Main

To obtain a high-resolution structure of the human GCGR transmembrane domain (TMD), a thermostabilized receptor (StaR) was generated6,7,8 containing 11 amino-acid substitutions. To facilitate crystallization further, the extracellular domain was removed from the N terminus (residues 2–135), and the C terminus was truncated by 60 residues (residues 418–477). Finally, T4-lysozyme (T4L) was inserted into intracellular loop (ICL)2 between Leu255 and Pro259 (Fig. 1a), resulting in the construct designated GCGR-StaR(136–417)–T4L. The structure was solved in the presence of the antagonist MK-0893 (Fig. 1b and Extended Data Table 1). The construct modifications did not alter the antagonist-binding properties of the receptor compared with wild type (Extended Data Table 2).

Figure 1: Structure of GCGR and the MK-0893 allosteric binding site.
figure 1

a, Crystallization construct showing stabilizing mutations (green), binding site residues (pink), disordered residues not located (grey), and the disulfide bond between Cys2243.29b and Cys294 (yellow line). b, Chemical structure of MK-0893. c, d, Ribbon representation of GCGR (blue), viewed parallel to the membrane (c) and from extracellular space (d). The position of the T4L insertion to ICL2 is indicated. MK-0893 in stick representation with carbon, nitrogen, oxygen and chlorine atoms coloured yellow, blue, red and green, respectively, Fo − Fc OMIT density contoured at 2.5σ. e, Surface representation of GCGR embedded within the membrane (20 ns molecular dynamics simulation), lipids in proximity of MK-0893 removed for clarity.

PowerPoint slide

Full size image

The core fold of the receptor features the canonical 7TM helices (TM1–TM7) (Fig. 1c) in a similar conformation to the previously published GCGR structure3. Continuous density is observed for intracellular loops ICL1 and ICL3, while extracellular loop (ECL)2 adopts a conformation capping the entrance to the orthosteric site. In contrast with the previously published GCGR structure3, the N terminus of TM5 unwinds by one helical turn, permitting ECL2 to stretch across to the central axis of the TM helical bundle, mediating interactions from TM3 across to TM6 and TM7 while maintaining the conserved disulfide bond between Cys2243.29b and Cys294 (numbers in superscript refer to the modified Ballesteros numbering system for class B GPCRs9,10,11). The highly conserved sequence motif GWGxP in TM4 of class B receptors has an important structural role supplying interactions stabilizing the configuration of TM2, TM3 and TM4 (ref. 5). In this GCGR structure, TM4 bulges at Gly2714.49b and, along with Pro2754.53b, they disrupt intra-helical hydrogen bonding and result in positioning of Trp2724.50b towards TM2 and TM3, which forms a hydrogen bond with Asn1792.52b on TM2 in an analogous fashion to CRF1R. In addition, a hydrogen bond from the side chain of Tyr2333.38b to the backbone carbonyl of Trp2724.50b further strengthens the TM4–TM3 interaction.

Unexpectedly, strong and unambiguous density is observed for the MK-0893 antagonist outside the 7TM helical bundle (Fig. 1c–e), straddling TM6 from within the lipid bilayer. TM6 then acts to divide the binding site into two distinct regions: a hydrophobic interface with TM5, and a polar cleft towards TM7. The different physicochemical properties of this bipartite antagonist pocket correspond to the dual hydrophilic/hydrophobic nature of the ligand. The apolar methoxynaphthalene moiety of the small molecule makes hydrophobic contacts in the TM5–TM6 interface with Leu3295.61b, Phe3456.36b, Leu3526.43b, Thr3536.44b and the alkyl chain of Lys3496.40b (Fig. 2a). On the opposite site, within the TM6–TM7 cleft, MK-0893 participates in a network of polar contacts: the ligand amide group hydrogen bonds with Lys3496.40b and Ser3506.41b, while the carboxyl group forms a salt bridge with Arg3466.37b. This moiety also makes additional polar interactions with Asn4047.61b, the backbone of Lys405 (located between TM7 and H8) and a water-mediated hydrogen bond with Ser3506.41b and Leu3997.56b (Fig. 2a). The phenylethylpyrazole core of the molecule makes further interactions with TM6, in particular with Thr3536.44b and Lys3496.40b. The position of the pyrazole moiety parallel to the membrane provides the two ligand vectors towards the bipartite antagonist sub-pockets. A third vector starting from the ligand pyrazole ring leads to the dichlorophenyl group, which makes a crystal contact to TM4 of a symmetry mate (see Extended Data Fig. 1). Molecular dynamics simulations of MK-0893 binding to the wild-type receptor (outside the constraints of the crystal system) demonstrate that the receptor–ligand interaction is stable, and MK-0893 remains at the cytoplasmic membrane boundary with the carboxyl moiety able to interact with intracellular solvent and with interactions between the ligand and receptor, involving Arg3466.37b, Lys3496.40b, Asn4047.61b and Lys405, maintained (Extended Data Fig. 2). Sequence conservation analysis demonstrates that, with the exception of Thr3536.44b and Phe3456.36b, the other key binding-site residues show good conservation across other members of human class B receptors (Extended Data Fig. 5).

