Main

Class B peptide G-protein-coupled receptors (GPCRs) regulate the control of glucose and energy homeostasis, bone turnover, and cardiovascular development and tone1. Several peptide agonists are clinically approved for disorders of energy and bone metabolism1; however, attempts to develop non-peptide, orally available analogues have yielded only limited success. Understanding the structural basis of class B GPCR activation is crucial to the rational development of peptidic and non-peptidic drugs. Recent structural determination of full-length, active class B receptors bound to peptide agonists2,3,4,5,6 confirmed that the N terminus of the peptide ligands, required for receptor activation, binds deep within the seven-transmembrane helical bundle. This is associated with an outward movement of the tops of transmembrane helices (TM) 6 and 7 (and interconnecting extracellular loop (ECL) 3) and a large kink in the centre of TM6 that opens up the intracellular face of the receptor to allow G-protein coupling2,3,4,7,8,9,10. In parallel, a conformational reorganization of ECL2 and an inward movement of TM1 facilitates peptide interaction and receptor activation.

The GLP-1 receptor (GLP-1R) is an established therapeutic target for type 2 diabetes and obesity11. Despite their clinical success, GLP-1R peptide drugs are suboptimal owing to their route of administration and side-effect profiles, most notably nausea and vomiting that reduce patient compliance11. For many years, oral GLP-1R agonists have been pursued, with recent studies reporting promising clinical trial data for oral semaglutide—a new formulation of the approved peptide semaglutide12,13. However, it induced slightly greater severity of nausea and gastrointestinal side effects than those observed with injectable GLP-1 mimetics13. Future development of non-peptide drugs could offer more traditional small molecule absorption characteristics that may assure better long-term patient compliance with the potential for reduced gastrointestinal liability, especially for patients who are co-administering with other medications.

Several non-peptidic GLP-1R agonists have been identified14. One class form covalent interactions with C3476.36 (in which the superscript denotes the Wootten class B GPCR numbering) and are predicted to allosterically disrupt polar networks at the base of the receptor, promoting activation15, whereas other small molecule compounds bind to unknown sites at the receptor extracellular face14,16,17. However, it is assumed that these molecules may need to mimic key interactions of the peptide N terminus deep within the transmembrane core to initiate receptor activation, as is seen for short stabilized 11-mer peptides, that occupy an overlapping site to full-length peptides18.

Here we investigate TT-OAD2 (Fig. 1a), a non-peptidic compound reported in the patent literature and part of the chemical series that contains the vTv Therapeutics investigational drug candidate, TTP273. TTP273, an orally administered GLP-1R agonist, successfully completed phase IIa efficacy trials for type 2 diabetes (ClinicalTrials.gov Identifier: NCT02653599), in which it met its primary endpoint, reducing levels of glycated haemoglobin in patients with type 2 diabetes, with no reported cases of nausea19, suggesting a potential clinical advantage for compounds of this series. Little has been disclosed about the molecular properties of this compound series; however, recent progression of TTP273 has been hampered by unexpected complexity in identifying optimal dosing that may be linked to a lack of understanding of its mechanism of action. Assessment of acute in vivo activity in humanized GLP-1R mice revealed that TT-OAD2 is insulinotropic and that this effect is dependent on the GLP-1R (Fig. 1b).

Fig. 1: Pharmacology exhibited by TT-OAD2 relative to GLP-1.
figure 1

a, Chemical structure of TT-OAD2. b, Plasma insulin induced by GLP-1 (10 μg kg−1), TT-OAD2 (3 mg kg−1) or gastric inhibitory polypeptide (GIP; 25 μg kg−1) in an acute IVGTT on humanized GLP-1R knock-in (KI) and GLP-1R knockout (KO) mice. c, Whole-cell binding assays showing the ability of GLP-1 and TT-OAD2 to displace 125I-exendin(9-39). d, cAMP accumulation, intracellular calcium mobilization, β-arrestin-1 recruitment and ERK1/2 phosphorylation (pERK1/2). e, Agonist-induced changes in trimeric Gs conformation in cell plasma membrane preparations for GLP-1 (left) and TT-OAD2 (middle). Rates (top right) and plateau (bottom right) at saturating concentrations (1 μM GLP-1, 10 μM TT-OAD2) were quantified by applying a one-phase association curve. f, Kinetics of cAMP production measured by an EPAC biosensor for GLP-1 (left) and TT-OAD2 (middle). Rates were quantified using approximate EC50 and Emax concentrations (1 nM and 0.1 μM for GLP-1, 0.1 μM and 10 μM for TT-OAD2) by applying a one-phase association curve. In e and f, arrows refer to the time at which ligand or vehicle was added. Parameters derived from kinetic data are represented as scatter plots with each individual experiment shown by black circles. All experiments were performed in GLP-1R expressing HEK293A cells. Data in b are mean + s.e.m. from 4–5 mice per treatment, representative of 3 independent experiments. Data in c–f are mean + s.e.m. of 4–5 independent experiments (in duplicate or triplicate). *P < 0.05, Student’s paired t-test.

Full size image

TT-OAD2 is a biased agonist with slow kinetics

In HEK293 cells that overexpress GLP-1R, TT-OAD2 only partially displaced the orthosteric probes 125I-exendin(9–39) and ROX-exendin-4 (Fig. 1c, Extended Data Fig. 1a), consistent with an allosteric mode of interaction16. Although GLP-1R signals to several cellular pathways, TT-OAD2 activated only a subset of these responses; it was a low-potency partial agonist for cAMP accumulation, with only weak responses detected for mobilization of intracellular Ca2+ and phosphorylation of ERK1/2 at very high concentrations (100 μM) (Fig. 1d) and no detectable recruitment of β-arrestin-1. These data are indicative of bias towards cAMP and away from these other pathways relative to endogenous GLP-1. There is considerable interest in exploiting biased agonism at GPCRs to maximize the beneficial effects of receptor activation, while minimizing on-target side-effect profiles.

