Introduction

Prostanoids are a class of oxygenated arachidonic acid metabolites that include prostaglandin F2α (PGF2α), prostaglandin D2 (PGD2), prostaglandin E2 (PGE2), thromboxane A2 (TXA2), and prostacyclin (PGI2). They provoke diverse biological actions in many tissues and cell types through direct interactions with nine lipid G-protein-coupled receptors (GPCRs), prostaglandin F2α receptor (FP), prostaglandin D2 receptors (DP1-2), prostaglandin E2 receptors (EP1-4), thromboxane receptor (TP) and prostacyclin receptor (IP1)1, which comprise a subfamily of class A GPCRs. FP is encoded in humans by the PTGFR gene2. Stimulated by PGF2α, FP plays a pivotal role in regulating inflammation, allergic responses, intraocular pressure, and blood pressure, making it a valuable target for therapeutic discovery and development1,3,4,5. FP is highly expressed in uterine myometrium, eye, smooth muscle, skin, and ovarian6,7. Upon agonist stimulation, FP is predominantly coupled to the Gq subtype of G proteins, which activation leads to subsequent PKC activation and a transient calcium signaling in response to the formation of inositol triphosphate8,9. In addition to Gq, FP activation also induces activation of small G protein Rho via G12/G1310 and activation of Raf/MEK/MAP kinase pathway through Gi11.

Given the diverse functions of FP, it has been targeted for therapeutic development. PGF2α, the endogenous FP activator, entered a clinical trial for glaucoma treatment12. However, its clinical application was limited by intolerable side effects, possibly caused by its low selectivity for FP receptor12. Since then, selective FP agonists have attracted extensive attention and have been developed for the treatment of glaucoma5, scalp alopecia13, and vitiligo14. From 1996 to 2012, several FP-selective prostaglandin analogs (PGAs) were approved by the United States Food and Drug Administration (FDA) for glaucoma treatment. As a prodrug of a selective FP receptor agonist, latanoprost (LTP) was approved to treat glaucoma for the first time in 1996. It has also been used to treat scalp alopecia15 and vitiligo14 in recent years. Latanoprost acid (LTPA), 17-phenyl-13,14-dihydro PGF2α, is an active metabolic form of LTP. Another fluorinated PGA prodrug, tafluprost (TFP), was first approved for the treatment of glaucoma in 201216. Tafluprost acid (TFPA), 15-deoxy-15,15-difluoro-16-phenoxy PGF2α, is the active metabolic form of TFP. Due to their high efficacy, these PGAs have been the first-line drug in clinics for the treatment of glaucoma. However, through post-marketing surveillance, 5–20% of patients suffered side effects such as conjunctival congestion and headache, including intolerance responses toward these PGAs17,18,19. Improving the selectivity of these drugs to the FP receptor and elucidating the molecular mechanisms underlying the functional selectivity of individual prostanoid receptor family members are highly important and clinically relevant.

Extensive efforts have been made to clarify how the binding of endogenous and synthetic ligands with various pharmacological profiles regulate FP’s downstream signaling8,9,10,11. However, the molecular details defining the binding modes of ligands remain largely unknown, which is partly attributed to the scarcity of the structural information on ligands bound FP complex. Understanding the mechanism of prostaglandin-FP signaling and identifying differences in the ligand selectivity of prostaglandin receptors may assist in the development of selective drugs with improved safety.

Here we present three cryo-EM structures of G protein-coupled FP in complex with its endogenous ligand PGF2α and with two synthetic agonists, LTPA and TFPA, at global resolutions of 2.67, 2.78, and 3.14 Å. Combined with functional characterizations of mutated receptors, these structures reveal conserved and divergent mechanisms of ligand binding, receptor activation, and G protein coupling by FP.

Results

Cryo-EM analysis and overall structure

To facilitate the expression of FP complexes, we introduced a BRIL tag to the N-terminus of the wild-type (WT) receptor20,21. A Gαq chimera was engineered based on the mini-Gαs scaffold with an N-terminal replacement of corresponding sequences of Gαi1 to facilitate the binding of scFv1622,23,24. Hereinafter, Gαq reference to Gαq chimera. The FP-Gq complex was further stabilized by the NanoBiT strategy25. Incubation of PGF2α/LTPA/TFPA with membranes from cells co-expressing receptors and heterotrimer Gq proteins in the presence of scFv16 and Nb35 enables efficient assembly of the PGF2α/LTPA/TFPA -FP-Gq complexes, which produces highly homogenous complex samples for structural studies26 (Supplementary Figs. 1–3, Table 1). The structures of the FP-Gq-scFv16-Nb35 complexes with PGF2α, LTPA, and TFPA were determined by cryo-EM to the resolutions of 2.67, 2.78, and 3.14 Å (Fig. 1a, Supplementary Figs. 1–4). The high-quality density map allowed unambiguous model building for the receptor structure containing residues 29–323, except for two invisible residues in the intracellular loop 3 (ICL3) (residues 238 and 239). The density maps are also clear for three agonists, most residues of the Gq heterotrimer, scFv16, and Nb35 (Fig. 1b, c Supplementary Fig. 4).

