Identification of Novel FZD4 Mutations in Familial Exudative Vitreoretinopathy and Investigating the Pathogenic Mechanisms of FZD4 Mutations

Abstract

Purpose: The purpose of this study is to report five novel FZD4 mutations identified in familial exudative vitreoretinopathy (FEVR) and to analyze and summarize the pathogenic mechanisms of 34 of 96 reported missense mutations in FZD4.

Methods: Five probands diagnosed with FEVR and their family members were enrolled in the study. Ocular examinations and targeted gene panel sequencing were conducted on all participants. Plasmids, each carrying 29 previously reported FZD4 missense mutations and five novel mutations, were constructed based on the selection of mutations from each domain of FZD4. These plasmids were used to investigate the effects of mutations on protein expression levels, Norrin/β-catenin activation capacity, membrane localization, norrin binding ability, and DVL2 recruitment ability in HEK293T, HEK293STF, and HeLa cells.

Results: All five novel mutations (S91F, V103E, C145S, E160K, C377F) responsible for FEVR were found to compromise Norrin/β-catenin activation of FZD4 protein. After reviewing a total of 34 reported missense mutations, we categorized all mutations based on their functional changes: signal peptide mutations, cysteine mutations affecting disulfide bonds, extracellular domain mutations influencing norrin binding, transmembrane domain (TM) 1 and TM7 mutations impacting membrane localization, and intracellular domain mutations affecting DVL2 recruitment.

Conclusions: We expanded the spectrum of FZD4 mutations relevant to FEVR and experimentally demonstrated that missense mutations in FZD4 can be classified into five categories based on different functional changes.

Familial exudative vitreoretinopathy (FEVR) is a hereditary eye disorder first identified by Criswick and Schepens¹ in 1969, characterized by aberrant vascular development in the retina. The hallmark pathological feature of FEVR is a delay in retinal vascular development, resulting in avascular zones in the peripheral retina, pathological neovascularization, hemorrhage, fibrovascular proliferation, and consequent retinal detachment, which often leads to blindness in childhood.² To date, research has identified 17 genes and one genetic locus implicated in FEVR.^3–19 Mutations in genes associated with the Norrin/β-catenin pathway, specifically NDP, FZD4, LRP5, TSPAN12, and CTNNB1, predominate in FEVR cases, with the other 11 genes all having no more than three reported cases of mutations. Two large-scale studies have shown the reported genes can explain nearly half of FEVR cases. LRP5 contribute to approximately 37% and FZD4 to 29% of the explainable FEVR cases, being the two genes with the highest proportions.^20,21 

The FZD4 gene encodes a 537-amino acid protein with distinct functional domains, including an N-terminal cysteine-rich domain (CRD), seven transmembrane domains, and a C-terminal S/T-X-V motif. As a member of the Frizzled family of transmembrane receptors, FZD4 plays a crucial role in regulating the Wnt signaling pathway.²² Studies using Fzd4 knockout models have successfully replicated the observed features of FEVR, mainly attributed to impaired Norrin/β-catenin signals. Restoration of these signals via β-catenin overexpression has been shown to reverse the conditions’ features.²³ 

After the identification of FEVR-associated mutations in the FZD4 gene in 2002,²⁴ a global body of case reports involving FZD4 mutations has emerged. Mutations in FZD4 encompass a spectrum of types, including missense, frameshift, nonsense, insertion, deletion, copy number variation (CNV), and splicing. In a recent report based on a large population, the two most prevalent FZD4 mutation types were found to be missense (45%) and deletion mutations (25.5%).²⁵ However, most reports are clinical case studies lacking detailed analysis of the mutations’ specific functional implications. The following studies have shed some light on the functional changes of FZD4 mutations in the Norrin/β-catenin signaling pathway. Smallwood et al.²⁶ identified the sequences in the CRD of FZD4 that binds to norrin through alanine scanning and block substitution. Additionally, studies have reported that Y58C and M105R mutations affect binding to norrin.^27,28 Furthermore, a series of mutations leading to defective trafficking of FZD4 have been found.²⁹ Finally, it has been reported that R253H and F328S reduce the ability of FZD4 to recruit DVL2,³⁰ and Y250 is also an important FZD-DVL interaction site.³¹ 

In this study, we identified five novel FZD4 missense mutations in FEVR patients. Through a comprehensive review of reported mutations linked to FEVR, we conducted an in-depth analysis of a diverse array of missense mutations spread across the FZD4 gene. Our focus was to unravel the distinct mechanisms of pathogenicity associated with mutations situated in various domains. We selected to analyze missense mutations because they alter single amino acids, allowing for a more precise examination of the functional roles of amino acids located within structural domains. 

