CTNND1 variants cause familial exudative vitreoretinopathy through the Wnt/cadherin axis

Familial exudative vitreoretinopathy (FEVR) is a hereditary disorder that can cause vision loss. CTNND1 encodes a cellular adhesion protein p120-catenin (p120), which is essential for vascularization with unclear function in postnatal physiological angiogenesis. Here, we applied whole-exome sequencing to 140 probands of FEVR families and identified 3 candidate variants in the human CTNND1 gene. We performed inducible deletion of Ctnnd1 in the postnatal mouse endothelial cells (ECs) and observed typical phenotypes of FEVR with reactive gliosis. Using unbiased proteomics analysis combined with experimental approaches, we conclude that p120 is critical for the integrity of adherens junctions (AJs) and that p120 activates Wnt signaling activity by protecting β-catenin from glycogen synthase kinase 3 beta–ubiqutin–guided (Gsk3β-ubiquitin–guided) degradation. Treatment of CTNND1-depleted human retinal microvascular ECs with Gsk3β inhibitors LiCl or CHIR-99021 enhanced cell proliferation. Moreover, LiCl treatment increased vessel density in Ctnnd1-deficient mouse retinas. Variants in CTNND1 caused FEVR by compromising the expression of AJs and Wnt signaling activity. Genetic interactions between p120 and β-catenin or α-catenin revealed by double-heterozygous deletion in mice showed that p120 regulates vascular development through the Wnt/cadherin axis. In conclusion, variants in CTNND1 can cause FEVR through the Wnt/cadherin axis.

Familial exudative vitreoretinopathy (FEVR) is a hereditary disorder that can cause vision loss. CTNND1 encodes a cellular adhesion protein p120-catenin (p120), which is essential for vascularization with unclear function in postnatal physiological angiogenesis. Here, we applied whole-exome sequencing to 140 probands of FEVR families and identified 3 candidate variants in the human CTNND1 gene. We performed inducible deletion of Ctnnd1 in the postnatal mouse endothelial cells (ECs) and observed typical phenotypes of FEVR with reactive gliosis. Using unbiased proteomics analysis combined with experimental approaches, we conclude that p120 is critical for the integrity of adherens junctions (AJs) and that p120 activates Wnt signaling activity by protecting β-catenin from glycogen synthase kinase 3 beta-ubiqutin-guided (Gsk3β-ubiquitinguided) degradation. Treatment of CTNND1-depleted human retinal microvascular ECs with Gsk3β inhibitors LiCl or CHIR-99021 enhanced cell proliferation. Moreover, LiCl treatment increased vessel density in Ctnnd1-deficient mouse retinas. Variants in CTNND1 caused FEVR by compromising the expression of AJs and Wnt signaling activity. Genetic interactions between p120 and β-catenin or α-catenin revealed by double-heterozygous deletion in mice showed that p120 regulates vascular development through the Wnt/cadherin axis. In conclusion, variants in CTNND1 can cause FEVR through the Wnt/cadherin axis.