Fig. 1: Cryo-EM structures of FP–Gq complexes.
figure 1

a Cryo-EM density of FP-Gq in complex with PGF2α, LTPA, or TFPA. FP in Medium Turquoise, Gαq in light Coral, Gβ in slate gray, Gγ in gold, ScFv16 in dark magenta, Nb35 in Olive, PGF2α in purple, LTPA in yellow, and TFPA in light coral. b, c Cartoon representation of the PGF2α-FP-Gq complex. FP in Green, Gαq in brown, Gβ in slate gray, Gγ in gold, ScFv16 in dark magenta, Nb35 in Olive, and PGF2α in purple. d, e Comparison of the FP–Gq complex with the EP2–Gs complex (PDB ID: 7CX2) and the EP4–Gs structure (PDB ID: 7D7M). FP shows a conserved conformation of ECL2 (green) with the other two receptors. The orientation of H8 of FP is different from the Gs-coupled receptor EP2 and EP3, and the ligand binding pose PGF2α to FP is different from PGE2 to EP2.

Full size image

The overall structure of the active FP receptor is highly similar to those of active EP2 (PDB code: 7CX2) and EP4 (PDB code: 7D7M), with root mean square deviation (RMSD) values of 1.27 and 1.23 Å, respectively. FP folds into a canonical seven-transmembrane helical domain (TMD). All three extracellular loops (ECLs 1–3) were well defined, where ECL2 forms a β-hairpin loop, which is stabilized by the highly conserved disulfide bond between C186ECL2 and C1083.25. The β-hairpin ECL2 of FP resembles those of EP2 and EP4, and tightly caps the extracellular region (Fig. 1d). We superpose the TMD structures of PGF2α-bound FP with PGE2-bound EP2 and EP4 to compare their overall receptor conformations and ligand-binding pockets. A notable structural difference occurs in the H8 of these receptors. The H8 of FP is almost perpendicular to that of EP2, with a rotation of 82.3°, and is closer to the cell membrane compared with the H8 of EP4. (Fig. 1e). In addition, although these ligands share a similar chemical scaffold and relatively conserved binding pocket in these three receptors, PGF2α displays a distinct binding pose from PGE2 in these complexes (Fig. 1e), as detailed below.

The PGF2α binding pocket of FP

The endogenous ligand PGF2α is mainly composed of three parts, a carboxyl group-containing α-chain, a five-membered ring (F ring) with two hydroxyl groups, and a hydrophobic ω chain with one hydroxyl group at ω6 position (Fig. 2a). The cryo-EM map enabled the unambiguous assignment of PGF2α within the receptor pocket. PGF2α is well resolved in the FP ligand-binding pocket by adopting an L-shape conformation with its carboxyl group-containing α-chain fitting into a hydrophilic sub-pocket near the top of the receptor (Fig. 2a), formed by residues from TM1, TM7, and ECL2 (Fig. 2b). The carboxyl group in the α-chain of PGF2α forms a salt bridge with R2917.40 and hydrogen bonds with T184EL2 and Y922.65. Mutations of residues R2917.40, T184EL2, and Y922.65, which are highly conserved in prostanoid receptors and participate directly in receptor binding, lead to decreased activity of PGF2α (Fig. 2b–d, Supplementary Fig. 5-6). Although electrostatic contacts are the major driving force for the interactions between the α-chain and the positively charged binding pocket of FP, several hydrophobic residues also play important roles. Particularly, we did see the side chain of M1153.23 forms lone pair-π interaction with ethylene linkage in the α-chain (Fig. 2b, c, Supplementary Fig. 6). Mutating this methionine in FP to alanine decreased the affinity for PGF2α by approximately 100-fold (Fig. 2d, Supplementary Fig. 5). The F ring is located in a sub-pocket formed by TM1, TM2, and TM7 (Fig. 2b). These two hydroxyl groups in F ring mainly participate in polar interactions with the receptor. Particularly, these hydroxyl groups form hydrogen bonds with S331.39 and T2947.43, which are not conserved among prostanoid receptor family members, indicating that this part mainly contributes to the selectivity of PGF2α to FP. Meanwhile, mutation of S331.39 or T2947.43 in FP to alanine significantly impaired the affinity to PGF2α (Fig. 2d, Supplementary Fig. 5). Besides α chain and F ring, the ω chain penetrates into the hydrophobic pocket formed by TM5, TM6, and TM7. This alkyl chain forms hydrophobic interactions with F2055.41, W2626.48, F2656.51, and L2907.39. Mutations of these residues in FP to alanine significantly impaired the affinity to PGF2α (Fig. 2d, Supplementary Fig. 5). Besides hydrophobic interaction, the ω6 hydroxy group forms a hydrogen bond with H812.54, which mutated to alanine also caused a dramatically reduced the activity of PGF2α to FP by over 1000 folds (Fig. 2d, Supplementary Fig. 5). Collectively, the α-chain and ω-chain that bind to the sub-pockets with highly conserved residues, mainly contributes to high receptor binding affinity, while the F ring could be important for receptor selectivity.