CRISPR-engineered HEK293 cells in which Gs/olf or Gi/o/z proteins were deleted revealed that Gs was essential for the production of cAMP; however, this response, for both ligands, was also dependent on the presence of Gi/o/z proteins. (Extended Data Fig. 1b). Assessment of proximal activation of Gs and Gi transducers using split luciferase NanoBit G-protein sensors (Extended Data Fig. 1c) determined GLP-1-decreased luminescence in a bi-phasic, concentration-dependent, manner for both G proteins with similar potencies in each phase. For TT-OAD2, the Gi sensor gave a similar decrease in luminescence to GLP-1; however, enhanced luminescence was observed for the Gs sensor, which suggests a different mechanism of Gs activation. To probe these differences further, we used membrane-based assays of bioluminescence resonance energy transfer (BRET) G-protein sensors to assess the rate and nature of the Gs conformational change. In contrast to the rates of change in the conformation of Gi, which were similar for both ligands (Extended Data Fig. 1), there was a marked distinction in kinetics for Gs coupling. GLP-1 promoted a rapid conformational change in Gs protein, whereas for TT-OAD2 this was very slow (Fig. 1e). However, both agonists induced a similar plateau of the measured response (Fig. 1e) that was reversed by excess GTP (Extended Data Fig. 1d), indicative of a similar overall conformational rearrangement. Together, this suggests that slower Gs conformational transitions, required for the exchange of GDP for GTP and Gs activation, would result in lower turnover of G protein and rate of cAMP production by TT-OAD2. Direct kinetic measurements of cAMP production validated this hypothesis (Fig. 1f, Extended Data Fig. 1e). Overall, these data revealed TT-OAD2 as a biased agonist that can only activate a subset of pathways with limited efficacy and with distinct activation kinetics relative to peptide agonists.

TT-OAD2 has an unexpected binding mode

To understand how TT-OAD2 binds and activates the GLP-1R, we determined the GLP-1R structure bound to TT-OAD2 and the transducer heterotrimeric Gs protein (Fig. 2). Complex formation was initiated in Tni insect cells by stimulation with 50 μM TT-OAD2, and complexes were then solubilized and purified (Extended Data Fig. 2a). Vitrified complexes were imaged by single-particle cryo-electron microscopy (cryo-EM) on a Titan Krios. Following 2D and 3D classification, the most abundant class was resolved to 3.0 Å (Extended Data Fig. 2c–f, Supplementary Table 1). The cryo-EM density map allowed unambiguous assignment of the TT-OAD2-binding site and pose, and clear rotamer placement for most amino acids within the receptor core and G protein (Fig. 2, Extended Data Figs. 3, 4a, b). The GLP-1R extracellular domain (ECD) and the Gαs α-helical domain were not resolved at high resolution, consistent with their greater mobility. Rigid body fitting of an available X-ray structure of the GLP-1R ECD domain (PDB code 3C5T)20 was performed into the density to generate a full-length model.

Fig. 2: TT-OAD2–GLP-1R–Gs cryo-EM structure reveals non-peptide binding site.
figure 2

Top, orthogonal views of the TT-OAD2–GLP-1R–Gs complex cryo-EM map (left) and the structure after refinement in the cryo-EM map (right), colour-coded to protein chains; GLP-1R (blue), TT-OAD2 (red), heterotrimeric Gs (α: gold, β: dark cyan, γ: purple, Nb35: salmon). Middle, TT-OAD2 interacts with the top of the GLP-1R bundle. Interacting residues of GLP-1R (blue) with TT-OAD2 (red). Bottom, TT-OAD2-mediated cAMP production by receptors containing alanine mutants of key residues assessed in ChoFlpIn cells. Data are mean + s.e.m. of four independent experiments performed in duplicate. WT, wild type.

Full size image

TT-OAD2 bound high up in the helical bundle interacting with residues within TM1, TM2, TM3, ECL1 and ECL2 (Fig. 2, Extended Data Fig. 4a). Most interactions are hydrophobic in nature (Fig. 2), including numerous π–π stacking interactions between receptor aromatic residues and phenolic regions within the ligand. Unexpectedly, TT-OAD2 adopts a ‘boomerang-like’ orientation within the binding site with the 3,4-dichloro-benzyl ring of TT-OAD2 protruding beyond the receptor core through transmembrane helices 2 and 3, interacting with W2032.73, and embedding in the detergent micelle, consistent with probable interactions with the lipid bilayer in a native system. F2303.33 and W297ECL2 interact with the 2,3-dimethyl-pyridin-4-yl-phenol region, Y220ECL1 forms a hydrogen bond with the 2,3-dimethyl-pyridine ring and K1972.67 forms a polar interaction with the propionic acid part of the ligand. Additional hydrophobic contacts are formed with TT-OAD2 by Y1451.40, L2012.71, I1962.69, A2002.70, L217ECL1, V2293.32 and M2043.36 (Fig. 2, Extended Data Fig. 4a). Molecular dynamics simulations of the TT-OAD2–GLP-1R–Gs complex predicted further transient interactions with TM1, TM2, TM3, ECL1, ECL2 and the ECD of GLP-1R (Extended Data Table 1). Assessment of TT-OAD2-induced cAMP production at alanine mutants of key receptor residues within the binding site revealed reduced potency (negative logarithm of the half-maximal effective concentration, pEC50), reduced maximal responses (Emax) or both relative to the wild-type receptor (Fig. 2, Supplementary Table 2). Application of the operational model of agonism revealed these mutations directly alter TT-OAD2 functional affinity (KA) and/or efficacy (τ) (Supplementary Table 2), which highlights the importance of these residues in TT-OAD2 function.