Material and Methods

Ethical Declarations, Clinical Evaluation, and Gene Sequencing

The study was approved by the institutional ethics committees of Xinhua Hospital Affiliated with Shanghai Jiao Tong University School of Medicine. All participants were recruited from this single instruction, and written informed consent was obtained either from each individual or from their legal guardian(s) in the case of minors. 

Each proband underwent ocular examinations, including wide-field fundus photography using either a RetCam (Natus Medical Incorporated, Middleton, WI, USA) or an Optos 200Tx (Optos, Dunfermline, UK), as well as indirect ophthalmoscopy with a 28D lens and scleral depression when required. When eligible, the proband and their parents underwent fundus fluorescein angiography (FFA). The diagnosis of FEVR was based on the presence of at least one of the following retinal vascular developmental anomalies: a lack of peripheral retinal vasculature with or without nonperfusion, vitreoretinal traction, subretinal exudation, retinal neovascularization at any age, or total retinal detachment with a fibrotic mass behind the lens.³² Patients with a history of premature birth were excluded. Targeted gene capture and sequencing procedures followed previously established protocols (MyGenostics), and Sanger sequencing was carried out to validate the variants as previously described.²¹ 

Cell Culture and Plasmids Construction

HEK293T cells, HEK293STF cells, and HeLa cells (American Type Culture Collection and Cell Systems) were cultured at 37°C in a humidified 5% CO₂ incubator in DMEM supplemented with 10% fetal bovine serum a 1% penicillin/streptomycin. 

The human FZD4 (NM_012193) coding sequence was subcloned into the pcDNA3.1 vector with a C-terminal HA tag. The mutations were introduced into the wild-type (WT) expression plasmid by site-directed mutagenesis with the Q5 Site-Directed Mutagenesis Kit (New England Biolabs, Ipswich, MA, USA). The FZD5, LRP5, LRP6, NDP, DVL2 and pDsRed2-ER plasmids were obtained from Youbio Biotechnology (Changsha, China). The FZD5 plasmid has a C-terminal HA tag, the NDP plasmid has a C-terminal Myc-FLAG tag, and the DVL2 plasmid has a C-terminal FLAG tag. The pGL4.1-Renilla was sourced from Promega (Madison, WI, USA). Plasmids were transfected by EntransterTM-H4000 transfection reagent (Engreen Biosystem Co, Ltd, Auckland, New Zealand). 

Western Blot and Co-Immunoprecipitation (Co-IP)

HEK293T cells were seeded into six-well plates and transfected with 1200 ng WT FZD4 or mutant plasmids. After 48 hours, total cell protein was extracted on ice with lysis buffer (150 mM NaCl, 50 mM Tris-HCl pH 8.0, 1% Triton X-100) in the presence of protease inhibitor cocktail. Cell lysates were mixed with 5 × loading buffer and then heated at 37°C for 30 minutes. For detection of glycosylation, protein samples were treated with EndoH (New England Biolabs) or PNGaseF (Yeasen, Shanghai, China) according to the manufacturer's instructions. 

For generation of norrin-conditioned medium, HEK293T cells were cultured in 10 cm dish and transfected with 10,000 ng NDP-Myc-FLAG plasmid. After 72 hours, the culture medium was collected and spun in a centrifuge at 900g for three minutes, after which the supernatant was transferred into new tubes. For Co-IP, after incubation of norrin conditioned medium for 45 minutes at 4°C, cells were washed three times with ice-cold PBS and lysed. Cell lysates were spun in a centrifuge for 15 minutes at 13,000 g at 4°C and 10% of the supernatants were collected into new tubes for input. The rest of soluble supernatants were rotated with anti-HA-magnetic beads at room temperature for two hours. After four washes with lysis buffer, the samples were resuspended in loading buffer and heated at 37°C for 30 minutes. The beads were discarded, and the supernatants were used for Western blotting. The following antibodies were used for immunoblots: rabbit anti-FLAG (20543-1-AP; Proteintech, Rosemont, IL, USA), rat anti-HA (3F10; Roche, Basel, Switzerland), rabbit anti-GAPDH (10494-1-AP; Proteintech), anti-rabbit and anti-rat secondary antibody conjugated with horseradish peroxidase (SA0001-2 and SA00001-15; Proteintech). 

