Cryo-EM structure of constitutively active human Frizzled 7 in complex with heterotrimeric G_s

Dear Editor,

The ten mammalian Frizzleds (FZD_1–10) belong to the class F of G protein-coupled receptors (GPCRs) and mediate WNT signaling through interaction with transducer proteins including Dishevelled (DVL) or heterotrimeric G proteins.¹ Their involvement in human disease has put FZDs at the forefront of drug targets, especially anti-cancer therapy.² However, no drugs have been developed for efficient pharmacological modulation of FZDs, partially owing to the limited understanding of FZD structure and activation mechanisms.^1,3 Among class F, FZD₇ is intensively pursued due to its relevance in various tumor models, particularly in intestinal cancers.⁴ Detailed structures of the receptor complexes would allow for structure-guided discovery of new drug candidates. FZD_1–10 share structural similarity with the related class F member Smoothened (SMO), which mediates Hedgehog signaling and is a validated target for cancer therapy.² In an effort to understand the structural basis of FZD activation and transducer interaction, we solved the structure of human FZD₇ in complex with heterotrimeric mini G_s (mG_s).⁵

Based on the evidence that FZD₇ interacts with G_s to mediate muscle hypertrophy,^6,7 we assessed its ability to activate heterotrimeric G_s independently of WNT stimulation. Co-expression of FZD₇ with a bioluminescence resonance energy transfer (BRET)-based G_s biosensor,⁸ reporting the rearrangement or dissociation of Gα_s and Gβγ following receptor engagement and G protein activation, revealed that FZD₇ exhibits constitutive activity similar to the class A β₂-adrenoceptor (Fig. 1a; Supplementary information, Fig. S1a, b). Using an analogous assay that measures activity-dependent Gα_s translocation (Supplementary information, Fig. S1c), we found that the constitutive activity of FZD₇ correlates with increased receptor expression (Supplementary information, Fig. S1d, e). Given the robust constitutive activity of FZD₇ towards G_s, we reconstituted purified, full-length human FZD₇, heterotrimeric mG_s and Nanobody35 (Nb35), which stabilizes the nucleotide-free Gα_s and Gβ subunits,⁹ in the absence of ligand and obtained pure complexes following size exclusion chromatography (Supplementary information, Fig. S2). The final complex was composed of FZD₇, mGα_s, Gβ, Gγ and Nb35, which could be clearly identified by 2D classification (Fig. 1b; Supplementary information, Fig. S2d). We used single-particle cryo-EM analysis to determine the 3D structure of this complex. After several rounds of classification and auto-refinement, the resolution of the final structure reached 3.2 Å allowing us to build an atomic model based on the density map (Fig. 1c; Supplementary information, Figs. S3–S5, Table S1).

Fig. 1: Structure of constitutively active FZD₇ in complex with heterotrimeric mG_s.

