Frizzled BRET sensors based on bioorthogonal labeling of unnatural amino acids reveal WNT-induced dynamics of the cysteine-rich domain

WNTs induce conformational changes in the extracellular CRD of Frizzleds irrespective of their signaling profile.


INTRODUCTION
The class Frizzled (FZD) of G protein-coupled receptors (GPCRs), also known as class F, comprises 10 FZD subtypes (FZD 1-10 ) and Smoothened (SMO) (1). These cell surface membrane proteins share the structural hallmarks of GPCRs [extracellular N terminus, seven membrane-spanning helices, transmembrane domain 1 (TM1) to TM7, connected via three extracellular and three intracellular loops, ECL1 to ECL3 and ICL1 to ICL3, and an intracellular C terminus]. FZDs also display a class-typical cysteine-rich domain (CRD) at an extended N terminus, which is crucial for the engagement of their endogenous ligands with the receptor (2).
While SMO regulates Hedgehog signaling, FZDs bind extracellular, secreted lipoglycoproteins of the Wingless/Int-1 family (WNTs) and mediate the vital effects of these 19 mammalian WNT paralogues on cellular proliferation, migration, differentiation, tissue polarity, tissue homeostasis, and cancer development (3). On the one hand, FZDs mediate signaling through dishevelled (DVL1 to DVL3)dependent pathways resulting in either the stabilization and nuclear translocation of -catenin or planar cell polarity-like signaling (PCP) (4,5). On the other hand, they mediate DVL-independent signaling through heterotrimeric G proteins (6)(7)(8). Despite the enormous physiological relevance of the WNT/FZD signaling system in human health and disease, little is known about ligand/receptor selectivity, the molecular details that underlie receptor activation or the initiation of intracellular signaling. Thus, it remains unclear how distinct WNT/FZD complexes achieve pathway selectivity (9)(10)(11).
Given the lack of structural information on WNT/FZD complexes and the fact that WNTs can bind to purified FZD-CRDs without the presence of the transmembrane core of the receptor (2), uncertainty remains about the structural basis for agonist-induced, FZDdependent signaling. As of today, two main models emerge, explaining how FZDs translate WNT/CRD association into different cellular signaling branches. One hypothesis is based on WNT-mediated crosslinking of FZDs with co-receptors [e.g., with low-density lipoprotein receptor-related protein 5 or 6 (LRP5/6), reversion-inducing cysteinerich protein with Kazal motifs (RECK), tyrosine-protein kinase transmembrane receptor ROR2, or the adhesion GPCR ADGRA2, also known as GPR124; (12)], determining which intracellular transducer protein is recruited to subsequently convey the stimulus to the cell interior. This "signalosome" concept is widely accepted for signaling of FZDs through -catenin, which depends on WNT-induced interaction of FZDs with LRP5/6 (2,13,14), and would allow for conformational flexibility of the extracellular linker region between the CRD and the receptor's transmembrane core (15). On the other hand, WNT-induced conformational changes in FZDs could also independently from WNT co-receptors promote distinct cellular signaling pathways (8,(16)(17)(18)(19)(20). The latter model mirrors the concept of how class A and B GPCRs activate, for instance, heterotrimeric G proteins in response to a ligand-induced opening of the intracellular receptor surface and subsequent GPCR/G protein coupling (21). However, this process would require a constrained conformational mobility of the linker region to allow for mutual allosteric regulation of WNT binding and transducer coupling according to the ternary complex model (21,22).
Although WNT-3A/-catenin signaling occurs independently from heterotrimeric G proteins in human embryonic kidney (HEK) 293 cells (6), it remains obscure whether and how the signalosome and "ternary complex" model are mechanistically intertwined in FZDs, hampering a rational development of FZD subtype-specific and intracellular pathway-selective drugs to treat WNT/FZD-dependent disorders such as colon and pancreatic cancer.
Recent studies provided intriguing but controversial insights into different aspects of the WNT/FZD system including the question of ligand/receptor selectivity and the underlying activation mechanism of FZDs (9,15,20,23). For instance, mutagenesis-based approaches suggested that FZD 5 does not undergo conformational changes while signaling through DVL/-catenin (15), molecular dynamics (MD) simulations of FZD 4 (19), and experiments with conformational FZD biosensors revealed structural flexibility and its importance for downstream signaling mediated by these receptors (8,16,18,20). These conflicting findings highlight the need for advanced biophysical approaches and molecular tools to investigate WNT/FZD interaction and the conformational landscape of FZDs to better understand the FZD mode of action.
The assessment of conformational dynamics of GPCRs in living cells was facilitated by the design of fluorescence resonance energy transfer (FRET)-and bioluminescence resonance energy transfer (BRET)-based biosensors already in the early 2000s (24,25). Their development has not only allowed to monitor the structural rearrangements of various GPCRs-including FZD 5 and FZD 6 (16,18)in real time and single cells, but refinements of the sensor design have further enabled the establishment of screening-compatible assay formats (26,27). More recently, a distinct approach based on conformationally sensitive, circularly permutated fluorescent proteins, which were introduced to various GPCRs to monitor receptor activation in living animals (28), provided unprecedented insights into WNTinduced conformational dynamics of FZDs (20). While these conformational GPCR sensors exclusively detect the receptors' structural dynamics at the intracellular side, several FRET-based GPCR sensors have been devised to study the extracellular conformational rearrangements of mainly class C GPCRs labeled using SNAP-and CLIP-tag technology (29)(30)(31). In addition, genetic code expansion and labeling of unnatural amino acids (uaas) enabled the investigation of tethered agonist ("Stachel") exposure in class adhesion GPCRs (32).
In class F GPCRs, MD simulations based on an inactive SMO crystal structure (including a resolved CRD) revealed moderate CRD flexibility in the absence of a ligand, which is further restrained upon cholesterol binding to the CRD (33). Similarly, MD studies on an active SMO structure show only minor CRD movement when bound to cholesterol on the CRD and the receptor core, allowing the receptor to adapt an active TM7 conformation to promote intracellular signaling (34). However, the CRD-binding ligands of SMO and FZDs are distinctively different (cholesterol versus WNTs, respectively), and the linker domain between the CRD and the 7TM core of the receptor is notably shorter in SMO than in FZDs (1), indicating that the functional dynamics of the CRD might also differ. FZD structures including a fully resolved CRD are not available, and techniques to investigate extracellular dynamics of GPCRs in intact cells are limited by the size of fluorescent tags. Inserting, for instance, a fluorescent protein in one of the receptor's ECLs would most likely impair a proper folding of the receptor and thereby abolish its expression at the cell surface or could sterically interfere with ligand/receptor association.
To overcome these limitations of conventional fluorescentlabeling techniques, we linked small fluorescent probes to modified receptor residues using a minimally invasive labeling procedure based on genetic code expansion and incorporation of uaas serving as anchors for a subsequent bioorthogonal coupling reaction [strain-promoted inverse electron-demand Diels-Alder cycloaddition (SPIEDAC)] (35)(36)(37)(38). With the aim to monitor and understand the WNT-induced dynamics of the extracellular CRD and its role for signal initiation, and to avoid interference with WNT-binding upon incorporation of bulky tags, we developed a set of conformational biosensors based on BRET between the small nanoluciferase (Nluc) and a fluorescently labeled uaa. The minimal size of the energy acceptor maintained functionality and ligand binding to the receptor and allowed assessment of WNT-induced dynamics of the FZD extracellular domains in living cells, providing intriguing insights into the mechanistic and kinetic details of receptor activation preceding WNT/FZD signaling.

