Quantitative Profiling of WNT-3A Binding to All Human Frizzled Paralogues in HEK293 Cells by NanoBiT/BRET Assessments

The WNT signaling system governs critical processes during embryonic development and tissue homeostasis, and its dysfunction can lead to cancer. Details concerning selectivity and differences in relative binding affinities of 19 mammalian WNTs to the cysteine-rich domain (CRD) of their receptors—the ten mammalian Frizzleds (FZDs)—remain unclear. Here, we used eGFP-tagged mouse WNT-3A for a systematic analysis of WNT interaction with every human FZD paralogue in HEK293A cells. Employing HiBiT-tagged full-length FZDs, we studied eGFP-WNT-3A binding kinetics, saturation binding, and competition binding with commercially available WNTs in live HEK293A cells using a NanoBiT/BRET-based assay. Further, we generated receptor chimeras to dissect the contribution of the transmembrane core to WNT-CRD binding. Our data pinpoint distinct WNT-FZD selectivity and shed light on the complex WNT-FZD binding mechanism. The methodological development described herein reveals yet unappreciated details of the complexity of WNT signaling and WNT-FZD interactions, providing further details with respect to WNT-FZD selectivity.

T he ten mammalian Frizzleds (FZD 1−10 ) are G proteincoupled receptors (GPCRs) and formtogether with Smoothened (SMO)the class F of GPCRs. 1,2 The 19 different WNT lipoglycoproteins are the main macromolecular ligands of FZDs, interacting with the extracellular cysteine-rich domain (CRD) of the receptor. WNT-FZD signaling orchestrates multiple processes during embryonic development, stem cell regulation, and adult tissue homeostasis. 2 Additionally, aberrant WNT signaling is implicated in tumorigenesis and other pathologies. 3,4 Whereas recent advances have resulted in a better understanding of the underlying mechanisms controlling WNT-induced FZD activation and signal initiation, the relative binding affinities and ligand−receptor selectivity remain largely unknown. 5−13 The quantitative assessment of WNT binding has been limited by the strong lipophilicity of WNTs, which makes their purification challenging and necessitates detergents and serum for solubilization and stabilization of WNTs, respectively. 14,15 Nevertheless, WNT-FZD interactions were studied using biochemical and biophysical assays as well as through the use of in silico calculations. 16−22 These studies generally reported WNT-FZD binding affinities in the range of 1−100 nM, which is reasonable when considering the known affinities of proteinaceous ligands to other GPCRs. 23,24 Nevertheless, it remains unclear how these values translate into the physiological reality, since the local concentration of WNTs at the receptors in vivo remains unknown and is likely to be highly context-dependent. 25 Until recently, the assessment of ligand binding was based on WNT binding to the CRD rather than the full-length FZD, or the reported assays were not performed in live cells. However, progress has been made with the use of FRET-and cpGFP-based biosensors to demonstrate WNT-induced FZD conformational dynamics and receptor activation, 7,12,26 where WNT binding to a full-length receptor was reflected by an outward movement of the transmembrane domain TM6. Furthermore, the generation of a functional eGFP-tagged WNT-3A provided for the first time a biologically relevant FZD ligand that could be used as a probe in quantitative binding assays in real-time using living cells. 27,28 Here, using live cell analysis of transiently transfected HEK293A cells overexpressing HiBiT-tagged FZDs, we provide a comparative assessment of binding affinities of eGFP-WNT-3A to all human FZD paralogues using kinetic and saturation binding formats. Furthermore, using a competition binding assay, we have assessed binding affinities of unlabeled, commercially available WNT proteins to FZD 4 . Finally, we have also explored the contribution of the FZD transmembrane core for the binding of WNTs to the primary FZD-CRD binding site. 29 Compared with the previously described BRET-based assay for Nluc-FZD 4 and Nluc-FZD 6 , 28 we have used a nanoluciferase complementation-based BRET binding approach here, termed NanoBiT/BRET. In this BRET assay format, the fluorescent WNT-3A binds FZDs that are Nterminally tagged with the 11-amino-acid HiBiT peptide. The addition of the complementary LgBiT to the system allows rapid and high-affinity association to the HiBiT peptide, forming a stable NanoBiT moiety with a luciferase activity. This setup allows targeted analysis of cell surface receptors due to the cell impermeability of LgBiT, thereby providing a system with less intracellular background luminescence. This method is a modification of a well-established NanoBRET binding assay to study ligand−receptor association, 30 and has been lately employed to study ligand binding to Class A GPCRs and receptor tyrosine kinases. 31−34 Our results demonstrate that eGFP-WNT-3A interacts with full-length human FZDs transiently overexpressed in live HEK293A cells in a paralogue selective manner. This concept was expanded to unlabeled WNTs in competition binding experiments suggesting a complex WNT-FZD selectivity profile. The binding data based on full-length FZDs, FZD-CD86, and FZD-FZD chimeras underline the complexity of the WNT-FZD interaction and suggest that the core regions of FZDs may contribute to receptor selectivity.
