of the isolated TM core of the CRF1 receptor (CRF1R) and the glucagon receptor (GCGR) (Hollenstein et al., 2013; Siu et al., 2013). This structural data, along with structure-activity studies, support the proposed two-domain model for peptide binding to class B GPCRs, with the a-helical C terminus binding to the receptor N-terminal ECD and the peptide N terminus interacting with the extracellular face of the TM bundle (this includes the top of the TMs and the extracellular loops [ECLs]) (Pal et al., 2012). However, there is very limited information available to define these N-terminal peptide interactions with the extracellular face of the receptor core and even less to indicate how this engagement drives receptor activation. Photoaffinity and mutagenesis data highlight the significance of the core domain in both peptide binding and receptor activation, including residues within the three ECLs and their juxtamembrane regions of class B GPCRs (Barwell et al., 2011; Bergwitz et al., 1997). These studies suggest

1632 Cell 165, 1632-1643, June 16, 2016 © 2016 The Authors. Published by Elsevier Inc.

This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).

CrossMark

that the extracellular face of the TM bundle forms a significant site of receptor interaction and/or plays an important role in stabilizing active receptor conformations in the presence of agonists, allowing for activation of intracellular signaling.

The GLP-1R couples to multiple effectors, and in vivo data support this as important for normal physiology in both glucose and energy homeostasis (Baggio and Drucker, 2007). The GLP-1R is an important target for treatment of type 2 diabetes mellitus, and there are multiple endogenous peptides that activate this receptor. These include four forms of GLP-1 and the related peptide oxyntomodulin (Baggio and Drucker, 2007). In addition, there are clinically approved peptides for treatment of type 2 diabetes, including exendin-4 and metabolically stabilized forms of GLP-1 (Reid, 2013). N-termi-nally truncated forms of these peptides are antagonists, for example exendin-4(9-39). In previous studies, we identified exendin-4 and oxyntomodulin as biased agonists relative to GLP-1 (GLP-1(7-36)NH2) (Koole et al., 2010; Wootten et al., 2013a). The phenomenon of biased agonism describes the ability of different ligands acting at the same receptor to promote distinct cellular responses (Kenakin and Christopoulos, 2013). Intriguingly, a biased GLP-1R peptide agonist, P5, that maintains G protein signaling, while exhibiting attenuated b-arrestin recruitment, induced adiposity and was more effective at correcting hyperglycaemia in diabetic animals than exendin-4, despite having markedly lower insulinotropic properties (Zhang et al., 2015). This highlights the potential utility of biased agonists as novel GLP-1R therapeutics.

Biased agonism is currently of great interest for drug discovery, with the potential to sculpt cellular responses to favor therapeutically beneficial signaling pathways over those leading to harmful effects. However, the mechanistic basis underlying biased signaling needs to be understood if this is to be exploited for rational drug design. Pleiotropic coupling of the GLP-1R leads to cAMP production, Ca2+ mobilization, and phosphorylation of ERK1/2 (pERK1/2) (Koole et al., 2010), each of which are physiologically important (Baggio and Drucker, 2007). The contribution of these signaling pathways and the extent to which one is activated relative to another is therefore important for optimal development of therapeutics. Existing data demonstrate that biased signaling does indeed occur at the GLP-1R; however, the mechanistic basis for this is unknown (Koole et al., 2010; Wootten et al., 2013a).

Using a combination of alanine-scanning mutagenesis, followed by pharmacological quantification of the effects of mutation on peptide agonist affinity and three distinct signaling pathways, we have identified critical regions within the extracellular face of the receptor core both for peptide agonist affinity and for driving receptor coupling to distinct signaling pathways, extending our initial work on ECL2 (Koole et al., 2012). We used a GLP-1R model in conjunction with experimental data to generate comparative heatmaps of the contribution of the extracellular surface to agonist affinity and signaling efficacy. These revealed distinct elements of the extracellular face of the GLP-1R that are engaged to activate individual signaling pathways in a ligand-dependent manner. Collectively, the work allows us to yield novel molecular insights into the initiation of receptor activation and the mecha-

nistic basis for biased agonism at this important class B GPCR. This provides a framework to enable future design of agonists with tailored signal bias for this receptor.

RESULTS

To understand the functional interface at the GLP-1 R extracellular surface, we completed alanine-scanning mutagenesis of the ECLs and adjacent TM residues, coupled with analysis of ligand affinity and signaling for three key pathways that are involved in GLP-1R function and rely on different effector engagement (Figure S1). We assessed three peptides (GLP-1, oxyntomodulin, and exendin-4) with highly conserved N-ter-minal sequences that display biased agonism (Figure S2). This biased agonism can be observed in both the recombinant cells used in this mutagenesis study and natively expressing insulinoma cells that display key features of b islets, where both exendin-4 and oxyntomodulin were biased away from GLP-1 in promotion of cellular proliferation and reducing apoptosis, compared to cAMP signaling (Figure S2). In addition to measurements of agonist binding affinity, the effects on signaling efficacy for each of the pathways were quantified using an operational model of agonism (Black and Leff, 1983). This enables comparison of effects of mutations across the different signaling pathways and reveals how individual pep-tide ligands interact with the receptor surface to elicit signaling. To understand the importance of residues in ligand binding and function, we developed a full-length, GLP-1-bound, GLP-1 R model (Active model S1) (Wootten et al., 2016). Residues located within the ECL/TM boundaries of the N-terminal ECD are numbered based on their location in the protein sequence. Residues that are located within the TM bundle also contain, in superscript, the class B numbering described in Wootten et al. (2013b).

The predicted ECL1 and adjacent TM boundary comprises 23 residues from L201 to S223, ECL2, 23 residues from G285 to L307 and ECL3, 16 residues from D372 to E387. The results for the pharmacology of ECL2 mutants have been published previously (Koole et al., 2012) and are discussed here, along with novel data on ECL1 and ECL3, in context of the 3D surface map developed for the receptor.

All mutant receptors, with the exception of W306A, were expressed at the cell surface, and most were expressed at levels equivalent to wild-type (Table S1) (Koole et al., 2012). Of the ECL1 and ECL3 mutants, only three exhibited a change in antagonist binding: H374A, an effect specific to exendin-4 (9-39); K383A, which had global effects on peptide binding; and F381A, which had selective effects dependent on the peptide (Table S1). Given that only limited mutations grossly altered cell-surface expression and antagonist affinity, altered effects on receptor function (affinity and efficacy) for most engineered mutations are likely a result of loss of direct interactions with ligands (affinity) or of altering (either directly or indirectly) interactions between receptor side chains and disruptions to hydrogen bonding networks that are crucial for the receptor to explore its conformational landscape, thereby indicating residues that are important in the mechanism of signal propagation.