MD simulations of a FZD 6 model reveal CRD mobility
To assess the putative range of motion of the CRD relative to the receptor core, we used first in silico approaches. In the absence of a self-evident template for the extended linker region of FZDs, we prepared an inactive full-length model of FZD 6 using the iTasser server (39)(40)(41). FZD 6 was selected for the modeling as it has a shorter extended linker sequence than other representatives of the four FZD homology clusters (i.e., FZD 4 , FZD 5 , and FZD 7 ). The best-ranked model differs from the inactive SMO [Protein Data Bank (PDB) ID: 5L7D] only by this extended linker region (Fig. 1A) and was then used to initiate all-atom MD simulations (250 ns in seven independent replicas). In these simulations, the CRD probes a range of movements culminating to six main CRD orientations, which represent 62% of the simulation trajectory (Fig. 1C). Of these, clusters 2 and 5 mark the conformational extremes of the CRD movements of the whole trajectory. The 7TM core including the disulfide bondstabilized part of the linker and the extended TM6 are stable throughout the simulation. Furthermore, the CRD remains stably folded, and only its location relative to the receptor core is changing (Fig. 1, C and D). Together, the MD data suggest that the CRD of FZD 6 is capable to occupy distinct-conformationally restricted-orientations. To investigate whether these findings relate to WNT binding, we set out to study how the overall orientation of the CRD relative to the receptor core is affected by WNT stimulation using BRET technology.

Incorporation of TCO*K into class F GPCRs and bioorthogonal labeling
As representatives of class F, we chose FZD 6 and FZD 5 , which belong to different homology clusters of class F (1). The receptors were fused to an N-terminal Nluc epitope, following the 5-HT 3 receptor signal peptide, and a C-terminal 1D4 epitope. By using the amber codon suppression technology, the uaa trans-cyclooct-2-ene-l-lysine (TCO*K) was introduced at distinct positions in the linker domain and the ECL3 of FZDs. We selected suitable positions for incorporation of TCO*K from the FZD 6 simulation trajectory ( Fig. 1B  and fig. S1). The residues that were intended to be mutated were found to be in a distance toward the FZD's N terminus that allows BRET analysis.
For amber codon suppression, HEK293T cells were transfected with the amber-mutated receptor, and the corresponding orthogonal suppressor transfer RNA (tRNA)/aminoacyl tRNA synthetase pair in the presence of the uaa TCO*K ( Fig. 2A). Western blot analysis using the monoclonal antibody (mAb) 1D4, which recognizes a fused C-terminal epitope tag, showed that the amber codon suppression with the uaa TCO*K was efficient ( fig. S2A). Cell surface expression of all TCO*K-incorporated FZD mutants was determined using whole-cell enzyme-linked immunosorbent assay (ELISA) detecting the N-terminal Nluc epitope ( fig. S2B). All mutants were expressed at the cell surface of HEK293T cells, averaging 31 to 81% compared with the respective wild-type (WT) receptor ( fig. S2B).
In a next step, the receptor mutants were expressed in HEK293T cells, and living cells were labeled with the tetrazine (Tet)-bearing, membrane-impermeable fluorescent dye Tet-Cy3 ( fig. S3A) using the SPIEDAC reaction. As an extension, also the membrane-permeable dye Tet-BODIPY-FL (BDP-FL) was used to label selected receptor mutants ( fig. S4A). For quantification of the labeling efficiency of

of 15
the different receptor mutants, a plate reader assay was used to detect fluorescence intensities (figs. S3B and S4B). While Tet-Cy3 specifically labeled HEK293T cells expressing receptor amber mutants exclusively in the presence of the uaa TCO*K, Tet-BDP-FL showed more unspecific labeling properties, likely owing to its high lipophilicity and membrane permeability, resulting in off-target labeling of TCO*K residues in intracellular compartments (including TCO*K-tRNA), other proteins with amber stop codons, and mature  S4B). In general, cell surface expression levels of the receptor mutants correlated with the labeling efficiency.