Figure 2: Confirming the MK-0893-binding site in GCGR.
figure 2

a, Diagram of ligand interactions in the MK-0893-binding site. Hydrogen bonds are depicted as dashed red lines with distances between heavy atoms in Å. GCGR in ribbon representation is coloured blue, MK-0893 in stick representation is coloured as per Fig. 1. b, Comparison of pKd of wild type (WT) with the mutants. Data are average of three independent experiments and error bars represent standard error of the mean (s.e.m.). P values are calculated from a two-tailed t-test. NS, not significant. The data set for R346E and N404A did not fit the one-site binding unambiguously due to near complete loss of specific binding. ND, not determined.

PowerPoint slide

Full size image

Tritium-labelled MK-0893 was prepared and used to characterize the ligand-binding site. Membrane fractions prepared from HEK293T cells expressing the wild-type GCGR were used in saturation binding analysis. Saturation binding was monophasic and best fitted to a one-site model with a linear Scatchard plot, consistent with the presence of a single high-affinity binding site (Fig. 3a). Competition with [3H]MK-0893 was used to characterize the binding of a number of reported GCGR antagonists (see Extended Data Fig. 3 for details). These compounds were selected to represent molecules that exhibit chemical similarity to MK-0893 (for example, NNC0640, Cpd-01, 02, 03 and 04), as well as those that are chemically distinct (for example, Cpd-05 and 06). Consistently, NNC0640, Cpd-01, 02, 03 and 04 were able to fully compete with [3H]MK-0893 binding, indicating that these compounds share the same binding site (Fig. 3b–f). By contrast, Cpd-05 and Cpd-06 were not competitive with [3H]MK-0893, indicating that these compounds bind to a different site (Fig. 3g, h). Furthermore, [3H]-MK-0893 was not displaced by glucagon or the related peptide antagonist des-His1-[Glu9]-glucagon (Fig. 3i, j), which bind at the orthosteric site. Interestingly, NNC0640 was the ligand used in the crystallization of the first reported GCGR structure3; however, the position of the ligand was not resolved. A clear peak is observed in the electron density map of GCGR–NNC0640 between TM6 and TM7 towards the intracellular side of the receptor. Although this was modelled as a polyethylene glycol (PEG) molecule in the reported structure, superposition with the GCGR structure reported here demonstrates that the amide moiety and carboxyl function of MK-0893 directly overlays with this peak (Extended Data Fig. 4). Given the chemical similarity of these ligands, coupled with our competition data, it is likely that this constitutes residual signal from NNC0640 binding in an analogous position to MK-0893 on GCGR rather than being a PEG molecule.

Figure 3: Pharmacology of MK-0893.
figure 3

a, Saturation binding of [3H]MK-0893 to membranes containing wild-type GCGR. Inset shows the Scatchard plot. b–j, Representative competitive binding data for inhibition of [3H]MK-0893 binding to membranes containing wild-type GCGR in the presence of indicated compounds (for chemical structures, see Extended Data Fig. 3). Values represent an average of at least three independent experiments ± s.e.m. Kd, dissociation constant. pKi, negative logarithm of the inhibition constant. Bmax, total concentration of receptors. B and F in the Scatchard plot denote bound and free ligand concentrations, respectively.