Peptide versus non-peptide binding sites

The TT-OAD2-binding pose has very limited overlap with full-length peptides, GLP-1 and exendin-P5 (ExP5)3,6 (Fig. 3, Extended Data Fig. 5). Structural comparisons, combined with associated molecular dynamics simulations performed on models generated from the cryo-EM data, identified only 10 out of 29 residues that interact with both TT-OAD2 and GLP-1. Moreover, the persistence and nature of ligand interactions formed by common residues differed (Fig. 3c, Extended Data Table 1). In contrast to TT-OAD2, peptide ligands engage transmembrane helices 5–7 in addition to extensive interactions deep within the bundle in transmembrane helices 1–3 (Fig. 3, Extended Data Fig. 5, Extended Data Table 1).

Fig. 3: Comparisons of GLP-1R conformations induced by GLP-1 and TT-OAD2.
figure 3

a, b, Superimposition of the GLP-1R from PDB 5VAI (GLP-1R or G protein: orange, GLP-1: green) and the TT-OAD2 structure (GLP-1R or G protein: blue, TT-OAD2: red) reveals partial overlap of peptide- and TT-OAD2-binding sites and conformational differences in the receptor. a, Left, full complex; middle, close up of ECD and the top of the seven-transmembrane (7-TM) bundle; right, close up of the transmembrane bundle. b, Left, 16 Å, 7 Å and 6 Å differences occur in the location of TM6/ECL3, TM7 and TM1, respectively. Middle, a 4 Å shift in the location of the top of TM2 result in distinct conformations of ECL1. Right, the intracellular region of the GLP-1R helical bundles have similar overall backbone conformations. c, Comparison of the GLP-1R–TT-OAD2 and GLP-1R–GLP-1 contacts during molecular dynamics simulations performed on the GLP-1R–TT-OAD2–Gs and GLP-1R–GLP-1–Gs complexes. Top (left) and side (right) views of the GLP-1R transmembrane domain (ribbon representation, TT-OAD2 in red sticks, GLP-1 not shown). TT-OAD2 made contacts (red coloured ribbon) with ECL1 and residues located at the top of TM2 and TM3. GLP-1 was able to engage TM5, TM6 and TM7 of the receptor and side chains located deep in the bundle (blue coloured ribbon). Residues that are involved both in the GLP-1R–TT-OAD2–Gs and GLP-1R–GLP-1–Gs complexes are indicated by asterisks, and coloured according to the algebraic difference in occupancy (contact differences in percentage frames) between GLP-1R–TT-OAD2–Gs and GLP-1R–GLP-1–Gs. Red indicates regions more engaged by TT-OAD2 and blue more engaged by GLP-1. The ECD is not shown. Plotted data are summarized in Extended Data Table 1.

Full size image

The relatively limited overlap between the peptide- and TT-OAD2-binding sites suggests that this compound series may modulate peptide function in a physiological setting. To address this, we assessed the effect of TT-OAD2 on the signalling of two physiological ligands (Extended Data Fig. 6). TT-OAD2 inhibited GLP-1- and oxyntomodulin-mediated cAMP, calcium, pERK1/2 and β-arrestin responses in a concentration-dependent manner (Extended Data Fig. 6). This suggests that the profile of signalling observed from the GLP-1R when using TT-OAD2-like compounds as drugs may depend on the dose administered; at high concentrations, their presence would probably inhibit all endogenous peptide effects, biasing receptor responses primarily to cAMP formation mediated by the compound itself. However, at lower concentrations, some endogenous peptide signalling may still occur. Notably, TTP273 was reported to exhibit greater clinical efficacy at lower concentrations, indicating that maintenance of some aspects of physiological signalling may be important for clinical efficacy19.

GLP-1R conformational changes and activation

At a gross level, the TT-OAD2-complexed GLP-1R helical bundle displays the key hallmarks of activated, peptide-occupied, class B GPCRs2,3,4,5,6. At the extracellular face, this includes the large outward movement of TM6, ECL3 and TM7, inward movements of TM1, helical extensions within TM2 and TM3, a reordering of ECL1, and conformational transitions within ECL2 that increases upward towards the extracellular side (Extended Data Fig. 5). At the intracellular side, there is an equivalent large outward movement of TM6 away from the centre of the helical bundle, and the smaller outward movement of TM5. It is important to note that the fully active state is driven in part by allosteric conformational changes, including those in the extracellular face, linked to G protein binding21. Nonetheless, all the GLP-1R structures are solved with the same G protein yet reveal conformational differences at their extracellular face, including within the extent of movement of TM6, ECL3, ECL7 and the conformation of the ECD, TM2–ECL1 and ECL2 that are linked to the bound agonists (Fig. 3a, b, Extended Data Fig. 5b, c). This suggests that distinct receptor activation triggers converge to common changes at the intracellular face that allow coupling to transducers.

Although the low resolution of the receptor ECD for the TT-OAD2 complex indicates extensive mobility, it occupied a distinct orientation relative to the transmembrane core in comparison to peptide-bound complexes, whereas both GLP-1- and ExP5-bound receptors stabilized a similar conformation3,6 (Extended Data Fig. 5a). Similarly, the short 11-mer peptide HepP5 forms few interactions with the ECD18 and occupies a distinct orientation relative to GLP-1 and ExP5, but this conformation also differs from that stabilized by TT-OAD2 (Extended Data Fig. 5c). The cryo-EM map of the TT-OAD2-bound receptor complex supports extended interactions of the ECD with ECL1 and ECL2 (Extended Data Fig. 4c) and this is supported by molecular dynamics simulations that predicts interactions of R40ECD with D215ECL1 and E34ECD with R299ECL2 (Extended Data Table 2). This later interaction is particularly important as R299ECL2 directly, and stably interacts with peptide ligands, but in the TT-OAD2-bound receptor, stabilizes the N terminus of the ECD in a position that may have an analogous role to the peptide in stabilizing ECL2. Indeed, in our models, the position of the far N-terminal ECD helix overlapped with the location of the C-terminal region of GLP-1 and ExP5 when comparing the TT-OAD2- and peptide-bound structures (Fig. 3a). Thus, the ECD is likely to be important for both stabilizing the TT-OAD2-binding site and facilitating receptor activation, as previously proposed for different classes of peptide ligands22,23.