Fig. 2: The PGF2α binding pocket of FP.
figure 2

a Vertical cross-section of the PGF2α-binding pocket in FP. b Corresponding interactions that contribute to PGF2α binding in FP. The hydrogen bond is depicted as a red dashed line. c Schematic representation of the interactions between FP and PGF2α in 2D format and region division in FP–PGF2α binding pockets corresponding to the structure of PGF2α (F ring, ω chain, and α chain). d IP1 accumulation assay of key mutants in FP that bind to PGF2α (∆pEC50 = pEC50 of PGF2α to specific Mutant-pEC50 of PGF2α to WT, Yellow column means ∆pEC50 ≤ −1, Blue column means ∆pEC50 ≤ −2, Red column means ∆pEC50 ≤ −3). Data are presented as mean values ± SEM; n = 3 independent samples; significance was determined with a two-sided unpaired t-test; n.s. no significant; *p < 0.05; **p < 0.01; ***p < 0.001. Exact p values and Source data are provided as a Source Data file.

Full size image

Specific engagement of LTPA and TFPA with FP

Even though FP is an important therapeutic target for many diseases, the poor selectivity of PGF2α has hampered its clinical application. LTPA and TFPA, two synthetic relative selective FP agonists, have been widely used in clinical treatment for diseases including glaucoma. LTPA showed high affinity to FP with an EC50 of 3.6 nM but only moderate potency to EP1 and EP3, with an EC50 of 6.9 and 17 μM, respectively27. Compared to PGF2α, the selectivity was enhanced by more than 100 times27. TFPA is the most potent FP agonist (EC50: 0.4 nM) and has a fairly low potency to the other members of the prostanoid receptor family except for EP3 (EC50: 67 nM)28. LTPA, TFPA, and PGF2α all have shared α-chain and F ring, while LTPA and TFPA have bulky substitutions in the ω-chain compared to PGF2α (Fig. 3 a-e). LTPA and TFPA bond to FP in nearly identical binding poses as PGF2α did, displaying a similar “L”-shape configuration in these two solved structures (Fig. 3a). Likewise, the carboxylate groups of LTPA and TFPA form strong polar interactions with R2917.40, T184EL2, and Y922.65, the ethylene linkage in the α-chain forms a lone pair-π29 interaction with M1153.23, and the two hydroxy groups in the F ring form hydrogen bonds with S331.39 and T2947.43 (Fig. 3b, c). Consistent with this observation, mutations of these residues in FP to alanine significantly decreased the potency of LTPA and TFPA to FP.

Fig. 3: Specific engagement of LTPA and TFPA with FP.
figure 3

a Superimposition of the PGF2α bound FP with LTPA bound FP and TFPA bound FP aligned at the ligand binding pocket. b, c Detail interactions of LTPA-FP and TFPA-FP were shown. H-bonds were depicted as red dashed lines. d, e 2D representation of contacts in LTPA-FP and TFPA-FP. The polar bonds were presented by red dotted lines. Conserved residues that formed interactions with ligands are presented in a red oval frame. f Sequence alignment of prostanoid receptors. Hydrophobic residues are in yellow, polar charged residues in blue, and polar uncharged residues in green. g, IP1 accumulation assay of key mutants in FP that bind to LTPA or TFPA (∆pEC50 = pEC50 of agonists to specific Mutant FP-pEC50 of PGF2α to WT FP, Yellow column means ∆pEC50 ≤ −1, Blue column means ∆pEC50 ≤ −2). Data are presented as mean values ± SEM; n = 3 independent samples; significance was determined with two-side unpaired t-test; n.s. no significant; *p < 0.05; **p < 0.01; ***p < 0.001. Exact p values and Source data are provided as a Source Data file.

Full size image

The derivatized ω-chain in LTPA also directly binds to FP. The phenyl group in the ω-chain of LTPA packs against F2055.41, F2656.51, F187ECL2, L2907.39, and W2626.48 with hydrophobic interactions. These residues are highly conserved in FP, EP1, and EP3. Interestingly, besides forming a hydrogen bond with H812.54, the hydroxyl group in the ω chain forms an additional hydrogen bond with S1183.35, which only exists in FP (Fig. 3d–f), indicating that this residue is of vital importance for the selectivity of LTPA to FP. Consistent with this finding, mutation of this serine to alanine or asparagine significantly reduced the potency of LTPA to induce FP activation (Fig. 3g, Supplementary Fig. 5).

