Allosteric modulation and biased signalling at free fatty acid receptor 2

Abstract

Free fatty acid receptor 2 (FFA2) is a G protein-coupled receptor (GPCR) that is a primary sensor for short-chain fatty acids produced by gut microbiota. Consequently, FFA2 is a promising drug target for immunometabolic disorders^1-4. Here we report cryogenic electronic microscopy structures of FFA2 in complex with two G proteins and three distinct classes of positive allosteric modulators (PAMs), and describe noncanonical activation mechanisms that involve conserved structural features of class A GPCRs. Two PAMs disrupt the E/DRY activation microswitch⁵ and stabilize the conformation of intracellular loop 2 by binding to lipid-facing pockets near the cytoplasmic side of the receptor. By contrast, the third PAM promotes the separation of transmembrane helices 6 and 7 by interacting with transmembrane helix 6 at the receptor-lipid interface. Molecular dynamic simulations and mutagenesis experiments confirm these noncanonical activation mechanisms. Furthermore, we demonstrate the molecular basis for the G_i versus G_q bias, which is due to distinct conformations of intracellular loop 2 stabilized by different PAMs. These findings provide a framework for the design of tailored GPCR modulators, with implications that extend beyond FFA2 to the broader field of GPCR drug discovery.

Extended Data Fig. 1 ∣. Pharmacological characterization of FFA2 ligands.

a. β-arrestin 1 recruitment induced by FFA2 activators and modulators. b. β-arrestin 2 recruitment induced by FFA2 activators and modulators in wild-type HEK293T and G protein deficient (ΔG_s, ΔG₁₂, ΔG₁₃, ΔG_q, ΔG₁₁ + Pertussis toxin treatment to inactivate Gi-family G proteins) HEK293T cells. Data are means +/– S.E.M. n = 3 (three biologically independent experiments). c. Cooperativity of function between three FFA2 PAMs and the orthosteric agonist C3 (propionate). The ability of the indicated concentrations of 4-CMTB, compound 187, and AZ-1729 to modulate inhibition of forskolin-stimulated levels of cAMP via FFA2 is shown. Data are means +/– S.E.M. n = 3 (three biologically independent experiments). d. Binding characteristics of FFA2 modulators assessed in co-operativity studies. ^aAgonist refers to the compound used to generate concentration-response curve. ^bModulator is the compound used in defined concentrations. ^cpK_A represents values estimated for the agonist. ^dpK_B represents values estimated for the modulator. ^eGain in potency was calculated using the equation (pEC₅₀ agonist + highest concentration of modulator) - (pEC₅₀ agonist + vehicle). ^fGain in efficacy was calculated using the equation (Emax agonist + highest concentration of modulator) - (Emax agonist + vehicle). nd means not determined (in experiments in which the effect of compound 187 was largely manifest in terms of efficacy and thus affinity values could not be derived). Emax was constrained to the maximal possible system response and n was constrained to 1 (the slope factor). Statistical significance was determined using one-way ANOVA followed by Dunnett’s multiple comparisons test; *p < 0.05, **p < 0.01, ****p < 0.0001, ns denotes not significant. Cmpd187 refers to compound 187.

Extended Data Fig. 2 ∣. Cryo-EM data processing and analysis.