Distinctions from peptide-bound receptors observed within TM2/ECL1 and ECL2 (Fig. 3b) are probably driven by direct ligand interactions by TT-OAD2 (Fig. 2), whereas those within TM6 and TM7 by direct interactions formed by peptide agonists. Molecular dynamics simulations also support a role of membrane lipid interactions in directly stabilizing both these regions within the TT-OAD2-bound structure (Extended Data Fig. 7). Notably, the helical bundle of the TT-OAD2-complexed receptor is in a more open conformation than the peptide-occupied receptors, largely owing to the top of TM6/ECL3, TM7 and TM1 residing 16 Å, 6 Å and 7 Å further outwards relative to the GLP-1-bound structure (measured from the Cα atoms of D3726.62/ECL3, F3817.37 and P1371.32, respectively (Fig. 3b). The orientation of TM6, ECL3 and TM7 also differs between ExP5- and GLP-1-bound structures, with ExP5 adopting a more open conformation3; however, the outward positioning of ECL3 induced by TT-OAD2 is much larger (Extended Data Fig. 5b). Peptide-bound structures of all solved class B GPCRs revealed direct interactions of the engaged peptide with residues within TM5, TM6, TM7 and ECL3 with the peptide volume (minimally) presumed to actively contribute to the outward conformational change in this region2,3,4,8,9,24. In the apo-state of the glucagon receptor, interactions occur between ECL3 and the ECD that contribute to maintenance of receptor quiescence7,8,25,26. Molecular dynamics simulations on the GLP-1R structures, performed after the removal of either TT-OAD2 or GLP-1, predict that the GLP-1R ECD also adopts both open and closed conformations in the apo-state, in which it can form transient interactions with both ECL2 and ECL325 (Extended Data Fig. 8). Combining this information with the GLP-1R active structures suggests that interactions, with either peptide or non-peptide agonists, can release ECL3-ECD constraints, lowering the energy barrier for receptor activation. However, the degree of ligand interaction with TM6–ECL3–TM7 determines the extent to which the transmembrane bundle opens, and this in turn directly contributes to G-protein efficacy and biased agonism, as these regions (TM6–ECL3–TM7 and TM1) have been identified as key drivers for these phenomena, particularly for the GLP-1R3,27,28,29.

Despite the different binding modes, commonalities observed in interactions with TT-OAD2 and peptide with transmembrane helices 1–3 and stabilization of ECL2 are sufficient to initiate conformational transitions that propagate to a similar reorganization of the class B GPCR conserved central polar network that is linked to activation, albeit the mechanism for this differs for peptide agonists versus TT-OAD2 (Fig. 4a, Supplementary Video 1, Extended Data Fig. 9). Molecular dynamics simulations of the GLP-1-bound GLP-1R predicted persistent interactions between Y1521.47, R1902.60, Y2413.44 and E3646.53 and the N terminus of GLP-1 that directly engage the central polar network (Fig. 4a, Extended Data Tables 1, 2, Supplementary Video 1). By contrast, TT-OAD influences the central polar network allosterically via interactions with K1972.67, Y1451.40 and Y1481.43. TT-OAD2 also promotes unique hydrogen bond networks with crucial residues in TM2 (Fig. 4a, Extended Data Table 2) that result in different interaction patterns at the top of TM1 and TM2 relative to peptide-occupied receptors. These effects propagate to the polar network through transient contacts between TT-OAD2 with Y1481.43 and Y1521.47 that in turn interact with R1902.60 of the central polar network (Supplementary Video 2). When bound by GLP-1, the polar network is stabilized by ligand and a network of water molecules, whereas for TT-OAD2, this occurs via a distinct network of structural waters rather than by the ligand (Fig. 4b, Supplementary Video 1). These differences in the mechanism of conformational transitions and stabilization of conserved polar networks (summarized in Extended Data Fig. 9) may contribute to the different kinetic profiles of G-protein activation, as well as the full versus partial agonism for cAMP production.

Fig. 4: TT-OAD2 interactions lead to reorganization and stabilization of the central polar network via a distinct mechanism to GLP-1.
figure 4

Summaries of interactions observed in molecular dynamics simulations (Supplementary Video 2) on TT-OAD2- and GLP-1-bound GLP-1R that predict interactions stabilizing the active conformation of the central polar network. a, Left, GLP-1 (brown ribbon) residue D9 (brown stick) forms an ionic interaction (red dotted lines) with R1902.60, which is involved in key hydrogen bonds with N2403.43 (in turn interacting with S1862.56). At the top of TM2, K1972.67, D1982.68 and Y1451.40 are stabilized in polar interactions (red dotted lines). Right, TT-OAD2 (brown stick and transparent surface) forms ionic interaction (red dotted lines) with K1972.67 and hydrophobic contacts with Y1451.40 and Y1481.43 (cyan transparent surfaces) modifying the interaction network at the top of TM1. Y1481.43 transiently interacts with R1902.60 and partially reorients N2403.43 and S1862.56. TM6 and TM7 were removed for clarity. b, GLP-1R transmembrane helix sites are occupied by structural water molecules; blue spheres indicate receptor volumes occupied by low-mobility water molecules (occupancy more than 75% frames). Left, the GLP-1R–GLP-1–Gs complex stabilizes the central transmembrane polar residues by waters interacting with Y1521.47, T3917.46, R1902.60 and E3645.53 (Supplementary Video 1). Right, the GLP-1R–TT-OAD2–Gs complex is characterized by structural water molecules interacting with N3205.50 and E3646.53 (Supplementary Video 1).