Figure 1. Agonist Affinity Profiles of GLP-1R ECL Alanine Mutants Reveal the Importance of Individual Residues for Peptide Affinity

pKi values for each peptide were derived from radioligand inhibition-binding experiments. Bars represent differences in calculated affinity (pKi) values for each mutant relative to the wild-type receptor for GLP-1 (top), oxyntomodulin (middle), and exendin-4 (bottom). Statistical significance of changes in affinity in comparison with wild-type was determined by one-way analysis of variance and Dunnett's post-test, and values are indicated with an asterisk (*p < 0.05). Data that are statistically significant are colored based on the extent of effect. All values are ± SEM of four to six independent experiments, conducted in duplicate.

s* E -0.5-

8 -1.0-

a -2.0-

< -2.5-

Oxyritomodulin ß -1.0-

-1.5-a. -2.0' 4 -2.5--3.0J

1.0-t o.o-

£ -0.5E -1.0-Exendin-4 &

a. -2.0-< -2.5-3.0-

OOOOOOOT—*—T-T-T-T—*— T— (\ICM CM CN CM CM CNI ^ IN CM CM CM CM CM CM CM N CM CM CM CM CM N CM

yguigy^yw-^yy x y-yuyl

M CM CM CM CM CM CM CM CM CM CM CM CM CM CM c\.

yu| iu y U UL

OOOOOOOT—T—T—T—*—T—T—CMCMCMCNi

yyuyyuuuy uyyy

TM2-ECL1-TM3

SSSSSS^SSS&igSigiSSSSiSli

coco£ocooocT>o)0)aio}C)a>GiO)a>oooooooo

KiSS&igSigiSSSSSSS

foco^oconcoocorococoncofoco

q hi i a: o !-■-1 ££ u-—x -i u-hi

___________________010)0)00000000

T i i z

TM4-ECL2-TM5

TM6-ECL3-TM7

Reduced function

>30 fold

10-30 fold

5-10 fold

3-5 fold

Involvement of the Extracellular Surface in Agonist Peptide-Binding Affinity

Effects of mutations on agonist affinity were established by heterologous competition with the antagonist radioligand (125I-ex-endin-4 (9-39)) (Table S1) (Koole et al., 2012). The affinity measures for each mutant were compared to the wild-type to determine the relative importance of each individual residue in peptide agonist affinity (Figure 1). These were mapped onto the 3D model to provide a comparative heatmap of the contribution of the extracellular surface to agonist affinity (Figure 2, Active model S1).

Overall, there was a high degree of overlap in the impact of alanine mutation on binding of GLP-1, exendin-4, and oxyntomodulin. In 3D space, there is a continuum of residues from K288, E292, D293, R299, N300 within the proximal part of ECL2 that link to TM6/ECL3 membrane-proximal residues D372, E373 and L379, K383, L384 in ECL3 that are globally important for binding affinity, along with most residues in the distal segment of ECL2 (Table S1, Figures 1 and 2, Active model S1).

In addition, M204 at the TM2/ECL1 boundary lines the pep-tide-binding groove in our model and is important for the affinity of all peptides. There is an additional network of residues deeper

in the protein (C296, W297, R380) that are important for GLP-1 and exendin-4 affinity but have little role in oxyntomodulin binding (Table S1, Figures 1 and 2, Active model S1). L218 within ECL1 is also important for the affinity of GLP-1 and exendin-4 but not oxyntomodulin. Additionally, L201 lies deeper in the peptide groove of the protein and is important for GLP-1 and oxyntomodulin affinity but not exendin-4 (Table S1, Figures 1 and 2, Active model S1). Only a limited number of residues were selectively important for affinity of individual peptides: W214 (ECL1) and G377 (ECL3) for GLP-1; K202 (ECL1) and F381 (ECL3) for oxyntomodulin; and T378andT386 (both in ECL3) for exendin-4.

Of all the residues important for peptide affinity, most are likely to have indirect effects on peptide binding. Alterations to agonist affinity can be achieved by the mutation either altering the conformation of residues that directly interact with the peptide within the binding pocket or altering the shape of the binding pocket such that the peptide cannot bind in the same manner. The only ECL side chains that our modeling predicted to interact directly with GLP-1 are L201, W297, R299, N300, and R380 (Figure S3). This includes three residues that, when mutated, have differential effects on peptide affinity. Although all are important for GLP-1 affinity, alanine mutation of L201 had little effect on ex-endin-4 affinity, and mutation of W297 and R380 had no effect on

GLP-l B Oxyntomodulin C Exendin-4

Figure 2. Heatmap 3D Representation of the GLP-1R Extracellular Face Based on Affinity-Binding Data

Molecular model of the GLP-1R-GLP-1 complex showing the extracellular surface of the TM bundle. Residues that altered affinity of GLP-1 (A), oxyntomodulin (B), and exendin-4 (C) when mutated are highlighted. Teal indicates residues that were assessed and did not alter affinity; yellow (3- to 5-fold), pale orange (5- to 10-fold), orange (10- to 30-fold), and red (>30-fold) are residues that statistically altered affinity.

oxyntomodulin affinity (Figures 1 and 2, Table S1). This is particularly interesting, as the residues in the GLP-1 peptide that are predicted to interact with these side chains are absolutely conserved in the N terminus of the three peptide ligands (Figure S3). This implies that differential interactions of the C terminus of the peptides with the N-terminal ECD may differentially orient the N terminus of these peptides in the binding groove such that they form distinct interactions with the bundle.

Involvement of the Extracellular Surface in Ligand Efficacy

Agonist potency is a composite of efficacy and affinity and cannot be used to distinguish pathway-specific effects of mutations. In contrast, in the operational model, the efficacy term "t" relates receptor occupancy to magnitude of response for an individual pathway and is independent of ligand affinity, although not receptor expression levels. However, t values can be normalized to experimentally determined levels of cell-surface expression to provide a measure of pathway activation (tc) that is independent of both affinity and cell-surface expression levels (Kenakin and Christopoulos, 2013). Concentration response curves for each of the peptides for cAMP formation, pERK1/2, and Ca2+ mobilization were established for wild-type and each receptor mutation to determine EC50 and Emax and tc values for each pathway for all mutants (Tables S2, S3, and S4). As with affinity, tc estimates for each mutant receptor were compared to the wild-type to determine the relative importance of each residue for efficacy in each pathway (Figures 3,4, and 5). These were mapped onto the 3D model to provide a comparative heatmap of the contribution of the extracellular surface to efficacy for individual pathways (Figure 6, Active model S1). Overall, there was a significant correlation between residues identified as important for peptide affinity, cAMP formation, and Ca2+ mobilization. Generally, there was less correlation between agonist af-

finity and pERK1/2 efficacy with a distinct pattern of residues in 3D space being important for transmitting efficacy down this pathway (Figures 6 and 7).