a, Representative cryo-EM micrograph (scale bar: 50 nm) and 2D class averages (scale bar: 5 nm). The micrograph shown is one example of 5535 micrographs for miniG_q-coupled FFA2 bound to TUG-1375 and 4-CMTB. b-m, Cryo-EM image processing workflow for TUG-1375-4-CMTB-FFA2-miniG_q complex (b-d), TUG-1375-compound 187-FFA2-miniG_q complex (e-g), TUG-1375-AZ-1729-FFA2-G_i complex (h-j), and TUG-1375-compound 187-FFA2-G_i complex (k-m), respectively. Detailed information on protein purification is provided in Supplementary Fig. 1. Angular distributions of the particles used in the final reconstruction are shown in c for TUG-1375-4-CMTB-FFA2-miniG_q complex; f for TUG-1375-compound 187-FFA2-miniG_q complex; i for TUG-1375-AZ-1729-FFA2-G_i complex; and l for TUG-1375-compound 187-FFA2-G_i complex. Gold-standard Fourier shell correlation (FSC) curves are shown in d for TUG-1375-4-CMTB-FFA2-miniG_q complex; g for TUG-1375-compound 187-FFA2-miniG_q complex; j for TUG-1375-AZ-1729-FFA2-G_i complex; and m for TUG-1375-compound 187-FFA2-G_i complex. n, Cryo-EM maps and models of the seven transmembrane helices (TM1-7) of miniG_q-coupled FFA2 bound to TUG-1375 and 4-CMTB. Representative cryo-EM maps of FFA2 in other complexes and ligands are shown in Supplementary Fig. 2.

Extended Data Fig. 3 ∣. TUG-1375 binding in the orthosteric binding pocket.

a, Left and middle panels show the comparison of the receptor interaction profiles of TUG-1375 and butyrate (C4) with FFA2. The binding profile of TUG-1375 is based on the structure of miniG_q-coupled FFA2 in complex with TUG-1375 and compound 187, while the C4 binding profile is derived from our previously published structure (PDB ID: 8T3S). Polar interactions are shown as black dashed lines. Right panel shows the outward movement of TM4 in the structure of FFA2 with TUG-1375 compared to that with C4 (butyrate) indicated by the red arrow. The red circle indicates the group of TUG-1375 that causes such movement. B, Effects of mutations within the orthosteric pocket on the function of TUG-1375 measured by β-arrestin 2 recruitment. Detailed calculations based on the concentration-response curves (upper) are summarized in the table (lower). pEC₅₀ of TUG-1375 and efficacy compared to that of TUG-1375 at wild type (WT) human FFA2 was assessed for the indicated point mutants of the receptor by β-arrestin 2 recruitment assays. For the indicated mutants, the ability of compound 187 to inhibit forskolin stimulated levels of cAMP in cells stably expressing eYFP-tagged forms of the receptor was used to confirm expression and function of an appropriately processed and cell surface delivered form of FFA2. Data are means +/– s.e.m. n = 3 (three biologically independent experiments). Statistical significance was assessed by one way ANOVA followed by Dunnett’s multiple comparison test. * p < 0.05, ** p < 0.01, *** p < 0.001, ****p < 0.0001. NR means no detectable response. c, MD simulations on TUG-1375 binding. The upper panel shows a representative simulation frame depicting TUG-1375 in the orthosteric binding pocket, with key interacting residues shown as sticks. Residue carbon colors and stick thickness represent average electrostatic and van der Waals interaction energies between each residue and the ligand, respectively. Dashed lines indicate hydrogen bonds. The lower panel shows violin plots indicating the distribution of distances between the guanidyl fragment of R180^5.39 and the carboxylic group of E166^ECL2 (left) and between the Cα atoms of Y165^ECL2 and Y90^3.33 (right) across three 1 μs replicates of FFA2/4-CMTB/TUG-1375/miniG_q (Full complex), FFA2/TUG-1375 (TUG-1375 only, without the G protein), and empty FFA2 (Empty, without the ligand and G protein). The results revealed that TUG-1375 affects the ECL2 conformation and the top of TM4. Upon ligand removal, the orthosteric pocket shrinks significantly due to a major conformational change in ECL2, with E166^ECL2 and Y165^ECL2 shifting towards R180^5.39 and Y90^3.33, respectively, forming extensive electrostatic and aromatic contacts. Detailed analysis of TUG-1375 binding can be found in Supplementary Information. Cmpd187 refers to compound 187.

