Molecular basis of Wnt biogenesis, secretion and Wnt7 specific signaling

Summary:

Wnt proteins are enzymatically lipidated by Porcupine in the ER and bind to Wntless for intracellular transport and secretion. Mechanisms governing the transfer of these low solubility Wnts from the ER to the extracellular space remain unclear. Through structural and functional analyses of Wnt7a, a crucial Wnt involved in central nervous system angiogenesis and blood-brain barrier maintenance, we have elucidated principles of Wnt biogenesis and Wnt7-specific signaling. The Wnt7a-Wntless complex binds to calreticulin revealing that calreticulin functions as a chaperone to facilitate Wnt transfer from Porcupine to Wntless during Wnt biogenesis. Our structures, functional analyses, and molecular-dynamics simulations demonstrate that a phospholipid in the core of Wnt-bound Wntless regulates the association and dissociation between Wnt and Wntless, suggesting a lipid-mediated Wnt secretion mechanism. Finally, the structure of Wnt7a bound to RECK, a cell surface Wnt7 co-receptor, reveals how RECK^CC4 engages the N-terminal domain of Wnt7a to activate Wnt7-specific signaling.

Graphical Abstract

In brief:

This study demonstrates that calreticulin facilitates the transfer of Wnt from Porcupine to Wntless during Wnt biogenesis and proposes a lipid-mediated Wnt release mechanism, along with the characterization on how Wnt7 co-receptor RECK binds to Wnt7a for ligand-specific signaling.

Introduction

Wnt signaling is a fundamental signal cascade that plays a crucial role in human development and stemness.^1-4 Aberrant Wnt signaling is frequently and closely linked to various cancers.^5-9 Wnt signaling is triggered by the interactions between the secreted lipid-modified signaling glycoproteins Wnt and their Frizzled (FZD) receptors, as well as their co-receptors.^10-12 Wnt is synthesized in the endoplasmic reticulum (ER) and the ER-resident membrane-bound O-acyltransferase Porcupine (PORCN) transfers a mono-unsaturated palmitoleate moiety (PAM) to a conserved serine residue on the hairpin 2 of Wnt proteins.^13-15 A previous study showed that the palmitoylation is indispensable for Wnt secretion and either mutation of the crucial serine or loss of PORCN activity causes the retention of Wnts in the ER.¹⁶ After lipidation, the membrane protein Wntless (WLS) employs its extracellular domain and transmembrane region to engage with the modified Wnt.¹⁷ This interaction facilitates Wnt transport across intracellular compartments.^18-20

Nineteen human Wnt ligands have been identified to activate different signaling pathways through the paracrine and autocrine systems.²¹ Human Wnt proteins contain at least 22 cysteine residues, which form multiple disulfide bonds contributing to the stability and folding of Wnts. Lipidated Wnt ligands, owing to their hydrophobic nature, often exhibit low solubility in aqueous environments, therefore necessitating protein partners during Wnt production. The crucial processes of Wnt lipidation and intracellular transport in all metazoans are mediated by two membrane proteins, PORCN and WLS, respectively. This implies the existence of a universal principle governing Wnt production.

One possible model for Wnt secretion involves direct release into the extracellular space. To start, the Wnt-WLS complex moves from the ER to the plasma membrane through the Golgi apparatus (Figure 1A). Once the Wnt-WLS complex reaches the cell surface, Wnt molecules bind to their receptors on the cell that excreted them (autocrine signaling) or to neighboring cells (paracrine signaling). This process may involve the transfer of Wnt molecules to various entities such as cell-surface proteoglycans and soluble transport proteins.^21-23 Another model suggests that the Wnt-WLS complex, after localizing to the plasma membrane, undergoes endocytosis and subsequent dissociation within an endosomal compartment. This then allows the Wnt to be loaded onto exosomes for secretion.²⁴ Various hypotheses regarding the dissociation of Wnt from WLS have proposed: that low pH regulates Wnt release,²⁵ that specific lipids compete with the palmitoleoyl moiety of Wnt in WLS,^17,26 or that alterations in membrane curvature promote Wnt release.²⁴ However, limited experimental evidence supports these proposed mechanisms; therefore, the molecular determinants of Wnt release from WLS remain elusive.

Figure 1. Structural analysis of Wnt7a-WLS-CALR complex.

(A) The general production pathway of Wnt ligands. In the ER, Wnts (purple) are modified by PORCN (green). Secretion of the modified Wnts is mediated by WLS (blue). It remains unclear how Wnts are transferred from PORCN to WLS and how Wnts are released from WLS. Created with BioRender.com.

(B) Overall structure of Wnt7a-WLS-CALR complex. The interaction details between PAM and WLS are indicated. The interaction area between the Wnt7a and CALR is shown in the right panel.

(C) Cryo-EM map of N-glycan of Wnt7a and its linked loop (contour level: 0.44).

(D) Interaction details between CALR and hairpin3, N-glycan of Wnt7a. The hydrophilic interactions are indicated by dashed lines.

(E) Structural comparison of Wnt7a-WLS-CALR complex with the PLC complex (PDB: 7QPD). The clash between Wnt7a and ERp57 is zoomed-in in the right box.

We selected Wnt7a as a Wnt representative to examine the process of Wnt production using biophysical and cell biological approaches. Wnt7 ligands, including Wnt7a and Wnt7b, are crucial for central nervous system (CNS) angiogenesis and the maintenance of blood-brain barrier (BBB) integrity.²⁷ GPR124 and RECK act as essential co-receptors of Wnt7 for stimulating the ligand-specific canonical Wnt signaling, since an enhanced activation of Wnt signaling is not observed with the other 17 mammalian Wnts or Norrin in the presence of GPR124 and RECK.^28-30 GPR124 contains seven transmembrane helices (TMs) and a >70kD extracellular domain (including a leucine-rich repeat (LRR) domain and Ig-like domain). RECK consists of five repeats of an uncommon domain with six cysteines (the CC domain) and a Frizzled-like cysteine-rich domain (Fz-like CRD), two epidermal growth factor-like domains (EGF-like 1 and 2), and three Kazal-like motifs (Kazal-like 1 to 3).³¹ Moreover, low-density lipoprotein receptor-related proteins 5 and 6 (LRP5 and LRP6) activate Wnt signaling as co-receptors of Wnt ligands (Figure S1A).^32,33 Recently, several studies have identified the interaction residues between RECK and Wnt7a and proposed a model of the RECK-Wnt7-FZD-GPR124-LRP5/6 complex^31,34-36; however, obtaining structural evidence is necessary to reveal the molecular mechanism of Wnt7 signaling by this complex.

Here, we identified calreticulin (CALR) as a glycan chaperone for Wnt proteins in the ER. Our findings also indicate that the incorporation of a phospholipid into WLS enhances the recruitment of Wnt proteins. In the apo state of WLS, the conformation of TMs 6-7 changes, obstructing the binding sites for the central phospholipid in WLS and Wnt protein, suggesting a mechanism by which lipids mediate the release of Wnts. Moreover, we have determined the structure of RECK bound to Wnt7a, revealing that the CC4 domain of RECK is essential for binding Wnt7a to activate the Wnt ligand-specific signal.

Reconstitution of Wnt7a-WLS complexes

A previous study showed that RECK plays a pivotal role in the Wnt7 signaling process within receiving cells, but it also participates in the biogenesis and/or release of Wnt7 from secreting cells.³⁷ Thus, we co-expressed human Flag-tagged WLS with C-terminal Strep-tagged Wnt7a and untagged full-length RECK (RECK^FL) in HEK293 cells to capture how Wnt7a binds to WLS and RECK. The lysate was incubated with anti-Flag M2 resin, and the affinity-purified material was then applied to gel filtration (Figure S1B). Notably, RECK^FL can be detected on the SDS-PAGE; however, the stoichiometry of Wnt7a and RECK^FL did not appear to be 1:1. Nevertheless, we subjected this sample to cryo-EM analysis and, after 3D classification, obtained two classes of Wnt7a-bound WLS at an overall resolution of 3.2-3.6 Å (Figure S1C). In Class-1, a four-helix bundle binds to the N-terminal domain (NTD) of Wnt7a, while Class-2 exhibits a globular protein bound to the C-terminal domain (CTD) of Wnt7a (Figure S1D). There is no 3D class that concomitantly contains both extra densities, indicating that two different states of the Wnt7a-WLS complex were captured in this dataset and that binding of the two extra ligands is mutually exclusive.

Since the previously determined crystal structure of the mouse RECK CC4 domain (mRECK^CC4)³¹ fits well into the density of the four-helix bundle in the Class-1 map, we speculated that the extra density corresponds to RECK^CC4, which was previously shown to be vital for the interaction with Wnt7a.^34,35 Since the extra density of Class-2 could not be assigned to a RECK region, we employed mass spectrometry to analyze the purified Wnt7a-WLS complex from the SDS-PAGE. Unexpectedly, an ER glycan-chaperone CALR is present at high abundance in the sample. Since the molecular weights of WLS and CALR are similar, the ~50 kD band represents both proteins (Figure S1B and E). CALR assists in the folding process in the ER and regulates the quality of newly synthesized glycoproteins via binding oligosaccharides with terminal glucose residues before these glycoproteins are transported to different subcellular compartments.³⁸ Although the AlphaFold-predicted model³⁹ and previously determined structures^40,41 of CALR could be docked into the density map of Class-2, the interaction details between Wnt7a and CALR were limited owing to the lower local resolution. We used the crosslinker BS³ to increase complex stability for cryo-EM study (Figure S1F). Using this method, we were able to determine the Wnt7a-WLS-CALR complex at 3.1-Å resolution revealing the interface details between Wnt7a and CALR (Figures 1B, S2A-D and Table S1).