Involvement of the Extracellular Surface in Peptide-Mediated cAMP Formation

Consistent with binding studies, there was a high degree of overlap in the impact of alanine mutation on cAMP-mediated signaling by GLP-1, exendin-4, and oxyntomodulin (Figures 3 and 6; Tables S1 and S2). These residues were concentrated mainly within ECL2 and the more buried, membrane-proximal regions of ECL1 and ECL3 that included residues deep within the binding groove important for affinity of these peptides (Figures 3 and 6A-6C, Active model S1). However, when mutated, these residues had a smaller effect on cAMP efficacy than they did on affinity (Figures 1,2,3, and 6A-6C). An additional residue, Y205, proximal to the binding groove at the TM2/ECL1 boundary was important for cAMP efficacy by all three peptides. Furthermore, two residues within the peptide binding groove (W297 and R380) that were important for GLP-1 and exendin-4, but not oxyntomodulin, affinity were required for all three peptides to activate this signaling pathway. Here the effect of mutation was larger for oxyntomodulin (no appreciable cAMP response) (Figures 2,3, and 6; Table S2) (Koole et al., 2012). GLP-1 and ex-endin-4 also engage a large proportion of ECL2 and ECL3 with only minor contributions from ECL1 for transmission of efficacy to the cAMP pathway (Figures 3, 6A, and 6C). In contrast, a large proportion of all three loops contributed to cAMP signaling via the peptide oxyntomodulin, with more involvement of ECL1 but less involvement of ECL2 compared with GLP-1 and ex-endin-4 (Figures 3 and 6B). The heatmaps indicate that the lower-affinity ligand oxyntomodulin engages regions (albeit not necessarily the same residues) in the extracellular surface similar to those of GLP-1 and exendin-4 upon binding to the receptor

Figure 3. Peptide-Dependent Effects of ECL Mutations on cAMP Efficacy

Differences in the coupling efficiency (logtc) for cAMP formation of ECL mutations, compared to the wild-type receptor, by GLP-1 (top), oxyntomodulln (middle), and exendin-4 (bottom). Statistical significance of changes in coupling efficacy was determined by one-way analysis of variance and Dunnett's post-test, and values are indicated with an asterisk (*p < 0.05 compared with wild-type). Data that are statistically significant are colored based on the direction and extent of effect. All values are logtc ± SEM of four to six independent experiments, conducted in duplicate.

GLP1 (7-36)NH2

Exendin-4

Oxyntomodulin o

1.0-^ 0.5-

I 0.0 tf -°-5H

O -1.0

1.0-1 0.5 0.0-0.5'

O -1.0-1

1.0 0.50.0 -0.5-I

o -1.0'

<<<<<<<<<<<<<<<<<<<<

OOOOOOOT—T—T—T-T—T-T—CMCMCMCM ------------------------------csicmcmçycml.

»>-aQi«

cooSoi cmc*S$ i?5coi 0§a5oic5ci^î2i tSS ooeocoooooo)cjicj)0>o)aïc)0)aioïoooooooo

gag^P^sjSSiSuSP^SSSS^S-

iasgissliäissssiiäääissä

o o o o o o o t-t-t-t-t-t-cm cnicvj cm

cm cm cm cm cm cm cm cm cm cm cm cm cm cm cm cm cm r

SSSäääiSSSSSiiSSiSiäSliS^

cooocoScociaxnaxnœœoKnœoooooooo cm cv cm cm cm cm cm cm cm cm cm cm cm co m m m m m m m cd" >2 >- -">- hjq ujoo ^h qüzwz 5z>- <»

Uum yiyy^yuuy

TM2-ECL1-TM3

JUP u p n y HPV

* * * *T J **L * * .*. *

u UUW II will 1 il m

ooooooorT-T-rT-T-^^cviNNNfflioiaiDMOiojtnffidiffiäffläoigoogooöo

aag^pispissïssaîiiiapisssss^gis^îi^îiasisôspissssssps*1 ! h h

TM4-ECL2-TM5

N-h-N-h-I^N-I^OOCOcOOOCOOOOOOO

r^h-r^N-r^N-r^cocooococococnco

h-h-r^h-h-N-r^-COCOcOCOCOOOCOQO

TNI6-ECL3-TM7

Reduced function

> 30 fold

10-30 fold

5-10 fold

3-5 fold

Enhanced function

> 4 fold

but requires distinct regions of this surface to promote conformational transitions leading to formation of cAMP (Figures 2 and 6, Active model S1).

Despite the critical importance of ECL2 and the membrane-proximal region of ECL3 for all three peptides to couple to cAMP, the relative contribution of each individual residue within this region varied considerably between oxyntomodulin and the other two peptides (Figure 7). This included a number of residues within ECL2 (Y291, E294, T298, S301, and M303) that were required for GLP-1- and exendin-4-mediated cAMP accumulation but not for oxyntomodulin and two residues, C296 in ECL2 and E387 in ECL3, that had the reverse profile. In addition, there were a number of residues that had global effects across all three peptides but with different magnitudes in the extent of effect (Figures 3,6, and 7; Table S2). Only a very limited number of residues were selectively important for cAMP efficacy between GLP-1 and exendin-4. K202 (ECL1) was selective for GLP-1 only, and Q211 (ECL1), D372, and I382 (ECL3) affected both GLP-1 and oxyntomodulin but not exendin-4. D222 (ECL1) and L384 (ECL3) were selective for exendin-4 only, whereas mutation of S223 (ECL1), D293, Y305 (ECL2), and K383 (ECL3) altered cAMP signaling by both exendin-4 and oxyntomodulin but not GLP-1 (Figures 3, 6, and 7; Table S2).

Involvement of the Extracellular Surface in Peptide-Mediated Intracellular Calcium Mobilization

Due to the low efficacy of oxyntomodulin for promoting Ca2+ mobilization, this pathway was only assessed for GLP-1 and ex-endin-4 (Figures 4,6D, and 6E, Active model S1). Consistent with binding and cAMP data, a continuum of residues in 3D space within ECL2 (K288, E292, D293, C296, W297, R299, N300) linking the membrane-proximal residues in ECL3 (D372, E373, L379, R380, K383), along with most residues in the distal segment of ECL2, were required for both ligands to promote Ca2+ mobilization. However, mutation of these residues had a larger impact on exendin-4 than on GLP-1, with more mutant receptors unable to produce a detectable exendin-4-mediated Ca2+ response (Figures 4, 6D, and 6E). In addition, residues within the TM2/ECL1 membrane-proximal region (L201, K202, M204, Y205, T207) were globally important for both peptides. There were also additional residues within the TM4/ECL2 (I286, V287, Y289, L290, Y291) and the ECL3/TM7 (L384, T386) membrane-proximal portions of ECL2 and ECL3, respectively, that were important for ex-endin-4-mediated signaling to this pathway, but with little role for GLP-1 coupling (Figures 4, 6D, and 6E). These residues extend within 3D space from the continuum of residues that are globally important.