Extended Data Fig. 4 ∣. Binding properties of FFA2 ago-PAMs at Site 1 and 2.

a, Loss of agonist activity and cooperativity with TUG-1375 for 4-CMTB at N²³⁰D FFA2, as measured by cAMP reduction assays. The data in the top row show that 4-CMTB lacks agonist action at N²³⁰D FFA2 whilst TUG-1375, compound 187 and AZ-1729 all activate this mutant as effectively as wild type. The data in the left and the middle panel of the bottom row indicate that 4-CMTB does not produce co-operativity with TUG-1375 at N²³⁰D FFA2 (left), while such effects of both AZ-1729 and compound 187 are retained at this mutant (middle). The data in the right panel indicate that the human FFA2 antagonist [³H]GLPG0974 displays high affinity binding to both wild type and N²³⁰D FFA2. Data are means +/– S.E.M. n = 3 (three biologically independent experiments). b, Structural comparison of FFA2 and FFA3 at Site 1. FFA2 and FFA3 are shown in green and pink, respectively. The space in FFA3 (middle panel, based on the structure PDB ID 8J21) corresponding to Site 1 in FFA2 (left panel) adopts a distinct conformation, which would result in a steric clash with 4-CMTB if it were to adopt a similar binding pose (right panel). c, Co-solvent simulations of AZ-1729 and compound 187. The middle figure shows the orthosteric ligand C4 and the probes as they are placed before the simulation, along with the cylinder (semi-transparent, blue) confining the probes around the area of interest, around ICL2. The top left figure shows AZ-1729 and the probe that allowed to identify its interaction with E106^3.49. The top right figure shows compound 187 and its probe molecule. The bottom left and right figures show the snapshots of the probe-AZ and probe-187, respectively, in MD simulations with the semi-transparent surface denoting the volume occupied by the probe for > 20% of the simulation time. d, Structural alignment of Site 2 for compound 187 in the structures with miniG_q and G_i. The interaction profile of compound 187 and the overall conformation of Site 2 including ICL2 are almost identical in these two structures. e, Effects of mutations in Site 2 on the potency of FFA2 modulators measured by [³⁵S]GTPγS binding assays. Data are means +/− s.e.m. for at least 3 experiments. pEC₅₀ and E_max values are listed in Extended Data Table 2. f, Undetectable β-arrestin-2 recruitment by AZ-1729 at Site 2 mutations in FFA2. β-arrestin-2 recruitment assays were performed with varying concentrations of TUG-1375 or AZ-1729 at wild type, E106G, and G102V FFA2. Data are means +/− s.e.m. n = 3 (three biologically independent experiments). Cmpd187 refers to compound 187.

Extended Data Fig. 5 ∣. Molecular dynamics analysis of FFA2 conformational changes and interactions upon different ligand binding and G protein coupling.