By contrast, crosslinking with BS³ did not improve the quality of the interface between Wnt7a and the presumed RECK^CC4 domain of Class-1. Thus, we co-expressed Wnt7a and WLS without RECK and incubated purified Wnt7a-WLS with recombinant RECK^CC4 purified from HEK 293 cells. The resulting Wnt7a-WLS-RECK^CC4 structure was determined at 3.2-Å resolution, confirming our previous finding using full-length RECK that Wnt7a employs its N-terminal domain to recruit RECK^CC4 (Figures S3A-D and Table S1). In addition, we expressed Flag-tagged human WLS alone from HEK293 cells and obtained the structure of WLS in the apo state at 3.8-Å resolution (Figure S3E-H and Table S1). The overall structure of WLS is at a moderate resolution level, but the side chains of most TMs could be assigned unambiguously (Figure S3H).

Structure of the Wnt7a-WLS-CALR complex

The conformation of WLS in the Wnt7a-WLS-CALR complex is similar to that in the previously reported Wnt8a-WLS and Wnt3a-WLS complexes^17,26 with Cα R.M.S.D. values of 0.6 Å and 1.1 Å, respectively. The lectin domain of CALR is well resolved in the cryo-EM map, while the P-loop (residues 208-294) and the C-terminal region of Å-helix 3 (residues 378-417) were not built due to the low local resolution (Figures 1B, S2C and S4A-B). Our previous observation on PORCN shows PAM adopting a C shape in the catalytic core of PORCN,¹⁴ whereas in the Wnt7a-WLS complex, PAM presents a more extended conformation, inserted into a cavity between TMs 4-5 of WLS. Several residues (e.g., Ile302, Ile306, and Phe350) are responsible for creating a hydrophobic pocket that accommodates the PAM (Figure 1B). In our complex structure, Wnt7a is well folded, and the CALR binds to its N-glycan at Asn295 and hairpin 3 of the CTD (Figures 1B and S4C).

The high quality of our cryo-EM map allowed us to build a model of the N-glycan of Wnt7a-Asn295 binding to the lectin domain of CALR (Figure 1C). Residues Tyr109, Lys111, Tyr128, Asp135, His145, and Asp 317 of CALR form hydrogen bonds with the sugar moieties of the Glc₁Man₃ tetrasaccharide of Wnt7a (Figure 1D).⁴⁰ This interaction mode is analogous to that in the complex of CALR and major histocompatibility complex class I (MHC I) in the peptide-loading complex (PLC) (Figure 1E).⁴² Moreover, residues 152-157 form β-strand 9 (β9) to interact with the hairpin 3 of Wnt7a, further stabilizing the Wnt7a and CALR complex (Figures 1B and S4B). Another chaperone, ERp57, stabilized the complex of MHC I and CALR through interactions with the end of the P-loop (Figure 1E). In our cryo-EM map at a low threshold, the P-loop in the Wnt7a-WLS-CALR complex contacts the NTD of Wnt7a (Figure S4D-E). Superimposing the complex structure to the PLC shows that ERp57 would clash with Wnt7a, implying that ERp57 is not involved in the Wnt-WLS-CALR complex assembly. It is possible if ERp57 and CALR modulate the Wnt synthesis, then ERp57 might dissociate from Wnt after Wnt is completely synthesized ready for lipidation. Since ERp57 forms a complex with CALR to help glycoprotein folding within the ER lumen,⁴³ it is also possible that the flexibility of the P-loop in CALR and the co-chaperone ERp57 may not allow this interaction to be fully visualized in the cryo-EM map.

The role of CALR in Wnt biogenesis

The conservation of the glycosylation sequence, Asn-X-Ser/Thr (X is any amino acid residue except proline), among at least 14 human Wnt proteins is evident in the sequence alignment (Figure S5A). In order to explore whether CALR serves as a general chaperone during Wnt biogenesis, we co-expressed WLS and different Wnt proteins in HEK293 cells and performed coimmunoprecipitation (Co-IP) assays. Our results showed that the Wnt proteins without the conserved glycosylation site in the CTD, including Wnt8a, Wnt8b, Wnt9a, and Wnt9b, do not bind to CALR (Figure 2A). Interestingly, although there is no glycosylation site in the CTD of Wnt2b, it still associates with CALR (Figure 2A). The sequence alignment and mass spectrometry analysis showed that the residue Asn283 at the NC-linker of Wnt2b is modified with N-glycan and may serve as CALR binding site (Figures. S4F and S5A-B).

Figure 2. Structure of Wnt3a-WLS-CALR complex.

(A) Anti-Flag coimmunoprecipitation (co-IP) assays in total HEK293 cell lysates of cells which expressed C-terminal Flag-tagged WLS with C-terminal Strep-HA-tagged Wnt proteins. Endogenous CALR is associated with WLS in the presence of Wnt1/2b/3/3a/5a/7a/11, but not Wnt8a/8b/9a/9b.

(B) Overall structure of Wnt3a-WLS-CALR complex viewed from the side of the membrane and cryo-EM map of its N-glycan at Asn298 (contour level: 0.285).

(C) The comparison of the overall structures of Wnt3a-WLS-CALR complex and Wnt7a-WLS-CALR complex.

(D) The comparison of the CALR structures of Wnt3a-WLS-CALR and Wnt7a-WLS-CALR complexes. The shifts of the CALR structural elements and the N-glycans are indicated by arrows.

To further study the CALR binding mode in the other Wnt proteins, we conducted co-expression experiments involving non-tagged Wnt3a and Flag-tagged WLS in HEK293 cells. The resulting complex was subsequently purified using anti-Flag M2 resin, and crosslinking was achieved using BS³ (Figure S6A). Notably, we obtained a structure of the Wnt3a-WLS-CALR complex at a resolution of 3.5-Å in one of the 3D classes (Figures 2B, S6B-E and Table S1). The local resolution of CALR within the context of the Wnt3a complex was lower than that of Wnt7a, however, it is clear that the N-glycan on Asn298 inserts into the cavity of CALR (Figure 2B). Structural analysis revealed a similar interaction mode between CALR and Wnt3a, as observed in the CALR-Wnt7a complex, but without a contact between β9 of CALR and the hairpin 3 of Wnt3a (Figure 2C-D). These findings strongly suggest that CALR also facilitates the production of other Wnts, albeit with slightly different interaction modes.

To validate the importance of the glycosylation, we individually mutated the glycosylation site of Wnt3a and Wnt7a. When we co-expressed WLS and Wnt variants in HEK293 cells, Co-IP revealed that mutation of the glycosylation site of either Wnt3a or Wnt7a disrupted the interaction between Wnt and endogenous CALR in the Wnt-WLS-CALR complex (Figure 3A and B). Furthermore, the secretion levels of both Wnt3a and Wnt7a variants in HEK293 cells considerably decreased compared to the wild-type Wnts (Figure 3C). Owing to the low secretion yield, the mutations result in reduced autocrine Wnt signaling, underscoring the crucial role of this modification for Wnt production (Figure 3D).

Figure 3. Validation of the roles of CALR and the Wnt glycosylation site in Wnt production and signaling.

(A) and (B) Anti-Flag co-IP assays in total HEK293 cell lysates of cells which expressed C-terminal Flag-tagged WLS with or without C-terminal Strep-HA-tagged Wnt3a (A) or Strep-tagged Wnt7a (B). Endogenous CALR is coimmunoprecipitated with WLS in the presence of Wnt3a^WT and Wnt7a^WT but not Wnt3a^N298T and Wnt7a^N295T.

(C) The secreted level of Wnt3a^WT, Wnt7a^WT, Wnt3a^N298T and Wnt7a^N295T.

(D) Signaling activity for different Wnt variants. Activity was measured by TOPFlash luciferase reporter assay. Normalized activity for wild-type Wnt (WT) is taken as 100%, and activity of Wnt mutants is shown as percentage activity compared to WT. n=4 biological repeats. Data are mean ± s.d. Two-sided t-test was performed between the wild-type and mutant groups by GraphPad Prism 9. ***, p<0.001; ****, p<0.0001.

(E) and (F) Anti-Flag co-IP assays in total HEK293 cell lysates of cells which expressed N-terminal Flag-tagged PORCN^L335A with or without C-terminal Strep-HA-tagged Wnt3a (E) and Strep-tagged Wnt7a (F). Endogenous CALR and WLS were coimmunoprecipitated with PORCN^L335A in the presence of Wnt3a and Wnt7a.

(G) The hypothetical model of CALR in Wnt transfer from PORCN to WLS. While the hairpin 2 is inserted into PORCN for lipidation, the rest of Wnt binds to WLS under the facilitation of CALR.