WNT-5A induced BRET changes in the FZD 5 and FZD 6 CRD sensors
We took advantage of the N-terminally fused Nluc and the fluorescent dye, incorporated site-specifically in the linker region or ECL3 of the receptor, to establish BRET biosensors that can detect WNTinduced conformational rearrangements of the CRD (Fig. 2B). Initially, we tested all FZD 6 CRD sensors, comprising four receptor mutants in the linker region and three receptor mutants in ECL3, in a ligand-free condition in terms of basal energy transfer between Nluc and the incorporated Cy3 or BDP-FL, respectively ( fig. S5). Bioluminescence emission spectra of Nluc-tagged FZD 5 and FZD 6 CRD sensors, labeled with Tet-BDP-FL (green) or Tet-Cy3 (red), were recorded and normalized to the maximal Nluc emission. To exclude an unspecific fluorescent labeling of HEK293T cells expressing the receptor mutant sensors, we subtracted the bioluminescence signal obtained in cells expressing the WT receptor lacking incorporated TCO*K. We obtained emission peaks at ~510 nm for BDP-FL and ~570 nm for Cy3, respectively, with highest peaks for the FZD 6 -K466TCO*K mutant.
A distant member of the class F GPCRs is the Hedgehog signalmediating SMO, which is structurally related to FZD but, in contrast, cannot be activated by WNTs. We successfully generated the FZD 6  fluorescence emission, confirming that the WNT-induced BRET signals detected with FZD 5/6 -based, but not SMO-based, biosensors are not due to the environmental sensitivity of Cy3.
The dynamic BRET signal illustrating the conformational rearrangement of the CRD can be induced by WNT-5A in both the FZD 6 -K466K-Cy3 and the FZD 5 -Q493K-Cy3 sensors with similar median effective concentration (EC 50 ) values (850.3 ± 84.4 ng/ml and 22.4 ± 2.2 nM for FZD 6 -K466K-Cy3 and 807.1 ± 94.6 ng/ml and 21.2 ± 2.5 nM for FZD 5 -Q493K-Cy3; Fig. 2F). Notably, the potency of WNT-5A measured at FZD 5 -Q493Amb is about fourfold lower compared to the affinity of the same recombinant WNT-5A protein when binding to the purified CRD of FZD 5 [5.1 ± 1.6 nM; (9)]. This difference resembles the affinity-to-potency shift observed previously with conformational GPCR biosensors (27). We also tested labeling of the two ECL3 mutant sensors FZD 6 -K466TCO*K and FZD 5 -Q493TCO*K with Tet-BDP-FL and detected WNT-5Ainduced BRET responses for both sensors ( fig. S8, A to C). In contrast to Cy3-labeled receptors, BDP-FL-labeled sensors show positive BRET amplitudes, which can be explained by changes in relative dipole orientation as observed with other intramolecular GPCR biosensors (43,44). More specifically, the two negatively charged sulfonate groups in Cy3 will likely be fully hydrated, resulting in high mobility of the chromophore. By contrast, the high lipophilicity of BDP could result in anchoring the chromophore in the lipid bilayer and a confinement of its mobility most likely explaining the opposite BRET changes in response to the same stimulus.

Intramolecular versus intermolecular BRET responses
Little is known about FZD dimerization, but there is evidence that FZD dimerization through a CRD-CRD interaction contributes to WNT-induced -catenin signaling as it was shown in Xenopus FZD 3 (47) or FZD 5 and FZD 7 (48). FZD 6 exists as homodimer under basal conditions and undergoes dissociation and reassociation upon WNT stimulation (49).
The existence of FZD dimers could result in intermolecular "cross-talk" between Nluc and the FZD 5 and FZD 6 sensors of different monomers. To quantify the contribution of intermolecular BRET to the total BRET response of the FZD CRD sensors, we cotransfected an Nluc-tagged FZD WT, which is not per se able to act as a sensor, and an Nluc-lacking FZD amber mutant. While cotransfection of Nluc-FZD 5 -WT and the FZD 5 -Q493Amb mutant did not result in a detectable WNT-induced BRET response in contrast to the BRET response detected with the intramolecular Nluc-FZD 5 -Q493K-Cy3 sensor ( fig. S9), cotransfecting Nluc-FZD 6 -WT and the Nluc-lacking FZD 6 -K466Amb mutant sensor did. The WNT-5A-induced BRET response obtained with the intermolecular BRET pair setup, however, was smaller compared to the intramolecular Nluc-FZD 6 -K466K-Cy3 sensor ( fig. S10, A and B).
Although stoichiometric differences of the BRET partners in these distinct experiments hamper a direct comparison of the BRET amplitudes, the response detected in the intermolecular BRET setup indicated that a minor part of the total BRET change originates from agonist-induced FZD 6 dimer dissociation. While these findings support the previous data on WNT-induced dimer dissociation (49), the small contribution of the intermolecular BRET does not affect the conclusions about the intramolecular BRET changes with regard to CRD rearrangements. The time course data clearly indicated a kinetic difference between the two assay setups, further underlining that distinct mechanistic processes account for the recorded BRET changes. To provide further support of this assumption, we calculated the rate constant k by fitting the BRET amplitudes for each of the experimental paradigms with intra-and intermolecular sensors. The higher k values for the intramolecular BRET changes compared to the intermolecular sensor argue for a chronological order of the events with faster extracellular conformational changes followed by FZD 6 dimer dissociation ( fig. S10C).
In addition, we made use of the FZD 6 dimerization-deficient triple Ala mutant D365A/R368A/Y369A (49) by introducing three Ala mutations into the Nluc-FZD 6 -K466 amber construct, resulting in the Nluc-FZD 6 -K466TCO*K dimer mutant. The FZD 6 -K466TCO*K dimer mutant maintained cell surface localization even though the surface expression was reduced compared to the FZD 6 -K466TCO*K .74 ± 0.68%). Especially in light of the reduced surface expression of the dimer-deficient FZD 6 sensor, the reduced but yet significant BRET signal argues for dimer dissociationindependent, intramolecular BRET within one monomer between the Nluc fused to FZD's N terminus and the introduced fluorescent dye in ECL3 ( fig. S11C). Furthermore, the basal BRET ratio of the FZD 6 -K466K-Cy3 dimer mutant was significantly reduced compared to the native FZD 6 -K466K-Cy3 sensor, arguing for substantially diminished dimerization tendency of the dimer mutant control ( fig. S11D).
Thus, these dimer control experiments indicate that the BRET amplitudes as a consequence of WNT-induced extracellular conformational changes in FZD 6 present a composite response of both intra-and intermolecular BRET events. In contrast, intermolecular BRET events have no impact on the FZD 5 CRD sensor, which solely detects intramolecular BRET as a consequence of WNT-induced extracellular conformational changes.