a-c, Molecular interactions between FFA2 ICL2 and G protein. a, Hydrophobic contacts between ICL2 (P114-R121) and G protein α5 helix. Data show mean number of carbon-carbon contacts ( ≤ 4 Å) per frame ± s.e.m. from three independent 1 μs simulations (n = 3). Individual replicate values shown as black dots. AZ-1729/G_i complex shows enhanced contacts compared to TUG-1375 alone, with significant reduction in Y117A mutant. b, Hydrogen between ICL2 and G protein α5 helix. Data show mean number of hydrogen bonds (D-A distance ≤3.5 Å, angle 180 ± 30°) per frame ± s.e.m. from three independent 1 μs simulations (n = 3). Individual replicate values shown as black dots. AZ-1729 enhances hydrogen bonding with G_i, while R121A mutation shows the strongest reduction in these interactions. c, Hydrogen bonds between R121^ICL2 and D350^H5.22 of G_i. Data show mean number of hydrogen bonds per frame ± s.e.m. from three independent 1 μs simulations (n = 3). Individual replicate values shown as black dots. Enhanced interactions in AZ-1729 complex and complete loss with R121A mutation. d-e, Analysis of R^3.50 interactions with Y^5.58 and the distance between R^3.50 and D/NP^7.50xxY motif in FFA2. d, Sidechain polar contacts between R107^3.50 and Y199^5.58. Data show percentage of frames with nitrogen-to-oxygen distance <4.5 Å ± s.e.m. from three independent 1 μs simulations (n = 3). Individual replicate values shown as black dots. e, Proximity between R107^3.50 guanidyl carbon and P270^7.50 Cα. Data show percentage of frames with distance <14 Å ± s.e.m. from three independent 1 μs simulations (n = 3). Individual replicate values shown as black dots. f-k, Conformational stability analysis of ICL2. f-g, Root-mean-square deviation (RMSD) of ICL2 residues P114-R121. Data show mean RMSD ± s.e.m. from three independent 1 μs simulations (n = 3), calculated after alignment to minimized cryo-EM structures using 7-transmembrane bundle Cα atoms. Individual replicate values shown as points. h-i, Root-mean-square fluctuation (RMSF) of ICL2 residues P114-R121. Data show mean RMSF ± s.e.m. from three independent 1 μs simulations (n = 3), calculated after alignment using 7-transmembrane bundle Cα atoms. Individual replicate values shown as points. j, RMSD distributions of ICL2 (P114-R121) relative to FFA2/TUG-1375/AZ-1729/Gi1 complex. Violin plots show probability densities from 3 simulation replicates per complex. Horizontal bars indicate minimum, maximum, and median values. Percentages indicate proportion of frames with RMSD < 1.5 Å. k, Similar analysis using FFA2/TUG-1375/ compound 187/miniG_q complex as a reference. l-n, Analysis of polar contacts at the TM6-TM7 interface. Violin plots show distributions of inter-atomic distances from snapshots taken every 100 ps across three independent 1 μs simulations per complex. When multiple atoms are selected (e.g., O^δ1 and N^δ2 of N230), the shortest distance between residue atoms is shown. Horizontal bars indicate minimum, maximum, and median values. l, Distances between N230^6.43 and N265^7.45 (D/NP^7.50xxY motif) with percentage of frames ≤3.5 Å indicated. m, Oxygen-to-sulfur distances between N230^6.43 and C234^6.47 (C^6.47W^6.48xP^6.50 motif) with percentage of frames ≤4.5 Å indicated. n, Hydrogen bonds between cytosolic region of TM6 (Q215-N230) and TM7 (R255-Y274). Data show percentage of frames with at least one hydrogen bond present ± s.e.m. from three independent 1 μs simulations (n = 3). Individual replicate values shown as black dots. Conformational stability analysis of FFA2 receptor states can be found in Supplementary Fig. 3. Detailed RMSD and RMSF data for all simulated systems and residue-ligand interaction energy values can be found in Supplementary Tables 1 and 2. Cmpd187 refers to compound 187.

Extended Data Fig. 6 ∣. G protein-coupling to FFA2.

a, Interactions between FFA2 and the α5 helix of G_i (from the structure of FFA2 with AZ-1729 and G_i) or miniG_q (from the structure of FFA2 with compound187 and miniG_q). b, Interactions between FFA2 and the αN helix of G_i or miniG_q. Polar interactions are shown as black dashed lines. Specifically, in both structures, D350^H5.22 (G protein numbering) in G_i, or E242^H5.22 in mG_αi/s/q forms a polar interaction with R121^ICL2 in FFA2. Additionally, L348^H5.20 and L353^H5.25 in G_i form hydrophobic interactions with V111^3.54, F202^5.61, M206^5.65, and L223^6.36 of FFA2, while the corresponding residues, L240^H5.20 and L245^H5.25 in mG_αi/s/q, also form direct interactions with these residues in FFA2. Notably, residues P114^ICL2, V115^ICL2, and L119^ICL2 in ICL2 of FFA2 inserts into a hydrophobic pocket formed by G_i residues L194^S3.01, F336^H5.08, I343^H5.15, and I344^H5.16, or mG_αi/s/q residues F228^H5.08, I235^H5.15, and L236^H5.16, stabilizing a significant hydrophobic environment. c, Structural comparison of the α5 and αN helices in the TUG-1375-AZ-1729-FFA2-G_i (yellow), TUG-1375-compound187-FFA2-miniG_q (yellow green), and TUG-1375-compound-FFA2-G_i (pink) complexes. The α5 helix adopts a highly similar conformation across all three structures, whereas the αN helix exhibits greater variability. d, Polar interactions between FFA2 and the α5 helix of miniG_q (right panel), which are missing in the structures with G_i (left panel). Polar interactions are shown as black dashed lines.