The interaction between CALR and Wnt proteins during the lipidation process catalyzed by PORCN has remained elusive. To shed light on this process, we co-expressed Wnt7a, PORCN^L335A (L335A blocks Wnt release thus maintaining the Wnt-PORCN complex,⁴⁴ the residue numbering corresponds to the isoform B of PORCN), and CALR in HEK293 cells, followed by purification through gel filtration. Mass spectrometry analysis confirmed the presence of all components in the resulting complex, irrespective of the presence of a crosslinker (Figure S7A-B). While the structural determination of this complex was unsuccessful due to particle heterogeneity, our result strongly suggests the formation of a complex involving Wnt7a, PORCN, and CALR before Wnt is transferred to WLS. Owing to the limited solubility, a direct transfer of lipidated Wnt between PORCN and WLS is unfavorable, but the palmitoleoyl moiety is likely protected by proteins or embedded in the lipid environment. To investigate this transfer, we co-expressed PORCN^L335A with Wnt3a or Wnt7a in HEK293 cells. A Co-IP assay showed that endogenous CALR and WLS associate with PORCN when Wnt3a or Wnt 7a is expressed, but no association between PORCN and WLS is detected in the absence of Wnt protein (Figure 3E and F). This finding reveals an assembly of an intermediate complex comprised of WLS, PORCN, CALR, and Wnt during Wnt modification in ER.

The cryo-EM map of the Wnt7a-WLS-CALR complex visualized two interfaces between Wnt7a and CALR: 1) the N-glycan on Asn295 of Wnt7a interacts with β-strands of CALR (Figure 1D); 2) the NTD of Wnt7a forms a complex with the P-loop of CALR (Figure S4D-E). Our previous molecular dynamics (MD) simulations indicate that the interaction between hairpins 1 and 3 of the Wnt protein and PORCN are highly dynamic.¹⁴ However, in the structure of the Wnt7a-WLS-CALR complex, the hairpins 1 and 3 appear to stably bind to the WLS (Figure 1B). Since hairpin 3 of Wnt presents a flexible nature (Figure S4C), we speculate that the presence of CALR enforces the entire Wnt molecule into a rigid and extended state, thereby facilitating the Wnt-WLS interaction, while hairpin 2 is inserted into PORCN (Figure 3G). After lipidation, hairpin 2 is released from PORCN, and its palmitoleoyl moiety may be embedded in the lipid environment during this transfer to reach the TMs of WLS. The structure of the Wnt7a-WLS-CALR complex shows that the N-glycan on Asn295 becomes the primary binding site for CALR when Wnt7a forms a complex with WLS. It is conceivable that CALR may dissociate from the Wnt7a-WLS complex after ER-resident glucosidases trim the sugar moiety, allowing this complex to traffic to the Golgi. Since there is a high degree of conservation among Wnts, and PORCN, WLS and CALR are conserved across metazoans^17,45 (Figure S4B), our model might be applicable to the majority of Wnt transfers from PORCN to WLS in ER.

Lipid-mediated Wnt release from WLS

Curiously, we have found a phospholipid density located in the central cavity of WLS, which is open to the extracellular/luminal leaflet of the plasma/ER membrane when it binds to Wnt7a (Figures 4A and S6F). A similar density in this cavity was found in the cryo-EM maps of Wnt8a-WLS and Wnt3a-WLS.^17,26 In our cryo-EM maps of Wnt7a-WLS and Wnt3a-WLS complexes, this density exhibits a distinct shape that closely resembles the characteristics of a phosphocholine molecule but not of any other common phospholipid (Figure S6G). Moreover, the environment of this lipid binding pocket is suitable for the choline head of phosphocholine, so we tentatively assigned the electron density to a 1-palmitoyl-2-oleoyl-glycero-3-phosphocholine (POPC) (Figure 4A and B). In contrast, we did not find an analogous density in the center of apo WLS, potentially due to a decrease in the binding pocket size (Figure 4C). The fatty acid chain of the POPC is directly exposed to the lipid bilayer, implying this association may be regulated by the membrane environment. In the complex, residues Lys238 and Arg303 bind the phosphate group of phosphocholine and the choline head of POPC forms a π-cation interaction with residue Trp234 (Figure 4B). When we mutated Phe230, Trp234, and Phe474, our co-IP assay shows that the interaction between each mutant and Wnt7a is extensively reduced compared to that in wild type (Figure 4D). Notably, these residues are far away from the Wnt binding site, suggesting an important role of POPC in the central cavity for Wnt engagement. A previous work showed that the WLS mutation Y478C reduces the secretion of Wnt3a causing developmental defect in mouse embryos.⁴⁶ Our structural analyses show that the hydroxyl group of Y478 forms a hydrogen bond with the side chain of Asn281 to stabilize the central cavity of WLS (Figure 4B).

Figure 4. Structural analysis of Wnt7a-WLS complex and the apo WLS.

(A) Overall structure of Wnt7a-WLS complex. Electrostatic surface representation of WLS is shown. The cryo-EM map of putative POPC in the central cavity is shown. The central cavity opens to the extracellular/luminal leaflet and is indicated by a dashed rectangle.

(B) The interaction details between POPC and WLS. The residue Ser206 of Wnt7a is underlined.

(C) Electrostatic surface representation of the apo WLS. The closed central cavity is indicated by a red oval. A potential drug binding pocket is indicated by a dashed rectangle and zoomed-in on the right.

(D) Anti-Flag co-IP assays in total HEK293 cell lysates of cells which expressed C-terminal Strep-tagged Wnt7a with C-terminal Flag-tagged WLS. 3M: F230A, W234A and F474A (colored in olive green in panel B).

(E) Structural comparison between Wnt7a bound WLS (blue) and the apo WLS (gray).

(F) The conformational changes of TM6 and TM7 in the apo state disrupt the POPC and PAM binding.

Compared to the structure of the Wnt7a-WLS complex, the TMs 5-7 of WLS in the apo state undergo a substantial conformational change. The TM7 rotates counterclockwise when viewed from the luminal/extracellular face, pivoting around Met436 (Figure 4E). This conformational change leads the extracellular half of TM7 to move towards the center of WLS, disrupting the POPC and Wnt hairpin 2 binding sites (Figure 4E). TM5 and TM6 in the extracellular leaflet are shifted 3 and 5 Å, respectively, toward the center of WLS. In particular, Phe347 of TM5 rotates by 90° to obstruct the PAM-binding site (Figure 4F). Thus, we term the conformations of WLS in the Wnt7a-bound and apo states as “open” and “closed”, respectively. TM5 in the closed conformation narrows the cavity that accommodates PAM in the complex and residues Phe230, Trp234, and Lys238 on TM2 in the closed conformation interfere with binding of the head group of POPC (Figure 4F). The >10-Å movement of the extracellular half of TM7 drastically collapses the cavity that engages the POPC and hairpin 2 of Wnt7a, potentially causing the release of the Wnt ligands (Figure 4F). Importantly, the conformation of WLS in its complexes with Wnt3a, Wnt7a and Wnt8a is quite similar, and the central phospholipid is present in all three complexes.^17,26 Therefore, it is possible that the central phospholipid may control the transition of WLS between the open and closed conformations, further regulating the majority of Wnts’ association with and dissociation from WLS.

To test our hypothesis, we conducted MD simulations of the Wnt7a-WLS complex structure (Videos S1-2). The results showed notable differences in the interaction energy between Wnt7a and WLS with and without the POPC. During a 150 ns simulation, we observed that the interaction energy between hairpin 2 and the PAM region of Wnt7a, as well as their neighboring residues within a 3-Å distance, was −72.804 kcal/mol without POPC and −104.661 kcal/mol with POPC (Figure 5A). Similarly, the interaction energy between hairpin 3 of Wnt7a and its neighboring residues within 3-Å distance was −54.136 kcal/mol without POPC and −70.432 kcal/mol with POPC (Figure 5A). Furthermore, our simulations indicated that Phe347 and Phe352 in TM5 exhibited different conformations, illustrating the dynamic behavior of TM5 in the absence of POPC, where it was not bound to the center of WLS (Figure 5B). This conformational variability resulted in a wider cavity that accommodates the PAM region, which aligns with our analysis showing that hairpin 2 and the PAM region of Wnt7a binding to WLS is energetically less stable in the absence of POPC (Figure 5A). Conversely, in the presence of POPC, hydrophobic contacts within the TMs were observed (Figure 4E); as a result, the conformation of TM5 remained stable (Figure 5C). Compared to the results without the POPC, the cavity did not exhibit further opening during the MD simulations (Figure 5B and C). These findings suggest that Wnt tends to dissociate from WLS when the central pocket lacks POPC. This highlights the crucial role of this central lipid in facilitating the interaction between Wnt and WLS.

Figure 5. The analysis of the interaction energy between WLS and Wnt7a by MD simulations at 150 ns scale.

(A) The interaction energy between WLS and Wnt7a at the hairpin 3 and hairpin 2-PAM binding sites. The energy from the complex with the POPC is colored in orange and the energy from the complex without the POPC is colored in gray.

(B) The conformational change of TM5 without POPC at 150 ns scale. F347 and F352 rotate towards the edge of WLS and trigger the opening of cavity which accommodates the PAM. The surface representations of the PAM binding cavity from the side of the membrane are shown.

(C) The conformational change of TM5 with POPC at 150 ns scale. F347 and F352 retain a similar position in the presence of the POPC. The surface representations of the PAM binding cavity from the side of the membrane are shown.