Kinetic insights into WNT-induced conformational changes of FZDs
WNT-induced FZD dynamics were previously investigated using intracellular fluorescence-based conformational sensors (20). We aimed to compare the speed of the WNT-induced conformational rearrangements in distinct domains of FZDs by quantifying and comparing the reaction rates of the extracellular and intracellular conformational FZD sensors (Fig. 4A). Therefore, we stimulated HEK293T cells expressing the FZD 5 -Q493K-Cy3 CRD sensor (Fig. 4, B and C) or HEK293A cells stably expressing the FZD 5circularly permutated green fluorescent protein (cpGFP) intracellular sensor (Fig. 4, B to D) with WNT-3A and WNT-5A (3 g/ml) and recorded the resulting BRET or fluorescence response of the two different sensors over time. The resulting rate constant k was found to be significantly higher for the FZD 5 -Q493K-Cy3 CRD sensor for both WNT-3A and WNT-5A (Fig. 4B). The higher rate constant implies that the conformational changes at the extracellular part of FZD occur faster than the intracellular detected rearrangements, highlighting a sequence of events, where WNT-induced conformational changes in the extracellular domain precede those in the core of the receptor.
In contrast, a surrogate WNT (named as WNT-surrogate), an artificial construct composed of a FZD-and a LRP5/6-binding moiety, initiates -catenin signaling through a forced cross-linkage of FZD with LRP5/6 (52). The WNT-surrogate applied to the FZD 5 -Q493K-Cy3 sensor yielded a concentration-response curve for the induction of a TCF transcriptional response (TOPFlash assay) with an EC 50 value of 57 pM (95% confidence interval, 51 to 63 pM; Fig. 5D). However, even at a 10× higher concentration (500 pM), the WNT-surrogate was neither able to induce any BRET responses in the FZD 5 CRD sensor (Fig. 5E) nor fluorescence responses in the cpGFP sensor ( fig. S12), arguing for the absence of extracellular and, in addition, intracellular conformational changes. These findings indicate that WNT-3A induces FZD 5 /LRP5/6 cross-linkage and extracellular conformational changes in FZDs, but the latter are not required to initiate -catenin signaling.