Extended Data Fig. 7 ∣. Compound3 and ligand selectivity at Site 2.

a, Effect of various concentrations of compound3 on the concentration-response profiles of 4-CMTB and C3. Assays were performed as described in Methods, with the following adjustments: (1) 2,000 cells were seeded per well; (2) cAMP accumulation was induced by 0.1 μM forskolin; (3) cells were incubated for 45 min with indicated agonists. Error bars represent mean ± s.e.m. from 3 biologically independent experiments, each performed in triplicate. p-values refer to changes in potency and are reported according to one-way ANOVA test with Dunnet’s post hoc analysis. b, Comparison of Site 2 in FFA2 and similar pockets in FFA1, FFA3, and FFA4. AZ-1729 and compound187 were docked to the other FFAs by aligning their structures to FFA2 structures with these two PAMs. Severe steric clashes as indicated by red circles are observed between AZ-1729 or cmpd187 and other FFAs, indicating a high selectivity of these two PAMs for FFA2. c, Comparison of Site 2 in FFA2 and the similar allosteric site in C5aR. Left: FFA2 (blue) in complex with the PAM AZ-1729 (orange); Middle: FFA2 (green) in complex with the PAM compound187 (pink); Right: C5a receptor (C5aR, brown) in complex with the NAM avacopan (light blue). The crystal structure of the C5aR-avacopan complex (PDB ID: 6C1R) illustrates the binding mode of this allosteric antagonist.

Fig. 1 ∣. Structure and function of FFA2 activators and modulators.

a, Schematic of how FFA2 senses SCFAs to activate multiple signalling partners, including β-arrestins and the G_i/o and G_q/11 families of G proteins. b, Chemical structures of FFA2 activators and modulators. c, Quantification of the activation of G_i3 and G_q TRUPATH sensors induced by FFA2 activators and modulators. BRET, bioluminescence resonance energy transfer. d, Quantification of β-arrestin 2 recruitment and inhibition of forskolin-stimulated levels of cAMP induced by FFA2 activators and modulators. For c and d, data are the mean ± s.e.m. n = 3 (3 biologically independent experiments).

Fig. 2 ∣. Cooperativity of function between various FFA2 modulators and overall structures of FFA2 with diverse modulators.

a, Cooperativity of function between three FFA2 PAMs and the orthosteric agonist TUG-1375. The ability of the indicated concentrations of 4-CMTB, compound 187 and AZ-1729 to modulate the inhibition of forskolin-stimulated levels of cAMP through FFA2 is shown. The cooperative effects of AZ-1729 and TUG-1375 displayed reciprocity. b, Cooperativity of function between the three FFA2 ago-PAMs. The ability of the indicated concentrations of 4-CMTB to modulate the function of AZ-1729 and compound 187 is displayed. Different concentrations of compound 187 do not modulate the effects of AZ-1729. For a and b, data are the mean ± s.e.m. n = 3 (3 biologically independent experiments). c, Cryo-EM maps (top row) and structural models (bottom row) of TUG-1375-bound FFA2 signalling complexes, including 4-CMTB–FFA2 (forest green) and compound 187–FFA2 (green) in complex with miniG_q, and AZ-1729–FFA2 (blue) and compound 187–FFA2 (turquoise) in complex with G_i. mGα_i/s/q, Gα_i, Gβ and Gγ subunits are coloured in dark yellow, green-yellow, purple and pale turquoise, respectively. ScFv16 is in dark grey. The clear cryo-EM density maps of all ligands are shown as meshes in the insets of the top row. d, Superimposition of FFA2 bound to three ago-PAMs showing site 1 and the upper and lower regions of site 2 for 4-CMTB, AZ-1729 and compound 187, respectively.