RT-qPCR

Total RNA was extracted from HEK293T cells transfected with WT, mutant, or vector plasmids using TrizolUP (TransGen Biotech, Beijing, China). Subsequently, 1 µg of total RNA was reverse-transcripted into cDNA with TranScript All-in-one First-Strand cDNA Synthesis SuperMix (TransGen Biotech). The cDNA was then amplified and detected with the 7500 Fast Real-Time PCR System (Applied Biosystems, Foster City, CA, USA). The fold changes of FZD4 were normalized to GAPDH. Primers: GAPDH (Forward: 5′- CTGACTTCAACAGCGACACC-3′ and Reverse: 5′- TGCTGTAGCCAAATTCGTTG-3′) and FZD4 (Forward: 5′- GTCTCAGTCTGGGGTTGCTC-3′ and Reverse: 5′- GTCACGTTGTAGCCGAGGTT-3′). 

Immunofluorescence Staining

After transfection with the corresponding plasmids, cells cultured on poly-l-lysine–coated coverslips were first washed twice with PBS and then fixed with 4% PFA (BioSharp, Tallinn, Estonia) at room temperature for 15 minutes. To label the cell membrane, cells were incubated with Wheat Germ Agglutinin (WGA)-647 (W32466; ThermoFisher, St. Louis, MO, USA) for 10 minutes before fixation. To analyze the norrin-binding ability of FZD4 mutants, cells were incubated at 4°C in norrin-conditioned medium for 45 minutes before fixation, followed by three washes with ice-cold PBS. After fixation, the coverslips underwent three washes with PBS and were subsequently blocked using a buffer (5% FBS and 0.2% Triton X-100 in PBS) for one hour. Next, the coverslips were incubated with primary antibodies diluted in blocking buffer overnight at 4°C. On the next day, the coverslips were washed with PBS three times and then incubated with fluorescently labeled secondary antibodies and DAPI (4083; Cell Signaling Technology, Danvers, MA, USA) for one hour at room temperature. Fluorescence was observed using an LSM900 confocal microscope (Zeiss, Oberkochen, Germany) equipped with a 63 × oil-immersion objective. Colocalization analysis was performed using the Pearson correlation coefficient method in Fiji. 

To evaluate the norrin-binding affinity of FZD4 mutants accurately, the fluorescence intensity of FZD4 was standardized during cell selection to maintain consistency in FZD4 expression levels across the analyzed cells. We identified notable variances in FZD4 fluorescence intensity on individual cell membranes. To mitigate this variability, we selected regions with moderate fluorescence intensity, marking a 1 µm line segment and measuring FZD4 and NDP fluorescence intensity at that location. The fluorescence values along the line segment were averaged, and then the relative strength of NDP was calculated by dividing NDP's fluorescence intensity by FZD4's fluorescence intensity. The following antibodies were used for immunofluorescence: rat anti-HA (3F10; Roche), rabbit anti-FLAG (14793S; Cell Signaling Technology), Alexa fluor 488 donkey anti-rat (A-21208; ThermoFisher), Alexa fluor 647 donkey anti-rabbit (A-31573; ThermoFisher). 

Dual-Luciferase Reporter Assay

To assess the Norrin/β-catenin activation capacity of the FZD4 mutants, HEK293STF cells were transfected with 100 ng pGL4.1-Renilla, 25ng LRP5, and increasing amounts of WT FZD4, mutant FZD4, or vector plasmid (5 ng, 15 ng, and 25 ng). After 24 hours, the cells were incubated in norrin-conditioned medium for another 24 hours. To assess the canonical WNT pathway activation capacity of the FZD4 mutants in the absence of a ligand, HEK293STF cells were transfected with 100 ng pGL4.1-Renilla, 50 ng LRP6, and 50 ng WT FZD4, mutant FZD4, or vector plasmid and incubated for 48 hours. The transfected cells were then subjected to luciferase activity assays using the Dual-Luciferase Reporter Assay System (Yeasen) according to the manufacturer's protocols. Reporter activity was calculated as the ratio of firefly luciferase activity to Renilla luciferase activity. 