Structure of Wnt7a-bound RECK

The structure of the Wnt7a-WLS-RECK^CC4 complex reveals a substantial interface of over 700 Ų between Wnt7a and RECK^CC4 (Figure 6A). RECK^CC4 is stabilized by three disulfide bonds, and the loop connecting helices 3 and 4 of CC4 (termed L3 loop) plays a crucial role in mediating this interaction (Figure 6B-C). Compared to the previously determined crystal structure of the CC4 domain in the apo state,³¹ the L3 loop moves toward the N-terminal domain of Wnt7a (Figure 6B). Three glutamine residues of CC4 (Gln253 and Gln257 in the L3 loop as well as Gln262 in helix 4) form hydrogen bonds with four residues of Wnt7a, including Gly63, Glu64, Glu89, and Val92 (Figure 6C). Residue Asp244 in helix 3 likely forms a salt bridge with Lys40 of Wnt7a. In addition to these hydrophilic interactions, Pro254/Leu255/Pro256 in the L3 loop contribute hydrophobic contacts with Wnt7a and Ile60 of Wnt7a has hydrophobic interaction with helix 4 (Figure 6C). A disulfide bond (Cys38-Cys52) confers conformational stability to its N-terminal region, ensuring a specific conformation for interaction with RECK^CC4 (Figure 6C). The residues of Wnt7a that are responsible for RECK^CC4 recognition are conserved in Wnt7b but not in the other 17 Wnt proteins, highlighting the specificity of the interaction between Wnt7 and RECK (Figure S5A).

Figure 6. Structural and functional analysis of the RECK^CC4-Wnt7a complex.

(A) Overall structure of RECK^CC4-Wnt7a-WLS complex.

(B) Structural comparison of RECK^CC4 in the complex with Wnt7a to crystal structure of mRECK^CC4 in the apo state (PDB: 6WBJ). The conformational change of L3 loop which binds to Wnt7a is indicated by a red arrow.

(C) Interactions between RECK^CC4 and Wnt7a. The disulfide bonds in RECK^CC4 are indicated. The hydrophilic interactions are indicated by dashed lines.

(D) and (E) Signaling activity for Wnt7a variants (D) and RECK variants (E). Activity was measured by TOPFlash luciferase reporter assay. Normalized activity for wild-type protein (WT) is taken as 100%, and activity of mutants are shown as percentage activity compared to WT. n=3-4 biological repeats. Data are mean ± s.d. Two-sided t-test was performed between the wild-type and mutant groups by GraphPad Prism 9. **, p<0.01; ****, p<0.0001.

(F) A structural model of RECK^CC4-Wnt7a-FZD8^CRD-LRP6^E1E2 complex. The structure of RECK^CC4-Wnt7a was docked to the structure of Wnt8-FZD8^CRD-LRP6^E1E2 (PDB: 8CTG).

To further test our structural findings, we introduced mutations in the full-length Wnt7a and RECK. We mutated Lys40 and Ile60 of Wnt7a to serine and proline respectively, according to the sequence of Wnt3a. Although the secretion yield of these mutants was high, they showed reduced signaling activity (Figure 6D). Each mutation resulted in a reduction of signal activity compared to wild-type RECK, consistent with previous alanine-scanning studies on RECK^CC4 to map the interface between RECK and Wnt7a (Figure 6E).³⁵ A previous study purported to show that the NC-linker of Wnt7a (residues 238-266) plays a key role in binding RECK³⁴, which is inconsistent with our finding (Figures 6A and C and S4C). A more recent study demonstrated that the NC-linker is crucial for binding to LRP6^E2 domain, which is essential for Wnt7-mediated signaling.⁴⁷ Therefore, the NTD of Wnt7a functions as the primary binding site of RECK. Based on our structural findings, we generated a model of CC4-bound Wnt7a using the previously determined structures of the Wnt8-FZD8^CRD-LRP6^E1E2 (Figure 6F).⁴⁷ This model shows that the interaction between RECK and Wnt7a does not interfere with the Wnt-FZD^CRD and Wnt-LRP6 interface. Collectively, our work provides a structural framework for understanding the physiology of Wnt7 signaling and paves the way for further studies on the Wnt ligand-specific signalosome.

Discussion

Nineteen human Wnt ligands activate various signaling pathways through both paracrine and autocrine signaling mechanisms.²¹ Owing to the PAM modification, Wnt is extremely insoluble; therefore, molecular chaperones are required for its production and trafficking. Although its extracellular chaperones (e.g., secreted FZD-related proteins, Swim, and Afamin) have been identified^48-50, its intracellular chaperone(s) have remained unclear. Here, we identify CALR as an intracellular binding partner of Wnts that may function as a general Wnt chaperone for modified Wnt proteins in the lumen of the ER, ensuring their proper folding and transfer between PORCN and WLS (Figure 7). Supported by structural and functional data, we propose a Wnt transferring complex (WTC), consisting of Wnt-PORCN-CALR-WLS, which avoids direct diffusion and the exposure of PAM to the aqueous environment during its ER transport (Figures 3G and 7). The interaction between CALR and Wnt may furthermore serve as a quality control for Wnt production ensuring the efficient transfer from PORCN to WLS, since abolishing the glycosylation site affects Wnt secretion substantially (Figure 3C).

Figure 7. Working model of Wnt production and Wnt7 specific signaling.

The hairpin 2 of Wnt interacts with the lumen cavity of PORCN (light green) to undergo lipidation. Then, hairpins 1 and/or 3 of Wnt bind to WLS (blue), with the assistance of CALR (wheat). Once the lipidation process is complete, the hairpin 2 of Wnt dissociates from PORCN and the TMs of WLS recruit a phospholipid (yellow), which facilitates the opening of the hairpin 2 binding cavity in WLS. Then the hairpin 2 of Wnt binds to the TMs of WLS, forming a complete Wnt complex. ER-resident glucosidases will trim the sugar moiety on Wnt. At this point, CALR can disengage from the Wnt-WLS complex, allowing the complex to traffic to the Golgi apparatus. Upon reaching the cell surface, the central phospholipid may dissociate from WLS, resulting in the secretion of Wnt into the extracellular space. Outside the cell, Wnt7 can bind to FZD, and interact with the CC4 domain of RECK (green) along with other co-receptors. This interaction triggers the Wnt ligand specific signaling. Created with BioRender.com.

When the Wnt-WLS complex arrives at the plasma membrane, Wnts may directly dissociate from WLS and interact with extracellular chaperones. Alternatively, the complex may be endocytosed into an endosomal compartment, where the released Wnt can bind to another carrier or become embedded in micelles in exosomes for secretion. Our work suggests a lipid-mediated mechanism for Wnt-WLS complex assembly and dissociation (Figure 7). Specifically, we found a phosphocholine in the central cavity of WLS that may play a critical role in the interaction between WLS and Wnt. In the ER membrane, the phosphocholine (PC) concentration exceeds 55% much more than that in plasma membrane (< 20%).⁵¹ These PC levels imply that phosphocholine diffusion into the lipid-binding pocket of WLS is a plausible process in the ER, facilitating Wnt binding. PC release into the plasma membrane may lead to subsequent conformational changes in WLS, ultimately resulting in the release of Wnt into the extracellular space or exosomes. This regulation bears a resemblance to how PIP2 binding triggers the conformational change of the TMs of an ion channel.⁵² Additionally, membrane curvature may provide a mechanical force to extrude the lipid, further destabilizing the interactions between Wnt and WLS.²⁴ It is tempting to speculate that any factors that disrupt the interactions between the central lipid and WLS could accelerate the release of Wnt.

Finally, our findings have implications for interfering with Wnt signaling for therapeutic purposes. Specifically, the apo WLS structure reveals a central pocket within its transmembrane region (Figure 4C) which could be targeted by a small molecule. Such a compound may stabilize WLS in the closed conformation, prevent the lipid-induced conformational change and subsequent recruitment of Wnt proteins. This may reduce Wnt signaling, which would be potentially beneficial for treating Wnt-dependent tumors.

Limitations of the study

Owing to the transient interactions between PORCN, Wnt, CALR and WLS, we were not able to capture the structure of this intermediate state. Other approaches, e.g., signal molecule fluorescence and in situ cryo-electron tomography, may be required. Moreover, our cryo-EM map suggests that phosphocholine binds to the center of WLS and regulates its interaction with Wnt. We attempted to identify this lipid using mass spectrometry; however, phospholipid is highly abundant in the lipid bilayer and consistently associates with membrane proteins, making it impossible to distinguish this lipid unambiguously. There it is possible that an alternative lipid species resides within the core of WLS, exerting influence over the regulation of Wnt release. While the specific kind of the lipid may not markedly affect our proposed lipid-mediated Wnt secretion mechanism, further work is required to refine this model.

STAR★Methods

RESOURCE AVAILABILITY

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Xiaochun Li (xiaochun.li@utsouthwestern.edu).

Materials availability

Plasmids generated in this study and the raw cryo-EM dataset are available upon request from the lead contact.

Data and code availability

• The 3D cryo-EM density maps have been deposited in the Electron Microscopy Data Bank (EMDB) under the accession numbers EMD-41764, EMD-41765, EMD-41767 and EMD-41768. Atomic coordinates for the atomic model have been deposited in the Protein Data Bank (PDB) under the accession numbers 8TZ0, 8TZP, 8TZR and 8TZS.

• This paper does not report original code.

• Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Experimental Model and Study Participant Details

Cell lines

Spodoptera frugiperda Sf9 cells (ATCC) were used to generate baculovirus and maintained in Sf-900 III SFM medium (Gibco) at 27°C with shaking. HEK 293S GnTI⁻ cells (ATCC) were used for protein expression and co-immunoprecipitation assays and maintained in Freestyle 293 expression medium (Gibco) at 37°C and 8% CO₂ with shaking. SuperTopFlash HEK293 (293STF, ATCC) were used for Wnt signaling assays and maintained in DMEM/F12 50/50 Mix medium (Corning) at 37°C in a humidity and CO₂ controlled incubator.

METHOD DETAILS

Protein expression and purification

WLS alone, WLS-Wnt7a, WLS-Wnt3a and WLS-Wnt7a-RECK^FL complexes

The complementary DNA (cDNA) encoding human WLS was cloned into the pEZT-BM vector with a C-terminal Flag tag. The cDNA encoding human Wnt7a was cloned into pEG BacMam vector with a GG linker followed by a Strep tag or a 6xHis tag at the C-terminus. The cDNA encoding human Wnt3a was cloned into the pEZT-BM vector without a tag. The cDNA encoding human RECK was cloned into pEG BacMam vector without a tag. Site-directed mutagenesis was carried out by QuikChange II kit (Agilent) or two-step overlapping PCR. All proteins were expressed using baculovirus-mediated transduction of mammalian HEK 293 S GnTI⁻ cells (ATCC). Baculoviruses were produced by transfecting Sf9 cells with the bacmids using Cellfectin II (Gibco). Totally 50 ml P2 baculoviruses were used to infect 1 L HEK293S GnTI⁻ at a density of 2-3 × 10⁶ cells/ml. For WLS-Wnt7a or WLS-Wnt3a complex, the volume ratio of baculoviruses of WLS and Wnt7a (Strep tagged)/Wnt3a is 1:4. For WLS-Wnt7a-RECK^FL complex, the volume ratio of baculoviruses of WLS, Wnt7a (6xHis tagged) and RECK is 1:4:1.5. 8 h after infection, 10 mM sodium butyrate was supplemented to the cells, and the cells was moved to 30 °C. For WLS-Wnt7a, WLS-Wnt3a and WLS-Wnt7a-RECK^FL complexes, 20 nM Bafilomycin A1 (Fisher Scientific) was added to the cells with sodium butyrate. Cells were collected at 72 h post-infection for protein purification.

The pellet from 1L cells was resuspended in 20 ml buffer A (20 mM HEPES, 150 mM NaCl, pH 7.5) supplemented by 10 μg ml⁻¹ leupeptin, 1 mM Phenylmethylsulfonyl fluoride (PMSF) and disrupted by sonication. After low-speed centrifugation, the resulting supernatant was then incubated with 1% (w/v) lauryl maltose neopentyl glycol (LMNG, Anatrace) at 4 °C for 1 h. The insoluble fraction was removed by centrifugation (20,000g, 4°C, 30 min), and the supernatant was incubated with 1 ml anti-Flag M2 resin (Sigma Aldrich) at 4 °C for 1 h. The beads were then washed with 20 column volume (CV) buffer B (20 mM HEPES, 150 mM NaCl, 0.01% LMNG, pH 7.5) by gravity flow. Proteins were eluted in 5 CV buffer B supplemented by 100 μg ml⁻¹ 3×Flag peptide. The eluate was concentrated and further purified by size-exclusion chromatography (SEC) (Superose 6 Increase, 10/300 GL column, GE Healthcare) in buffer C (20 mM HEPES, 150 mM NaCl, 0.06% digitonin, pH 7.5) for cryo-EM or crosslinking studies. For WLS-Wnt7a-RECK^FL complex, an additional cryo-EM sample was prepared in buffer A supplemented 0.003% glycol-diosgenin (GDN).

RECK^CC4 protein

The cDNA encoding human RECK^CC4 (206-270) was cloned to pHL-sec vector which contains a N-terminal signal peptide and C-terminal 6xHis tag. 0.6 mg plasmid was transfected to 0.5 L HEK293S GnTI⁻¹ cells by 1.6 mg PEI Max (Polysciences) for protein expression. The medium was collected at 72 h post-transfection/infection and loaded on the Ni-NTA and flowed through by gravity. The Ni-NTA was then washed with 20 CV buffer A containing 20 mM imidazole by gravity flow. Proteins were eluted in 5 CV buffer A containing 250 mM imidazole. The eluate was concentrated and further purified by SEC (Superdex^® 200 Increase, 10/300 GL column, GE Healthcare) in buffer A for complex assembly.

PORCN^L335A-Wnt7a-CALR complex

The cDNA encoding human PORCN (isoform B) and containing a L335A mutation was cloned into the pEG-BacMam vector with a N-terminal Flag tag. The cDNA encoding human calreticulin (CALR) was cloned into pEG BacMam vector with a C-terminal 8xHis tag. Baculoviruses were produced as described above. The volume ratio of P2 baculoviruses for infection of PORCN^L335A, Wnt7a (Strep tagged) and CALR is 3:7:2. After 8 h, 10 mM sodium butyrate was supplemented to the cultures, and the cells was moved to 30 °C. Cells were collected at 72 h post-infection for protein purification.

The pellet of 1 L post-transduction cells was resuspended in 20 ml buffer A supplemented by 10 μg ml⁻¹ leupeptin, 1 mM PMSF and disrupted by sonication. After low-speed centrifugation, the resulting supernatant was then incubated with 1% (w/v) LMNG at 4 °C for 1 h. The insoluble fraction was removed by centrifugation (20,000g, 4 °C, 30 min), and the supernatant was incubated for 1 h with 1 ml anti-Flag M2 resin at 4 °C. The beads were then washed with 20 CV buffer B supplemented by 10 μg ml⁻¹ 1-palmitoyl-2-oleoyl-sn-glycero-3 -phospho-l-serine (POPS, Anatrace) by gravity flow. Proteins were eluted in 5 CV buffer B supplemented by 100 μg ml⁻¹ 3×Flag peptide. The eluate was concentrated and purified by SEC (Superose 6 Increase, 10/300 GL column, GE Healthcare) in buffer C.

Crosslinking of WLS-Wnt7a-CALR, WLS-Wnt3a-CALR and PORCN^L335A-Wnt7a-CALR complex

The purified WLS-Wnt7a-RECK^FL complex and WLS-Wnt3a complex samples, containing endogenous CALR which forms a complex with ER located WLS-Wnt7a/Wnt3a, were incubated with 2 mM BS³ (Thermo Scientific) at 4 °C for 1.5 h, respectively. For PORCN^L335A-Wnt7a-CALR complex, the purified PORCN^L335A-Wnt7a-CALR complex was incubated with 5 mM Glutaraldehyde at 4 °C for 3 h. The crosslinking reaction was quenched by 20 mM Tris-HCl, pH 7.5 at room temperature for 10 min. The crosslinked complexes were further purified by SEC (Superose 6 Increase, 10/300 GL column, GE Healthcare) in buffer C.

Cryo-EM sample preparation and data acquisition

For WLS-Wnt7a-RECK^FL/WLS-Wnt7a-CALR (Digitonin, non-crosslinking), WLS-Wnt7a-RECK^FL/WLS-Wnt7a-CALR (GDN, non-crosslinking), WLS-Wnt7a-RECK^FL/WLS-Wnt7a-CALR (crosslinked), WLS-Wnt3a-CALR (crosslinked) complexes, Protein samples were concentrated to 5-10 mg ml⁻¹ for grid preparation. For WLS-Wnt7a-RECK^CC4 complex, the purified WLS-Wnt7a and RECK^CC4 were mix at 1:3 molar ratio and concentrated to 5-10 mg ml⁻¹ for grid preparation. For apo WLS, the protein was concentrated to 10-15 mg ml⁻¹ for grid preparation. Aliquots of 3 μl protein samples were applied to glow discharged Quantifoil R1.2/1.3 400 mesh Au holey carbon grids (Quantifoil). The grids were blotted and plunged into liquid ethane using a Vitrobot Mark IV (FEI) and loaded onto a 300 kV Titan Krios transmission electron microscope for data collection. For apo WLS, raw movie stacks were collected with a Falcon 4i camera at 0.738 Å per pixel and a nominal defocus range of −0.8 to −1.8 μm. The exposure time for each micrograph was 4 s with a total dose of ~60 electrons per Ų. For other samples, raw movie stacks were collected with a K3 camera at 0.83 Å per pixel and a nominal defocus range of −1to −2 μm. The exposure time for each micrograph was 5 s, dose-fractionated into 50 frames with a total dose of ~60 electrons per Ų.