DISCUSSION
The WNT/FZD system represents a primary component of myriad vital biological processes in human physiology and disease. Binding of WNTs to the CRD of FZDs constitutes the initial step of WNT morphogen signaling (2), triggering diverse intracellular signaling cascades through cross-linking of FZDs with WNT co-receptors (most prominently with LRP5/6) (2, 13-15) and by inducing intramolecular, conformational dynamics in FZDs independently from co-receptor interaction (20). Seminal work in models of Drosophila melanogaster has allowed to infer WNT/FZD interaction from intracellular signaling-dependent readouts (23,53,54), and more recent studies with conformational receptor biosensors have provided valuable insights into WNT-induced structural dynamics at the intracellular parts of FZDs (16,18,20). However, how WNT engagement with the extracellular CRD is mechanistically connected to these intracellular receptor conformational changes or FZD/WNT-coreceptor cross-linking remained unclear. The flexibility of the CRD relative to the core of FZDs and its relevance for signal initiation has been a matter of debate, and it remains unclear how the information flow from WNT binding to the CRD is transduced to the core of the receptor provided that the linker domain is freely flexible. The ligand-free CRD of FZD 7 contributes to constitutive receptor activation toward heterotrimeric G s proteins (55). However, the underlying structural aspects of CRD-mediated constitutive activity remain obscure.
Our study provides structural and kinetic insights into this core event of WNT/FZD signaling. We describe the development and validation of optical biosensors that unveil extracellular conformational rearrangements in FZDs in real time in living cells. These biosensors rely on BRET between N-terminally fused Nluc and a fluorescent dye, incorporated in the receptor's linker region or ECL3, using a minimally invasive technique based on site-directed insertion of uaas and bioorthogonal coupling chemistry in live cells to minimize interference with receptor functionality and ligand/ receptor interaction. Although we cannot entirely exclude that sterically induced Nluc displacement triggered by the engagement of a bulky WNT ligand with the CRD contributes to the detected BRET changes, two of our observations argue for a distinct molecular mechanism underlying the optical signals recorded with these biosensors: First, comparing the WNT potencies (EC 50 values) obtained with these biosensors to previously determined WNT binding affinities reveals a four-to sevenfold shifted concentration-response correlation, similar to what has been described for intramolecular GPCR sensors that report on conformational rearrangements at the cytoplasmic side of these membrane-spanning proteins (27). Second, the lack of BRET response seen with the WNT-surrogate ligand, which exhibits an even higher molecular weight than WNT-3A and WNT-5A and is composed of the CRD-binding region of vantictumab, a FZD-targeting antibody blocking WNT-FZD binding (52,56), implies that sterically induced Nluc relocation cannot be the only trigger of the observed BRET changes. These two observations suggest that the BRET signals revealed by these biosensors reflect global conformational changes occurring in FZDs' extracellular domains upon WNT binding.
The biosensors present an important step toward understanding the conformational dynamics of the extracellular domains of FZDs, by shedding light on the stepwise processes occurring between WNT-CRD binding and FZD/transducer coupling. Using a set of distinct BRET-and fluorescence-based assays, we show that CRD movements take place before the rearrangement of the receptors' intracellular domains is initiated and, in the case of FZD 6 , before ligand-induced dissociation of receptor homodimers occurs. Although we cannot provide a causal connection between the two molecular events, our observations indicate that the extracellular CRD rearrangement in FZDs presents an early event underlying signalosome-independent WNT/FZD signaling and confirm-using an unprecedented live cell biosensor system-the central modulatory role of the CRD in class F GPCRs (33). It remains to be defined, what molecular movement in fact determines the agonist-induced changes in BRET using the FZD CRD sensors. In an attempt to better understand the consequences of ligand association with the CRD, we overlaid the WNT structure with the different CRD position clusters extracted from the MD simulations. This schematic overlay ( fig. S1D) suggests that one mechanism resulting in the detected BRET changes could be the restriction of the range of CRD motion through WNT binding, which could be accompanied by a ligandinduced equilibrium shift toward distinct receptor core conformations. More experiments are required to dissect these structural details of FZD activation.
One important mechanistic finding of our study is that the FZD CRD rearrangements detected by our BRET sensors are not required to initiate -catenin signaling. Our experiments with a WNT-surrogate known to mediate -catenin signaling through cross-linkage of FZDs with LRP5/6 (57) showed that (i) WNTsurrogate/FZD 5 /LRP5/6 assembly does not provoke conformational changes in the FZD 5 CRD sensor and (ii) -catenin-dependent signaling can be mediated by the FZD 5 biosensor without the type of extracellular conformational changes detected by our FZD CRD sensor. Likewise, blocking WNT-3A-mediated FZD 5 /LRP5/6 assembly with DKK1 had no effect on the CRD dynamics of FZD 5 . These observations support previous notions of a receptor tyrosine kinase-like functionality of FZDs that exclusively relies on clustering FZD and LRP5/6 in a signalosome (15,20).
Furthermore, we found that WNT-3A and WNT-5A induced very similar BRET responses at saturating concentrations, arguing for analogous CRD rearrangement upon ligand binding. Although we cannot exclude that WNT-3A and WNT-5A promote subtly distinct CRD movements that are not resolved by our FZD CRD sensors, the finding of similar BRET responses is somewhat unexpected in light of the distinct modes of action and intracellular signaling pathways initiated by these two endogenous FZD ligands (FZD/ co-receptor cross-linking and -catenin-dependent signaling by WNT-3A versus FZD conformational changes and -cateninindependent signaling by WNT-5A). This analogy poses the question of whether WNT-3A, concurrent to co-receptor-dependent signal propagation, mediates similar functionalities of FZD 5/6 as WNT-5A by inducing conformational changes in the receptor. FZD 5 alanine mutants of either R 6.32 or W 7.55 , two residues that stabilize the inactive receptor conformation through -cation interaction (8), are completely deficient in mediating WNT-3A-or WNT-surrogateinduced -catenin signaling (15), arguing for receptor conformational control over WNT-3A-dependent intracellular signal propagation. This also implicates that FZD activation and pathway selectivity are governed by two distinct mechanisms: WNT-mediated signalosome formation feeding into WNT/-catenin signaling on the one hand and GPCR-like signal propagation triggered by intramolecular conformational changes that link extracellular WNT-FZD interaction to intracellular FZD-transducer coupling on the other. Supporting the hypothesis of two coexisting activation mechanisms in FZDs, WNT-3A mediates phosphorylation of extracellular signalregulated kinases 1 and 2 (ERK1/2) in primary mouse microglia (45) and regulates small guanosine triphosphatase activity in platelets (58), processes that are often associated with GPCR/G protein signaling. WNT-3A signaling through the WNT/-catenin pathway and -catenin-independent signaling to ERK1/2 occurs simultaneously in mouse primary microglia, albeit with different kinetics, regulating the proinflammatory status of these brain macrophages. While the -catenin-dependent signaling pathway is sensitive to DKK1 treatment, phosphorylation of ERK1/2 is not affected by DKK1-mediated inhibition of FZD-LRP cross-linkage (45). In line with these previous findings, DKK1 pretreatment affects neither the WNT-induced BRET response in FZD CRD nor the fluorescence response in the FZD-cpGFP sensors (20). In summary, these results underline that FZDs can respond to WNTs with signalosome-dependent and signalosome-independent, FZD conformation-dependent signal initiation, which provides a molecular and mechanistic basis for FZD functional selectivity.
In summary, the sensor design described here presents a universal approach to develop cell-based optical probes for membrane-spanning proteins, including other classes of GPCRs, receptor tyrosine kinases, or cytokine receptors, aiding in the mechanistic exploration of fundamental biological processes underlying ligand initiation, kinetics of receptor conformational changes, and cellular signaling in general.

Experimental design
The objectives of this study were to (i) develop a generalizable biosensor design that allows the assessment of the extracellular conformational dynamics of FZDs upon stimulation with WNTs and to (ii) use these sensors to explore the mechanistic details underlying WNT-FZD signaling in living cells. The conformational FZD sensors were expressed in parental and FZD 1-10 HEK293T cells and stimulated with recombinant, commercially available WNTs in at least three independent experiments. The WNT-induced changes in FZD conformation and signaling activity were recorded in 96well microtiter plates and corrected for vehicle responses.
The MD simulations were run using GROMACS 2020.3 (59). FZD 6 was oriented by aligning it to the FZD 4 from the Orientations of Proteins in Membranes (OPM) database database (https://opm. phar.umich.edu/) and embedded in the phosphatidylcholine (POPC) lipid bilayer (150 lipids per leaflet) by CHARMM-GUI server (www. charmm-gui.org/) with TIP3p water molecules and 0.15 M NaCl. The system was minimized for approximately 2000 steps and then equilibrated with gradually decreasing position restraints on protein and lipid components. In the last 50 ns of the equilibration run, the harmonic force constants of 50 kJ mol −1 nm −2 were applied on the protein atoms only.
Seven (250 ns) independent isobaric and isothermic (NPT) ensemble production simulations were initiated using the CHARM-M36m force field (60) and a 2-fs time step. First, replica 1 was run starting from the equilibrated structure and random velocities. Then, six other replicas were simulated starting from the snapshots of replica 1 at time points t = 0 ns, t = 50 ns, t = 100 ns, t = 150 ns, t = 200 ns, and t = 250 ns and random velocities. The temperature at 303.15 K was maintained with a Nose-Hoover thermostat (61), and the pressure at 1 bar was maintained with a Parrinello-Rahman barostat (62). Potential-shift-Verlet was used for electrostatic and van der Waals interactions with a 12-Å cutoff, and the bonds between hydrogen and other atoms were constrained by the Linear Constraint Solver for Molecular Simulations (LINCS) algorithm (63). The data were analyzed using VMD (visualization, measurement of RMSDs and distances, and RMSD clustering) (64) and visualized in PyMol. For RMSD clustering, the size of the trajectory was reduced to contain one frame per every 10 ns of the simulation, the frames were superimposed on the 7TM core of the receptor at the first frame of the first replica, and similarity cutoff was set to 5 Å and maximum number of clusters to 30. The cluster seeds were used as the representative models of each cluster. The MD data will be deposited to GPCRmd (an open-access MD database for GPCRs; www.gpcrmd.org).