Figure 4. Peptide-Dependent Effects of ECL Mutations on Efficacy for Ca2+ Mobilization

Differences in the coupling efficiency (logTc) for Ca2+ mobilization of ECL mutations, compared to the wild-type receptor, by GLP-1 (top) and exendln-4 (bottom). Statistical significance of changes in coupling efficacy was determined by one-way analysis of variance and Dunnett's post-test, and values are indicated with an asterisk (*p < 0.05 compared with wild-type). Data that are statistically significant are colored based on the extent of effect. All values are logTc ± SEM of four to six independent experiments, conducted in duplicate.

GLP1 (7-36)NH2

Exendin-4

1.5 1.00.5 0.0-0.5-1.0-1.5-2.0

1.5 1.00.50.0 -0.5-I -1.0 -1.5-2.0-

OOOOOOT—T—T—^T—T—

« t-cn-.

l^-COOJO-----

T-T-T-CMCMCMCM CM CM CM CM CM CM CM CM

Q_u«>-OQM

coeocoooooo)0)0)0)a)0)oaso)0)oooooooo

OOgOOOOT-T-r-rr-r-rT-T-NNNN

ii_£L

yi vyyyyvvv ■ yu

TM2-ECL1-TM3

b b bb b

aiswnsaiaoiaaffiaaianoogogooc

£zbzjz>-5-

b bb bb bb bbb b

II lju y U^l

TM4-ECL2-TM5

SSSSSSSSSSSSSSii

r^f-f-h-r-f-r-oococococoooaoco «2 co co eg co oo <r> co co <r> co co co co co

r^-h-r^-s-h-r^-h-.c&cococococomco

TM6-ECL3-TM7

Reduced function

> 30 fold

10-30 fold

5-10 fold

3-5 fold

Enhanced function

>4 fold

Interestingly, mutation of two residues (D215 in ECL1 and T378 in ECL3) lying outside of the predicted peptide-binding groove enhances the ability of GLP-1 to promote Ca2+ mobilization, and one of these residues (T378) also had the same effect for exendin-4 (Figures 4 and 6).

Involvement of the Extracellular Surface in Peptide-Mediated pERK1/2

Mapping mutational effects for coupling to pERK1/2 onto the 3D model revealed a strikingly distinct pattern in regions of the GLP-1R extracellular face that were involved in coupling to this pathway, in comparison to those important for affinity, cAMP, and Ca2+ mobilization. Whereas these latter aspects of receptor function required a large area of the protein's extracellular surface for transmission of signal, residues important for pERK1/2 were localized mainly to membrane-proximal residues of ECL3 (Figures 5 and 6F-6H, Active model S1) with very little involvement (at least for GLP-1 and exendin-4) of ECL2 (the most critical domain for all other assessed aspects of receptor function). Only one residue throughout the entire extracellular surface, Y205 in ECL1, was globally important for coupling all three peptides to cAMP, iCa2+, and pERK1/2 (Figures 3, 4, 5, 6, and 7).

Despite all three peptides utilizing ECL3 for coupling receptor activation to pERK1/2, the importance of individual residues within this loop varied between the ligands. D372, T378 and R380, and T386 were globally important for signaling by all three peptides, but for T378 and R380, the effect of mutation varied. Interestingly, T378A increased the efficacy for GLP-1 and exen-

din-4 but had the opposite effect on oxyntomodulin, reducing its efficacy. The reverse effect was observed for R380A, where oxyntomodulin efficacy was increased and GLP-1 and exen-din-4 efficacies were both impaired (Figures 5, 6F-6H, and 7). This implies that these residues may be important for conforma-tional switching of the receptor, altering the ensemble of conformations that allow for coupling to this pathway. In addition, within this loop, E373 was required for both GLP-1 and exen-din-4 but played little role in the ability of oxyntomodulin to activate this pathway. Furthermore, mutation of R376, L379, F381, and I382 significantly altered signaling by GLP-1, with a similar trend displayed by exendin-4 but little role in oxyntomodulin-mediated pERK1/2 (Figure 5).

Although, compared to GLP-1, exendin-4 utilized a larger proportion of the extracellular surface for coupling the GLP-1R to iCa2+ mobilization, mutational effects on pERK1/2 were even more confined to ECL3 than those for GLP-1. Intriguingly, despite requiring a very large portion of ECL2 for intracellular Ca2+ mobilization, this domain played no role in coupling exen-din-4 binding to pERK1/2. ECL2 played a limited role for GLP-1 coupling to this pathway with E292 and N300 being important. These residues were also important for oxyntomodulin coupling to pERK1/2 along with D293, N302, and Y305 (Figures 5, 6F, and 6G).

Additional residues selectively important for pERK1/2 by individual peptides included M204 and Q213 (ECL1) for GLP-1; L201 (ECL1) and F385 (ECL3) for oxyntomodulin; and W203 (ECL1) for exendin-4.

Reduced function >30 fold 10-30 fold 5-10 fold 3-5 fold Enhanced function >4 fold

Figure 5. Peptide-Dependent Effects of ECL Mutations on pERK1/2 Efficacy

Differences in the coupling efficiency (logtc) to pERK1/2 of ECL mutations, compared to the wild-type receptor, by GLP-1 (top), oxyntomodulln (middle), and exendln-4 (bottom). Statistical significance of changes in coupling efficacy was determined by one-way analysis of variance and Dunnett's post-test, and values are indicated with an asterisk (*p < 0.05 compared with wild-type). Data that are statistically significant are colored based on the direction and extent of effect. All values are logtc ± SEM of four to six independent experiments, conducted in duplicate.