■ RESULTS AND DISCUSSION eGFP-WNT-3A/FZD Binding Kinetics. In order to establish a nanoluciferase complementation-dependent Nano-BiT/BRET binding assay format to study all human FZD paralogues, we generated constructs for all 10 receptors carrying an N-terminal HiBiT tag ( Figure S1A,B). Upon transient overexpression in HEK293A cells, all receptor constructs were detected at the cell surface, albeit with varying cell surface expression levels ( Figure S1C). Additionally, HiBiT-tagged FZD 1 , FZD 2 , FZD 4 , FZD 5 , FZD 7 , FZD 8 , and FZD 10 mediated WNT-3A-induced β-catenin-dependent signals as assessed by the TOPFlash reporter assay performed in HEK293T cells devoid of endogenous FZDs (ΔFZD 1−10 Figure 1. eGFP-WNT-3A binding kinetics. A. The scheme depicts the experimental setup of the NanoBiT/BRET analysis of association kinetics between the HiBiT-tagged FZD and the eGFP-WNT-3A. Created with BioRender.com. B. Association kinetics of the eGFP-WNT-3A to human HiBiT-FZDs were determined by the detection of NanoBiT/BRET in transiently overexpressing living HEK293A cells over time. BRET was sampled once per 90 s for 240 min. Data points are presented as means ± SEM from n = 3 individual experiments, fitting a two or more hot ligand concentrations kinetics model. Experiments were performed with eGFP-WNT-3A batch 1. HEK293T cells, Figure S1D 25 ). In contrast, FZD 3 , FZD 6 , and FZD 9 could not transduce WNT-3A-induced activation of this pathway, similar to what has been reported previously, yet with differing results for FZD 9 . 28,35−38 Having verified the sequence and functionality of the HiBiT-tagged FZD constructs, HEK293A cells transiently overexpressing these FZDs were used in kinetic binding experiments. For these experiments, cells were first incubated with the complementary LgBiT protein and the luciferase substrate vivazine, and after 1 h, eGFP-WNT-3A was added to final concentrations of 2.1, 4.2, or 8.3 nM, and BRET readings were taken over a 4 h period at 37°C ( Figure 1A). The concentrations used for eGFP-WNT-3A were dictated by the maximal concentration that could be obtained for the eGFP-WNT-3A preparations and the assay format. We detected a saturable net BRET ratio indicative of eGFP-WNT-3A specific binding to HiBiT-tagged FZD 1 , FZD 2 , FZD 4 , FZD 5 , FZD 7 , and FZD 10 , with K d values varying from 2.3 to 29.9 nM ( Figure 1B, Table 1). In line with the TOPFlash data, no concentration-dependent increase in receptor−ligand BRET was detected for FZD 3 and FZD 9 . Interestingly, FZD 8 , which maintained a strong WNT-3Ainduced TOPFlash activity, and to a lesser extent FZD 6 , displayed very low but detectable binding that could be fitted to association curves over time ( Figure S2A).