Cloning of FZD constructs
The synthetic gene constructs for FZD 5 and FZD 6 were designed on the basis of the amino acid sequences of the full-length receptor lacking the native signal peptide [residues 27 to 585 of Uniprot Q13467 (FZD5_HUMAN) and residues 19 to 706 of Uniprot O60353 (FZD6_HUMAN)] and codon-optimized for expression in human cells using the GeneArt online tool while avoiding a set of motifs corresponding to several restriction sites (Nhe I, Hind III, Nco I, Eco RI, Sbf I, Mfe I, Kpn I, Not I, and Xba I). We extended the 5′ end of the genes with the nucleotide sequence 5′-AAG CTT GCC GCC ACC ATG GCG CTG TGT ATC CCT CAA GTT CTG CTG GCC CTG TTC CTG AGC ATG CTG ACA GGA CCT GGC GAG GGC TAC CCT TAC GAT GTG CCT GAC TAC GCC GAA TTC GCT CCT GCA GGG AGT CAA TTG-3′ that adds the restriction site Hind III used for cloning and the protein sequence MALCIPQVLLALFLSMLTGPGEGYPYDVPDYAEFAPAGSQL to the N terminus of the receptor. This sequence is derived from the mouse serotonin 5-HT 3 receptor-cleavable signal sequence carrying the R2A mutation to enable usage of a strong Kozak consensus sequence (GCCGCCACCATGG, start codon underlined) and a hemagglutinin tag (YPYDVPDYA) followed by a linker EFAPAGSQL corresponding to a nucleotide sequence with Eco RI, Sbf I, Mfe I, and Mly I sites (65). We extended the 3′ end with the sequence 5′-GGT ACC GCC TCC TCG GAT GAG GCC AGC ACA ACC GTG TCT AAG ACC GAG ACA TCT CAG GTG GCC CCT GCC TAA GCG GCC GCT CTA GA-3′ containing a Kpn I site at the 5′ end and Not I and Xba I sites at the 3′ end. The C-terminal 18-residuelong sequence is the rhodopsin 1D4 mAb epitope tag (66). The constructs were synthesized and cloned into a plasmid derived from pcDNA3.1(+) modified to eliminate an internal Mfe I site.
All FZD 5 and FZD 6 amber mutants were generated using the GeneArt Site-directed Mutagenesis System (Thermo Fisher Scientific) with the following primers: FZD 6  To boost the amber suppression, we introduced the respective amber mutant-bearing FZD into an expression plasmid carrying four repeats of the orthogonal suppressor tRNA (plasmid as a gift from S. Elsässer, Addgene number: 140008).
For cloning the BRET extracellular sensors, an Nluc tag in combination with FZD 5 or FZD 6 was subcloned into the tRNA expression vector using the NEBuilder HiFi DNA Assembly Kit (New England Biolabs). Briefly, the tRNA expression vector was digested with Nhe I and Bam HI. The Nluc tag was obtained from the Nluc-FZD 6 construct (16) and contains the N-terminal protein sequence MRLCIPQVLLALFLSMLTGPGEGSRKL (signal peptide derived from the 5-HT 3 receptor, followed by the Hind III restriction site and Nluc). The Nluc tag was cloned N-terminally of FZD 5 or FZD 6 , both connected via a linker (EFAPAGSQL, corresponding to a nucleotide sequence with Eco RI, Sbf I, Mfe I, and Mly I), with the following primers: Nluc, 5′-TCC AAG CTG TGA CCG GCG CCT ACT CTA GAG CTA GCC ACC ATG CGG CTC TGC-3′ (forward) and 5′-TGC AGG AGC GAA TTC CGC CAG AAT GCG TTC GCA C-3′ (reverse); FZD 5/6 , 5′-GAA CGC ATT CTG GCG GAA TTC GCT CCT GCA GGG AGT C-3′ (forward) and 5′-GCA GAC AGC GAA TTA ATT CCA GCG GCC GCG GAT CCG GCC GCT TAG GCA GGG GC-3′ (reverse). For the dimer control construct, FZD 5 -Q493Amb or FZD 6 -K466Amb was subcloned without the N-terminally Nluc tag into the tRNA expression vector using the NEBuilder HiFi DNA Assembly Kit with the following primers: Cell culture, transfection, and treatments HEK293T cells cultured in Dulbecco's modified Eagle's medium (DMEM) with 1% penicillin/streptomycin and 10% fetal bovine serum (all from Thermo Fisher Scientific) in a humidified 5% CO 2 incubator at 37°C. Cells were transfected 24 hours after seeding with Lipofectamine 2000 according to the supplier's information (Invitrogen). The absence of mycoplasma contamination was routinely confirmed by polymerase chain reaction using 5′-GGC GAA TGG GTG AGT AAC ACG-3′ (forward) and 5′-CGG ATA ACG CTT GCG ACT ATG-3′ (reverse) primers detecting 16S ribosomal RNA of mycoplasma in the media after 2 to 3 days of cell exposure.