a Normalized BRET₀ values of ΔFZD_1–10 HEK293 cells transiently co-transfected with the G_s BRET sensor along with either negative control (mock), the β₂-adrenoceptor (β₂AR) or FZD₇. Data are represented as the means ± SEM of raw BRET₀ that were obtained from simple linear regression of five independent experiments measured in quadruplicates shown in Supplementary information, Fig. S1A and normalized to the negative control. *P < 0.05; **P < 0.01 (one-way ANOVA followed by Sidak’s multiple comparison). b An annotated 2D class average of FZD₇–mG_s–Nb35 complex. c Overall density map and atomic model of FZD₇–mG_s–Nb35 complex (CRD was omitted due to linker flexibility). FZD₇, blue; mGα_s, orange; Gβ, green; Gγ, yellow; Nb35, gray. d Insertion of the α5-helix (mGα_s, orange) into FZD₇ helical bundle represented as surface (ICL1, blue; ICL2, pink; ICL3, yellow; TM7/H8, green). e Schematics of interactions between FZD₇ and α5-helix. Hydrogen bonds are shown as red dashed lines. The red circle represents the hydrophobic interaction network. Yellow shades indicate residues that reside in TM5/6/ICL3; pink, TM3/4/ICL2; blue, TM1/2/ICL1; green, TM7/H8. f Superposition of FZD₇ (blue) and FZD₄ (yellow) structures, viewed from the intracellular side (bottom view). g Superposition of the active FZD₇ structure (blue) with the inactive FZD₄ (PDB: 6BD4, yellow), inactive FZD₅ (PDB: 6WW2, light pink), active SMO (PDB: 6OT0, gray) and inactive SMO (PDB: 5V57, green) structures. h Comparison of the cytoplasmic portion of TM6 (from K^6.28 to P^6.43) in FZD₇, FZD₄, active SMO and inactive SMO structures. i R^6.32, F^6.36, W^7.55 network in FZD₇, FZD₄ and active SMO structures. Blue dashed lines indicate the distance of F^6.36–W^7.55 and F^6.36–R^6.32 in FZD₇. Gray dashed lines indicate the distance of R^6.32–W^7.55 and W^7.55–F^6.36 in active SMO structure. j Normalized BRET₀ values of ΔFZD_1-10 HEK293 cells transiently co-transfected with rGFP-CAAX and Gα_s-67-RlucII, along with either negative control (mock), wild-type FZD_7, ΔCRD-FZD₇ or the indicated FZD₇ mutants. Data are represented as the means ± SEM of raw BRET₀ that were obtained from simple linear regression of four independent experiments measured in quadruplicates shown in Supplementary information, Fig. S11b and normalized to the negative control. **P < 0.01; ***P < 0.001 (one-way ANOVA followed by Tukey’s multiple comparison). k Normalized FRET₀ values of ΔFZD_1–10 HEK293 cells transiently co-transfected with the FRET-based cAMP biosensor along with either negative control (mock), wild-type FZD_7, ΔCRD-FZD₇ or the indicated FZD₇ mutants. Data are represented as the means ± SEM of raw FRET₀ that were obtained from simple linear regression of five independent experiments measured in quadruplicates shown in Supplementary information, Fig. S11c and normalized to the negative control. **P < 0.01; ****P < 0.0001 (one-way ANOVA followed by Sidak’s multiple comparison).

In accordance with the functional evidence for constitutive activity, the FZD₇–mG_s complex structure provides the structural basis for ligand-independent G protein coupling (Fig. 1c). The interface between FZD₇ and mG_s is dominated by the distal C-terminal segment of the α5-helix in mGα_s (Fig. 1d, e). The C-terminal leucine residues (L393^H5.25, L394^H5.26; superscripts refer to the residue position in the common Gα numbering scheme for G proteins/GPCRdb) are inserted into the helical bundle of the receptor. L393^H5.25 and L394^H5.26 establish extensive interactions with FZD₇ residues yielding a locally converged network that stabilizes the complex (Fig. 1e). The terminal carboxyl group of L394^H5.26 in mGα_s forms an ionic bond with K466^6.28, and residues R281^ICL1, K552^8.49 and R470^6.32 of FZD₇ are located in close proximity (superscript numbers refer to the Ballesteros and Weinstein numbering system). Y391^H5.23 forms a hydrogen bond with the backbone of W369^ICL2. Residues I450^5.72, I453^5.75 and M454^5.76 of FZD₇ form a hydrophobic cleft accommodating L388^H5.20. Furthermore, R385^H5.17 forms an ionic bond with D457 in ICL3, further strengthening the interaction between the α5-helix and FZD₇. In summary, the recognition of Gα_s by FZD₇ is primarily governed by a network of hydrogen bonding and electrostatic interactions contributed from the C-terminal segment of the α5-helix (D381^H5.13-L394^H5.26), among which, interactions with L394^H5.26 lock the α5-helix tail in an uncoiled, elongated conformation (Fig. 1e).