eGFP-WNT-3A/FZD Binding Affinity at Equilibrium. To define saturation binding affinity of eGFP-WNT-3A, we incubated human HiBiT-FZDs with a full concentration range (16.7 pM to 16.7 nM) of eGFP-WNT-3A for 240 min at 37°C (Figure 2A). The net BRET ratio representing ligand−receptor binding increased in a clear, concentration-dependent manner  Figures 1B and 2B) and shown as a best-fit value ± SEM; n.d. = not determined. for FZD 1 , FZD 2 , FZD 4 , FZD 5 , FZD 7 , and FZD 10 . Unfortunately, using transient overexpression of HiBiT-FZDs in HEK293A cells and the eGFP-WNT-3A with a limited maximal concentration in the conditioned medium, binding curves did not reach maximal asymptotic values but only came to near-saturable levels ( Figure 2B). Again, detection of binding of eGFP-Wnt-3A to FZD 6 and FZD 8 was only marginally above background levels, as the net BRET values were low ( Figure S2B). Similar to the kinetic binding assays, no quantifiable eGFP-WNT-3A binding was detected for FZD 3 or FZD 9 . The affinities of eGFP-WNT-3A/FZD interactions were determined from linear regression curves showing near-saturable binding, 39 and the K d values are shown in Table 1. The reported saturation binding affinity values range from 4.9 to 48.6 nM, and they are in good agreement with the K d values determined with kinetic binding for FZD 1 , FZD 4 , and FZD 6 . The degree of agreement is, however, only fair for FZD 5 , FZD 8 , and FZD 10 and relatively poor for FZD 2 and FZD 7 . Taken together, these kinetic and saturation binding data are in line with our previous results using fluorescence microscopy analysis, where no eGFP-WNT-3A association could be observed with C-terminally mCherry-tagged FZD 6 , and only a very weak association with FZD 8 and FZD 9 (FZD 3 was not used). 28 Although this fluorescence imaging-based method could provide an estimate of the relative ability of eGFP-WNT-3A to associate with different FZDs, accurate quantification of the binding affinities was not possible. Also, in that study, ΔFZD 1−10 HEK293 GFP-free cells overexpressing FZD 8 -mCherry (but not FZD 6 -mCherry), showed very faint binding of eGFP-WNT-3A, and this is also in agreement with the HiBiT-tagged system used here, which can detect very low level, but specific, binding to FZD 8 ( Figure S2). This is in agreement with a recent report claiming that ligand−receptor interaction using the HiBiT-tagged system allows detection of very weak interactions. 32 Furthermore, in the case of FZD 6 , there are differences in reports of its ability to bind or respond to WNT-3A. Biochemical experiments failed to detect any association between WNT-3A and FZD 6 -CRD-IgG, 16 and no eGFP-WNT-3A/Nluc-FZD 6 interaction was detected in our previous NanoBRET study. 28 However, it should be noted that, compared with the HiBiT-tagged systems, binding analyses with Nluc-tagged receptors can display reduced sensitivity for detection of weak interactions, as recently discussed. 32 On the other hand, recombinant human WNT-3A induced a conformational change in FZD 6 12 and affected the mobility of the receptor in the cell membrane as assessed by fluorescence recovery after photobleaching assay. 40 Finally, the findings with respect to FZD 8 are particularly intriguing, given the existing structural information on the WNT-3A-FZD 8 CRD complex. 41 It is currently unclear why such discrepancies exist for these two FZD paralogues.
