Molecular recognition and activation mechanism of short-chain fatty acid receptors FFAR2/3

Dear Editor,

Saturated short-chain fatty acids (SCFAs) (with carbon chains containing 1–6 carbon atoms) a aliphatic organic acids derived from gut bacterial fermentation products. Different bacterial taxa are associated with the production of distinct SCFAs, which mediate diverse biological effects (Supplementary information, Table S1). SCFAs act as agonists for free fatty acid receptors 2 and 3 (FFAR2/3), which belong to G protein-coupled receptor (GPCR) superfamily.¹ FFAR2/3 serve as messengers connecting gut microbiota and host, playing vital roles in regulating metabolism, inflammation and hormone homeostasis (Supplementary information, Fig. S1). They are also valuable targets for treating diabetes, obesity, asthma, allergies and inflammatory bowel disease (Supplementary information, Table S2).^1,2 FFAR2 is activated by acetate (AA), propionate (PA) and butyrate (BA), while FFAR3 is primarily activated by C3–C6 saturated fatty acids, including PA, BA, valerate (VA) and caproate (CA).¹ FFAR2 couples to Gα_i and Gα_q pathways, while FFAR3 primarily couples to Gα_i pathway. Considering the vital physiological functions of SCFAs, a paucity of knowledge about the recognition patterns of C2–C4 and C3–C6 SCFAs by FFAR2 and FFAR3 results in the urgent need to elucidate the molecular mechanisms behind FFAR2/3 signaling.

To obtain stable and homogeneous FFAR2/3–Gα_i complexes, we optimized human FFAR2/3 using the consensus design method,³ denoting them as FFAR2^C/3^C. We also added a BRIL tag to the N terminus of FFAR2^C/3^C. The active state structures of FFAR2^C/3 ^C in complex with either the G_i heterotrimer, endogenous agonist or synthetic agonists were further stabilized using the NanoBiT strategy⁴ and by the addition of single-chain antibody scFv16 (Supplementary information, Fig. S2a–c). We found that SCFA potency was similar for the modified and wild-type receptors (Supplementary information, Fig. S2d, e). All complexes were purified to homogeneity for single-particle cryo-electron microscopy (cryo-EM) analysis (Supplementary information, Fig. S3). The structures of FFAR2^C–Gα_iβ₁γ₂ in complex with AA or TUG-1375 were determined with resolutions of 2.6 and 3.2 Å (Fig. 1a; Supplementary information, Figs. S4, S5 and Table S3), respectively. A similar method was applied to obtain the BA- or VA-AR420626-bound FFAR3^C–Gα_iβ₁γ₂ complex structures at resolutions of 3.3 and 3.2 Å, respectively (Fig. 1a; Supplementary information, Figs. S6, S7 and Table S3).

Fig. 1. Structure basis of FFAR2/3 activation and ligand selectivity.

a Cryo-EM density maps of the FFAR2^C/3^C–Gα_iβ₁γ₂ complexes. b–d Interactions between AA (yellow) with FFAR2^C (cyan) (b), BA (red) with FFAR3^C (gold) (c) and VA (lightseagreen) with FFAR3^C (purple) (d). e Structural comparison of AA-bound FFAR2^C complex (cyan) with BA-bound FFAR3^C complex (gold). The distances between the α-carbon of SCFA ligands and γ-sulfur of C^4.57 are highlighted as red dashed lines. f Effects of the FFAR2 Y90^3.33F, FFAR2 L183^5.42M, FFAR3 F96^3.33Y and FFAR3 M188^5.42L mutations in responses to AA-, BA-, VA- or CA-induced cAMP reduction. g Extracellular view of the superimposed AA-bound (cyan) and TUG-1375-bound (blue) FFAR2^C structures. h The interactions between TUG-1375 (light blue) and residues in the binding pocket of FFAR2^C (blue) in the TUG-1375-bound FFAR2^C structure. i Sequence alignment of FFAR2 with other fatty acid receptors. j Side view of the allosteric binding pocket. AR420626 is colored chartreuse. k Side view of TM5, TM6 and TM7 of VA-AR420626-bound (purple) and BA-bound (gold) FFAR3^C structures. AR420626-induced the conformational rearrangements of TM5, TM6 and TM7. Polar interactions are indicated with the black dashed lines.