The placement of the α5-helix of mGα_s in the core of FZD₇ stabilizes an open FZD₇ conformation. We compared the FZD₇–mG_s structure with the available inactive-state FZD₄ crystal structure (PDB: 6BD4) and the inactive-state FZD₅ cryo-EM structure (PDB: 6WW2) and observed a clear outward bending of TM6 and an inward shift of TM5 at the cytoplasmic side (Fig. 1f–h)—a conformational change characteristic of active-state class A and B GPCRs. This helical rearrangement is achieved through interaction of TM6 and TM5 with mG_s and opening of the molecular switch between TM6 and TM7 (R^6.32/W^7.55; Fig. 1i).⁷ Comparing inactive FZD₄ with FZD₇–mG_s reveals that the extracellular portion of TM6 of FZD₇ extends above the surface of the lipid bilayer at an angle of 45° (Supplementary information, Figs. S6, S7), similar to what we have predicted in previous models¹⁰ and in contrast to the almost 90° bending in the FZD₄ structure.¹¹ Moreover, conserved cysteines within the hinge domain form disulfide bonds to both stabilize its structure and to link it with ECL1 (C210–C230; C234–C315^ECL1) (Supplementary information, Fig. S6).¹²

To better understand the activation mechanism of FZD₇ and G protein coupling to class F receptors, we compared the FZD₇–mG_s structure with the agonist (24(S), 25-epoxycholesterol)-bound structure of SMO–G_i¹³ (PDB: 6OT0). The helical arrangement at the upper portion of the FZD₇ transmembrane core is more compact, presumably due to the absence of ligand (Supplementary information, Fig. S7). At the lower portion of TM6, substantially distinct conformations are observed between the SMO–G_i and FZD₇–mG_s structures. Most strikingly, TM6 in SMO–G_i undergoes a parallel outward movement compared to inactive SMO, whereas TM6 in the FZD₇–mG_s complex accomplishes a similar displacement of the cytoplasmic portion through a kink in the helix (Fig. 1h). The ionic interactions between TM6, ICL3 and the α5-helix of mGα_s (K466^6.28–L394^H5.26 and D457–R385^H5.17) are likely to be the main contributors in maintaining this kink. In addition, Y478^6.40 forms π–π interaction with W354^3.43 to further maintain the bent TM6 conformation (Supplementary information, Fig. S7).

While the most evident structural rearrangements relate to TM6, additional positional shifts of TM2, TM3, TM4 and TM5 in the FZD₇–mG_s complex are observed when compared to the SMO–G_i complex. These four helices constitute a more compact bundle in the FZD₇–mG_s structure, partially stabilized by a network of π interactions (Supplementary information, Fig. S7e). In a cooperative manner, these interactions promote the cytoplasmic portion of TM4 shifting inward by ~2 Å (comparing the Cα of L383^4.47 in FZD₇–mG_s with corresponding L362^4.47 in SMO–G_i complex structures) (Supplementary information, Fig. S7f, black arrow).

A conserved molecular switch between TM6 and TM7 was previously identified for all class F GPCRs, maintaining the receptor in an inactive conformation (observed as a hydrogen-bonding distance between R^6.32 and the backbone of W^7.55) in all inactive class F receptor structures.⁷ The polar interactions between R^6.32 and W^7.55 are broken in active SMO–G_i and the FZD₇–mG_s complexes, resulting in a 6.4 Å distance between R470^6.32 and W547^7.55 in the FZD₇–mG_s complex (Fig. 1f, h).⁷