In order to directly compare NanoBiT/BRET and Nano-BRET binding assay formats, we used Nluc-FZD 4 and HiBiT-FZD 4 constructs for saturation binding ( Figure S3A−C). NanoBiT/BRET experiments were performed in two different experimental paradigms, where LgBiT protein was added either directly after (setup 1, used throughout this study Figure  S3B and Figure 2A) or for 10 min before (setup 2, Figure  S3C) the 4 h incubation with eGFP-WNT-3A. This allowed us to test for any potential steric hindrance of WNT binding to FZD caused by the presence or complementation of nanoluciferase. In these comparisons, although the differences in K d values were apparent, they did not reach statistical significance (Nluc-FZD 4 K d (nM) ± SEM = 7.6 ± 3.6; HiBiT-FZD 4 setup 1 K d (nM) ± SEM = 11.9 ± 3.6; HiBiT-FZD 4 setup 2 K d ± SEM (nM) = 20.4 ± 7.6). Furthermore, eGFP-WNT-3A binding to Nluc-FZD 4 ( Figure S3A) resulted in lower maximal BRET (BRET max ) compared to either of the two HiBiT-FZD 4 binding setups ( Figure S3B,C) at similar luminescence levels (Nluc-FZD 4 BRET max ± SEM = 0.026 ± 0.006 vs HiBiT-FZD 4 setup 1 BRET max ± SEM = 0.049 ± 0.005, P = 0.0484; vs HiBiT-FZD 4 setup 2 BRET max ± SEM = 0.049 ± 0.004, P = 0.0349). This suggests that intracellular luminescence originating from receptors that are not accessible for the ligand reduces the assay's dynamic range (see Figure  S3D for expression analysis). In support of our choice to change from NanoBRET to the NanoBiT/BRET experimental setup, a recent study has shown that affinity measurements obtained from NanoBRET binding assays were generally less consistent in comparison with NanoBiT/BRET analyses. 32 eGFP-WNT-3A Competition Binding with Untagged WNTs at FZD 4 . With the aim to understand the competitive nature of eGFP-WNT-3A binding to FZD, we combined eGFP-WNT-3A with increasing concentrations of several commercially available and purified untagged WNT proteins: WNT-3A, WNT-5A, WNT-5B, WNT-10B, WNT-11, and WNT-16B. Again, we have used HiBiT-FZD 4 as our model receptor. In this assay setup, HEK293A cells transiently overexpressing HiBiT-FZD 4 were preincubated with the untagged WNTs for 30 min before addition of eGPF-WNT-3A to a final concentration of 0.4 nM. Cells were then incubated for a further 4 h to allow a competitive equilibrium to be reached before addition of LgBiT and subsequent BRET measurement ( Figure 3A). The results of these experiments are shown in Figure 3B and summarized in the Table 2, and show that the untagged WNTs competitively displaced eGFP-WNT-3A from FZD 4 with different affinities and capacities. Interestingly, WNT-10B and WNT-11 had the highest affinities, but they showed moderate BRET signal decrease as indicated by remaining residual BRET. Intriguingly, WNT-3A presented higher binding affinity (lower K d ) but caused a lower reduction of BRET than WNT-5A and WNT-5B. WNT-16B did not compete with eGFP-WNT-3A in this FZD 4 -based assay. These results are in fair agreement with the recently published potencies and efficacies of WNTs in eliciting conformational changes in FZD 4 -cpGFP biosensor except for WNT-5B. 12 Nevertheless, a similar rank order of affinities was obtained for WNT-3A, WNT-5A, and WNT-5B in binding to the isolated FZD 4 CRD. 17 Obviously, the insights into the mechanism of WNT−WNT competition for the primary binding site at FZDs remain obscure. We can only speculate that the process of functional ligand binding is more complex than for small-molecule ligands and other GPCRs. Along these lines, FZD oligomers associate or dissociate upon ligand addition, 42−44 adding to the complexity of ligand binding analysis. 45 Additionally, FZD coreceptors and various regulators could alter WNT interactions with FZD in HEK293 cells. 25 Binding of eGFP-WNT-3A to FZD Chimeras. WNTs directly engage the CRD through protein−protein and protein−lipid interactions. 41,46 However, it remains unclear whether non-CRD domains of FZDs contribute to WNT-FZD interaction. 20 It has been hypothesized that the CRD simply serves the purpose of binding WNTs in order to bring them in close proximity with the receptor for additional binding/ activation mechanisms. 47 Indeed, the long and flexible nature of the linker connecting the CRD of FZDs to TM1 would tend to support such a hypothesis. 48 The role of the FZD transmembrane domains in ligand binding, complex formation, receptor conformational changes, and signal transduction has Data are based on n = 3−6 individual experiments presented in Figure 3B. pK i values are presented as a best-fit value ± SEM; n.d. = not determined.