Immunoblotting
The day prior transfection, 100,000 HEK293T cells per well were seeded in 24-well plates. The cells were transfected with a defined transfection ratio of 9:1 with 0.45 g of the indicated constructs and 0.05 g of tRNA/synthetase (Addgene number: 140023; control conditions were balanced with pcDNA) per well and were cultured in the absence or presence of 0.1 mM TCO*K (Sichem, SC-8008). Because FZD 5 -Q493TCO*K was weakly expressed, 300,000 HEK293T cells per well were seeded in 12-well plates. The cells were transfected with 0.9 g of the FZD 5 -Q493Amb construct or 0.18 g of FZD 5 -WT/0.72 g of pcDNA3.1 and 0.1 g of tRNA/synthetase (control conditions were balanced with pcDNA). Cells were lysed 48 hours after transfection in 2× Laemmli buffer containing 200 mM dithiothreitol (Merck). Lysates were sonicated and separated by SDSpolyacrylamide gel electrophoresis/immunoblotting using 7.5% gels. Transfer to a polyvinylidene difluoride membrane was done with the Trans-Blot Turbo Transfer System (Bio-Rad). After transfer, membranes were incubated in 5% low-fat milk/TBS-T [25 mM tris-HCl, 150 mM NaCl, and 0.05% Tween 20 (pH 7.6)] and subsequently in primary antibodies overnight at 4°C. The next day, the membranes were washed four times in TBS-T, incubated with goat anti-mouse or goat anti-rabbit secondary antibody conjugated to horseradish peroxidase (Thermo Fisher Scientific, 1:5,000), washed, and developed using Clarity Western ECL Substrate (Bio-Rad) according to the supplier's information. Primary antibodies were as follows: anti-1D4 (National Cell Culture Center, mouse; 1:1000), anti-Nluc (R&D Systems, MAB100261-SP, mouse; 1:500), and glyceraldehyde-3-phosphate dehydrogenase (Cell Signaling Technology, 2118, rabbit, 1:4000).

Whole-cell ELISA
For quantification of cell surface receptor expression, 15,000 HEK293T cells were plated in 96-well plates precoated with poly-d-lysine (PDL). Next day, cells were transfected with 0.09 g of the indicated constructs and 0.01 g of tRNA/synthetase and were cultured in the absence or presence of 0.1 mM TCO*K. After 48 hours, cells were incubated with an anti-Nluc antibody (R&D Systems, MAB100261-SP, mouse; 1:500) in 1% BSA/DPBS for 1 hour at 4°C. Following incubation, cells were washed five times with 0.5% BSA/DPBS and probed with a horseradish peroxidase-conjugated goat anti-mouse antibody at a 1:2500 dilution in 1% BSA/DPBS for 1 hour at 4°C. The cells were washed five times with 0.5% BSA/ DPBS, and 100 l of the peroxidase substrate 3,3′,5,5′-tetramethylbenzidine (Merck) was added (30 min at room temperature). After acidification with 100 l of 2 M HCl, the absorbance was read at 450 nm using a POLARstar Omega plate reader (BMG Labtech).

Live cell imaging
The day prior transfection, 15,000 HEK293T cells per well were seeded into PDL-precoated black 96-well glass bottom plates. The cells were transfected with 0.09 g of the indicated constructs and 0.01 g of tRNA/ synthetase and were cultured in the absence or presence of 0.1 mM TCO*K. Forty-eight hours after transfection, cells were washed in DPBS, kept for 2 hours in DMEM to remove remaining TCO*K, and labeled with 1 M Tet-Cy3 (Jena Bioscience, CLK-014-05) or Tet-BDP-FL (Jena Bioscience, CLK-036-05) for 30 min. Cells were washed again with DPBS and kept for an additional 30 min in DMEM. Last, DMEM was exchanged with 0.1% BSA/Hanks' balanced salt solution (HBSS), and cells were imaged using a Zeiss LSM880 confocal microscope.

Assessment of fluorescence labeling efficacy
For quantification of the fluorescence labeling of the receptor mutants, 15,000 HEK293T cells were plated in black PDL-precoated 96-well plates. Next day, cells were transfected with 0.09 g of the indicated constructs and 0.01 g of tRNA/synthetase and were cultured in the absence or presence of 0.1 mM TCO*K. Forty-eight hours after transfection, cells were washed in DPBS, kept for 2 hours in DMEM, and labeled with 1 M Tet-Cy3 (Jena Bioscience, CLK-014-05) or Tet-BDP-FL (Jena Bioscience, CLK-036-05) for 30 min. Cells were washed again with DPBS and kept for an additional 30 min in DMEM. Last, DMEM was exchanged with 0.1% BSA/HBSS, and fluorescence intensities were read using a POLARstar Omega plate reader (BMG Labtech, Ortenberg, Germany) equipped with filters for Cy3 (excitation, 544 nm/emission, 590 nm) and BDP-FL (excitation, 485 nm/emission, 520 nm).

Nluc-FZD BRET measurements
For BRET measurements with the Nluc-FZD 5/6 CRD sensors, 15,000 HEK293T cells were plated in PDL-precoated white 96-well plates. Next day, cells were transfected with 0.09 g of the indicated constructs and 0.01 g of tRNA/synthetase and were cultured in the presence of 0.1 mM TCO*K. Forty-eight hours after transfection, cells were washed in DPBS, kept for 2 hours in DMEM, and labeled with 1 M Tet-Cy3 or Tet-BDP-FL for 30 min. Cells were washed with DPBS and kept for an additional 30 min in DMEM. Next, cells were again washed with DPBS and incubated with 90 l of a 1/1000 dilution of furimazine stock solution (Promega) in 0.1% BSA/ HBSS. After 5 min of incubation, the basal BRET ratio was measured in three consecutive reads, and 10 l of a WNT-3A or WNT-5A solution (3 g/ml; in 0.1% BSA/HBSS) or vehicle control was applied per well. Subsequently, the BRET ratio was recorded for an additional 25 to 60 min. For experiments with a higher temporal resolution, eight baseline BRET reads were recorded within 2 min prior manual addition of compounds or vehicle control, followed by at least 40 reads. All experiments were conducted using a CLAR-IOstar plate reader (BMG Labtech, Ortenberg, Germany) equipped with monochromators to separate Nluc (450/80 nm), BDP-FL (520/40 nm), and Cy3 (580/30 nm), respectively.