Cryo-EM Data Processing

For apo WLS, motion correction and contrast transfer function (CTF) estimation were performed using cryoSPARC.⁵³ For other data sets, motion correction was performed using MotionCor2⁵⁴ and performed in RELION-3.^54,55 The CTF was estimated using CTFFIND4.⁵⁶ For all data sets, Particle picking was performed with crYOLO using the general model.⁵⁷ Subsequent 2D classification, ab-initio, heterogenous 3D refinement, non-uniform refinement, CTF refinement, local 3D refinement and particle subtraction were performed in cryoSPARC. For WLS-Wnt7a-RECK^FL/WLS-Wnt7a-CALR (Digitonin, non-crosslinking) data set, 2D classes with clean background were selected to generate initial models for the heterogenous 3D refinement. The best 3D class showing features of WLS, Wnt7a and two extra densities was subjected to non-uniform refinement. The WLS-Wnt7a-RECK^FL/WLS-Wnt7a-CALR (GDN, non-crosslinking) data set was processed in a similar way. The final particles from these two data sets were combined and 3D classified in RELION-3 without alignment. The resulting good maps show either a 4-helix bundle like extra density (3D Class-1) or a globular protein like extra density (3D Class-2) (Figure S1C-D). For WLS-Wnt7a-RECK^FL/WLS-Wnt7a-CALR (crosslinked) data set. Two good models (3D Class-1 and 2) and three bad models from the non-crosslinking data sets were used for heterogenous 3D refinement. The best 3D class containing the CALR density was selected for the non-uniform refinement, CTF refinement and local refinement (Figure S2A). For WLS-Wnt7a-RECK^CC4 data set, the same five models were used for heterogenous 3D refinement. The best 3D class containing the RECK^CC4 density was selected for the further heterogenous 3D refinement, non-uniform refinement, CTF refinement and local refinement (Figure S3A). For WLS-Wnt3a-CALR (crosslinked) data set, good classes in the initial 2D classification were selected to generate initial models for the heterogenous 3D refinement. The best 3D class was selected for the secondary 2D classification and heterogenous 3D refinement, followed by the further heterogenous 3D refinement, non-uniform refinement, CTF refinement and local refinement (Figure S6B). For apo WLS data set, good classes in the initial 2D classification were selected to generate three initial models for four rounds of heterogenous 3D refinement. The resulting best class was subjected to non-uniform refinement and local refinement, followed by the particle subtraction to remove the signal of the detergent micelle. The final local refinement was performed with a mask covering the WLS protein (Figure S3E). For all data sets, the mask-corrected FSC curves were calculated in cryoSPARC. and reported resolutions are based on the 0.143 criterion. Local resolution estimations were performed in cryoSPARC.

Model building and refinement

For WLS-Wnt7a-CALR complex, WLS structures from PDB: 7KC4 ,AlphaFold2³⁹ predicted structures of human Wnt7a and CALR, N-Glycans structure from 7QPD, POPC and PAM structures generated by Phenix⁵⁸ were docked into the cryo-EM map as an initial model. For WLS-Wnt7a-RECK^CC4 complex, WLS, Wnt7a, N-Glycans, POPC and PAM structures from WLS-Wnt7a-CALR complex and AlphaFold2 predicted structure of human RECK^CC4 (209-270) were docked into the cryo-EM map as an initial model. For WLS-Wnt3a-CALR complex, WLS-Wnt3a structure from PDB: 7DRT, CALR, N-Glycans, POPC and PAM structures from WLS-Wnt7a-CALR complex were docked into the cryo-EM map as an initial model. For apo WLS, WLS structure from WLS-Wnt7a-CALR complex were docked into the cryo-EM map as an initial model. For all structures, the models were manual build in Coot⁵⁹ and refined by Phenix and ISOLDE⁶⁰ in ChimeraX.⁶¹ In WLS-Wnt7a-CALR and WLS-Wnt7a-RECK^CC4 complexes, residues 1-36 and 134-145 in Wnt7a, residues 1-2 and 499-541 in WLS were not build. In WLS-Wnt3a-CALR complex, residues 1-39 in Wnt3a, residues 1-2 and 495-541 in WLS were not build. In WLS-Wnt7a-CALR and WLS-Wnt3a-CALR complexes, residues 1-17, 208-294 and 378-417 in CALR were not build. In apo WLS, residues 1-2, 414-418, 457-467 and 499-541 in WLS were not build. Structural model validation was performed using Phenix. Figures were prepared using PyMOL, Chimera⁶² and ChimeraX.

Construction of model

The Wnt-WLS complex structure was modeled as follows. The atomic coordinates were constructed from Wnt7a-WLS-CALR complex. For the PAM ligand, the hydrogen atoms were constructed using Avogadro (http://avogadro.cc/); the force field parameters were obtained using the CGenFF server.⁶³ Hydrogen atoms were added to the protein molecules using H-build from CHARMM.⁶⁴ An initial protonation pattern at pH 7 was obtained by calculating the pKa values of all titratable residues. For this, electrostatic energy computations were carried out with PROPKA.⁶⁵ The following thirteen disulfide bonds were constructed: in Wnt7a Cys38-Cys52, Cys73-Cys84, Cys123-Cys131, C133-C152, Cys200-Cys214, Cys202-Cys209, Cys278-Cys309, Cys294-Cys304, Cys308-Cys348, Cys324-Cys339, Cys326-Cys336, Cys331-Cys332; in WLS, Cys50-Cys71 and Cys162-Cys179.

Next, using the CHARMM-GUI⁶⁶ and OPM database⁶⁷ the protein complex was inserted in a lipid bilayer consisting of a cytosolic leaflet and an extracellular leaflet. The WLS-Wnt7a complex was simulated in the plasma membrane (PM) in which both leaflets contain 40% cholesterol. In the cytosolic leaflet, the remainder is comprised of 20% 3-palmitoyl-oleoyl-D-glycero-1-Phosphatidylethanolamine (POPE), 20% 3-palmitoyl-2-oleoyl-D-glycero-1-Phosphatidylserine (POPS), and 20% 3-palmitoyl-2-oleoyl-D-glycero-1-phosphatidylinositol (POPI); the extracellular leaflet contains 30% sphingomyelin and 30% palmitoyl-2-oleoyl-D-glycero-1-Phosphatidylcholine (POPC). The protein-membrane system was next solvated in explicit TIP3 water molecules⁶⁸ with 150 mM KCl concentration to neutralize charge. The total system was simulated in a rectangular box of dimension ~100 Å × ~100 Å × ~163 Å.

Geometry optimizations and molecular dynamics

The initial geometry of each solvated protein-membrane complex was optimized with a combination of 1500 steps of steepest descent (SD) energy minimization and 1500 adopted basis Newton-Raphson (ABNR) steps to remove any close contacts. Energy minimizations and geometry optimizations were carried out with NAMD⁶⁹ using the all-atom CHARMM36 parameter set for the protein and lipid molecules,⁷⁰ the TIP3P model for water molecules,⁶⁸ and the PAM force field parameters as obtained in the previous step using the CGenFF server.⁶³

The solvated ligand-protein-membrane complex was simulated with molecular dynamics (MD) at 310 K according to the following protocol: 1) equilibration MD with Langevin dynamics (time step of 1 fs) for 50 ps followed by CPT dynamics (time step 2 fs) for 350 ps; 2) production MD with CPT dynamics (time step 2 fs) for 150 ns. To simulate a continuous system, periodic boundary conditions were applied. Electrostatic interactions were summed with the Particle Mesh Ewald method⁷¹ (grid spacing ~0.92.; fftx 108, ffty 108, fftz 180). A nonbonded cutoff of 12.0 was used.

Interaction energy calculations

Interaction energies (electrostatic and van der Waal energies) between amino acids were calculated using the INTER module in CHARMM with the CHARMM36 parameter set. The energies were tabulated at regular intervals of the 150 ns simulation and averaged. Amino acids within a 3 Å radius were selected for two regions of Wnt7a: residues 195-219 including PAM, and residues 327-336.

Co-Immunoprecipitation

For endogenous CALR co-immunoprecipitation with WLS-Wnt complex by anti-Flag M2 resin, HEK293S GnTI⁻ cells were co-infected with P2 baculoviruses of WLS and Wnt7a (Strep tagged), Wnt7a N295T mutant (Strep tagged), other Wnts or their variants (Strep-HA tagged) or empty pEZT-BM vector. For endogenous CALR co-immunoprecipitation with PORCN^L335A-Wnt7a/Wnt3a complex by anti-Flag M2 resin, HEK293S GnTI⁻ cells were co-infected with P2 baculoviruses of PORCN^L335A and Wnt7a (Strep tagged) or Wnt3a (Strep-HA tagged) or empty pEZT-BM vector. Expression and purification of WLS-Wnt7a and PORCN^L335A-Wnt7a complexes were performed as described above with a small scale. For Wnt7a co-immunoprecipitation with WLS variants by anti-Flag M2 resin, HEK293S GnTI⁻ cells were co-infected with P2 baculoviruses of Wnt7a (Strep tagged) and WLS wild-type or mutants. Expression of WLS-Wnt7a complex was performed as described above with a small scale but without adding Bafilomycin A1. Anti-Flag M2 resin was used for the purification. The Flag-tagged recombined WLS and PORCN^L335A, Strep-tagged recombined Wnt7a, Strep-HA tagged Wnt3a (Figure 3A and 3E), Strep-HA tagged Wnts (Figure 2A), endogenous CALR and endogenous WLS were detected by western blot using anti-Flag antibody (MBL International, Cat# M185-3L), anti-Wnt 7a antibody (Abcam, Cat# ab274321), anti-Wnt3a antibody (Cell Signaling Technology, Cat# 2721S), anti-HA antibody (Cell Signaling Technology, Cat# 3724S), anti-CALR antibody (Cell Signaling Technology, Cat# 12238S) and anti-WLS (GPR177) antibody (Thermo Scientific, Cat# 50-173-1894), respectively. The endogenous Calnexin (CNX) was used as an internal control and detected by an anti-CNX antibody (Cell Signaling Technology, Cat# 2433S). Each experiment was performed at least three times with similar results.