been a subject of debate. 5,8,9,49−52 Here, we seek to obtain a more mechanistic insight into the contribution of the transmembrane core to eGFP-WNT-3A binding. To this end, we generated two chimeric proteins fusing the N-terminal domain (NTD; CRD + linker) of one FZD with an unrelated CD86 single transmembrane domain spanning protein ( Figure   4A). Specifically, we generated FZD 4 -CD86 and FZD 8 -CD86 chimeric proteins ( Figure S4A). In this manner, we aimed to study the effect of a FZD core on eGFP-WNT-3A binding to the CRD. We validated the chimeras with regard to proper membrane trafficking upon transient overexpression in HEK293A cells and detected no difference in surface  Figure 2B and parts C−D). Data are presented as means ± SEM from n = 3−4 individual experiments. Expression data of FZD 4 , FZD 8 , and pcDNA are also depicted in SI Figure 1C. C. Saturation binding of eGFP-WNT-3A at FZD 4 -CD86 and FZD 4 (data also present in Figure 2B) was determined by the detection of NanoBiT/ BRET in transiently overexpressing living HEK293A cells following 240 min incubation. Data points are presented as means ± SEM from n = 3−4 individual experiments. eGFP-WNT-3A batch 1 was used in these experiments. D. Saturation binding of eGFP-WNT-3A at FZD 8 -CD86 and FZD 8 (data also present in Figure 2B) was determined by the detection of NanoBiT/BRET in transiently overexpressing living HEK293A cells following 240 min incubation. Data points are presented as means ± SEM from n = 3−4 individual experiments. eGFP-WNT-3A batch 1 was used in the experiments. Linear scale data are fitted to a one-site specific binding model. Logarithmic-scale data are fitted to a normalized three-or fourparameter model. The right plot in every panel shows data normalized between 0% (BRET min ) and 100% (BRET max ) for each studied construct. expression levels between FZD-CD86 and wild-type (WT) FZDs ( Figure 4B).
Next, we performed a similar analysis for a FZD 8 -CD86 chimera. eGFP-WNT-3A binding to the chimera FZD 8 -CD86 compared to FZD 8 at similar levels of receptor surface expressions (P = 0.8340; Figure 4B) did not differ in affinity in the analysis of non-normalized data (FZD 8 -CD86 K d ± SEM (nM) = 22.9 ± 10.6, P = 0.6668; Figure 4D), but the difference reached statistical significance when comparing the normalized values (P = 0.0390; Figure 4D). Furthermore, the NanoBiT/BRET signal increased significantly when the FZD 8 core was replaced by CD86. The NanoBiT/BRET signal (BRET max ) was in fact comparable to other WNT-3A-binding competent FZDs (FZD 8 -CD86 BRET max ± SEM = 0.056 ± 0.006 vs FZD 8 BRET max ± SEM = 0.003 ± 0.0003, P = 0.0002; Figure 4D and Figure 2). Intriguingly, these findings are the opposite of what we observe for FZD 4 , where replacing the receptor core with CD86 visibly reduced maximal BRET ( Figure 4C). The efficiency of resonance energy transfer depends on both orientation and distance between BRET donor and BRET acceptor. 53 Thus, in the NanoBiT/BRET binding setup, the differences in BRET max can be interpreted as distinct ligand−receptor conformations. This further suggests that the cores of FZD 4 and FZD 8 can differently contribute to WNT-FZD binding.