Nluc-FZD and Nluc-SMO fluorescence measurements
For fluorescence measurements with the Nluc-FZD 5 -Q493K-Cy3 and Nluc-SMO-E508K-Cy3 CRD sensors, 15,000 HEK293T cells were plated in PDL-precoated black 96-well plates. Next day, cells were transfected with 0.09 g of the indicated constructs and 0.01 g of tRNA/synthetase and were cultured in the presence of 0.1 mM TCO*K. Forty-eight hours after transfection, cells were washed in DPBS, kept for 2 hours in DMEM, and labeled with 1 M Tet-Cy3 for 30 min. Cells were washed with DPBS and kept for an additional 30 min in DMEM. Next, cells were again washed with DPBS and incubated with 90 l of 0.1% BSA/HBSS. Baseline fluorescence was recorded in three consecutive reads, followed by application of 10 l of a WNT-3A or WNT-5A solution (3 mg/ml; in 0.1% BSA/HBSS) or vehicle control per well, and the resulting fluorescence intensity was recorded for an additional 45 min using a CLARIOstar plate reader (BMG Labtech, Ortenberg, Germany) equipped with filters to excite Cy3 (580/30 nm).

FZD-cpGFP experiments
HEK293A cells stably expressing FZD 5 -cpGFP (20) were seeded at a density of 80,000 cells per well onto PDL-precoated, black-wall, black-bottomed 96-well plates. Twenty-four hours later, all wells were washed with HBSS and incubated with 0.1% BSA/HBSS. Baseline fluorescence was recorded in eight consecutive reads within 2 min, 10 l of 10-fold WNT solution or vehicle control was applied per well, and the resulting fluorescence intensity was recorded for an additional 40 reads. All experiments were conducted using a CLARIOstar plate reader (BMG Labtech, Ortenberg, Germany) equipped with filters to excite cpGFP (470/15 nm) and record its emission intensity (515/20 nm). Forty flashes were applied per data point.
TOPFlash reporter gene assay FZD 1-10 HEK293T cells (400,000 cells/ml) were transfected in suspension with 400 ng of M50 Super 8xTOPFlash, 100 ng of pRL-TK Luc, 450 ng of Nluc-FZD 5 -Q493Amb, and 50 ng of tRNA/synthetase per milliliter of cell suspension; supplemented with 0.1 mM TCO*K; and seeded onto PDL-precoated white-wall, white-bottomed 96-well plates (50,000 cells per well). Twenty-four hours after transfection, cells were washed with 100 l of HBSS and incubated for 4 hours in fetal bovine serum-reduced (0.5%) DMEM (72 l per well) supplemented with 10 nM C59. Thereafter, 8 l of recombinant WNT-3A (10 g/ml) and/or DKK1 (in 0.1% BSA/HBSS) and varying concentrations of WNT-surrogate or the respective vehicle controls were added. Twenty-four hours after stimulation, cells were washed with HBSS and lysed in 30 l of Promega's dual luciferase passive lysis buffer. Subsequently, 20 l of luciferase assay reagent (LARII) was added to each well, and -catenin-dependent firefly luciferase (Fluc) intensity was measured using a CLARIOstar microplate reader (580/80 nm; 1-s integration time). Next, 20 l of Stop&Glo Reagent was added to quantify Renilla luciferase (Rluc) emission intensity (480/80 nm; 1-s integration time) to control for variations in cell number and transfection efficiency.

Statistical analysis
All immunoblot experiments are representative of three independent experiments. Statistical and graphical analysis was done using GraphPad Prism 9 software. For analyzing the surface expression, the basal absorbance detected in pcDNA-transfected HEK293 cells was subtracted from all data, and mean values were normalized to WT FZD 5 or FZD 6 , which was set to 100%. Differences among the TCO*K-untreated and TCO*K-treated groups were analyzed by one-way analysis of variance (ANOVA) with uncorrected Fisher's least significant difference (LSD) test. Significance levels are given as follows: *P < 0.05, **P < 0.0196, ***P < 0.001, and ****P < 0.0001. All data points represent normalized values, each performed in triplicate. Bars show means ± SEM of three to four independent experiments.
Differences in fluorescence labeling among the TCO*K-untreated and TCO*K-treated groups were analyzed by one-way ANOVA with uncorrected Fisher's LSD test. Significance levels are given as follows: *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. All data points represent mean values, each performed in triplicate. Bars show means ± SEM of three to six independent experiments.
BRET ratios were interpreted as acceptor emission/donor emission. At least three individual BRET reads were averaged before ligand/vehicle application. For quantification of ligand-induced changes, BRET was calculated for each well as a percentage over basal BRET. Subsequently, the average BRET of the vehicle control was subtracted. All data points represent mean values, each performed in triplicate or quadruplicate, ± SEM of three to five independent experiments. For analyzing the BRET data, the WNT-induced averaged BRET responses were fitted with a plateau followed by one-phase decay equation. Data from concentration-response experiments were fitted using a four-parameter fit. All data are represented as means ± SEM of at least three independent experiments. Data from TOP-Flash experiments were expressed as ratios of Fluc over Rluc luminescence intensity to correct for distinct transfection efficiencies in the different samples. The resulting TOPFlash ratios were subsequently normalized for the average TOPFlash ratio of vehicle-treated wells to express WNT-induced changes as increases over baseline.