Statistical Analysis

Data were expressed as mean ± SD. One-way ANOVA were used. All statistical evaluations and graphs were generated in GraphPad Prism 8.0 software. Statistical significance was set at P < 0.05. 

Results

Identification of Five Novel FZD4 Missense Mutations in FEVR patients

Five novel FZD4 mutations were identified in FEVR patients diagnosed at Xinhua hospital, all of which were predicted as pathogenic through in silico analysis (Table 1). Further validation of these mutations was conducted using Sanger sequencing, for which all exhibited genotype-phenotype cosegregation (Fig. 1A). The fundus of all probands manifested typical FEVR phenotypes, including peripheral avascular area or retinal folds (Figs. 1B–F). Consistent with previous reports,^33,34 we found that clinically affected father or mother within the same family exhibited milder phenotypes, mostly manifesting only peripheral supernumerary branching or avascular areas on FFA (Fig. 1G). 

Table 1.

 

Novel FZD4 Mutations Found in This Study

 

Figure 1.

 

The pedigrees of the families with mutations in FZD4 and the fundus images of probands and family members. (A) FEVR pedigrees and Sanger sequencing analysis of five FEVR-associated families. Affected individuals are represented in black, with the proband of each family indicated by a black arrow. The altered nucleotides are denoted by red arrows. (B) Fundus images of the proband in family 1 showing bilateral dragged discs. (C) Fundus images of the proband in family 2 showing dragged disc and temporal peripheral exudation in right eye and peripheral avascular area. (D) Fundus images of the proband in family 3 showing bilateral retinal folds. (E) Fundus images of the proband in family 4 showing dragged disc in left eye. (F) Fundus images of the proband in family 5. Because of retinal detachment, the left eye is not amenable to visualization. (G) FFA of the mother in family 5 showing supernumerary branching.

Assessment of Mutations in Different Domains of FZD4 on Protein Expression and Norrin/β-catenin Signaling Activity

In order to investigate the pathogenic mechanisms of five novel mutations, we conducted a comprehensive literature review up to June 2023 using a search on PubMed with keywords “FZD4 and FEVR,” gathering all documented missense mutations. FZD4 variants previously classified as benign in the literature were excluded from our analysis. From this review, we identified a total of 96 reported FZD4 missense mutations in FEVR patients and classifying them by structural domains based on the reported crystal structure of the Norrin-Fz4_CRD complex and FZD4 in the ligand-free state (Figs. 2A, 2B, Table 2).^35,36 The four mutations S91F, V103E, C145S, and E160K are located in the extracellular domain (ECD), which contains three loops that come into contact with norrin (Fig. 2B). Notably, novel mutations at positions 91 and 103 are located between loop I and loop II, and 145 is located between loop II and loop III. No previous analysis on whether mutations located between the loops can also affect norrin binding currently exists. 

Figure 2.

 

Classification of all reported missense mutations in FZD4 in FEVR patients and assessment of expression levels and Norrin/β-catenin signaling activity of 34 mutations. (A) Classification of FZD4 functional domains and disease-associated mutations. ICL, intracellular loop; ECL, extracellular loop; ICD, intracellular domain. (B) Lateral depiction of the Norrin-Fz4_CRD complex, with coloring denoted as blue (Fz4_CRD), yellow (loop I-III), and pink (Norrin) (PDB ID = 5BQC). (C) Western blot of HA-tagged FZD4 protein level in HEK293T cells transfected with WT or mutants. (D) Relative STF luciferase activity after transfecting increasing dosage of WT or mutant FZD4 (5 ng, 15 ng, and 25 ng) together with LRP5 into HEK293STF cells, followed by incubation with norrin conditioned medium. Normalization was performed based on the values from the 5 ng dose of the vector group (n = 3). Error bars: SD. One-way ANOVA with Dunnett's multiple comparisons test was performed to compare each mutant and WT within each dosage group. ****P < 0.0001; ns, not significant.

Table 2.