Full size image

Collectively, our work provides key advances in understanding the activation of class B GPCRs and Gs protein efficacy, identifying a non-peptide binding site within the GLP-1R that can promote distinct efficacy and biased signalling relative to peptide ligands, and this may extend to other class B GPCRs. The demonstration that non-peptide agonists of the GLP-1R are not required to mimic the extensive receptor contacts formed by peptides within the transmembrane cavity to promote receptor activation will advance the pursuit of non-peptide agonists for therapeutically important class B receptors.

Methods

Data reporting

No statistical methods were used to predetermine sample size. The experiments were not randomized and investigators were not blinded to allocation.

TT-OAD2 synthesis

Several azoanthracene-based derivatives are reported as potent agonists of the GLP-1R (WO10114824), and a compound from this series known as OAD2 was selected for our studies (WO14113357). OAD2, (S)-2-{[(3S,8S)-3-[4-(3,4-dichloro-benzyloxy)-phenyl]-7-((S)-1-phenyl-propyl)-2,3,6,7,8,9-hexahydro-[1,4]dioxino[2,3-g]isoquinoline-8-carbonyl]-amino}-3-[4-(2,3-dimethyl-pyridin-4-yl)-phenyl]-propionic acid, was synthesized using procedures previously described (see example 179 in WO10114824), and a dihydrochloride salt form (OAD2.2HCl) was prepared by standard methods from the free base. Therefore, TT-OAD2 is the dihydrochloride salt of OAD2 in patent WO14113357. The purity of TT-OAD2 was determined by liquid chromatography–mass spectrometry (LC–MS) to be 98.62%.

Constructs

GLP-1R was modified to contain either a 2xcMyc-N-terminal epitope tag (for signalling and radioligand-binding assays) or a Nanoluc tag (with a 12xGly linker; for NanoBRET binding studies) after the native signal peptide. For β-arrestin recruitment assays, a C-terminal Rluc8 was fused to the C terminus of the receptor. For G-protein conformational assays, a Nanoluc flanked by SGGGGS linkers was inserted into Gαs and Gαi2 after G(h1ha10) in Gαs or E(HA.03) in Gαi2 as previously described30,31. These were used in conjunction with an N-terminally Nluc-labelled Gγ2. For G-protein steady-state assays, G-protein NanoBit-split luciferase constructs were generated by fusing the LgBIT after G(h1ha10) in Gαs or E(HA.29) in Gαi2 and the SmBIT to Gγ2. For structural studies, human GLP-1R in the pFastBac vector was modified to include an N-terminal Flag tag epitope and a C-terminal 8×histidine tag; both tags are removable by 3C protease cleavage. These modifications did not alter the pharmacology of the receptor3. A dominant-negative Gαs construct was generated previously by site directed mutagenesis to incorporate mutations that alter nucleotide handling, stabilize the G0 state and interactions with the βγ subunits30.

Insect cell expression

GLP-1R, human dominant-negative Gαs, His6-tagged human Gβ1 and Gγ2 were expressed in Tni insect cells (Expression systems) using baculovirus. Cell cultures were grown in ESF 921 serum-free media (Expression Systems) to a density of 4 million cells per ml and then infected with three separate baculoviruses at a ratio of 2:2:1 for GLP-1R, dominant-negative Gαs and Gβ1γ2. Cells were obtained by centrifugation 60 h after infection and the cell pellet was stored at −80 °C.

Purification of the TT-OAD2–GLP-1R–Gs complex

Cell pellet was thawed in 20 mM HEPES, pH 7.4, 50 mM NaCl, 2 mM MgCl2 supplemented with cOmplete Protease Inhibitor Cocktail tablets (Roche). Complex formation was initiated by addition of 50 μM TT-OAD2, Nb35–His (10 μg ml−1) and apyrase (25 mU ml−1, NEB) to catalyse hydrolysis of unbound GDP and allow for stabilization of the G0 state; the suspension was incubated for 1 h at room temperature. Membrane was solubilized by 0.5% (w/v) lauryl maltose neopentyl glycol (LMNG, Anatrace) supplemented with 0.3% (w/v) cholesteryl hemisuccinate (CHS, Anatrace) for 2 h at 4 °C. Insoluble material was removed by centrifugation at 30,000g for 30 min and the solubilized complex was immobilized by batch binding to M1 anti-Flag affinity resin in the presence of 3 mM CaCl2. The resin was packed into a glass column and washed with 20 column volumes of 20 mM HEPES pH 7.4, 100 mM NaCl, 2 mM MgCl2, 3 mM CaCl2, 1 μM OAD, 0.01% (w/v) MNG and 0.006% (w/v) CHS before bound material was eluted in buffer containing 5 mM EGTA and 0.1 mg ml−1 Flag peptide. The complex was then concentrated using an Amicon Ultra Centrifugal Filter (molecular mass cut off 100 kDa) and subjected to size-exclusion chromatography on a Superdex 200 Increase 10/300 column (GE Healthcare) that was pre-equilibrated with 20 mM HEPES pH 7.4, 100 mM NaCl, 2 mM MgCl2, 1 μM OAD, 0.01% (w/v) MNG and 0.006% (w/v) CHS to separate complex from contaminants. Eluted fractions consisting of receptor and G-protein complex were pooled and concentrated. Final yield of purified complex was approximately 0.15 mg per litre of insect cell culture.