Similar to LTPA, the phenyl group in the ω-chain of TFPA forms extensive hydrophobic interactions with FP. Moreover, the oxygen atom forms a hydron bond with the side chain of Q297, and the two fluoride groups in the carbon 12-position of TFPA form three hydrogen bonds with several residues in FP30, which may explain its high potency to FP. One fluoride group in TFPA forms hydrogen bonds to H812.54 and S1183.35, as the hydroxyl group in LTPA did. Notably, the other fluoride group forms a hydrogen bond with N842.57(Fig. 3e, Supplementary Fig. 6). This residue is highly diverse among prostanoid receptor family members. Intriguingly, among the other 8 prostanoid receptors, only EP3 and TP harbor a similar polar uncharged residue, threonine (Fig. 3f). Through structural analysis, mutation of N842.57 to threonine in FP could still form a hydrogen bond with the fluoride group (Supplementary Fig. 6). This explains its strong affinity for EP3. Consistent with this prediction, mutation of this residue to threonine does not significantly reduce ligand binding of TFPA, but mutating to alanine does (Fig. 3g, Supplementary Fig. 5). This hydrogen bond formed between T2.57 in EP3 and TFPA may contribute to the high affinity of TFPA toward EP328.

The active structure of FP

Structural comparison of the PGF2α-bound FP-Gq complexes with the antagonist-bound TP (PDB: 6IIU)31 supports the notion that FP in these structures is in the active state, featured by the outward displacement of the cytoplasmic end of TM6, the hallmark of class A GPCR activation, and concurrently inward shift of TM7. In addition, TM5 of FP laterally shifts relative to that of antagonist-bound TP. These conformation changes largely resemble that of the Gq-coupled 5-HT2AR complex (PDB: 6WHA)22 (Fig. 4a), but the outward amplitude of TM6 of FP is smaller than that of 5-HT2AR (Fig. 4a).

Fig. 4: The active structure of the FP.
figure 4

a Superimposition of Gq-coupled FP with the Gq-coupled 5-HT2AR complex and antagonist-bound TP (PDB: 6IIU). FP in Green, 5-HT2AR in brown, and TP in purple. b Superimposition of Gq-coupled FP with Gq-coupled 5-HT2A aligned at the PIF motif. c Superimposition of Gq-coupled FP with antagonist-bound TP aligned at the LLW motif. The left panel is a magnified view of the LLW motif. The right panel is the top view of the LLW motif. Residues are shown in the sticks, with the correspondent cryo-EM density represented in the mesh. d Superimposition of Gq-coupled FP with antagonist-bound TP aligned at the D(E)RY(C) motif and RQR motif. The left panel is a magnified view of the D(E)RY(C) motif’ and RQR motif. The right panel is the top view of D(E)RY(C) and RQR motifs. Residues are shown in the sticks, with the correspondent cryo-EM density represented in the mesh.

Full size image

Structure comparison of the active FP in complex with PGF2α and the inactive TP bound to an antagonist provide clues for understanding the activation mechanism of FP. Upon PGF2α binding, the ω-chain of PGF2α approaches the toggle switch residue W2626.48 to trigger the downward displacement of W2626.48 by 2.4 Å (Fig. 4b, c). The movement of W2626.48 further constitutes a hydrophobic LLW core triad, which is comprised of L1233.40, L2135.49, and W2626.48, to fasten TM3, TM5, and TM6. The importance of the LLW core triad is functionally supported by the decreased potency of PGF2α to activate L1233.40A and L2135.49A FP mutants (Fig. 4c, Supplementary Fig. 7). Notably, the canonical P5.50 I3.40 F6.44 motif in majority of class A GPCRs26,32 is replaced by G2145.50 L1233.40 S2586.44. In our structure of active FP, G2145.50 is sterically apart from S2586.44 and L1233.40, leading to the lack of the conserved PIF hydrophobic triad, thus indicating that FP employs an activation mechanism not mediated by the traditional PIF motif (Fig. 4b).