 

All Missense Mutations of FZD4 in FEVR Patients

 

Additionally, FZD4 mutations have been primarily reported in clinical case studies, lacking systematic analyses of their potential effects across various domains. To understand the impact of mutations at various domains within FZD4, we generated a total of 34 mutants selected distributed across the gene, encompassing 5 novel mutants reported in this study (Table 2). First, we assessed the impact of these mutations on protein expression in HEK293T cells. Under equivalent RNA expression levels (Supplementary Fig. S1), mutants at positions 145, 204, 223, 226, 228, 234, 377, and 488 exhibited significantly reduced protein expression levels (Fig. 2C, Supplementary Fig. S2). Furthermore, they showed decreased molecular weights which were later confirmed to be caused by lack of post-translational modifications (Fig. 3A). Given that there is only one FZD4 transcript in humans, the presence of multiple bands for Y58C, V103E, I114S, W139S, and M157T suggests the potential occurrence of distinct modifications (Fig. 2C). Subsequently, using luciferase assays to evaluate the ability of these FZD4 mutants to activate Norrin/β-catenin signaling, we found that, with the exception of E180K, I232V, and S245T, which showed no significant differences compared to the wild-type, all other mutations compromised the activation ability (Fig. 2D). E180K, I232V, and S245T exhibited similar behavior to WT in all subsequent experiments. Further literature review revealed that these mutations lack explicit hereditary information and in silico analysis indicated low likelihood of pathogenicity, suggesting they are more likely benign variants rather than mutations (Supplementary Table S1). 

Figure 3.

 

Membrane localization analysis of FZD4 Mutants. (A) Western blot analysis of Endo H or PNGaseF treated cell lysates of HEK293T cells transfected with WT or mutant FZD4. (B) Representative immunofluorescence images of FZD4 (green), ER (magenta), WGA (light blue), DAPI (blue) in HeLa cells transfected with FZD4 (WT or mutants) and pDsRed2-ER. Scale bar:10 µm. (C) Densitometry analysis of Endo H resistant proportion of WT or mutant FZD4 protein expressed in HEK293T cells (n = 3). Error bars: SD. (D) Co-localization analysis using the Pearson correlation coefficient (n = 20). Error bars: SD. (E) Alignment of all human Frizzled receptor homologs based on the amino acids at position 226 and 488 of FZD4. (F) Immunofluorescence images of FZD5 (green), ER (magenta), WGA (light blue), DAPI (blue) in HeLa cells transfected with FZD5 (WT or mutants) and pDsRed2-ER. Scale bar:10 µm. (G) Western blot analysis of Endo H or PNGaseF treated cell lysates of HEK293T cells transfected with WT or mutant FZD5. One-way ANOVA with Dunnett's multiple comparisons test was performed to compare each mutant and WT for C and D. ****P < 0.0001.

Analysis of Membrane Localization of FZD4 Mutants

To investigate the mechanisms underlying decreased activation of the Norrin/β-catenin pathway due to FZD4 mutations, we initially examined their impact on membrane localization, which is prerequisite for the functional role of FZD4 as a receptor. We used selective enzymes (EndoH and PNGaseF) to evaluate the cell surface proportion and immature form of FZD4 before Golgi processing.³⁷ Mature glycoproteins passing through the Golgi apparatus exhibit complex N-glycans, while high-mannose type glycans are primarily displayed in the endoplasmic reticulum (ER). Treatment with EndoH distinguishes between these types, as it cleaves high-mannose and hybrid glycans but not complex N-glycans. Compared to the wild type, the proportion of C145S, C204R, M223K, W226R, S228R, T234I, C377F, and G488D mutants resistant to EndoH was significantly reduced (Figs. 3A, 3C), whereas the other mutations showed no significant differences (Supplementary Fig. S3). Furthermore, the expression levels of these mutants were also affected (Supplementary Fig. S2), implying potential degradation by the ER quality control system. Through overexpressing mutant FZD4 and DsRed2-ER plasmid in HeLa cells and using WGA staining for cell membrane labeling, we observed that C145S, C204R, M223K, W226R, S228R, T234I, C377F, and G488D exhibited reduced membrane localization and enhanced co-localization with ER, whereas the other mutations did not show significant differences compared to WT (Figs. 3B, 3D, Supplementary Fig. S4). 