Despite differences in ligand structures, the active-state orthosteric pockets in the FFAR2^C/3^C complex structures share similar components, comprised of TM3, TM4, TM5, TM6 and TM7. Although the small size of AA, we observed clear EM density for AA in the orthosteric pocket of the AA-bound FFAR2^C structure. Molecular dynamics (MD) simulations revealed stable interactions between AA and the binding pocket (Supplementary information, Fig. S8a–d). AA was anchored to FFAR2^C through hydrogen bonds between the carboxylate group of AA and residues Y165^ECL2 and Y238^6.51 (superscripts indicate Ballesteros—Weinstein numbering for GPCRs) in FFAR2^C as well as through ionic interactions with the residues R180^5.39, H242^6.55 and R255^7.35 (Fig. 1b). Additionally, the methyl group of AA formed hydrophobic interactions with the residues C141^4.57, V144^4.60 and L183^5.42 (Supplementary information, Fig. S9a). These interactions were further confirmed by mutagenesis analysis (Supplementary information, Fig. S9d). To our knowledge, AA is the smallest agonist observed in GPCR complex structures. In the FFAR3^C structures, the carboxylate groups of BA and VA engaged the residues R185^5.39, H245^6.55 and R258^7.35 through ionic interactions and formed hydrogen bonds with the residues Y241^6.51 (Fig. 1c, d). MD simulations indicated that the carboxylic groups of BA and VA established persistent interactions with residues R185^5.39, R258^7.35 and Y241^6.51 (Supplementary information, Fig. S8e–l). These ligands were further consolidated by interactions between the aliphatic portions of BA/VA and residues in TM4 (V150^4.60 and C147^4.57), as well as the interaction of BA with TM3 (Y100^3.37) (Supplementary information, Fig. S9b, c). Mutations of these residues impaired the cAMP response of FFAR3^C (Supplementary information, Fig. S9e, f). Unlike the persistent contacts of the carboxylic acid groups of BA and VA with Y241^6.51, R185^5.39 and R258^7.35, the aliphatic chains of BA and VA in the FFAR3^C structures exhibited greater dynamic movements (Supplementary information, Fig. S8h, l). Interestingly, we observed that the carboxylate groups of SCFAs interacted with the residues Y90^3.33 in FFAR2^C and F96^3.33 in FFAR3^C through anion–π interactions,⁵ a previously unreported feature in GPCR structures. As a result, it is evident that FFAR2/3 exhibit a shared motif for SCFA recognition (Supplementary information, Fig. S9g).