GLP1 (7-36)NH2 I

Oxyntomodulin

Exendin-4

1.5-1 1.00.50.0-0.5-

& -1.0-

< -1.5-2.0-

1.51.00.50.0-0.5-

-1.5-2.01.5-1 1.00.50.0-0.5-

& -1.0-

< -1.5-2.0-

____ ________

OOPpoOOT-T-r-T-T-T-T-T-T-fgCMÇJCM

:5>-Mi-ooiogQ_i_im>-aQio

n n n.npn

OOOOOOOri-TT-i-ri-T-NNOjtM

-------------------------------

N CM CM CM CM CM CM CM ¿1 CM CM CM CM CM CM CM

s>-«i-ooio5Q-|-|w>-aQw

TM2-ECL1-TM3

Uy u ^

cirt^cpiaSaii

S.r^l^.r^r^r^-I^öcococococococc

&oocococoa»a»o»o}i------------------

tm cm cm cm cm cm cm cm n cm cm cm n si n co co co £0 co co so co

3—l>-UJQUl(50gf-l£zcnZ5Z>-g-

¿ùûix

yy gggyg

TM4-ECL2-TM5

r-~r^-r--r--r--r«-r--cocoooaocococooo

TM6-ECL3-TM7

DISCUSSION

Ligands binding to GPCRs modify the conformational landscape and thus stabilize a subset of conformational ensembles, providing the basis for both differential efficacy and biased ago-nism. There is limited information linking the dynamic events of receptor activation to engagement of specific effector proteins, and this is particularly true for class B GPCRs. The current study explores the molecular determinants for ligand affinity and engagement of signaling. Specifically, we highlight crucial surface residues within a class B GPCR that link initial peptide agonist interactions to distinct intracellular signaling pathways and biased agonism.

Peptide interactions with the extracellular surface and TM domains of class B GPCRs promote conformational transitions required to allow the binding of signaling effectors at the intracellular surface of these receptors. The ability of receptor mutants to affect signaling at only a single pathway highlights that different elements of the extracellular face are required for coupling to different effectors. In addition, differential effects on signaling by the three peptide agonists following mutation of individual residues supports the notion that the extracellular face of the receptor is important for initiating a switch in the

conformational landscape that the receptor explores, with different ligands capable of promoting/stabilizing alternative subsets of ensembles that lead to biased agonism.

Importance of the ECL Regions of the GLP-1R for Peptide Binding

Large portions of ECL2, ECL3 and the juxtamembrane positions of TM2/ECL1, ECL2/TM5 and TM6/ECL3 were important for molecular recognition of all peptide agonists but not for binding of the N-terminally truncated antagonist exendin-4(9-39). Despite the separation in sequence of these residues, they are all located together in 3D space. In addition, there is a network of residues provided by all three loops lining the cavity entrance in the TM bundle, and these residues are important for GLP-1 affinity, extending the peptide-binding groove from the N-terminal ECD into the TM domain cavity. These residues include L201, M204 (TM2/ ECL1), E294, W297, T298, R299, N300, Y305 (ECL2), R380, L384 (ECL3) (Figure S4). Interestingly, most of these residues appear to have indirect effects on agonist affinity as only five of these residues interact directly with GLP-1 in our model (L201, W297, R299, N300 and R380) (Figure S3).

The peptide-binding groove in the molecular model extends from the ECLs down into the TM bundle, forming a deep cavity

GLP-l Oxyntomodulin Exendin-4

Figure 6. Heatmap 3D Representation of the GLP-1R Extracellular Face Based on Efficacy Data from Three Different Signaling Assays

Molecular model of the GLP-1R-GLP-1 complex showing the extracellular surface of the TM bundle. All residues assessed In this study are shown In the center box; the locations of residues in ECL1, ECL2, and ECL3are highlighted in purple, orange, and blue, respectively. Residues that when mutated altered efficacy are highlighted in (A)-(H). (A-C) cAMP efficacy of GLP-1 (A), oxyntomodulin (B), and exendin-4 (C); (D and E) Ca2+ efficacy of GLP-1 (D) and exendin-4 (E); (F-H) pERK1/2 efficacy of GLP-1 (F), oxyntomodulin (G), and pERK1/2 (H). Teal indicates residues that were assessed and did not alter efficacy; yellow (3- to 5-fold), pale orange (5- to 10-fold), orange (10- to 30-fold), and red (>30-fold) are residues that statistically altered efficacy. The 3D heatmaps can be found in Active model S1 (Data S1).

Affinity

Figure 7. 3D Model Illustrating GLP-1R ECL Loop Residues that Are Globally or Selectively Important for GLP-1, Oxyntomodulin, and Exendin-4

Based on statistical significance (p <0.05) of effect when mutated to alanine, experimentally observed effects on peptide affinity and efficacy can be mapped onto the molecular model to clearly highlight similarities and differences between the three peptide agonists. Residues highlighted in red reduce function (affinity [A] or efficacy [B and C]) of all three peptides, those in pink selectively reduce GLP-1 only, those in yellow selectively reduce exendin-4 only, and those in green selectively reduce oxyntomodulin only. A large number of residues are important for both GLP-1 and exendin but not oxyntomodulin, and these are highlighted in orange. Other colors represent either enhanced function (GLP-1 only/oxyntomoduin only or GLP-1 and exendin) or existence of opposite effects when mutated on oxyntomodulin compared to GLP-1 and exendin-4.

lined by residues in all TMs except TM4 (Figure S4). In addition to the identified residues within the ECLs, published information on the requirement of other residues within this proposed cavity for GLP-1 affinity support our molecular model wherein GLP-1 enters into this cavity upon binding, with its N terminus residing deep within the helical bundle in the final ligand-docked model (Figures S3, S4, S5, and S6, Active model S1). Four of the residues that reside at the bottom of this pocket (R190260, N240343, E3646 53, and Q394749) form part of a hydrogen bond network in the inactive, unliganded receptor and are important for the binding and function of GLP-1 and exendin-4, though these residues also have roles in the biased agonism of these peptides (Wootten et al., 2013b, 2016). K197267 sits below L201 and W297 in 3D space and is important for GLP-1 binding and activation (Coopman et al., 2010). R310540 resides below N300 and also displays reduced potency in cAMP when mutated to alanine. Our model is also consistent with recent extensive studies on the GCGR (Siu et al., 2013) and CRF1R (Coin et al., 2013).

Despite the conservation of N-terminal sequence across all three peptides, oxyntomodulin appears to engage with the

GLP-1R in a manner that is significantly different from that of ex-endin-4 and GLP-1. Although this ligand also requires large portions of ECL2 and, although to a lesser degree, the residues L201 and L384 deep in the extracellular surface of the protein, it does not require some key residues lining the entry to the cavity, including C296/W297 in ECL2 and R380 in ECL3, that are crucial for affinity of the other two peptides. The deeper membrane-proximal residues of ECL2 and ECL3 are also less important, and other residues such as L218 in ECL1 play no role in the affinity of oxyntomodulin, and this implies that oxyntomodulin, which has a lower affinity than GLP-1 and exendin-4, may not bind in the same manner as the other two peptides. This is supported by previous data that show only a limited role of residues at the bottom of the binding cavity (R190 2 60, N2403 43, E3646 53, and Q394749) in oxyntomodulin affinity or cAMP formation (Wootten et al., 2013b, 2016). Indeed, a key predicted interaction in the GLP-1-GLP-1R model occurs between E9 (position 3) of the peptide and R190260 of the receptor. However, oxyntomodulin contains a Q at this position that would not be expected to form a salt bridge with R1902 60. Modified GLP-1 and oxyntomodulin peptides where the residue at position 3 is swapped

converts the behavior of these two peptides such that R1902 60 is required for cAMP production by the modified oxyntomodulin but not for the modified GLP-1 (Figure S3). This provides strong evidence validating the positioning of the N-terminal segment of GLP-1 in our molecular model, and extended molecular dynamics (MD) simulation indicates that this interaction is stable (Figures S5E-S5H).