Wnt signaling assay

Wnt signaling activities were measured by Dual-Luciferase^® Reporter Assay System (Promega) using SuperTopFlash HEK293 (293STF, ATCC) cells. The pRL-SV40 Renilla luciferase (Promega) was used as a control to normalize the signal. Human Frizzled-4 and GPR124 were cloned to pcDNA3.1 vector with a C-terminal Flag tag, respectively. Human Wnt3a and Wnt7a were cloned to pcDNA3.1 vector with a C-terminal GG linker followed by a Strep tag, respectively. Human LRP6 (20-1613) was cloned to pHL-sec vector. Human RECK was cloned to pcDNA3.1 vector without a tag. Site-directed mutagenesis was carried out by QuikChange II kit (Agilent) or two-step overlapping PCR. For signaling assay, the 293STF cells were setup in 24-well plates and transfected by FuGENE6 Transfection Reagent (Promega). To measure the signaling activity of Wnt3a variants, the cells were co-transfected with pRL-SV40 Renilla luciferase (20 ng/well), Frizzled-4 (60 ng/well), LRP6 (60 ng/well) and pcDNA3.1 or Wnt3a^WT or Wnt3a^N298T (20 ng/well). To measure the signaling activity of Wnt7a variants, the cells were co-transfected with pRL-SV40 Renilla luciferase (20 ng/well), Frizzled-4 (60 ng/well), LRP6 (60 ng/well), RECK (60 ng/well), GPR124 (60 ng/well) and pcDNA3.1 or Wnt7a^WT or Wnt7a^N295T (20 ng/well). To measure the signaling activity of RECK or Wnt7a variants, the cells were co-transfected with pRL-SV40 Renilla luciferase (20 ng/well), Frizzled-4 (60 ng/well), LRP6 (60 ng/well), GPR124 (60 ng/well), Wnt7a wild-type or Wnt7a mutants (20 ng/well) and pcDNA3.1 or RECK wild-type or RECK mutants (60 ng/well). The Dual-Luciferase assays were performed 48 h post-transfection. The Wnt7a, Wnt3a and RECK in the cells were detected by western-blot using anti-Wnt 7a antibody (Abcam, Cat# ab274321), anti-Wnt 3a antibody (Cell Signaling Technology, Cat# 2721S) and anti-RECK antibody (Thermo Scientific, Cat# MA5-14781). CNX was used as an internal control and detected by an anti-CNX antibody (Cell Signaling Technology, Cat# 2433S). Data analysis was performed using Prism 9 (GraphPad Software). Results are shown as mean±s.d. from three biologically independent experiments. To detect the protein level of Wnt3a/7a variants in the medium, 293STF cells were setup in 10 cm dish and transfected with Wnt3a/7a variants (10 ug/dish) by FuGENE6. 48 h post-transfection, the medium was collected and incubated with Blue Sepharose resin (GE Healthcare) over night to enrich Wnt proteins. The resin was then washed three times with buffer B and the protein was eluted by buffer B supplemented with 1M NaCl for western blot. Wnt3a and Wnt7a were detected by the antibodies mentioned above. Each experiment was performed at least three times with similar results.

Reproducibility

All functional experiments were repeated at least three times on different days. Similar results were obtained.

QUANTIFICATION AND STATISTICAL ANALYSIS

Statistical analyses of data were performed using GraphPad Prism 9. Quantification methods and tools used are described in each relevant section of the methods or figure legends.

Supplementary Material

1

Figure S1 Biochemical and cryo-EM analyses of Wnt7a-WLS complex, related to Figures 1 and 6

(A) Schematic of domain structures of human GRP124, RECK and LRP5/6.

(B) Representative gel-filtration chromatogram of Wnt7a-WLS-RECK^FL complex.

(C) Summary of cryo-EM data processing procedures of Wnt7a-WLS-RECK^FL complex. The extra densities are indicated by dashed circles.

(D) The extra densities after 3D classification. The extra densities are indicated by dashed circles.

(E) Mass spectrometry analysis of Wnt7a-WLS-RECK^FL complex. The top candidates from the mass spectrometry analyses are highlighted in yellow.

(F) Representative gel-filtration chromatogram of the crosslinked Wnt7a-WLS-RECK^FL complex.

2

Figure S2 Biochemical and cryo-EM analyses of Wnt7a-WLS-CALR complex, related to Figure 1

(A) Summary of cryo-EM data processing procedures of the crosslinked Wnt7a-WLS-RECK^FL complex. The final map was shown at contour level 0.514.

(B) Fourier shell correlation (FSC) curves between two half maps of Wnt7a-WLS-CALR complex.

(C) Local resolution of cryo-EM map of Wnt7a-WLS-CALR complex. Maps are colored according to local resolution, estimated using cryoSPARC.

(D) Cryo-EM map of the major structural elements of Wnt7a-WLS-CALR complex.

(E) Structural comparison of Wnt7a-WLS-CALR complex to Wnt8a-WLS complex (PDB: 7KC4).

(F) Structural comparison of Wnt7a-WLS-CALR complex to Wnt3a-WLS complex (PDB: 7DRT).

(G) Structural comparison of CALR in Wnt7a-WLS-CALR complex to crystal structure of CALR lectin domain (PDB: 3O0V).

3

Figure S3 Cryo-EM analyses of Wnt7a-WLS-RECK^CC4 complex, related to Figures 4 and 6

(A) Summary of cryo-EM data processing procedures of Wnt7a-WLS-RECK^CC4 complex. The final map was shown at contour level 0.418.

(B) FSC curves between two half maps of Wnt7a-WLS-RECK^CC4 complex.

(C) Local resolution of cryo-EM map of Wnt7a-WLS-RECK^CC4 complex. Maps are colored according to local resolution, estimated using cryoSPARC.

(D) cryo-EM map of the major structural elements of Wnt7a-WLS-RECK^CC4 complex.

(E) Summary of cryo-EM data processing procedures of the apo WLS. The final map was shown at contour level 0.117.

(F) FSC curves between two half maps of the apo WLS.

(G) Local resolution of cryo-EM map of the apo WLS. Maps are colored according to local resolution, estimated using cryoSPARC.

(H) cryo-EM map of the major structural elements of the apo WLS.

4

Figure S4 The structures of CALR and Wnt proteins and potential interaction between Wnt7a and P-loop, related to Figures 1, 2 and 3

(A) AlphaFold-predicted structure of human CALR. Each domain is indicated.

(B) Sequence alignment of CALR from different species. The residue number and secondary structure of human CALR are on the top of sequences. The conserved residues among all the CALR are highlighted in yellow and brown. HS: Homo sapiens; XL: Xenopus laevis; DR: Danio rerio; DM: Drosophila melanogaster. Sequences of CALR were aligned using Clustal Omega server (https://www.ebi.ac.uk/Tools/msa/clustalo/).

(C) Structure of Wnt7a in the Wnt7a-WLS-CALR complex. Each domain is indicated.

(D) Cryo-EM density of CALR P-loop in the resulting 3D class and refinement. In the class 2, the feature of P-loop could be detected in the low threshold.

(E) The modeled P-loop into the Wnt7a-WLS-CALR complex based on the cryo-EM map. The P-loop is colored in gray. The potential interaction area is indicated by a dashed circle.

(F) The AlphaFold predicted structure of Wnt2b. Each domain is indicated.

5

Figure S5 Sequence alignment of human Wnts and mass spectrometry analyze of Wnt2b glycosylation, related to Figures 2 and 6

(A) Sequence alignment of 19 human Wnts. The residue number and secondary structure of Wnt7a are on the top of sequences. The key residues for glycosylation in the CTD are highlighted in red box and the asparagine residue is colored in cyan. The Wnt2b glycosylation site in the NC-linker is colored in magenta. The conserved residues among all the Wnts are highlighted in yellow and brown. The key residues of Wnt7a/7b for RECK engagement are colored in green. Sequences of Wnt were aligned using Clustal Omega server (https://www.ebi.ac.uk/Tools/msa/clustalo/).

(B) The purification of Wnt2b-WLS-CALR complex and the mass spectrometry analyses of Wnt2b from SDS-PAGE. The top candidate from the mass spectrometry analyses is highlighted in yellow. The mass spectrometry analyses of Wnt2b glycosylation are shown on the right. #PSMs: Number of Peptide Spectrum Matches, or the number of spectra assigned to peptides that contributed to the inference of the protein.

6

Figure S6 Biochemical and cryo-EM analyses of Wnt3a-WLS-CALR complex, related to Figures 2 and 4

(A) Representative gel-filtration chromatogram of the crosslinked Wnt3a-WLS-CALR complex.

(B) Summary of cryo-EM data processing procedures. The final map was shown at contour level 0.287.

(C) FSC curves between two half maps.

(D) Local resolution of cryo-EM map. Maps are colored according to local resolution, estimated using cryoSPARC.

(E) Cryo-EM map of the major structural elements.

(F) Cryo-EM maps of the central lipid density in Wnt7a-WLS and Wnt3a-WLS complexes. PC has been modeled into the density. The contour levels are 0.785, 0.316 and 0.303, respectively.

(G) Model of the different phospholipids in the cryo-EM map of Wnt7a-WLS-CALR complex at contour level 0.785.