In addition to the FZD-CD86 chimeras detailed above, we also generated FZD-FZD chimeras. Specifically, we constructed FZD 4 -FZD 6 , FZD 4 -FZD 8 , FZD 5 -FZD 6 , FZD 5 -FZD 7 , and FZD 6 -FZD 4 chimeric proteins ( Figure S4B). In this manner, we have used at least one FZD paralogue from every FZD homology cluster. The transmembrane cores of FZD 6 and FZD 8 were selected to test whether they negatively affect eGFP-WNT-3A binding to the CRD of FZD 4 and FZD 5 . On the other hand, the FZD 7 core was chosen to assess if it positively modulates ligand binding to the CRD of FZD 5 . The rationale for this selection was based on the very weak NanoBiT/BRET signal seen for eGFP-WNT-3A binding to FZD 6 and FZD 8 , and strong NanoBiT/BRET signal of eGFP-WNT-3A/FZD 7 association in the two assay paradigms described in this study ( Figure 1B and Figure 2B). Additionally, the FZD 6 core was replaced with the FZD 4 core to assess whether eGFP-WNT-3A binding to FZD 6 CRD would increase upon insertion of a core from an eGFP-WNT-3A binding-competent FZD paralogue ( Figure 1B and Figure 2B). We validated the chimeras with regard to their proper membrane trafficking upon transient overexpression in HEK293A cells and detected that the FZD-FZD chimera proteins are relatively poorly expressed on the cell surface compared to WT receptors ( Figure S4C). Binding affinities of fluorescent propranolol to HiBiT-tagged β2-adrenergic receptors vary depending on the protein expression levels, with higher expression levels generally resulting in slightly elevated K d values (lower affinity). 32 Furthermore, binding affinities of DKK1-eGFP proteins to the WNT coreceptor LRP6-mCherry measured by dual-color axial line-scanning FCS (axial lsFCS) were higher for lower, more physiologically relevant receptor expression levels. 54 Our data for transient overexpression of FZD 4 in HEK293A cells mostly support these notions, with some exceptions ( Figure S4D). Overall, it needs to be emphasized that although the K d is a thermodynamic parameter that should be constant for various expression levels, differences in the cellular context may lead to different functionalities of overexpressed receptors present on the cell surface. This in turn can affect the comparative analysis and interpretation of real-time ligand binding data. 54 However, provided the availability of cell surface expression data in the experimental paradigm of the NanoBiT/BRET binding assay, we shed light on a potential role of the receptor core for WNT-CRD binding. Focusing on the FZD 4 chimeras, eGFP-WNT-3A bound with a significantly higher affinity (lower K d ) to the FZD 4 -FZD 6 and the FZD 4 -FZD 8 chimeras compared to fulllength FZD 4 (FZD 4 -FZD 6 K d ± SEM (nM) = 3.5 ± 1.8, P = 0.0255; FZD 4 -FZD 8 K d ± SEM (nM) = 2.9 ± 1.3, P = 0.0371; Figure S4E). Interestingly, BRET max values were visibly or significantly lower for both weakly expressed FZD 4 -FZD chimeras than for the intact FZD 4 (FZD 4 -FZD 6 BRET max ± SEM = 0.021 ± 0.003, P = 0.0828; FZD 4 -FZD 8 BRET max ± SEM = 0.010 ± 0.001, P = 0.0192).