Fig. 3 ∣. Binding of 4-CMTB in site 1.

a, Superimposition of 4-CMTB–FFA2 with A₁R bound to the PAM MIPS521 (Protein Data Bank (PDB) accession 7LD3). Compared with MIPS521, 4-CMTB binds more superficially. b, Details of the interactions between 4-CMTB and FFA2 at site 1. c,d, Simulation snapshot depicting the (S) isomer of 4-CMTB (c) and the less active (R) isomer of 4-CMTB (d) in site 1, with protein residues shown as sticks. 4-CMTB is shown as green sticks. Carbon colours and stick thickness represent the simulation-averaged energies of electrostatic and van der Waals interactions of each residue with the ligand, respectively. Classical and nonclassical hydrogen bonds with N230^6.43 are indicated by dashed lines. The results illustrate the inability of (R) 4-CMTB to form a nonclassical hydrogen bond with N230^6.43 and the poor steric fit of its isopropyl (i-Pr) fragment. e, Effects of various mutations in site 1 of FFA2 on the function of FFA2 ligands measured in [³⁵S]GTPγS-binding assays. 4-CMTB is unable to promote the binding of [³⁵S] GTPγS at N230D and N230S FFA2, but other ligands are unaffected. Data are the mean ± s.e.m. n = 3 (3 biologically independent experiments). WT, wild type.

Fig. 4 ∣. Binding of AZ-1729 and compound 187 at site 2.

a,b, Receptor interaction profiles of AZ-1729 (a) and compound 187 (b) in the upper and lower regions, respectively, of site 2. c, Different binding modes of AZ-1729 and compound 187. The area of each pocket was calculated using PyMol. d, Effects of mutations in site 2 on the function of FFA2 ligands measured by cAMP inhibition. Although these mutations affect the function of AZ-1729 and compound 187, they have little effect on the function of TUG-1375 and 4-CMTB. Data are the mean ± s.e.m. n = 3 (3 biologically independent experiments). e, Simulation snapshots of AZ-1729 (left) and compound 187 (right) bound to FFA2. Protein residues are shown with carbon colours and stick thickness denoting the simulation-averaged energies of electrostatic and van der Waals interactions of each residue with the ligand, respectively. Transparent iso-surfaces illustrate water presence in >50% of the simulation trajectory. Hydrogen bonds are shown with dotted lines. f, Distances between the E106^3.49 side-chain oxygens and R107^3.50 showing the disruption of their contact only in the presence of a G protein, AZ-1729 or compound 187 (mean ± s.e.m., n = 3 (3× 1-μs simulations)). g, Distribution of distances between the backbone and side-chain oxygen of S120^ICL2 and the donor nitrogen of AZ-1729 or the acceptor hydrogens of compound 187. Violin plots show probability densities from three simulation replicates per complex. Horizontal bars indicate minimum, maximum and median values. The dotted line at 3.5 Å shows the hydrogen-bond distance cut-off. Percentages indicate the proportion of frames with distances ≤3.5 Å.

Fig. 5 ∣. Noncanonical FFA2 activation mechanisms revealed by MD simulations.