Although FFAR2/3 share common agonists, such as PA and BA, structural comparison of the FFAR2^C/3^C complexes revealed distinct conformations of the active-state orthosteric pocket and the orientation of the ligand-interacting residues at the extracellular vestibule (Fig. 1e). Sequence alignments of FFAR2, FFAR3 and FFAR1 highlighted highly conserved residues within the orthosteric pocket (Supplementary information, Fig. S10). Furthermore, the residue^4.57 and residue^4.60 of TM4 in FFARs play crucial roles in determining the fatty acid length. The small-sized residue substitution of these two residues in FFAR2 and OR51E2 enables accommodation of longer-chain fatty acids.^6,7 In the AA-bound FFAR2^C structure, the hydrogen-bonding interaction between residues Y90^3.33 and R255^7.35 was observed in the AA-binding pocket, creating a tight orthosteric pocket. In contrast, in the BA-bound FFAR3^C structure, the corresponding residue F96^3.33 did not form a hydrogen-bonding interaction with R258^7.35, leading to the outward movement of Y241^6.51 (relative to Y238^6.51 in the FFAR2^C structure) and an expanded ligand binding pocket (Fig. 1e; Supplementary information, Fig. S9h). Consequently, the distance between the α-carbon of the AA and γ-sulfur of C^4.57 in the AA-bound FFAR2^C structure was 5.7 Å, shorter than the corresponding 7.9 Å distance in the BA-bound FFAR3^C structure. Additionally, the side chain of residue M188^5.42 in FFAR3^C underwent an outward movement, away from the orthosteric pocket to accommodate the longer fatty acids compared to L183^5.42 in FFAR2^C (Fig. 1e). Functional assays revealed that FFAR2 Y90^3.33F and L183^5.42M mutants reduced the binding affinity to AA and increased the binding affinities to BA, VA and CA. Conversely, FFAR3 F96^3.33Y and M188^5.42L mutants increased the binding affinity to AA and decreased the binding affinities to BA, VA and CA. (Fig. 1f). Furthermore, the side chain of V144^4.60 in the AA-bound FFAR2^C structure resided between TM3 and TM4, blocking the extension of longer-chain length SCFAs (Fig. 1e). In contrast, in the BA-bound FFAR3^C structure, the side chain of V150^4.60, which corresponds to residue V144^4.60 in the AA-bound FFAR2^C structure, rotated ~50° towards TM5, creating additional space between TM3 and TM4 to accommodate SCFAs with longer chain lengths. These conformational rearrangements in TM3, TM4, TM6 and TM7 resulted in a larger orthosteric binding cavity in FFAR3^C, compared to FFAR2^C (Fig. 1e; Supplementary information, Fig. S9h). The orthosteric binding pocket of FFAR3^C had a significantly larger surface-accessible volume compared to FFAR2^C, measuring 148 ų for FFAR3^C and 68 ų for FFAR2^C (Supplementary information, Fig. S9i, j). Comparison of FFAR1, FFAR2^C and FFAR3^C structures revealed that FFAR1’s ligand selectivity mechanism was like that of FFAR3 (Supplementary information, Fig. S11a), but a larger space between TM3 and TM4 in FFAR1 allowed for the accommodation of larger ligands (Supplementary information, Fig. S11b, c).⁸

To uncover molecular recognition mechanism of synthetic agonist selectivity of FFAR2 and facilitate the design of selective and potent FFAR2 agonists, we determined the cryo-EM structure of the TUG-1375-bound FFAR2^C–Gα_i complex. Compared to the structure of AA-bound FFAR2^C, the TUG-1375-bound FFAR2^C exhibited a sharp kink around the residue G139^4.55, resulting in a dramatic outward movement of TM4 for ~ 5.2 Å (relative to the α-carbon of I146^4.62), creating a significantly larger orthosteric binding pocket (Fig. 1g). These results suggest that the FFAR2^C binding pocket has high plasticity for accommodating large agonists. TUG-1375 fitted into the hydrophobic pocket created by residues Y90^3.33, V144^4.60, V175^5.34, V176^5.35, V179^5.38, L183^5.42 and Y238^6.51 (Supplementary information, Fig. S12a). The residues R180^5.39 and R255^7.35 formed polar interactions with TUG-1375, respectively (Fig. 1h; Supplementary information, Fig. S12a). These interactions were confirmed by MD simulations (Supplementary information, Fig. S8m–o) and cAMP response assay (Supplementary information, Fig. S12b). Compared with natural SCFAs, the distinct binding mode of TUG-1375 suggested that more structurally diverse compounds could be exploited as selective FFAR2 agonists. Sequence alignment indicated that at least three residues of TM3–TM7, including Y90^3.33, V175^5.34 and V176^5.35, are distinct compared to other GPCRs known to sense fatty acids (Fig. 1i). Mutation of these residues in FFAR2 to allelic residues in FFAR1, FFAR3, FFAR4, GPR84 and OR51E2 receptors (such as Y90^3.33F/T, V175^5.34I/S/G/Q/P and V176^5.35A/L/E/N) resulted in decreased FFAR2 activity in response to TUG-1375 (Supplementary information, Fig. S12b).