Although the extreme N terminus of GLP-1 is predicted to interact in the deep cavity within the TM bundle, MD simulations where the peptide N terminus is placed in a superficial position in a model of the open inactive receptor predict that the peptide ligand initially makes interactions with the extracellular surface of the GLP-1R prior to movement of the peptide deeper into the cavity driven by E9 (Figure S5). In the open conformation, ECLs 2 and 3 reside further apart in 3D space (Figures S5A-S5D), suggesting that there is also a reorganization of the ECLs in response to peptide binding with ECL2 and ECL3 moving closer together in 3D space in the activated, ligand-occupied receptor (Figures S5E-S5H). Mapping of mutational data (affinity and efficacy) onto this surface formed by ECL2 and ECL3 reveals a continuous surface illustrating that this 3D surface is critical for stabilization of peptide binding and for activation of downstream effectors (Figures 6 and 7, Active model S1). Taken together with the extensive crosslinking/cysteine-trapping studies on other class B GPCRs (Coin et al., 2013; Dong et al., 2012), this supports a role for both interactions of peptide ligands with the extracellular loops and deeper interactions within the TM bundle, which are both important for peptide binding, leading to propagation of signaling in class B GPCRs.

Importance of the ECL Regions of the GLP-1R for Efficacy

Overall, there was a very high correlation between residues important for peptide affinity and those linked to efficacy for cAMP formation and Ca2+ mobilization (Figures 2, 6, and 7). This is perhaps not surprising as both cAMP and Ca2+ mobilization are predominantly G protein-mediated pathways (Figure S1), and the ternary complex of the agonist-occupied receptor and effector (e.g., G protein) provides thermodynamically reciprocal regulation of agonist binding (De Lean et al., 1980). As such, the heatmaps of mutant effects on agonist affinity are a composite of direct effects on binding and those allosterically imposed via the effects of effector coupling, in particular, G protein coupling. For GLP-1 and exendin-4, unlike effects on affinity, almost the entire region of ECL2 is required for transmission of signal. Moreover, the contribution of individual residues varies between the different functional measures and the two ligands. Exendin-4 and GLP-1 display a similar efficacy for coupling to cAMP; however, exendin-4 is less efficient than GLP-1 at coupling to Ca2+ mobilization (Figure S2). Interestingly, the Ca2+ response mediated by exendin-4 is more sensitive to mutations within ECL2, ECL3, and TM2 membrane-proximal regions of ECL1 than that mediated by GLP-1, perhaps suggesting that subtle differences in the interactions formed by these li-gands account for the small distinctions in signaling bias that are observed experimentally. These subtle differences in bias and the effect of mutations may be reflective of the nature of effector coupling that drives stimulation of individual signaling

pathways. This is observed in inhibitor studies where relatively subtle differences were observed between GLP-1 and exen-din-4, most notably in the relative contribution of Gpg subunits to pERK1/2 and iCa2+ signaling (Figure S1).

In addition to distinctions in the pattern of residues required for oxyntomodulin affinity compared to GLP-1 and exendin-4, there are also significant differences in the pattern of residues important for coupling the receptor to cAMP. Like the other two peptides, oxyntomodulin utilizes ECL2 and membrane-proximal regions of ECL3 for its function, and there is also evidence for the involvement of deeper residues in ECL2 and ECL3 that were not required for its affinity. However, in contrast to GLP-1 and exendin-4, there is a large involvement of residues in ECL1 and no requirement for residues in the proximal region of ECL2. Oxyntomodulin displays a very distinct signaling profile to GLP-1 (Figure S2), and collectively, the affinity and cAMP data support the notion that oxyntomodulin does not interact in the same manner, requiring a much larger portion of the extracellular surface to engender conformational transitions linking peptide interactions to signaling inside the cell. Furthermore, in contrast to GLP-1 and exendin-4, a component of the oxyntomodulin-mediated cAMP production is dependent on Gpg subunits, suggesting perhaps that a different subset of adenylate cyclases are activated to generate this cAMP response.

In addition to the network of interconnected residues along the extracellular surface of the receptor, a number of more distal residues were also identified to contribute to signaling efficacy. Studies of the GCGR have suggested that interactions between the receptor ECD and ECLs can occur and that these can influence conformational transitions required for signaling (Koth et al., 2012). It is likely that similar interactions also occur for the GLP-1R, and these may account for the observed effects of some residues distal to the interconnected networks. Extended MD simulation of the full-length receptor is consistent with the potential for such interactions to occur (Movies S1 and S2).

Interactions Determining Activation of Distinct Signaling Pathways and Promotion of Biased Signaling

Regardless of the ligand, ECL3 is essential for coupling the pep-tide-receptor interaction to pERK1/2, whereas ECL2 is critical for coupling of these interactions to cAMP and Ca2+. Using inhibitors to disrupt various G protein- and b-arrestin-mediated signaling pathways revealed that whereas cAMP and Ca2+ mobilization are predominantly driven by G protein-mediated signaling, pERK1/2 is a composite of both G protein- and p-ar-restin-driven events, with approximately 30%-60% (depending on the peptide) of the signal attributed to this latter mechanism (Figure S1). As the signaling events leading to pERK1/2 are partially independent of G proteins, this may explain why there is a very distinct region of the receptor required for signaling to this pathway in comparison to Ca2+ and cAMP signaling, which are predominantly G protein mediated.

In addition to heatmapping, the data from this study can also be mapped onto models depicting global importance of residues across sets of ligands or importance of residues for individual ligands. In this way, it is easy to visually observe distinct regions of

the receptor that can achieve biased agonism. In Figure 7, which summarizes the three measures of function with data for all three peptides, the differential importance of regions in the extracellular face for oxyntomodulin compared to GLP-1 and exendin-4 can be clearly observed. Oxyntomodulin is a highly biased ligand compared to GLP-1 (and exendin-4), with bias toward both pERK1/2 signaling and regulatory protein recruitment (including p-arrestins) (Figure S2), biases that may be linked given the greater contribution of p-arrestins to the pERK1/2 response of oxyntomodulin relative to the other two peptides (Figure S1). Therefore, it is not surprising that quite a different pattern of residues are required for transmission of signal for this peptide. Compared to GLP-1, exendin-4 only has very minor bias in its signaling profile for the pathways assessed in the current study, and therefore it is not necessarily surprising that the residues important at the extracellular face for each of these peptides to signal to various pathways are similar, even if (as the heatmapping suggests) the extent to which each of these residues contributes may be different.