a, Cartoons showing cryo-EM structures of FFA2–TUG-1375–G_i (PDB accession 8J22; grey), FFA2–4-CMTB–miniG_q (dark green), FFA2–compound 187–miniG_q (cyan) and FFA2–AZ-1729–G_i (blue) overlaid with surfaces representing FFA2 main chain atoms occupied at >75% of simulation time. Solid surfaces show cryo-EM-based simulations; meshes show docking-based simulations to structure 8J22. Top row, empty FFA2 (PDB accession 8J23; orange), FFA2–TUG-1375 (beige) and FFA2–TUG-1375–G_i (8J22; grey) demonstrate G protein but not orthosteric ligand stabilization of ICL2. Middle row, G-protein-free simulations with 4-CMTB (dark green), compound 187 (Cmpd187) (cyan) and AZ-1729 (blue) show ICL2 stabilization by AZ-1729 and compound 187, with compound 187 causing deviation of ICL2 from the cryo-EM structure. Bottom left, miniG_q-bound simulations with C4 (PDB accession 8T3S; olive), TUG-1375–4-CMTB (dark green) and TUG-1375–compound 187 (cyan), showing similar ICL2 conformations. Bottom middle, comparison of G_i-bound FFA2–TUG-1375 (grey) with FFA2–TUG-1375–AZ-1729 (blue) and FFA2–TUG-1375–compound 187 (red), revealing distinct ICL2 conformations compatible with G_i coupling. Bottom right, contrast of G_i-bound and G-protein-free simulations of FFA2–compound 187, showing the minimal impact of G_i binding on compound 187-induced ICL2 conformation. b, Distribution of distances between N230^6.43 side-chain nitrogen (N^δ2) and D269^7.49 carboxyl oxygens (O^δ1,δ2) across simulated complexes. Labels in parentheses indicate PDB accession numbers of the complexes. Violin plots show probability densities from three simulation replicates per complex. Horizontal bars indicate minimum, maximum and median values. The dotted line at 4.5 Å shows the interaction distance cut-off. Percentages indicate the proportion of frames with distances ≤4.5 Å. c, Snapshots of site 1 from MD simulations with 4-CMTB (top) and without (bottom), highlighting hydrogen bonds (dashed lines) among N230^6.43, C234^6.47, N265^7.45 and D269^7.49 residues critical for allosteric modulation.

Fig. 6 ∣. Distinct ICL2 conformations induced by AZ-1729 and compound 187 confer differential G-protein-subtype selectivity.

a, ICL2 conformations in FFA2–G_i complexes with AZ-1729 (orange) and compound 187 (light green), and FFA2–miniG_q complex with compound 187 (pink). Gα5 helices are coloured for Gα_i with AZ-1729 (green-yellow), Gα_i with compound 187 (light orange) and mGα_q/s/i (dark yellow). The red arrow indicates the uplifted ICL2 conformation with AZ-1729. Y117^ICL2 faces C351^H5.23 of Gα_i or Y243^H5.23 of mGα_q/s/i (Y356^H5.23 in wild-type Gα_q). b, Kernel density plots in two dimensional principal component analysis (2D-PCA) of ICL2 (P114–R121) Cα atoms for select simulations, with colours showing the iso-proportion density from 20%. Top left, FFA2–TUG-1375–AZ-1729–G_i (blue), FFA2–TUG-1375–AZ-1729 without G protein (red) and FFA2–TUG-1375 alone (grey). Top right, similar simulations for FFA2–TUG-1375–4-CMTB–miniG_q. Bottom left, simulations for FFA2–TUG-1375–compound 187–miniG_q. Bottom right, FFA2–TUG-1375–compound 187–miniG_q (purple), FFA2–TUG-1375–compound 187–G_i (yellow) and FFA2–TUG-1375–compound 187 without G protein (cyan). Coloured dots show simulation-average ICL2 positions; distances d₁ and d₂ connect the G_i-bound simulation to G-protein-free and miniG_q-bound states, respectively. ICL2 conformations show greater similarity between G_i-bound and G-protein-free states (d₁ = 1.42) than with G_q-bound states (d₂ = 3.41). c, Distribution of Cβ atom distances between Y117^ICL2 and Y243^H5.23 (mGα_q/s/i) or C351^H5.23 (Gα_i) across simulated complexes. Violin plots show probability densities from three simulation replicates per complex. Horizontal bars indicate minimum, maximum and median values. The dotted line at 10 Å shows the interaction distance cut-off. Percentages indicate the proportion of frames with distances ≤10 Å. d, TRUPATH assays using the G_q Y356^H5.23C sensor (Y243H5.23 in mGα_q/s/i) and wild-type FFA2, showing the potency and efficacy of indicated ligands. Data are the mean ± s.e.m. n = 3 (3 biologically independent experiments).