The receptor activation also accompanies the arrangement of ionic lock (D/E3.49 R3.50 Y3.51, E1323.49 R1333.50 C1343.51 in FP), leading to the broken of the salt bridge between E1323.49 and R1333.50 and the stretching of the R1333.50 side chain towards TM6 and the latter’s outward displacement of 6.7 Å compared with that of inactive TP (measured at Cα of residue 6.30). C1343.51 in the ERC motif moves inwards, forming interactions with I2225.58 and T2235.59 in TM5. The importance of motif in FP activation is evidenced by the decreased PGF2α activity on FP mutants of E1323.49, R1333.50, and C1343.51 (Supplementary Fig. 7). Intriguingly, unlike its cognate residue R3128.47 in TP, the side chain of R3088.47 in FP undergoes a large-scale upward rotation and forms a hydrogen bond with Q2506.36, which leads to the inward shift of the cytoplasmic end of TM7 and the upshift of H8 (Fig. 4d). The importance of this hydrogen bond in FP activation is evidenced by the loss of PGF2α activity on FP mutants of R1333.50A, Q2506.36A, and R3088.47A. (Supplementary Fig. 7) Noteworthily, the arginine at position 8.47 is highly conserved across prostaglandin receptors, while Q6.36 only exists in several class A receptors, including EP1, EP3, and FP, indicating that this interaction network could also exist in active structures of EP1 and EP3.

FP-Gq coupling

The notable outward displacement of TM6 at the cytoplasmic side opens a cavity to accommodate the Gαq subunit. Structural comparisons of FP-Gq with Gq-coupled CCKAR and 5HT2AR reveal a difference in the conformations of TM1, TM6, and Gαq subunits among these Gq-coupled GPCR complexes (Fig. 5a). The cytoplasmic end of FP TM1 and TM6 undergoes a remarkably inward displacement relative to CCKAR and 5HT2AR. Consequently, the C-terminus of the α5 helix of the Gαq subunit in the FP-Gq complex rotates toward TM7 and H8 to avoid clashes with TM6 and forms extra hydrophobic interactions with side chains of residues in H8, accompanied by the rotation of the entire Gαq subunit (Fig. 5a). This alteration triggers a 14° tilt of the αN helix of Gαq, bringing it closer to the cytoplasmic end of TM4 in the FP-Gq complex relative to CCKAR-Gq and 5HT2AR-Gq complexes (Fig. 5a). In addition to coupling Gq by interactions with H8 and TM4, the engagement of Gq is also maintained by interactions with FP from TM1, TM2, TM3, TM5, TM4, TM6, ICL1, and ICL2. The side chain of E1323.49 of the E3.49 (D) R3.50C3.51 (Y) motif in TM3 makes a direct hydrogen bond interaction with the side chain of the Y356 of α5 (Fig. 5b, c).

Fig. 5: FP–Gq coupling.
figure 5

a The structures of Gq-coupled FP, 5-HT2AR (PDB ID: 6WHY), and CCK1R (PDB ID: 7MBY) complexes were superimposed based on TM2, TM3 and TM4, FP is shown in light sea green, 5-HT2AR in blue and CCK1R in yellow. This panel is shown with views an orthogonal view (left) and a cytoplasmic view (right); the red arrows indicate the tilt of the α5 helix of Gαq from the FP–Gq complex compared to the 5-HT2AR–Gq or CCKA1–Gq complexes. b the residues in FP, 5-HT2AR, and CCKA1 that contact Gq. c–e The detailed interactions of ICL2 with Gαq (c), TM6 with the α5 helix of Gαq (d) and TM1, TM2, and ICL1 with the αN and α5 helices of Gαq (e). The hydrogen bonds are depicted as dashed lines.

Full size image

Like other Gq-coupled GPCR33, the ICL2 of FP facilitates broad interactions with Gq to stabilize the complex. Typically, ICL2 adopts a helix conformation and inserts into the groove formed by the αN, β2–β3 loop, and α5 helix of Gαq. Interestingly, we identified several distinct interactions between TM6 and Gαq. For instance, the H2446.30 forms a hydrogen bond interaction with the side chain of Q350 of Gαq, and E2466.32 forms a hydrogen bond interaction with the side chain of N357 of Gαq (Fig. 5d). Mutating of H2446.30 and E2466.32 to alanine destabilizes the complex and reduces the IP1 accumulation of FP (Fig. 5d, Supplementary Fig. 6). Most notably, unlike other solved Gq-coupled GPCR structures, the extensive interactions of TM1, ICL1, and TM2 of FP with Gq play an important role in stabilizing the complex (Fig. 5e). F58 packs against L358 and V359 of Gαq stabilized by hydrophobic interaction, and K63 packs against R37 of Gαq. S62 forms a hydrogen bond with E355 of Gαq (Fig. 5e). Mutating S62 to alanine almost abolished the IP1 accumulation of FP (Fig. 5e, Supplementary Fig. 7), supporting that the direct interactions of TM1, ICL1, and TM2 with G protein are important for FP to couple with G protein.