Electron microscopy

Samples (3 µl) were applied to a glow-discharged Quantifoil R1.2/1.3 CuRh 200 mesh holey carbon grid (Quantifoil GmbH) and were flash frozen in liquid ethane using the Vitrobot mark IV (Thermo Fisher Scientific) set at 100% humidity and 4 °C for the prep chamber. Data were collected on a Titan Krios microscope (Thermo Fisher Scientific) operated at an accelerating voltage of 300 kV with a 50 μm C2 aperture at an indicated magnification of 105 K in nanoprobe EFTEM mode. Gatan K3 direct electron detector positioned post a Gatan Quantum energy filter, operated in a zero-energy-loss mode with a slit width of 25 eV was used to acquire dose fractionated images of the GLP-1R TT-OAD2-bound sample without an objective aperture. Movies were recorded in hardware-binned mode (previously called counted mode on the K2 camera) yielding a physical pixel size of 0.826 Å pixel−1 with an exposure time of 3.715 s amounting to a total dose of 65.6 e− Å−2 at a dose rate of 12.2 e− pixel−1 s−1, which was fractionated into 62 subframes. A second dataset of 1,568 micrographs was also recorded using the same microscope but in ‘super-resolution’ mode on the K3 detector, the physical pixel size was 0.413 Å with an exposure time of 4.015 s amounting to a total dose of 63.5 e− Å−2, which was fractionated into 67 subframes. Defocus range was set between −0.7 and −1.5 μm. A total of 3,158 plus 1,568 movies were collected in two data collection sessions.

Electron microscopy data processing

Movies were motion-corrected with UCSF MotionCor232 (movies collected in super-resolution mode were Fourier scaled by a factor of ×2 to match the pixel size of the larger data set). This was followed by CTF estimation sing the GCTF software packag33. Particles were picked from the micrographs using the automated reference-free procedure in RELION34,35. Reference free 2D and 3D classification (by generating multiple ab initio models with no structural identity enforced) was carried out in CryoSPARC (v.2.5.0)36. A homogeneous subset of particles was then subjected to movie refinement and Bayesian particle polishing as implemented in RELION (v.3.0). This homogeneous subset of polished particles was used in a 3D refinement in RELION and then further classified into 3D classes with alignment of Euler angles not taken into account. Particles belonging to the 3D class that yielded the best resolved map were then subjected to signal subtraction to subtract density due to the detergent micelle and the alpha domain of the G protein. Final 3D refinement was performed in RELION (3.0) yielded a map of resolutions 3.01 Å. Local resolution estimations were performed using the ResMAP software packag37.

Atomic model refinement

Fitting the model to the cryoEM electron density map was achieved using the MDFF routine in namd38. The fitted model was further refined by rounds of manual model building in coot39 and real space refinement as implemented in the Phenix software package40, the model restraints for the TT ligand were prepared by using the coordinates generated from Chem3D and the ELBOW software package41. The ligands were fitted after the first round of real-space refinements, manually first in coot39, then refined using Phenix real-space refinement42. Ramachandran, rotamer and secondary structure restraints were applied for the first round of real-space refinement, and after manual inspection and adjustment of the model in coot further real-space refinements were carried out with only Ramachandran and rotamer restraints applied and the model/data weight was allowed to freely refine. The density around the extracellular domain was poorly resolved (local resolution estimated at >8 Å) and was not modelled.

Modelling methods for preparation of molecular dynamic simulations

The two missing receptor loops, namely the stalk region and ICL3, were generated using PLOP43; ICL3 was also minimized in the presence of Gα to eliminate steric clashes. On the basis of the electron density of our structures, TM1 for the GLP-1-bound 5VAI structure6 was replaced by TM1 from the P5-bound structure (PDB code 6B3J)3 by the method of molecular superposition. The missing residues in the stalk region were reconstructed using Modeller44 subject to the constraint that the high variability positions45 in the GLP-1R multiple sequence alignment (E133–R134) faced outwards. The missing loops in the G protein were generated by molecular superposition, using VMD46, of the corresponding loops in the β2-adrenergic receptor–G protein complex47, PDB code 3SN6 to the flank either side of the gap, since this particular X-ray structure (with 99% identity to the G protein used in this study) generally gave a lower root mean squared deviation value on molecular superposition than plausible alternative G-protein structures (for example, PDB 5VAI). The joining point was taken as the closest atom pairs (usually separated by approximately 0.2 Å) that maintained an appropriate Cα–Cα distance (3.7–3.9 Å) across the join; selected residues spanning the join were minimized using PLOP where additional refinement was deemed necessary. The exception to this was the loop between A249–N264, which was completed using the shorter loop from the adenosine A2A receptor–G-protein complex, PDB code 5G5348. The helical domain, between residues G47 and G207, which is not visible in the cryo-EM structure, was omitted as in earlier work.

Molecular dynamics methods

Four GLP-1R complexes (GLP-1R–TT-OAD2–Gs; GLP-1R–TT-OAD2; GLP-1R–GLP-1–Gs; and GLP-1R–GLP-1; Supplementary Table 3) and two apo GLP-1R structures (obtained by removing both the Gs protein and the ligands; Supplementary Table 3) were prepared for simulation with the CHARMM36 force field49, through use of in-house python htmd50 and TCL (Tool Command Language) scripts. The pdb2pqr51 and propka52 software were used to add hydrogen atoms appropriate for a pH of 7.0; the protonation of titratable side chains was checked by visual inspection. The coordinates were superimposed on the corresponding GLP-1R coordinates from the OPM database53 so as to orient the receptor before insertion54 in a rectangular pre-built 125 Å × 116 Å 1-palmitoyl-2-oleyl-sn-glycerol-3-phosphocholine (POPC) bilayer; lipid molecules overlapping the receptor were removed. TIP3P water molecules were added to the 125 Å × 116 Å × 195 Å simulation box using the VMD Solvate plugin 1.5 (Solvate Plugin, v.1.5; http://www.ks.uiuc.edu/Research/vmd/plugins/solvate/). Overall charge neutrality was maintained by adding Na+ and Cl− counter ions to a final ionic concentration of 150 mM using the VMD Autoionize plugin 1.3 (Autoionize Plugin, v.1.3; http://www.ks.uiuc.edu/Research/vmd/plugins/autoionize/). CGenFF force field parameters55,56,57 and topology files for TT-OAD2 were retrieved from the Paramch56 webserver. No further optimization was performed because the obtained parameters were associated to low penalty scores.