C145S, C204R, and C377F occur at positions crucial for the formation of intramolecular disulfide bonds, and substitutions of cysteine at these sites may exert an influence on the proper protein folding. M223K, W226R, S228R, T234I are all located with TM1, whereas G488D is located with TM7. Notably, our sequence alignment analysis revealed a relatively high degree of conservation for the amino acids tryptophan (W) at position 226, threonine (T) at position 234, and glycine (G) at position 488 of FZD4 among human Frizzled receptor homologs (Fig. 3E). We were curious whether mutations at these specific positions might also affect the membrane localization of other Frizzled receptors. Therefore we generated corresponding mutations in human FZD5, specifically FZD5 W242R corresponding to FZD4 W226R, and FZD5 G514D corresponding to FZD4 G488D. Our subsequent EndoH analysis and immunofluorescence staining unequivocally revealed that these two mutations also led to a significantly impaired localization of FZD5 to the membrane (Figs. 3F, 3G), indicating the significance of these positions for the maturation of Frizzled receptors. 

Analysis of Binding Capability to Norrin of FZD4 Mutants

After excluding mutations that affect expression and membrane localization, we further explored the pathogenic mechanisms of the remaining mutations. The CRD domain of FZD4 is essential for norrin binding,^26,27 and we aimed to confirm whether each mutation in ECD would lead to weakened norrin binding and whether mutations outside the ECD could also affect norrin binding. By overexpressing WT FZD4 or mutant forms in HEK293T cells incubated with norrin conditioned medium, we conducted Co-IP experiments to analyze the ability of FZD4 to bind norrin in the supernatant (Fig. 4A). G115V showed a mild reduction in norrin binding, while G57R, Y58C, S91F, V103E, M105V, I114S, W139S, N152K, M157V, M157T, E160K, and G161R displayed severe reductions in binding capacity (Fig. 4B). Mutations in domains outside the ECD exhibited norrin binding capacities similar to WT. Additionally, through immunofluorescence staining, we quantified the amount of membrane-bound norrin in HeLa cells overexpressing WT FZD4 or mutants after incubation of norrin conditioned medium (Figs. 4C, 4D, Supplementary Fig. S5), with results consistent with the Co-IP findings. The degree of reduction in norrin binding ability because of mutations in ECD corresponds closely to the impairment of the ability to activate the Norrin/β-catenin signaling pathway. 

Figure 4.

 

Norrin binding analysis of FZD4 Mutants. (A) Western blot analysis of norrin conditioned medium co-immunoprecipitated with FZD4-HA (WT or mutants). Because of the 1:9 ratio of input to IP, the quantity of norrin in the input is insufficient to be detected. (B) The NDP band intensity in the immunoprecipitation was normalized to the pulldown HA (FZD4) band, and further standardized for each mutant relative to WT values (n = 3). Error bars: SD. (C) Representative immunofluorescence images of FZD4 (green), NDP (magenta), and DAPI (blue) in HeLa cells transfected with FZD4-HA (WT or mutants) and incubated with norrin conditioned medium. Scale bars: 10 µm. (D) Quantification of membrane signal intensity of NDP in HeLa cells transfected with FZD4-HA (WT or mutants) and incubated with norrin conditioned medium (n = 10). Error bars: SD. Experiments were performed independently at least three times. Normalization was performed against the WT. One-way ANOVA with Dunnett's multiple comparisons test was performed to compare each mutant and WT for B and D. *P < 0.05; ****P < 0.0001; ns, not significant.

Analysis of DVL2 Recruitment of FZD4 Mutants

Previous studies have indicated that the activation of the canonical β-catenin pathway requires FZD-mediated recruitment of DVL2,³⁸ with all intracellular domains of FZD being essential for FZD-DVL interaction.³⁹ In the case of FZD4, the R253H and F328S mutations have been found to impact the recruitment of DVL2,³⁰ and another study has highlighted the significance of the Y250 to recruit DVL2.³¹ This prompted us to investigate whether mutations in FZD4, previously shown to reach the membrane and bind norrin normally, could impact DVL2 recruitment. Overexpression of DVL2-FLAG and vector in HeLa cells resulted in punctate distribution within the cells (Fig. 5A). However, coexpression with wild-type FZD4 led to complete recruitment of DVL2 to the cellular membrane in 60% of the transfected cells, whereas partial recruitment to the membrane was observed in 36% of the transfected cells, retaining the punctate intracellular pattern (Fig. 5A). The mutations Y250C, R253H, M342V, and R417Q exhibited significantly reduced DVL2 recruitment, whereas T237R, L239P, W335C, T445P, G492R, and S497F almost entirely lost their ability to recruit DVL2. In contrast, mutations located in the ECD still retained recruitment abilities comparable to the wild type (Fig. 5B). To validate these findings, we explored the overexpression of FZD4 and LRP6 to evaluate the capacity of FZD4 mutants to activate the WNT/β-catenin signaling pathway in the absence of ligands like Norrin or WNT. As anticipated, FZD4 mutants incapable of effectively recruiting DVL2 displayed markedly diminished WNT signaling pathway activation through LRP6 (Fig. 5C). 