Discussion

Glaucoma is the leading cause of permanent eyesight loss in the world34. According to the World Health Organization (WHO), 79.6 million individuals suffered from glaucoma in 2020, and the number of patients may increase to 111.8 million by 204035. Activation of prostaglandin receptors like FP, EP1–EP4, and DP showed anti-glaucoma effects. Research for the discovery of pharmaceutical drugs selectively targeting each of these receptors has been extensively conducted, whereas only FP selective agonists are approved for treatment due to their minimal side-effect profile36. Elucidating the mechanism underlying the functional selectivity of individual FP receptors is of utmost importance for developing new drugs with higher selectivity towards individual prostaglandin receptors to avoid or reduce undesirable side effects.

In this study, we present the cryo-EM structures of FP-Gq in complex with its endogenous ligand PGF2α as well as two synthetic agonists LTPA and TFPA. Through structure analysis of the PGF2α–FP–Gq complex, the carboxylate-containing α-chain that binds to the sub-pocket with highly conserved residues contributes to the majority of the high potency of the receptor. In contrast, the F ring is important for receptor selectivity. Similar to PGE2 to EP3, the hydrophobic ω-chain directly interacts with the toggle switch W2626.48 in FP and forms hydrophobic interactions with a set of hydrophobic residues. The hydroxyl group in the ω-chain of LTPA forms an additional hydrogen bond with S1183.35 in FP in the LTPA-FP-Gq structure, which may be the key to LTPA’s higher selectivity to FP. Despite PGF2α containing the hydroxyl group at the same position as LTPA that the presence of the aromatic ring and lack of carbon-to-carbon double bonds in the ω-chain of LTPA may affect the geometry of the ligand, indicating that these modifications in the ω-chain of PGAs may improve their selectivity to FP. Compared with the LTPA, TFPA forms an extra hydrogen bond to N842.57 in FP with a fluoride group in the ω chain. EP3 harbors a similar polar uncharged residue, T2.57, which may form a hydrogen bond with the fluoride group of TFPA. Although TFPA had a higher affinity for FP than LTPA, its selectivity needs to be further improved. These structures revealed the ligand recognition of FP, which will serve as templates for the rational design of a new generation of potent agonists with desired selectivity profiles.

Through structural comparison and mutagenesis studies, we also elucidated the mechanisms of receptor activation and G protein coupling by FP. FP has neither the traditional ‘PIF’ core triad nor the DRY motif that is commonly involved in the activation of class A GPCRs, instead FP senses ligand binding by the conserved toggle switch W6.48 to tether TM3–TM5–TM6 through a hereby identified LLW core triad consisting of L1233.40, L2135.49. Intriguingly, mutation of the W2626.48 in FP to alanine did not entirely abolish these three agonists induced activation of FP. It is speculated that the ω chain of PGF2α, LTPA, or TFPA in W262A mutant still induces the activation of mutated FP because it may point downward in a similar manner to EP2 (Fig. 1e), which naturally lacks W6.48 at the toggle switch residue position37. In the FP receptor, the conserved D/ERY motif is replaced by ERC, which only exists in several class A GPCRs, like EP1, IP, and Neurotensin receptor type 2. Interestingly, in the active FP structure, Q2506.36 induces an upswing of the side chain of R3088.47 and forms a hydrogen bond with it that leads to the inward shift of the cytoplasmic end of TM7 and the upshift of H8. R8.47 is particularly conserved in prostaglandin receptors but rarely exists in other class A GPCRs, and Q6.36 only exists in several class A receptors, including EP1, EP3, and FP. All these key residues for activation in FP are harbored by EP1, which also primarily couples with Gq. Thus, this activation mechanism should be applicable to both FP and EP1, which have the same unique residues.

The distinct features of the active FP structures also define the way of Gq protein coupling. Specifically, the TM1, ICL1, and TM2 of FP form extensive interactions with Gq, which, to our knowledge, haven’t been shown in any other reported Gq-coupled GPCR structures. These observed characteristics in the FP-Gq complex structures, including the arrangement of the 7TM bundle, the ligand-binding mode, the ligand-induced receptor activation, and the manner of G protein coupling, expand the understanding of lipid recognition and GPCR-Gq coupling mechanism. Collectively, our results reveal conserved and diverse mechanisms of ligand binding, receptor activation, and G protein coupling by FP.

Methods

Constructs

The full-length human FP was modified to contain the N-terminal thermally stabilized BRIL20 to enhance receptor expression and the addition of an N-terminal Flag tag. LgBiT was inserted at the C-terminus of the human FP using homologous recombination. The modified FP was cloned into the pFastBac (Thermo Fisher Scientific) vectors using the ClonExpress II One Step Cloning Kit (Vazyme Biotech). An engineered Gαq chimera was generated based on the mini-Gαs scaffold with its aa1-18 replaced by corresponding sequences of Gαi1, aa348-359 replaced by corresponding sequences of Gαq, A33, H35, A87, V92, D107, V115, R137, N144, C151, F155, K158, V161, K163, D171, and D319 replaced by corresponding residues in Gαq, designated as mGαs/q/iN. Human wild-type (WT) Gβ1, human Gγ2, and a single-chain antibody scFv1638, as well as a Gβ1 fused with SmBiT at its C-terminus, were cloned into pFastBac vectors.