Combining this study with previously published quantitative information of residues that also contribute to conformational changes associated with activation provides additional context to how these surface interactions link with intramembranous networks to differentially control signaling (Figure S6). This type of information allows us to begin to understand how initial peptide interactions at the extracellular surface engage with distinct networks of intramembranous residues to link extracellular binding to engagement of intracellular effectors.

Mapping mutational effects on efficacy onto molecular models may also provide a basis for rational design of biased peptide agonists. If the required cellular efficacy for translation to therapeutic success is known, then this provides an ability to design peptides that exploit signaling bias therapeutically. This hypothesis can be tested to some extent using the metabolite of GLP-1, GLP-1(9-36)NH2. Whereas this ligand lacks the first two amino acids of GLP-1, including the N-terminal His7 that is critical for affinity and activation of cAMP (Adelhorst et al., 1994), it retains mid-regions of the peptide that, in our molecular model, interact with ECL3. Although unable to activate the cAMP pathway, the metabolite can still promote pERK1/2 (Figure S7). This suggests that peptides that maintain interactions with ECL3 while altering/removing interactions deeper in the bundle and perhaps with ECL2 could bias ligands toward pERK1/2; however, it remains to be seen whether ECL3 is a common activation domain for this subclass of receptors to promote coupling to p-arrestin-mediated signaling pathways such as pERK1/2.

Biased agonism is likely to be a crucial element of the function of class B GPCRs, as many of them can be activated by multiple endogenous ligands and their receptors are capable of activating multiple intracellular signaling pathways. Therefore, understanding the molecular determinants linking ligand interactions to activation of distinct signaling pathways, in addition to the physiological benefit of activating individual pathways, could have significant ramifications for future drug development and may provide the potential to rationally design future drug therapies, and this is highlighted by the in vivo actions of the biased GLP-1 agonist, P5 (Zhang et al., 2015).

EXPERIMENTAL PROCEDURES Molecular Biology

We used Quikchange (Stratagene) to introduce mutations into the GLP-1R cloned into the pEF5/Frt/V5-Dest vector.

Cell Culture

Stable FlpIn CHO cell lines were generated using Gateway technology. For all assays, cells were seeded in 96-well plates at a density of 30,000 cells per well.

Radioligand-Binding Assays

Whole-cell competition radioligand binding was performed using 125I-exendin-4(9-39) as the tracer ligand and competing with increasing concentrations of unlabeled peptide ligands as described previously (Koole et al., 2010).

Cell-Surface Expression

Cell-surface expression was detected either by using a cell-surface ELISA to detect a double c-Myc epitope label incorporated with the N-terminal region of the GLP-1R constructs or by calculation of the Bmax in the radioligand-binding experiments (Koole et al., 2012; Wootten et al., 2016).

Signaling Assays

For cAMP assays, cells were stimulated for 30 min in the presence of the phosphodiesterase inhibitor IBMX and then lysed. For pERK1/2, cells were stimulated for 6 min (the peak of the response) before lysis. Detection of cAMP and pERK1/2 in the lysates was performed using Alphascreen technology as previously described (Koole etal., 2010). Ca2+ mobilization was detected using a Fluo-4-AM dye immediately after ligand addition with an excitation wavelength of 485 nM and an emission wavelength of 520 nM with values derived from the peak response.

Data Analysis

Concentration response data were analyzed using a three-parameter logistic equation to determine affinity, EC50, and Emax values. Efficacy was calculated by applying the operational model of agonism:

Y = Bottom +

Em — Bottom

1 + ((IQ'osKa ) + (1QtogM ))/(10(/ogr +

where Bottom represents the y value in the absence of ligand, Em is maximal system stimulation, KA is the agonist-receptor dissociation constant, [A] is the ligand concentration, and t is the operational measure of efficacy in the system, which incorporates signaling efficacy and receptor density. Derived t values were corrected to cell-surface expression (tc) measured by ELISA, and errors were propagated from both t and cell-surface expression.

Molecular Modeling

Energy-based conformational modeling of the GLP-1R complex with GLP-1 was performed with Modeler 9.15, and peptide docking and energy optimization were guided by published experimental data (Tables S5 and S6), as previously described (Wootten et al., 2016).

Detailed procedures and analysis are reported in the Supplemental Experimental Procedures.

SUPPLEMENTAL INFORMATION

Supplemental Information includes Supplemental Experimental Procedures, seven figures, six tables, one data file, and two movies and can be found with this article online at http://dx.doi.org/10.1016Zj.cell.2016.05.023.

AUTHOR CONTRIBUTIONS

Conceptualization, D.W., J.S., P.M.S.; Methodology, D.W., C.A.R., K.J.S., J.C.M., M.L., P.E.T., P.M.S.; Software, C.A.R., K.J.S., J.C.M.; Formal Analysis,

D.W., C.K., A.C., P.M.S.; Investigation, D.W., C.A.R., K.J.S., J.C.M., C.K.,

E.E.S., K.P., M.L., J.S., R.S.; Writing - Original Draft, D.W., C.A.R., L.J.M., A.C., P.M.S.; Writing - Review and Editing, C.K., S.G.B.F.; Visualization,

D.W., P.M.S.; Supervision, D.W., C.A.R., P.E.T., P.M.S.; Project Administration, D.W., P.M.S.; Funding Acquisition, D.W., C.A.R., A.C., P.M.S.

ACKNOWLEDGMENTS

This workwas supported by National Health and Medical Research Council of Australia (NHMRC) project grants (1061044 and 1065410) and NHMRC program grant (1055134); Biotechnology and Biological Sciences Research Council of the United Kingdom project grant (BB/M006883/1); and Medical Research Council (UK) Research Leader Fellowship (G1001812) (C.A.R.). D.W. is a NHMRC Career Development Fellow. P.M.S. is a NHMRC Principal Research Fellow, and A.C. is a NHMRC Senior Principal Research Fellow. C.K. is a NHMRC CJ Martin Postdoctoral Fellow.