Systems equilibration and molecular dynamics simulation settings

ACEMD58 was used for both equilibration and molecular dynamics productive simulations. Isothermal-isobaric conditions (Langevin thermostat59 with a target temperature of 300 K and damping of 1 ps−1 and Berendsen barostat60 with a target pressure 1 atm) were used to equilibrates the systems through a multi-stage procedure (integration time step of 2 fs). Initial steric clashes between lipid atoms were reduced through 3,000 conjugate-gradient minimization steps, then a 2 ns molecular dynamics simulation was run with a positional constraint of 1 kcal mol−1 Å−2 on protein atoms and lipid phosphorus atoms. Subsequently, 20 ns of molecular dynamics simulations were performed constraining only the protein atoms. In the final equilibration stage, protein backbone alpha carbons constraints were applied for a further 60 ns.

Productive trajectories in the canonical ensemble (NVT) at 300 K (four 500-ns-long replicas for each GLP-1R complex; Supplementary Table 3) were computed using a thermostat damping of 0.1 ps−1 with an integration time step of 4 fs and the M-SHAKE algorithm61 to constrain the bond lengths involving hydrogen atoms. The cut-off distance for electrostatic interactions was set at 9 Å, with a switching function applied beyond 7.5 Å. Long-range Coulomb interactions were handled using the particle mesh Ewald summation method (PME)62 by setting the mesh spacing to 1.0 Å. Trajectory frames were written every 100 ps of simulations.

Molecular dynamics analysis

The first half (500 ns) of the molecular dynamics replicas involving GLP-1R–TT-OAD2, GLP-1R–GLP-1 complexes as well as the apo-GLP-1R (TT-OAD2), and apo-GLP-1R (GLP-1) systems (Supplementary Table 3) were considered as part of the equilibration stage and therefore not considered for analysis. Atomic contacts (atom distance less than 3.5 Å) were computed using VMD46. Hydrogen bonds were identified using the GetContacts analysis tool (https://getcontacts.github.io/), with the donor-acceptor distance set to 3.3 Å and the angle set to 150°. Videos were generated using VMD46 and avconv (https://libav.org/avconv.html). Root mean square fluctuation (RMSF) values were computed using VM46 after superposition of the molecular dynamic trajectories frames on the alpha carbon of the transmembrane domain (residues E1381.33–V4047.60). The orientation of the N-terminal helix of the ECD of GLP-1R was drawn in VMD considering a representative frame every 10 ns. To detect volumes within the transmembrane domain of GLP-1R occupied by water molecules with low mobility (structural water molecules), the AquaMMapS63 analysis was performed on 10-ns-long molecular dynamics simulations of the GLP-1R–TT-OAD2–Gs and GLP-1R–GLP-1–Gs complexes (coordinates were written every 10 ps of simulation); all the alpha carbons were restrained in analogy with the approach proposed previously64.

Whole-cell radioligand binding assays

HEK293 cells (confirmed mycoplasma negative) were seeded at 30,000 cells per well in 96-well culture plates and incubated overnight in DMEM containing 5% FBS at 37 °C, 5% CO2. Media was replaced with HBSS containing 25 mM HEPES and 0.1% (w/v) BSA with 0.1 nM 125I-exendin(9–39) and increasing concentrations of unlabelled agonist. Cells were incubated overnight at 4 °C, washed three times in ice-cold buffer and then solubilized in 0.1 M NaOH. Radioactivity was determined by gamma counting. Non-specific activity was defined using 1 μM exendin(9–39).

cAMP accumulation assays

HEK293 cells (confirmed mycoplasma negative) were seeded at a density of 30,000 cells per well into 96-well culture plates and incubated overnight in DMEM containing 5% FBS at 37 °C in 5% CO2. cAMP detection was performed as previously described in the presence of the phosphodiesterase inhibitor 3-isobutyl-1-methylxanthin65. All values were converted to cAMP concentration using a cAMP standard curve performed in parallel and data were subsequently normalized to the response of 100 μM forskolin in each cell line. In one series of experiments, vehicle or increasing concentrations of TT-OAD2 was added 30 min before assay of peptide response.

cAMP kinetics studies

HEK293A cells (confirmed mycoplasma negative) were transfected with an Epac-cAMP sensor (CAMYEL) and human GLP-1R at an optimized ratio. Ligand-mediated cAMP production was measured 48 h after transfection. In brief, culture media was replaced with assay buffer (1× HBSS, 10 mM HEPES, 0.1% BSA, pH 7.4). BRET signals were measured at 1 min intervals using a PHERAstar plate reader (BMG LabTech) in the absent or present of increasing concentration of ligands. Forskolin (100 μM) was used as a positive control, and data were normalized to the forskolin response.

β-arrestin recruitment assays

HEK293 cells (confirmed mycoplasma negative) were transiently transfected with GLP-1R-Rluc8 and β-arrestin1-Venus at a 1:4 ratio and seeded at a density of 30,000 cells per well into 96-well culture plates and incubated for 48 h in DMEM containing 5% FBS at 37 °C in 5% CO2. β-arrestin recruitment was performed as previously described66. In one series of experiments, vehicle or increasing concentrations of TT-OAD2 was added 30 min before assay of peptide response.