PowerPoint slide

Full size image

To confirm the allosteric pocket identified in the structure, single point mutations were made to residues in the binding site and the binding of [3H]MK-0893 was subsequently assessed. The mutations were introduced in the full-length wild-type receptor with a C-terminal enhanced green fluorescent protein (eGFP) tag. The presence of the eGFP tag has no impact on MK-0893 binding or glucagon activation of the receptor (data not shown). For this analysis, Arg3466.37b was mutated to alanine or glutamic acid. Lys3496.40b was mutated to either alanine or methionine. Ser3506.41b and Asn4047.61b were changed to alanine. These residues were selected as they participate in polar interactions with the ligand and thus were considered more likely to have a measurable effect on ligand binding (Fig. 2a). In addition, these residues interact with the amide and carboxyl functions of MK-0893, which have been demonstrated to be critical for the antagonist activity of the ligand4. Fluorescence-activated cell sorting (FACS) analysis using an antibody to the extracellular surface of receptor was performed to compare cell surface expression of mutants with the wild-type receptor. None of the mutations caused any reduction in the cell surface expression levels of the receptor nor had any effect on the binding affinity of glucagon peptide for the orthosteric site (Extended Data Fig. 6 and Extended Data Table 2). Mutation of Arg3466.37b to glutamic acid or Asn4047.61b to alanine reduced binding levels to close to undetectable above non-specific binding, while mutation of Lys3496.40b to alanine or methionine, and mutation of Arg3466.37b to alanine, significantly reduced the binding affinity of [3H]MK-0893 (Fig. 2b and Extended Data Fig. 7). Consistently, these mutations also reduced the ability of MK-0893 to antagonize the glucagon-mediated increase in cAMP (Extended Data Fig. 8).

The MK-0893 binding mode suggests that the ligand acts as a clamp holding TM6 in the inactive state and hampering receptor conformational changes required for G-protein coupling (Fig. 4a, c). In the case of class A receptors, activation results in a rotation of TM6 in response to agonist binding, which is transmitted through a rigid-body movement amplified along TM6, thereby altering the interface between TM5 and TM6 and leading to an outward movement of the cytoplasmic end of TM6 to enable G-protein binding12. So far, no active structures of class B receptors have been reported; however, the ability of MK-0893 to block signalling of GCGR points to a similar critical role of TM5 and TM6 in class B receptor activation. This is consistent with the observation that, despite divergence in the extracellular arrangement of TM helices between class A and B receptors, very good structural conservation exists on the intracellular sides in the inactive conformation5,11. An extra-helical binding site was recently described for 1-(2-(2-(tert-butyl)phenoxy)pyridin-3-yl)-3-(4-(trifluoromethoxy)phenyl)urea (BPTU), an allosteric antagonist of the P2Y1 receptor13, although in this case the binding site was located between TM1, 2 and 3 (Fig. 4b). It is likely that BPTU inhibits the movement of TM3 that is also critical in receptor activation14. Together, these structures demonstrate that modulation of helical movements from membrane-proximal surfaces represents an alternative way of modifying the activity of GPCRs.

Figure 4: MK-0893 allosteric mechanism of action.
figure 4

a, b, Ribbon representation of GCGR (blue) and P2Y1 receptor (P2Y1R; yellow) respectively, viewed parallel to the membrane. MK-0893 in stick representation is coloured as per Fig. 1. BPTU is shown in stick representation with carbon, nitrogen, oxygen and fluorine atoms coloured yellow, blue, red and grey respectively. Potential restrictions on TM movements are indicated. c, View of cytoplasmic side of GCGR (rotated 90° from a) superposed with the β2-adrenoceptor (β2-AR) in complex with carazolol (grey) (Protein Data Bank (PDB) accession 3NY9) and the β2-AR–Gs complex (orange) (PDB accession 3SN6). d, Schematic overview of known binding positions of class A, B and C GPCR ligands.

PowerPoint slide

Full size image

The identification of an allosteric binding site in the class B GCGR located outside the canonical helical bundle further adds to the diversity of interactions now known to occur between GPCRs and their ligands in modifying receptor activation states (Fig. 4d). The GCGR–MK-0893 structure provides insight into the activation mechanism of class B receptors, as well as facilitating the application of structure-based drug design strategies to discover compounds with improved qualities. Strong conservation of this binding site across human class B receptors indicates that the structural information provided here can be applied to other members of this medically relevant family of GPCRs. In future, drug design paradigms must consider that, in addition to the orthosteric binding pocket, extra-helical binding sites accessed from the membrane or within the cell may provide alternative strategies to modulate receptor function.