To explore the conformational dynamics around the open and active FZD₇ structure, we performed molecular dynamics (MD) simulations of FZD₇ in complex with mG_s393⁵ (Supplementary information, Fig. S8). Monomeric mG_s facilitated MD simulations due to its small size while minimizing the effect on receptor dynamics. These MD simulations allowed us to monitor general receptor integrity and the status of the molecular switch by assessing the angle of the kinked TM6 and the distance between R470^6.32 and W547^7.55. The overall hallmark of FZD₇ activation — the kink in TM6 — is maintained over the time course of the simulation (measured as an angle between the backbone nitrogen atoms of V485^6.47, P481^6.43 and E462^6.24). P^6.43 is fully conserved among the FZD paralogues, but not in SMO (F^6.43) (Supplementary information, Fig. S9). Analogous to P^6.50 and P^6.47 in class A and B receptors, respectively (Supplementary information, Fig. S10), P481^6.43 is likely to contribute to the observed outward movement of the lower part of TM6¹⁴ (Fig. 1f, h). In the MD trajectories, the conformational changes of TM6 are manifested by the disruption of the molecular switch and a rearrangement of an extended aromatic network stabilizing the active receptor conformation (Supplementary information, Fig. S8). R470^6.32 and the backbone oxygen atom of W547^7.55 remain at over 8 Å distance throughout the simulation, rendering hydrogen bonding impossible between these two residues. Instead, R470^6.32 is frequently bound with the carboxyl terminus of L^H5.26 of mGα_s. Interestingly, R470^6.32 remains within hydrogen-bonding distance to the carboxyl terminus of L^H5.26 more often than K466^6.28, indicating that these positively charged residues lock the carboxyl tail between them (Supplementary information, Fig. S8d). This could contribute to the observed non-helical conformation of the tail of the α5-helix.

To gather functional evidence for the FZD₇–mG_s interface and its role in maintaining the constitutive activity of FZD₇ towards G_s, we employed a mutagenesis-based approach in combination with assessment of Gα_s translocation and cAMP production as functional readouts of G_s-dependent signaling. We focused on D457 (in ICL3) and K466^6.28, which interact with the α5-helix of mGα_s. Mutating either D457 or K466^6.28 to alanine alone did not affect the constitutive activity of FZD₇ on G_s translocation or cAMP production (Fig. 1j, k; Supplementary information, Figs. S11, S12). However, the double mutant D457A/K466^6.28A abrogated FZD₇ constitutive activity towards G_s, suggesting that these mutations collectively interfere with G protein coupling. In contrast, the double mutant did not affect the ability of FZD₇ to mediate WNT-induced activation of the WNT/β-catenin pathway as assessed by the TOPFlash reporter assay (Supplementary information, Fig. S13), underlining the concept of conformational selection for DVL-dependent signaling over G protein coupling as has been suggested previously.^7,15

Although the CRD could not be resolved in the present structure, we observed that removal of the CRD (ΔCRD-FZD₇) resulted in the inability to reconstitute the receptor–mG_s complex in vitro to the same extent as that of full-length FZD₇ (Supplementary information, Fig. S14). Thus, we surmised that the CRD is required for FZD₇–mG_s complex stability and that removal of the CRD could decrease constitutive activity. Therefore, we assessed the ability of the ΔCRD-FZD₇ construct to functionally couple to G_s by assessing Gα_s translocation and cAMP production (Fig. 1j, k). Removal of the CRD blunted the constitutive activity towards G_s signaling as evidenced by the lack of Gα_s translocation and cAMP production. These data underline the requirement for the CRD to maintain constitutive activity of FZD₇ towards heterotrimeric G proteins through intramolecular allostery.

In conclusion, we report the cryo-EM structure of FZD₇–mG_s demonstrating how constitutive activity feeds into downstream signaling via heterotrimeric G proteins. With respect to the overall diversity among GPCRs, FZD₇ has evolved a unique way to maintain certain homologous movements consistent with class A and B GPCR activation, while adapting its class-specific architecture to mediate G protein activation. While the classical hallmarks of G protein engagement are present in our structure, several differences can be found at the interface between the receptor and the G protein suggesting that FZDs harbor their own selectivity determinants for heterotrimeric G proteins. In short, the present structure of constitutively active FZD₇–mG_s, alongside previously published inactive structures of FZD₄ and FZD₅, opens the door to more accurate modeling of other FZDs and a platform for in silico drug discovery, which will aid in the discovery of new treatments to help those afflicted with diseases of WNT-FZD signaling.