ERK1/2 phosphorylation assays

HEK293 cells (confirmed mycoplasma negative) expressing stably expressing the GLP-1R were seeded at a density of 30,000 cells per well into 96-well culture plates and incubated overnight at 37 °C in 5% CO2. Receptor-mediated pERK1/2 was determined using the AlphaScreen ERK1/2 SureFire protocol as previously described14. Data were normalized to the maximal response elicited by 10% FBS determined at 6 min. In one series of experiments, vehicle or increasing concentrations of TT-OAD2 was added 30 min before assay of peptide response.

Ca2+ mobilization assays

HEK293 cells (confirmed mycoplasma negative) stably expressing the GLP-1R were seeded at a density of 30,000 cells per well into 96-well culture plates and incubated overnight at 37 °C in 5% CO2, and receptor- mediated intracellular calcium mobilisation determined as previously described65. Fluorescence was determined immediately after ligand addition, with an excitation wavelength set to 485 nm and an emission wavelength set to 520 nm, and readings taken every 1.36 s for 120 s. The peak value was used to create concentration-response curves. Data were normalized to the maximal response elicited by 100 μM ATP. In one series of experiments, vehicle or increasing concentrations of TT-OAD2 was added 30 min before assay of peptide response.

Generation of stable cell lines containing wild-type and mutant GLP-1R

Mutant receptors were generated in a 2xc-Myc epitope-tagged receptor using QuikChange site-directed mutagenesis (Invitrogen) and sequences confirmed. Wild-type and mutant receptors were stably expressed in CHOFlpIn cells (confirmed mycoplasma negative) using the FlpIn Gateway technology system and selected using 600 μg ml−1 hygromyocin B.

NanoBRET ligand binding

HEK293A cells were transiently transfected with Nluc-hGLP-1R. Forty-eight hours after transfection, cells were collected and plasma membrane was extracted as described previously31. Cell membrane (1 μg per well) was incubated with furimazine (1:1,000 dilution from stock) in assay buffer (1× HBSS, 10 mM HEPES, 0.1% (w/v) BSA, 1× P8340 protease inhibitor cocktail, 1 mM DTT and 0.1 mM PMSF, pH 7.4). RhodamineX-Ex4 (Rox-Ex4) was used as fluorescent ligand in the NanoBRET binding assay. BRET signal between Nluc-hGLP-1R and Rox-Ex4 was measured using PHERAstar (BMG LabTech) at 10 s interval (25 °C), a 2 min baseline was taken before addition of Rox-Ex4 (Kd concentration 3.16nM, determined previously), the measurement continued for 15 min followed by adding increasing concentration of TT-OAD2, or unlabelled Ex4 as a control. Data were corrected for baseline and vehicle treated samples.

G-protein conformation assays

HEK293AΔS/Q/12/13 cells stably expressing GLP-1R (tested and confirmed to be free from mycoplasma) were transfected with a 1:1:1 ratio of Nanoluc-Gαs (Nanoluc inserted at position 72):Gβ1:Venus-Gγ2 24 h before collection and preparation of cell plasma membranes. Cell membrane (5 μg per well) was incubated with furimazine (1:1,000 dilution from stock) in assay buffer (1× HBSS, 10 mM HEPES, 0.1% (w/v) BSA, 1× P8340 protease inhibitor cocktail, 1 mM DTT and 0.1 mM PMSF, pH 7.4). The GLP-1R-induced BRET signal between Gαs and Gγ was measured at 30 °C using a PHERAstar (BMG LabTech). Baseline BRET measurements were taken for 2 min before addition of vehicle or ligand. BRET was measured at 15-s intervals for a further 7 min. All assays were performed in a final volume of 100 μl.

G-protein NanoBIT assays

HEK293A wild-type cells stably express human GLP-1R were transiently transfected with Gα-LgBIT, Gβ1, Gγ2-SmBIT (1:5:5) 48 h before the assays. Cells were then incubated with coelenterazine H (5 μM) for 1 h at room temperature. Luminescence signals were measured using a Clariostar plate reader (BMG LabTech) at 30 s intervals before and after ligand addition (25 °C). Data were corrected to baseline and vehicle treated samples.

In vivo IVGTT assays

Intravenous glucose tolerance tests were performed in male human GLP-1R knock-in and knockout mice (all on C57/BL6 background67). Catheters were placed in the right carotid artery and left jugular vein of mice 6–11 months of age. Approximately one week later, mice (n = 4–5 per group) were fasted overnight and the catheters were exteriorized as mice acclimated to test cages. Vehicle (5% DMSO, 20% Captisol in NaHPO4, pH 2, 1 ml kg−1), GLP-1(7-36)NH2 at 10 μg kg−1, GIP(1-42) at 25 μg kg−1, or OAD2 at 3 mg kg−1 was administered intravenously one minute before glucose load (0.5 g kg−1). Blood samples were collected at −10, 0, 2, 4, 6, 10, 20 and 30 min to determine blood glucose concentrations via glucometer (Roche, Aviva) and plasma insulin measurement (Alpco, 80-INSMSU-E10). All mouse experiments were performed in accordance with the Institutional Animal Care and Use Committee of Eli Lilly and Company and the NIH Guide for the Use and Care of Laboratory Animals.

Data analysis

Pharmacological data were analysed using Prism 7 (GraphPad). Concentration response signalling data were analysed using a three-parameter logistic equation, or via operational analysis. Changes in the rate of change in BRET kinetic data were fitted to one-phase association curve. Statistical analysis was performed with either one-way analysis of variance and a Dunnetts post-test or a paired t-test, and significance accepted at P < 0.05.

Graphics

Molecular graphics images were produced using the UCSF Chimera package from the Computer Graphics Laboratory, University of California, San Francisco (supported by NIH P41 RR-01081).

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this paper.