7

Figure S7 Biochemical and mass spectrometry analyses of Wnt7a-PORCN-CALR complex, related to Figure 3

(A) The purification of N-terminal Flag tagged PORCN^L335A, C-terminal Strep-tagged Wnt7a and C-terminal 8xHis-tagged CALR complex by anti-Flag M2 resin and the mass spectrometry analyses of each band from SDS-PAGE. The top candidates from the mass spectrometry analyses are highlighted in yellow.

(B) The mass spectrometry analyses of the crosslinked N-terminal Flag tagged PORCN^L335A, C-terminal Strep-tagged Wnt7a and C-terminal His-tagged CALR complex by glutaraldehyde. The top candidates from the mass spectrometry analyses are highlighted in yellow.

9

Supplementary Video 1 MD simulations of Wnt7a-WLS complex with POPC at 150 ns scale, related to Figures 4 and 5

The residues of the hairpins 2 and 3 of Wnt7a are shown. The PAM is colored in magenta and the POPC is colored in orange.

10

Supplementary Video 2 MD simulations of Wnt7a-WLS complex without POPC at 150 ns scale, related to Figures 4 and 5

The residues of the hairpins 2 and 3 of Wnt7a are shown. The PAM is colored in magenta.

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Mouse monoclonal anti-FLAG tag antibody	MBL International	Cat# M185-3L; RRID: AB_11123930
Rabbit monoclonal anti-Wnt7a antibody	Abcam	Cat# ab274321
Rabbit monoclonal anti-Wnt3a antibody	Cell Signaling Technology	Cat# 2721S; RRID: AB_2215411
Rabbit polyclonal anti-Calnexin antibody	Cell Signaling Technology	Cat# 2433S; RRID: AB_2243887
Rabbit monoclonal anti-Calreticulin antibody	Cell Signaling Technology	Cat# 12238S; RRID: AB_2275208
Rabbit monoclonal anti-RECK antibody	Thermo Scientific	Cat# MA5-14781; RRID: AB_11005172
Rabbit Polyclonal anti-WLS (GPR177) antibody	Thermo Scientific	Cat# 50-173-1894
Rabbit monoclonal anti-HA antibody	Cell Signaling Technology	Cat# 3724S; RRID: AB_1549585
Anti-Rabbit IgG, HRP-linked antibody	Cell Signaling Technology	Cat# 7074S; RRID: AB_2099233
Anti-mouse IgG, HRP-linked antibody	Cell Signaling Technology	Cat# 7076S; RRID: AB_330924
Bacterial and Virus Strains
E. coli DH5α Competent Cells	GoldBio	Cat# CC-101-TR
E. coli DH10Bac Competent Cells	Thermo Fisher	Cat# 10361012
Chemicals, Peptides, and Recombinant Proteins
LB broth	Fisher Scientific	Cat# BP1426-500
Sf-900 III SFM medium	Gibco	Cat# 12658027
Freestyle 293 expression medium	Gibco	Cat# 12338018
DMEM/F12 50/50 Mix medium	Corning	Cat# 10-090-CV
Fetal Bovine Serum (FBS)	Cytiva	Cat# SH30396.03
Penicillin-Streptomycin solution (100x)	Corning	Cat# 30-002-CI
HEPES	Gibco	Cat# 11344041
NaCl	Fisher Scientific	Cat# S271
Sodium butyrate	Millipore-Sigma	Cat# 303410
Phenylmethylsulfonyl fluoride (PMSF)	GoldBio	Cat# P-470-25
Leupeptin	vivitide	Cat# ILP-4041
Lauryl maltose neopentyl glycol (LMNG)	Anatrace	Cat# NG310
Digitonin	Thermo Scientific	Cat# 407560050
Ni-NTA agarose	Qiagen	Cat# 30250
Anti-Flag M2 resin	Millipore-Sigma	Cat# A2220
3X Flag peptide	ApexBio	Cat# A6001
FuGENE 6 transfection reagent	Promega	Cat# E2692
Cellfectin II reagent	Gibco	Cat# 10362-100
PEI Max	Polysciences	Cat# 24765-1
Phosphate buffered saline with Tween 20 (PBST)	Millipore-Sigma	Cat# 08057-100TAB-F
BS3 (bis(sulfosuccinimidyl)suberate)	Thermo Scientific	Cat# 21586
Glutaraldehyde solution	Sigma-Aldrich	Cat#G5882
Bafilomycin A1	Fisher Scientific	Cat#NC1351384
Critical Commercial Assays
Dual-Luciferase Reporter Assay System	Promega	Cat# E1910
Deposited Data
Cryo-EM structure of WLS-Wnt7a-CALR	This paper	PDB: 8TZO; EMDB: EMD-41764
Cryo-EM structure of WLS-Wnt3a-CALR	This paper	PDB: 8TZR; EMDB: EMD-41767
Cryo-EM structure of WLS-Wnt7a-RECKCC4	This paper	PDB: 8TZP; EMDB: EMD-41765
Cryo-EM structure of WLS	This paper	PDB: 8TZS; EMDB: EMD-41768
Experimental Models: Cell Lines
Sf9 insect cell	ATCC	Cat# CRL-1711; RRID: CVCL_0549
HEK-293S GnTI−	ATCC	Cat# CRL-3022; RRID: CVCL_A785
SuperTopFlash HEK293	ATCC	Cat# CRL-3249; RRID: CVCL_AQ26
Recombinant DNA
pEG-BacMam-Wnt7a-strep	This paper	N/A
pEG-BacMam-Wnt3a	This paper	N/A
pEG-BacMam-Wnt3a-strep	This paper	N/A
pEG-BacMam-Flag-PORCNL335A	This paper	N/A
pEG-BacMam-CALR-8His	This paper	N/A
pEG-BacMam-Wnt1-Strep-HA	This paper	N/A
pEG-BacMam-Wnt2b-Strep-HA	This paper	N/A
pEG-BacMam-Wnt3-Strep-HA	This paper	N/A
pEG-BacMam-Wnt3a-Strep-HA	This paper	N/A
pEG-BacMam-Wnt5a-Strep-HA	This paper	N/A
pEG-BacMam-Wnt7a-Strep-HA	This paper	N/A
pEG-BacMam-Wnt8a-Strep-HA	This paper	N/A
pEG-BacMam-Wnt8b-Strep-HA	This paper	N/A
pEG-BacMam-Wnt9a-Strep-HA	This paper	N/A
pEG-BacMam-Wnt9b-Strep-HA	This paper	N/A
pEG-BacMam-Wnt11-Strep-HA	This paper	N/A
pEZT-BM	Addgene	Cat# 74099
pEZT-BM-WLS-Flag	This paper	N/A
pRL-SV40 Renilla luciferase	Promega	Cat# E2231
pcDNA3.1	Invitrogen	Cat# V79020
pcDNA3.1-wnt7a-strep	This paper	N/A
pcDNA3.1-wnt3a	This paper	N/A
pcDNA3.1-FZD4	This paper	N/A
pcDNA3.1-RECK	This paper	N/A
pcDNA3.1-GPR124	This paper	N/A
pHL-sec-LRP6	This paper	N/A
pHL-sec-RECKCC4(206-270)	This paper	N/A
Software and Algorithms
MotionCor2	Zheng et al.54	https://emcore.ucsf.edu/ucsf-motioncor2
CTFFIND4	Rohou and Grigorieff56	http://grigoriefflab.janelia.org/ctffind4
RELION 3.1	Zivanov et al.55	https://www3.mrc-lmb.cam.ac.uk/relion/
SPHIRE-crYOLO v1.7.6	Wagner et al.57	http://sphire.mpg.de/
cryoSPARC v3.3.1 cryoSPARC v4.1.1	Punjani et al.53	https://cryosparc.com/
UCSF Chimera 1.13	Pettersen et al.62	http://www.cgl.ucsf.edu/chimera
UCSF ChimeraX 1.5	Pettersen et al.61	https://www.cgl.ucsf.edu/chimerax/
ISOLDE 1.4	Croll60	https://isolde.cimr.cam.ac.uk
Coot 0.9.6	Emsley et al.59	https://www2.mrc-lmb.cam.ac.uk/personal/pemsley/coot/
Phenix 1.16	Adams et al.58	http://www.phenix-online.org/
PyMOL 2.3.4	Schrödinger	https://pymol.org/2/
Prism 9	GraphPad	https://www.graphpad.com/
Avogadro	Avogadro Chemistry	https://avogadro.cc/
CHARMM	Brooks et al.64	https://www.charmm.org/
PROPKA	Li et al.65	https://github.com/jensengroup/propka
NAMD	Phillips et al.69	http://www.ks.uiuc.edu/Research/namd/
Other
Blue Sepharose High Performance	GE Healthcare	Cat# 90100374
R1.2/1.3 400 mesh Au holey carbon grids	Quantifoil	Cat# 1210627
Superose 6 Increase 10/300 GL column	Cytiva	Cat# 29091596
Superdex 200 Increase 10/300 GL column	Cytiva	Cat# 28990944

Highlights:

• Calreticulin binds most Wnt isoforms via their N-glycan in the ER.

• Calreticulin, Porcupine, and Wntless form a complex with Wnt during Wnt biogenesis.

• A phospholipid binds to the center of Wntless, modulating Wnt association.

• RECK binds to the N-terminal domain of Wnt7a for ligand-specific signal activation.

Inclusion and diversity

We support inclusive, diverse, and equitable conduct of research.