In addition, our data indicated that eGFP-WNT-3A bound to FZD 6 -FZD 4 with the same affinity (FZD 6 -FZD 4 K d ± SEM (nM) = 5.8 ± 3.4, P = 0.9695; Figure S4G) as to FZD 6 but with a significant, over 10-fold increase in the maximal BRET signal (FZD 6 -FZD 4 BRET max ± SEM = 0.030 ± 0.006 vs FZD 6 BRET max ± SEM = 0.002 ± 0.001, P = 0.056) arguing that the FZD 6 CRD can efficiently bind eGFP-WNT-3A in the context of a different receptor core. As mentioned before, it needs to be emphasized that receptor expression levels can affect ligand−receptor interaction. Along these lines, in classical BRET titration experiments, increasing BRET donor amounts (by increasing plasmid DNA amounts) with constant BRET acceptor levels (unchanged plasmid DNA amounts) leads to a decrease in BRET max signal for specific donor−acceptor interactions. 55 In contrast, no such relationship was found in our FZD-CD86 and FZD-FZD NanoBiT/BRET binding experiments, further supporting the notion that the FZD core has a differential role in WNT binding depending on the FZD paralogue. Here, we have used the NanoBiT/BRET system to analyze the contribution of the FZD seven-transmembrane-spanning core to eGFP-WNT-3A binding to the CRD. We show that swapping the receptor core can have a substantial effect on the affinity or maximal BRET of eGFP-WNT-3A binding in the NanoBiT/BRET read-out. Our data, particularly from the FZD-CD86 experiments, argue that the seven-transmembranespanning core contributes to ligand binding for the tested FZDs, even though the details on the molecular level remain obscure.
This study adds substantial methodological advance to the pharmacological toolbox suitable for the study of the class F GPCRs and their coreceptors. 28,54,56 We have demonstrated the vast potential of employing fluorescent WNTs and the NanoBiT/BRET binding technique for the pharmacological quantification of WNT-FZD interactions in live HEK293A cells despite the limitations that come with the low concentration of the tracer WNT in the conditioned medium preparation. The broader analysis of the selectivity of ligand− receptor interactions, WNT binding in the presence or absence of either FZD coreceptors, FZD-binding intracellular transducer proteins and at different FZD expression levels, can now be further investigated to understand the pluridimensionality of WNT-FZD system in a more physiologically relevant cell system.
Cells were first cleared from the HEK293F CM by centrifugation at 260 g (1200 rpm) for 10 min and then at 2800 g (4000 rpm) for 30 min to remove any remaining cellular debris and insoluble material. This "raw" CM then was concentrated 5-fold using Vivaspin turbo 15 centrifugal concentrators (30,000-molecular-weight-cutoff, Satorius AG, Goẗtingen, Germany) and exchanged to the desired cell culture medium using Sephadex G-25 PD10 desalting columns (GE Healthcare Bio-Science, Freiburg, Germany). The final concentration and integrity of eGFP-WNT-3A in the CM samples were determined using ELISA (GFP ELISA kit, #ab171581, Abcam) and SDS-PAGE/Western Blot analysis, respectively. Two eGFP-WNT-3A batches (eGFP-WNT-3A batch 1 final concentration: 16.7 nM; eGFP-WNT-3A batch 2 final concentration: 16.2 nM) were used in this study. Current WNT purification methods allow only limited WNT concentration to be obtained from CM. 58 For validation of the eGFP-WNT-3A batches, please see Figure S5.