Received: December 1, 2015 Revised: February 16, 2016 Accepted: May 3, 2016 Published: June 16, 2016

REFERENCES

Adelhorst, K., Hedegaard, B.B., Knudsen, L.B., and Kirk, O. (1994). Structure-activity studies of glucagon-like peptide-1. J. Biol. Chem. 269, 6275-6278. Baggio, L.L., and Drucker, D.J. (2007). Biology of incretins: GLP-1 and GIP. Gastroenterology 132, 2131-2157.

Barwell, J., Conner, A., and Poyner, D.R. (2011). Extracellular loops 1 and 3 and their associated transmembrane regions of the calcitonin receptor-like receptor are needed for CGRP receptor function. Biochim. Biophys. Acta 1813, 1906-1916.

Bergwitz, C., Jusseaume, S.A., Luck, M.D., Juppner, H., and Gardella, T.J. (1997). Residues in the membrane-spanning and extracellular loop regions of the parathyroid hormone (PTH)-2 receptor determine signaling selectivity for PTH and PTH-related peptide. J. Biol. Chem. 272, 28861-28868. Black, J.W., and Leff, P. (1983). Operational models of pharmacological ago-nism. Proc. R. Soc. Lond. B Biol. Sci. 220, 141-162.

Coin, I., Katritch, V., Sun, T., Xiang, Z., Siu, F.Y., Beyermann, M., Stevens, R.C., and Wang, L. (2013). Genetically encoded chemical probes in cells reveal the binding path of urocortin-I to CRF class B GPCR. Cell 155, 1258-1269. Coopman, K., Huang, Y., Johnston, N., Bradley, S.J., Wilkinson, G.F., and Willars, G.B. (2010). Comparative effects of the endogenous agonist glucagon-like peptide-1 (GLP-1)-(7-36) amide and the small-molecule ago-allosteric agent "compound 2'' at the GLP-1 receptor. J. Pharmacol. Exp. Ther. 334,795-808. Couvineau, A., and Laburthe, M. (2012). The family B1 GPCR: structural aspects and interaction with accessory proteins. Curr. Drug Targets 13, 103-115.

De Lean, A., Stadel, J.M., and Lefkowitz, R.J. (1980). A ternary complex model explains the agonist-specific binding properties of the adenylate cyclase-coupled beta-adrenergic receptor. J. Biol. Chem. 255, 7108-7117. Dong, M., Xu, X., Ball, A.M., Makhoul, J.A., Lam, P.C., Pinon, D.I., Orry, A., Sexton, P.M., Abagyan, R., and Miller, L.J. (2012). Mapping spatial approximations between the amino terminus of secretin and each of the extracellular loops of its receptor using cysteine trapping. FASEB J. 26, 5092-5105.

Fredriksson, R., Lagerstrom, M.C., Lundin, L.G., and Schioth, H.B. (2003). The G-protein-coupled receptors in the human genome form five main families. Phylogenetic analysis, paralogon groups, and fingerprints. Mol. Pharmacol. 63, 1256-1272.

Hollenstein, K., Kean, J., Bortolato, A., Cheng, R.K., Dore, A.S., Jazayeri, A.,

Cooke, R.M., Weir, M., and Marshall, F.H. (2013). Structure of class B GPCR

corticotropin-releasing factor receptor 1. Nature 499, 438-443.

Katritch, V., Cherezov, V., and Stevens, R.C. (2013). Structure-function of the

G protein-coupled receptor superfamily. Annu. Rev. Pharmacol. Toxicol. 53,

531-556.

Kenakin, T., and Christopoulos, A. (2013). Signalling bias in new drug discovery: detection, quantification and therapeutic impact. Nat. Rev. Drug Discov. 12, 205-216.

Koole, C., Wootten, D., Simms, J., Valant, C., Sridhar, R., Woodman, O.L., Miller, L.J., Summers, R.J., Christopoulos, A., and Sexton, P.M. (2010). Allosteric ligands of the glucagon-like peptide 1 receptor (GLP-1R) differentially modulate endogenous and exogenous peptide responses in a pathway-selective manner: implications for drug screening. Mol. Pharmacol. 78, 456-465. Koole, C., Wootten, D., Simms, J., Miller, L.J., Christopoulos, A., and Sexton, P.M. (2012). Second extracellular loop of human glucagon-like peptide-1 receptor (GLP-1R) has a critical role in GLP-1 peptide binding and receptor activation. J. Biol. Chem. 287, 3642-3658.

Koth, C.M., Murray, J.M., Mukund, S., Madjidi, A., Minn, A., Clarke, H.J., Wong, T., Chiang, V., Luis, E., Estevez, A., et al. (2012). Molecular basis for negative regulation of the glucagon receptor. Proc. Natl. Acad. Sci. USA 109,14393-14398.

Overington, J.P., Al-Lazikani, B., and Hopkins, A.L. (2006). How many drug targets are there? Nat. Rev. Drug Discov. 5, 993-996.

Pal, K., Melcher, K., and Xu, H.E. (2012). Structure and mechanism for recognition of peptide hormones by class B G-protein-coupled receptors. Acta Pharmacol. Sin. 33, 300-311.

Reid, T.S. (2013). Practical use of glucagon-like peptide-1 receptor agonist therapy in primary care. Clin. Diabetes 31, 148-157.

Siu, F.Y., He, M., de Graaf, C., Han, G.W., Yang, D., Zhang, Z., Zhou, C., Xu, Q., Wacker, D., Joseph, J.S., et al. (2013). Structure of the human glucagon class B G-protein-coupled receptor. Nature 499, 444-449.

Wootten, D., Savage, E.E., Willard, F.S., Bueno, A.B., Sloop, K.W., Christopoulos, A., and Sexton, P.M. (2013a). Differential activation and modulation of the glucagon-like peptide-1 receptor by small molecule ligands. Mol. Pharmacol. 83, 822-834.

Wootten, D., Simms, J., Miller, L.J., Christopoulos, A., and Sexton, P.M. (2013b). Polar transmembrane interactions drive formation of ligand-specific and signal pathway-biased family B G protein-coupled receptor conformations. Proc. Natl. Acad. Sci. USA 110, 5211-5216.

Wootten, D., Reynolds, C.A., Koole, C., Smith, K.J., Mobarec, J.C., Simms, J., Quon, T., Coudrat, T., Furness, S.G.B., Miller, L.J., et al. (2016). A hydrogen-bonded polar network in the core of the glucagon-like peptide-1 receptor is a fulcrum for biased agonism: Lessons from class B crystal structures. Mol. Pharmacol. 89, 335-347.

Zhang, H., Sturchler, E., Zhu, J., Nieto, A., Cistrone, P.A., Xie, J., He, L., Yea, K., Jones, T., Turn, R., et al. (2015). Autocrine selection of a GLP-1R G-protein biased agonist with potent antidiabetic effects. Nat. Commun. 6, 8918.