The outward sharp turn at the toggle switch of TM6 is a universal activation feature of GPCRs. In FFAR2/3, the toggle switch residue was the phenylalanine F^6.48 (Supplementary information, Fig. S13a). Mutation of these residues impaired the cAMP response (Supplementary information, Fig. S13b). The PIF triad connector (P191^5.50T97^3.40F231^6.44 in FFAR2 and P196^5.50A103^3.40F234^6.44 in FFAR3) formed tight packing interactions for maintaining the active state (Supplementary information, Fig. S13c). Moreover, the activation hallmarks in class A GPCRs were also observed in FFAR2/3, including the rearrangement of the E/DR^3.50Y and N^7.49P^7.50XXY^7.53 motifs.⁹ The residues R^3.50 of the E/DR^3.50Y motifs displayed similar conformations in both receptors (Supplementary information, Fig. S13d–g) and formed hydrogen bonds with the α5-helix from the Gα_i (Supplementary information, Fig. S14b, c). Furthermore, the side chains of R^3.50 extended towards TM7 and closely packed with the N^7.49P^7.50XXY^7.53 (D269^7.49PLLF in FFAR2 and D272^7.49PFVY in FFAR3) motifs (Supplementary information, Fig. S13d–g), leading to closer interaction between the receptor and Gα_i.

The structures of FFAR2^C/3^C–Gα_i complexes revealed that the interactions between Gα_i and FFAR2^C/3 ^C were mainly contributed by TM2, TM3, TM5, TM6 and ICL2 in FFAR2^C, and by TM3, TM5, TM6, ICL1 and ICL2 in FFAR3^C. Comparison of the FFAR2^C/3 ^C structures exhibited the TM5 and TM6 of FFAR3^C having 10.6° and 14.5° anticlockwise rotations, respectively, relative to those of FFAR2^C (viewed from intracellular to extracellular direction), resulting in a larger cytoplasmic cavity in FFAR3^C to accommodate the α5-helix from Gα_i (Supplementary information, Fig. S14a). While FFAR2^C/3^C share common structural features, they exhibit significant differences in their Gα_i coupling interfaces. In the AA-bound FFAR2^C–Gα_i complex structure, the hydrogen-bonding interaction was established between the residue R107^3.50 and the residue C351 in Gα_i’s α5-helix. Additionally, residue D350 in Gα_i’s α5-helix formed hydrogen-bonding and ionic interactions with H45^2.40 and R121^ICL2, respectively (Supplementary information, Fig. S14b). Furthermore, the hydrophobic side chains of L353, L348, I344 and I343 in Gα_i’s α5-helix were oriented towards the hydrophobic pocket (V111^3.54, F202^5.61, M206^5.65, A220^6.33, L223^6.36 and P114^ICL2) of FFAR2^C. Notably, mutations in these residues impaired the activity of AA on FFAR2 (Supplementary information, Fig. S14d). Similarly, in the BA-bound FFAR3^C–Gα_i complex structure, the residues L353, L348 and I344 of Gα_i’s α5-helix participated in extensive hydrophobic interactions with FFAR3^C, involving residues V117^3.54, Y204^5.58, L207^5.61, L211^5.65, V223^6.33, L226^6.36 and P120^ICL2. Furthermore, the residues C351 and D350 of Gα_i formed hydrogen-bonding and ionic interactions with R113^3.50 and R46^ICL1 in FFAR3^C (Supplementary information, Fig. S14c). These interactions between FFAR3 and Gα_i’s α5-helix have been confirmed through cAMP response assays (Supplementary information, Fig. S14e).