Insect cell expression

Human FP, Gq chimera, Gβ1, Gγ, and scFv16 were co-expressed in High Five insect cells (Invitrogen) using the baculovirus method (Expression Systems). Cell cultures were grown in ESF 921serum-free medium (Expression Systems) to a density of 2–3 million cells per mL and then infected with six separate baculoviruses at a suitable ratio. The culture was collected by centrifugation 48 h after infection, and cell pellets were stored at −80 °C.

Complex purification

Cell pellets were thawed in 20 mM HEPES pH 7.4, 150 mM NaCl, 10 mM MgCl2, and CaCl2 supplemented with Protease Inhibitor Cocktail (TargetMol). For the PGF2α/LTPA/TFPA-FP-Gq-scFv16 complexes, 10 μM PGF2α/LTPA/TFPA (MedChemExpress) and 2 mg Nb35 were added. The suspension was incubated for 1 h at room temperature, and the complex was solubilized from the membrane using 0.5% (w/v) lauryl maltose neopentyl glycol (LMNG) (Anatrace) and 0.1% (w/v) cholesteryl hemisuccinate (CHS) (Anatrace) for 2 h at 4 °C. Insoluble material was removed by centrifugation at 70,000g for 35 min, and the supernatant was immobilized on the Flag resin (SinoBiological). The resin was then packed and washed with 30 column volumes of 20 mM HEPES pH 7.4, 150 mM NaCl, 0.01% (w/v) LMNG, 0.002% CHS, and 10 μM ligand. The complex sample was eluted in buffer containing 20 mM HEPES pH 7.4, 150 mM NaCl, 0.01% (w/v) LMNG, 0.002% CHS, 10 μM ligand, and 0.2 mg/ml FLAG peptide (GenScript). Complex fractions were concentrated with a 100-kDa molecular weight cut-off (MWCO) Millipore concentrator for further purification. The complex was then subjected to size-exclusion chromatography on a Superdex 6 Increase 10/300 GL column (GE Healthcare) pre-equilibrated with size buffer containing 20 mM HEPES pH 7.4,150 mM NaCl, 0.00075% (w/v) LMNG, 0.00025% (w/v) GDN (Anatrace), 0.00025% digitonin (w/v), 0.00015% CHS, and10 μM ligand to separate complexes. Eluted fractions were evaluated by SDS-PAGE, and those consisting of receptor-Gq protein complexes were pooled and concentrated for cryo-EM experiments.

Cryo-EM data collection

Cryo-EM grids were prepared with the Vitrobot Mark IV plunger (FEI) set to 4 °C and 100% humidity. Three microliters of the sample were applied to the glow-discharged gold R1.2/1.3 holey carbon grids. The sample was incubated for 10 s on the grids before blotting for 3 s (double-sided, blot force −2) and flash-frozen in liquid ethane immediately. For FP–Gq–PGF2α complex, FP-Gq-LTPA complex, and FP-Gq-TFPA complex datasets, 3902, 7972, and 6513 movies were collected, respectively, on a Titan Krios equipped with a Gatan K3 direct electron detection device at 300 kV with a magnification of 105,000, corresponding to a pixel size 0.824 Å. Image acquisition was performed with EPU Software (FEI Eindhoven, Netherlands). We collected a total of 36 frames accumulating to a total dose of 50 e− Å−2 over 2.5 s exposure.

Cryo-EM image processing

MotionCor2 was used to perform the frame-based motion-correction algorithm to generate a drift-corrected micrograph for further processing, and CTFFIND4 provided the estimation of the contrast transfer function (CTF) parameters39,40. All subsequent steps, including particle picking and extraction, two-dimensional (2D) classification, three-dimensional (3D) classification, 3D refinement, CTF refinement, Bayesian polishing, post-processing, and local resolution estimation, were performed using Relion3.041.

For FP–Gq–PGF2α complex dataset, 115 aligned micrographs were deleted because of contaminations or bad ice quality. A total of 3,391,620 particles were extracted from the cryo-EM micrographs and followed by two rounds of reference-free 2D classification, yielding 895,825 particles after clearance. Mask 3D classification on the receptor part was used to separate out 479,164 particles which resulted in a clearer density of PTGFR. We refined these particles, which led to a structure at 3.24 Å global resolution. After the postprocessing, the particles were reconstituted to a 2.67 Å structure (Supplementary Fig. 1).