Figure 5.

 

DVL2 recruitment analysis of FZD4 Mutants. (A) Representative immunofluorescence images of FZD4 (magenta) and DVL2 (light blue) in HeLa cells transfected with FZD4-HA (WT, mutants or vector). Scale bar:10 µm. (B) Quantification of DVL2 recruitment to the membrane by WT FZD4 and mutants (n = 50). (C) Relative STF luciferase activity after transfecting WT or mutant FZD4 or vector with LRP6 into HEK293STF cells. Normalization was performed based on the values of the vector group (n = 3). Error bars: SD. One-way ANOVA with Dunnett's multiple comparisons test was performed to compare each mutant or vector and WT. ***P < 0.001; ****P < 0.0001.

Discussion

Pathogenicity Assessment of Variants

Upon scrutinizing all examined mutations, E180K, I232V, and S245T exhibited no significant deviations from WT in all assays. Upon literature review, we noted the absence of family information for these mutations in the original reports.^40,41 Use of in silico analyses, such as PolyPhen-2 and REVEL, indicated that these three mutations lean toward benign variants. Hence, we recommend reclassifying these three mutations as benign variants. In contrast, the five newly reported mutations S91F, p.V103E, C145S, E160K, and C377F all have available family information, are in line with genotype-phenotype cosegregation and are predicted as pathogenic by multiple prediction programs. This underscores the pivotal role of family history in evaluating the pathogenicity of newly identified variants. The predictions from PolyPhen-2 and REVEL closely match up with our wet laboratory experiments and can serve as a basis for assessing the pathogenicity of variants. 

Phenotypic variability is frequently observed in FEVR.^25,33 The five novel mutations identified in this study also align with the trend where probands exhibit more severe phenotypes than their family members. Notably, on individual proband phenotype analysis, patients with mutations affecting FZD4 protein expression, such as C145S and C377F, displayed more severe phenotypes compared to those with mutations (S91F, V103E, and E160K) that solely impact norrin binding without affecting protein expression. We speculate that mutations causing more pronounced protein functional impairment pose a higher risk of developing severe phenotypes. 

Importance of the ECD of FZD4 in Norrin Binding

The CRD domain is the primary domain in the extracellular region of FZD4 responsible for norrin binding, spanning from amino acid 45 to 158. Previous studies have identified certain regions within the CRD, including 55–60, 105–107, 125–130, 133–137, and 151–157, for norrin binding through rigorous and comprehensive substitution experiments.²⁶ The crystal structure of the Norrin-Fz4_CRD complex has unveiled three loops directly binding to norrin,³⁵ substantiating findings from earlier biochemical experiments. It is conceivable that mutations occurring in the crucial norrin-binding regions mentioned above are likely to lead to weakened norrin binding. In this study, we experimentally verified the importance of these regions and intriguingly discovered that mutations in the amino acids connecting these critical regions may also compromise norrin binding, such as S91F, V103E, G115V, and W139S. Notably, even a slight weakening of the interaction between FZD4 and norrin, as seen with G115V, is sufficient to cause approximately a 50% reduction in Norrin/β-catenin signaling activity, which can lead to the development of FEVR. 

Previous studies have underscored the importance of the FZD4 linker (between loop III and TM1, amino acids 164-203) in norrin binding and the essential role of the FZD6 linker domain in receptor localization.^40,42 Encountering only four reported linker mutations in FEVR, with two involving cysteine residues (C181R and C181Y), we assessed E180K and propose it to be a benign variant. Research indicated that the simultaneous presence of P168S and P33S result in a modest decline in Wnt activity.⁴³ This suggests that P168S may not affect the membrane localization of FZD4 but instead exerts an impact on norrin binding. 