Plasmids. Generation of HiBiT-FZD 4 , HiBiT-FZD 6 , and Nluc-FZD 4 has been described previously. 28 Gibson cloning was used to generate other HiBiT-tagged receptor constructs using HiBiT-tagged backbone from HiBiT-FZD 4 containing a 5-HT 3 A signal sequence. To generate chimeras, the N-terminal domains (NTD; CRD with a linker region) and the transmembrane cores were defined according to Frizzled structures predicted on GPCRdb (http://www.gpcrdb.org), and the constructs were generated with Gibson cloning. Nluc-CD86 used to generate FZD-CD86 chimeras was from Martin J. Lohse (Max-Delbrueck Center for Molecular Medicine, Berlin, Germany). Frizzled and CD86 signal peptides were defined with SignalIP-5.0 Server (http://www.cbs.dtu.dk/ services/SignalP/). The constructs were validated by sequencing (Eurofins GATC, Konstanz, Germany). The details of the constructs used in this study are presented in Figure S1A,B and Figure S4A. Subsequently, 50 μL of eGFP-WNT-3A conditioned medium or control medium supplemented with 5% FBS and 10 mm HEPES was added, and the BRET signal was measured every 90 s for 240 min at 37°C (161 measurements, no CO 2 ). In the saturation-binding experiments, the cells were incubated with different concentrations of eGFP-WNT-3A conditioned medium (90 μL) supplemented with 5% FBS and 10 mm HEPES for 240 min at 37°C with no CO 2 . In the competition binding experiments, the cells were preincubated for 30 min at 37°C with 80 μL of unlabeled WNT proteins at 37°C with no CO 2 . Subsequently, 10 μL of eGFP-WNT-3A at a concentration of 3.6 nM (final concentration of 0.4 nM) were added, and the cells were incubated for further 240 min at 37°C with no CO 2 . Next, for saturation and competition binding experiments, 10 μL of a mix of furimazine (1:10 dilution; TOPFlash Reporter Gene Assay. ΔFZD 1−10 HEK 293T cells were transfected in suspension (4 × 10 5 cells were transfected in 1 mL) with 700 ng of HiBiT-tagged receptor, 250 ng M50 Super 8× TOPFlash (#12456; Addgene, Watertown, MA, USA), and 50 ng pRL-TK Luc (#E2241, Promega, Fitchurg, WI, USA) and seeded (50 μL) onto a poly(D-lysine)-coated white 96-well cell culture plate with a solid flat bottom (Greiner BioOne). Next, 50 μL of complete DMEM medium was added to each well. Twenty-four hours after transfection, the medium was changed to starvation medium (DMEM without FBS) containing either 8.0 nM (300 ng/mL) WNT-3A or vehicle, and 10 nM C59. Twenty-four hours after stimulation, cells were lysed gently shaking with 20 μL 1× Passive Lysis Buffer (#E1910; Promega, Fitchurg, WI, USA) for 15 min. Subsequently, 20 μL of LAR II (Promega, E1910) was added to all wells after which luminescence (580− 80 nm) was read, and then 20 μL of Stop & Glo (Promega, E1910) was added to all wells after which luminescence (480− 80 nm) was read again with a CLARIOstar microplate reader (BMG, Ortenberg, Germany).
Data Analysis and Statistics. All data were analyzed in GraphPad Prism 8 (San Diego, CA, USA) using built-in equations. All data presented in this study come from n individual experiments (at least three biological replicates) with each individual experiment performed typically in duplicates (technical replicates) for each tested concentration/condition. Data points on the binding curves represent mean ± SEM. Saturation binding curves were fit using onesite-specific or total and nonspecific saturation nonlinear regression models (linear scale for eGFP-WNT-3A concentrations) or normalized three-parameter or normalized fourparameter nonlinear regression models (logarithmic scale for eGFP-WNT-3A concentrations with normalized net BRET ratio). The fitting models were selected based on an extra-sumof-squares F-test (P < 0.05). Kinetic binding data were analyzed using the association model with two or more hot ligand concentrations. Binding affinity values (K d ) are presented as a best-fit K d with SEM. K d values were compared using an extra-sum-of-squares F-test (P < 0.05). Competition binding curves were analyzed using a three-or four-parameter nonlinear regression model to obtain equilibrium dissociation constant values pK i with SEM of unlabeled ligands as per the Cheng−Prusoff equation. 59 Minimal BRET (BRET min ) and maximal BRET (BRET max ) were defined as the lowest and highest measured net BRET ratios, respectively. BRET max values were compared using unpaired t-test. TOPFlash and cell surface expression data are presented as mean ± SEM. TOPFlash and cell surface expression data were analyzed for differences with Brown−Forsythe and Welch one-way analysis of variance (ANOVA); ** P ≤ 0.01, * P ≤ 0.05.

Notes
The authors declare no competing financial interest.

■ ACKNOWLEDGMENTS
We thank Anna Krook for access to the CLARIOstar plate reader and Benoit Vanhollebeke, who kindly provided the ΔFZD 1−10 cells used in this work.