For FP–Gq–LTPA complex dataset, 1438 aligned micrographs were deleted because of contaminations or bad ice quality. A total of 5,478,774 particles were extracted from the cryo-EM micrographs and followed by two rounds of reference-free 2D classification, yielding 1,181,590 particles after clearance. The global 3D classification was used to separate out 437,740 particles. We then continued the processing in Relion3.0 and refined 437,740 particles, which led to a structure at 3.40 Å global resolution. After CTF refinement, Bayesian polishing, and postprocessing, then the particles were reconstituted to a 2.78 Å structure (Supplementary Fig. 2).

For FP–Gq–TFPA complex dataset, 279 aligned micrographs were deleted because of contaminations or bad ice quality. A total of 5,774,308 particles were extracted from the cryo-EM micrographs and followed by two rounds of reference-free 2D classification, yielding 2,418,674 particles after clearance. The global 3D classification was used to separate out 805,802 particles. Two rounds of mask 3D classification on the receptor part were used to separate out 578,962 particles which resulted in a clearer density of PTGFR. We refined these particles, which led to a structure at 3.35 Å global resolution. After CTF refinement, Bayesian polishing, and postprocessing, then the particles were reconstituted to a 3.14 Å structure (Supplementary Fig. 3). We also performed postprocessing of all three final maps with DeepEMhancer42.

Model building

A predicted FP structure from Alphafold2 was used as the starting reference model for receptor building43. Structures of Gαq, Gβ, Gγ, and the scFv16 were derived from PDB entry 7WKD44 and were rigid body fit into the density. All models were fitted into the EM density map using UCSF Chimera45, followed by iterative rounds of manual adjustment and automated rebuilding in COOT46 and PHENIX47, respectively. The model was finalized by rebuilding in ISOLDE48, followed by refinement in PHENIX with torsion-angle restraints to the input model. The final model statistics were validated using Comprehensive validation (cryo-EM) in PHENIX47 and provided in Supplementary Table 1. All structural figures were prepared using Chimera45, Chimera X49, and PyMOL (Schrödinger, LLC.).

Inositol phosphate accumulation assay

IP-One production was measured using the IP-One HTRF kit (Cisbio)50. Briefly, AD293 cells (Agilent) were grown to a density of 400,000–500,000 cells per mL and then infected with separate plasmids at a suitable concentration. The culture was collected by centrifugation 24 h after incubation at 37 °C in 5% CO2 with a Stimulation Buffer. The cell suspension was then dispensed in a white 384-well plate at a volume of 7 μl per well before adding 7 μl of ligands. The mixture was incubated for 1 h at 37 °C. IP-One-d2 and anti-IP-One Cryptate dissolved in Lysis Buffer (3 μl each) were subsequently added and incubated for 15-30 min at room temperature before measurement. Intracellular IP-One measurement was carried out with the IP-One HTRF kit and EnVision multi-plate reader (PerkinElmer) according to the manufacturer’s instructions. Data were normalized to the baseline response of the ligand. pEC50 Emin, and Emax for each curve were calculated by GraphPad Prism 8.0. ∆pEC50 equals pEC50 of agonists to specific Mutant minus pEC50 of agonists to WT. Data are presented as mean values ± SEM; n = 3 independent samples; n.s. no significant; *p < 0.05; **p < 0.01; ***p < 0.001.

Receptor surface expression

Cell-surface expression levels of WT or mutants FP were quantified by flow cytometry. AD293 cells were seeded at a density of 1.5 × 105 per well into 12-well culture plates. Cells were grown overnight and then transfected with 1.0 μg FP construct by FuGENE® HD transfection reagent in each well for 24 h. After 24 h of transfection, cells were washed once with PBS and then detached with 0.2% (w/v) EDTA in PBS. Cells were blocked with PBS containing 5% (w/v) BSA for 15 min at room temperature before incubating with primary anti-Flag antibody (diluted with PBS containing 5% BSA at a ratio of 1:300, Sigma-Aldrich, F3165) for 1 h at room temperature. Cells were then washed three times with PBS containing 1% (w/v) BSA and then incubated with anti-mouse Alexa-488-conjugated secondary antibody (diluted at a ratio of 1:1000, Thermo Fisher, A-11029) at 4 °C in the dark for 1 h. After another three times of washing, cells were collected, and fluorescence intensity was quantified in a Luminex flow cytometer system (Guava® easyCyte) through a Luminex guavaSoft 4.5 at excitation 488 nm and emission 519 nm. Approximately 10,000 cellular events per sample were collected, and data were normalized to the wild-type FP. Experiments were performed at least three times, and data were presented as means ± SEM.

Statistics

All functional study data were analyzed using GraphPad Prism 8.0 (Graphpad Software Inc.) and showed as means ± S.E.M. from at least three independent experiments in triplicate. The significance was determined with a two-sided, unpaired t-test, and *p  <  0.05 was considered statistically significant.

Reporting summary

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