Classification of FZD4 Missense Mutations

In summary of the above findings, we mapped mutations investigated in this study onto the full-length FZD4 amino acid sequence (Fig. 6). Classifying mutations that result in similar functional changes together, the following categories can be delineated: (1) Mutations in the signal peptide are generally predicted to affect protein maturation and trafficking. (2) Mutations involving cysteine residues forming intramolecular disulfide bonds, such as previously reported C45Y, C117R, C181R, C204R, C302Y, and newly identified C145S and C377F, are predicted to affect protein folding and maturation. (3) Mutations in the ECD, all of which affect norrin binding, lead to reduced Norrin/β-catenin activation. Notably, M105V and M157V were previously reported not to affect norrin binding, which differs from our findings.⁴⁴ We posit that this disparity may stem from differing detection methods, as our employment of Co-IP and immunofluorescence staining methods offers heightened sensitivity. (4) The majority of mutations located in TM1 and the 488 position in TM7 of FZD4 affect membrane localization. While one previous study found that M105V and M105T in the ECD also affect membrane localization,²⁹ we observed that M105V could localize normally, consistent with the earlier study by Xu et al.⁴⁴ (5) Mutations in the intracellular domain affect DVL2 recruitment. Previous reports have highlighted the extensive influence of the intracellular domain of FZD5 on DVL2 recruitment,³⁹ and our study corroborates that this finding also applies to FZD4. 

Figure 6.

 

Schematic of the full-length amino acid sequence of FZD4 (excluding the signal peptide, showing amino acids 37–537). Three loops in contact with norrin are denoted in green. Cysteine residues involved in the formation of intramolecular disulfide bonds are denoted in yellow. Mutation sites affecting norrin binding are denoted in orange, those affecting the membrane localization of FZD4 are denoted in magenta, and those affecting the recruitment of DVL2 are denoted in light blue.

For the future functional analysis of newly discovered FZD4 mutations, we devised the following analytical workflow: Mutants are first constructed on the expression vectors. Validation is performed to ascertain the effect on RNA and protein expression. Then, membrane localization is assessed through immunofluorescence staining. For mutations in the ECD region, the focus is on the norrin-binding capability, which can be evaluated via Co-IP or immunofluorescence staining methods. In the case of mutations in the intracellular domain region, priority is given to the analysis of DVL2 recruitment ability using immunofluorescence staining. 

Conclusively, we propose that all FZD4 missense mutations in FEVR can be classified into five categories based on different functional changes, presenting a domain-based map with detailed functional annotations of various FZD4 missense mutations. This map will provide assistance for prenatal counseling and molecular diagnosis for FEVR-associated patients with FZD4 mutations. Furthermore, this study provides molecular basis for the development of potential therapeutic approaches for not only FEVR but also other vascular diseases by targeting different domains of FZD4. 

The study's limitation lies in the fact that although all analyses are focused on missense mutations, a considerable portion of the mutations include frameshift, nonsense, insertion, deletion, CNV, and splicing.⁴⁵ The workflow outlined above may not be suitable for CNV and splicing mutations. Additionally, when dealing with frameshift, nonsense, insertion, and deletion mutations, the challenge arises in classifying alterations involving large segments of amino acids. 

Acknowledgments

Supported by the National Natural Science Foundation of China (82371070 to P.Z., 82121003, 82330030 to Z.Y. and 82101153 to M.Y.), the Sichuan Science and Technology program (2023NSFSC0036, 2022ZYD0059, 2023YFS0038 to M.Y., and 2023YFS0119 to Y.H.), the China Postdoctoral Science Foundation (2023M740518 to Y.H.), the Fundamental Research Funds for the Central Universities of Ministry of Education of China (ZYGX2022J023 to S.L.) and the Research Fund of Sichuan Academy of Medical Sciences and Sichuan Provincial People's Hospital (2023BH09 to R.Z. and 2022BH022 to Y.H.). Shanghai Science and Technology Committee (22015820200 to P.Z.). 

Disclosure: E. Dai, None; M. Liu, None; S. Li, None; X. Zhang, None; S. Wang, None; R. Zhao, None; Y. He, None; L. Peng, None; L. Lv, None; H. Xiao, None; M. Yang, None; Z. Yang, None; P. Zhao, None