The allosteric modulators have a high subtype selectivity and can fine-tune receptor activation.¹⁰ To date, only a few crystal and cryo-EM structures of GPCR-positive allosteric modulator (PAM) complexes have been determined. Notably, these PAM-binding sites are typically situated at the membrane-embedded binding site, the outer membrane-contacting surface or the extracellular side of GPCRs (Supplementary information, Fig. S15). AR420626 is a selective PAM of FFAR3.¹¹ The structure of VA-AR420626–FFAR3^C revealed an unusual PAM-binding pocket located at the cavity of the intracellular side, composed of TM3, TM5, TM6 and TM7 (Supplementary information, Fig. S16a). Intriguingly, the binding mode of AR420626 bears resemblance to the localization of PCO371 in PTH1R and SBI-553 in NTSR1.^12–14 The AR420626-binding pocket primarily comprised hydrophobic residues and resided beneath the P^5.50A^3.40F^6.44 triad connector (Fig. 1j). The dichlorophenyl ring of AR420626 pointed to the interval between TM5 and TM6, forming the π–π interaction with the residue Y204^5.58. Furthermore, AR420626 was involved in extensive hydrophobic interactions with L54^2.43, L106^3.43, L231^6.41 and V275^7.52. Additionally, the residues R113^3.50 and T230^6.40 established polar interactions with AR420626, respectively (Fig. 1j; Supplementary information, Fig. S16b). Consistent with our observations, mutations of these residues significantly diminished PAM activity on FFAR3 due to loss of the interactions between receptor and PAM (Supplementary information, Fig. S16c). Alignment of VA-AR420626- and BA-bound FFAR3^C structures revealed noticeable PAM-induced conformational changes of FFAR3^C, including the inward movement of TM5 and the outward movement of TM6 and TM7 in their intracellular parts (Supplementary information, Fig. S16d). The interaction of AR420626 with FFAR3^C lead to a 30° anticlockwise rotation of toggle switch F238^6.48. Moreover, AR420626’s binding induced an outward movement and side chain rotation of Y276^7.53, along with a 30° anticlockwise rotation of Y204^5.58. The side chains of L231^6.41 and T230^6.40 underwent an outward movement, away from the PAM-binding pocket to accommodate AR420626 (Fig. 1k). These findings indicate an induced-fit conformational change between PAM and FFAR3^C, involving the rearrangements of TM5, TM6 and TM7, ultimately maintaining FFAR3^C in a more active-like state. Importantly, we discovered that the residue L353 of Gα_i’s helix5 formed direct hydrophobic interaction with AR420626, which was also observed in previously determined PCO371-bound PTH1R–Gα_s structures.^12,13 The Gα_i’s helix5 further stabilized the interactions of AR420626 with the key interacting residues (R113^3.50 and T230^6.40) in the intracellular cavity (Supplementary information, Fig. S8p–t).

In our study, the cryo-EM structures reveal that FFAR2/3 utilize a delicate mechanism to differentiate between various SCFAs. The TUG-1375-bound FFAR2^C complex reveals an outward movement of TM4 around G139^4.55, indicating the orthosteric pocket’s plasticity for recognizing diverse synthetic ligands. This finding provides important insight for drug design targeting GPCRs containing residue G^4.55, including 5-HT_2A, 5-HT_2B, 5-HT_2C, A_2B, GPR35, NTR2, P2RY4, LPAR6 and δ opioid receptor. Furthermore, we identified an unexpected PAM-binding pocket in FFAR3 and unveiled a direct interaction between AR420626 with Gα_i protein, which provided evidence of a biased signalling mechanism that targets the receptor–transducer interface. The FFAR2/3 structures complete a final piece in the puzzle on how FFARs recognize fatty acids of various chain lengths^8,15 and reveal the molecular recognition mechanism of FFAR2/3 towards orthosteric and allosteric ligands, and facilitate rational drug design targeting FFAR2/3.

Author contributions

J.W. organized the project. J.W., F.L., L.T. and X.L. guided all the structural analyses. J.W. and F.L. designed all the mutants for functional analyses. F.L., Q.G., Z.Z., S.M., S.T. and T.W. developed the FFAR2/3 constructs and optimized protein expression. F.L. performed protein expression. F.L., X.S. and D.G. prepared samples for cryo-EM. F.L. prepared the cryo-EM grids and collected the cryo-EM data. L.T. performed cryo-EM map calculation, model building, and structure refinement. W.T. performed the MD simulations. Z.L., X.Z. and M.Y. performed the functional assays. J.W. and F.L. wrote the manuscript. All the authors have read and commented on the manuscript.

Data availability

The cryo-EM density maps have been deposited in the Electron Microscopy Data Bank (EMDB) and Protein Data Bank (PDB) under accession numbers EMD-35944 and 8J24 for AA–FFAR2^C complex; EMD-35942 and 8J22 for TUG-1375–FFAR2^C complex; EMD-35941 and 8J21 for BA–FFAR3^C complex; EMD-35940 and 8J20 for VA-AR420626–FFAR3^C complex.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41422-023-00914-z.