WES analysis of FEVR families identified variants in CTNND1.
To identify potential candidate pathogenic variants for FEVR, we applied WES analysis to 140 FEVR-associated families without known causative variants. Presumed autosomal dominant variants identified by WES that had a frequency of less than 0.001 in the dbSNP151, Exome Variant Server, ExAC, 1000 Genomes Project, and gnomAD databases and were absent from 1000 sequenced controls were selected as candidate genes. Genes that met the following 3 standards were then selected (20): first, a candidate gene should have at least 3 variants in unrelated families; second, at least 1 "disruptive" variant (nonsense or frameshift) should be among the candidate variants; and third, the candidate gene should be related to Wnt signaling or angiogenesis. Thus, 3 potentially novel candidate variants in the CTNND1 gene (NM_001085458) were identified in 3 Chinese families: a heterozygous variant c.949C>T (p.R317C) in Family-441, a heterozygous nonsense variant c.1867A>T (p.K623*) in Family-506, and a heterozygous missense variant c.2099G>A (p.R700Q) in Family-3313 ( Figure 1, A-C). The WES results also showed no convincingly pathogenic variants in the known FEVR-associated genes in these probands. The pathogenicity of the variants was predicted to be deleterious by PolyPhen-2, Likelihood Ratio Test, and Mutation Taster software. The protein sequence alignment of human p120 with homologs from humans and other species indicated that all affected amino acids were highly conserved ( Figure 1, A-C).
Clinical findings. For the heterozygous c.949C>T (p.R317C) variant in Family-441, the proband was diagnosed with FEVR at age 6, manifesting severe retinal exudates with a characteristic peripheral avascular retina in the left eye ( Figure 2A). His father was heterozygous and manifested a bilateral peripheral retinal avascular area, straightened peripheral vessels, and increased vessel branching, while his mother showed a normal fundus without this variant (Figure 2A). The second heterozygous variant, c.1867A>T (p.K623*), was found in a 1-month-old boy with macular ectopia directed toward the temporal periphery in the right eye and a peripheral avascular area in the left eye ( Figure 2B). The proband also manifested with features of BCD syndrome phenotypes, including syndactyly, and cleft lip/palate (CLP) ( Figure 2B). His affected mother was heterozygous and manifested macular ectopia and phthisis bulbi secondary to retinal detachment in the right eye, while his father manifested normal visual acuity and fundus without this variant ( Figure 2B). The mother of the proband also manifested distinct BCD syndrome phenotypes (syndactyly without CLP). The K623* was a nonsense variant that was predicted to result in premature transcription termination. We thus speculated that the K623* variant might be a major pathogenic variant in this family.
The third heterozygous variant, c.2099G>A (p.R700Q), was found in an 8-year-old boy with peripheral avascularization and severe retinal exudates in the right eye, whereas the vasculature of his left eye was undetectable due to band keratopathy ( Figure 2C). The younger brother of the proband was heterozygous for the variant, manifesting a similar phenotype in the right eye and retinal detachment in the left ( Figure 2C). Their mother was heterozygous for this variant, manifesting bilateral peripheral avascularization, exudation, and neovascularization ( Figure 2C). The father and maternal grandfather were normal without this variant.
Thus, these variants in CTNND1 cosegregated with the disease phenotype in their families ( Figure 1); pathogenesis of these variants that cause FEVR was further investigated by in vivo and in vitro experiments.
The development of mouse retinal vasculature initiates from the optic nerve; the tip cells at the leading edge (hereafter termed the angiogenic front) extend numerous filopodia and radially sprout to form the superficial vasculature (10,34). The primary vessel plexus behind the tip cells (hereafter termed the remodeling plexus) subsequently undergoes pruning to form a hierarchically mature vasculature (10,34). To evaluate whether deletion of p120 could cause FEVR-like phenotypes, we first applied immunofluorescence analysis on the P7 flat-mounted retinas and observed compromised radial outgrowth of retinal vessels upon deletion of endothelial Ctnnd1 in mice ( Figure 3, A and B). Interestingly, the vascular density of the Ctnnd1 iECKO/iECKO mouse retinas was reduced in the remodeling plexus ( Figure 3, A and C), whereas it was elevated in the angiogenic front ( Figure 3, A and D), demonstrating vascular hyperplasia at the peripheral retina (10,34). After the initial outgrowth of the superficial layers from P1 to P7, the retinal vessels sprout perpendicularly into the deep layers and ramify in the outer plexiform layers (OPLs) and inner plexiform layers, leading to the formation of a 3-layered retinal vascular architecture (35,36). At P9, vertical sprouts and OPL capillaries appeared in the control mice, whereas both were absent in the Ctnnd1 iECKO/iECKO mice ( Figure 3E).
During postnatal vascular development, the hyaloid vessels reach maximal development at approximately P3 and start to regress in the second week (34,36). Delayed hyaloid regression has previously been reported in several FEVR-associated mouse models (10,20,34,36,37). As predicted, we observed retarded regression of hyaloid vessels in the P9 Ctnnd1 iECKO/iECKO mouse retinas, as compared with their littermate controls ( Figure 3F). Given that regression of hyaloid vessels begins from P4 and finishes by P21 in mice (38), we further tested whether hyaloid vessel is persistent in P21 Ctnnd1 iECKO/iECKO mice. Hematoxylin and eosin staining of paraffin-embedded ocular sections showed remanent hyaloid vessels in P21 Ctnnd1 iECKO/iECKO mice, which were absent in wild-type mice (Supplemental Figure 2A). Additionally, we did not observe structural abnormalities in the eyes of Ctnnd1 iECKO/iECKO mice, as compared with those of littermate controls ( Figure 3E and Supplemental Figure 2A).
The integrity of the inner blood-retinal barrier (BRB) is critical for selective vascular permeability, and disruption of BRB integrity might lead to severe leakage, which is a typical phenotype of FEVR ( Figure 3G). However, a large abundance of erythrocytes was detected outside the vessel lumen in the retinas of P6 Ctnnd1 iECKO/iECKO mice ( Figure 3G), indicating the disruption of BRB upon depletion of Ctnnd1. Biocytin-TMR is an 869 Da molecule that ordinarily crosses the BRB when paracellular junctional integrity is disrupted (41). We therefore performed injection of biocytin-TMR to detect the effect of Ctnnd1 depletion on BRB integrity. As expected, loss of p120 in mouse ECs resulted in enhanced vascular permeability for biocytin-TMR at P6 (Supplemental Figure 2B), indicative of disrupted BRB integrity. p120 was previously reported to facilitate the anchoring of VE-cadherin to the plasma membrane (42). To evaluate the localization of VE-cadherin in the absence of p120, we stained the P6 retina flat mounts with antibodies against VE-cadherin. VE-cadherin outlined the retinal vasculatures of the P6 control mice, while it was abnormally accumulated and diffusely distributed with a less clear vessel boundary distribution in the retinas of Ctnnd1 iECKO/iECKO mice, indicating disruption of the AJ continuity (Supplemental Figure 2C). Meanwhile, the ultrastructure of P9 ECs under transmission electron microscopy showed a discontinuous EC-EC junction in the absence of p120 (Supplemental Figure 2D), confirming disruption of the endothelial junctions.
Since an increased level of glial fibrillary acid protein (GFAP) has previously been reported to indicate active astrocytes in Fzd4, Norrin, Tspan12, or Ctnna1 mutant mouse retinas (10,20,34,37), we stained GFAP in the P6 retinas of Ctnnd1 iECKO/iECKO mice and control littermates to evaluate the activation of astrocytes. The results showed prominently higher GFAP intensities in both the angiogenic front and remodeling plexus of Ctnnd1 iECKO/iECKO mouse retinas compared with littermate controls, indicating overactivation and gliosis of astrocytes ( Figure 4, A and B). Astrocytes and microglia are resident immune cells that act as primary defenses in the retina. Reactive astrogliosis and microgliosis are involved in immune responses (43,44). We thus performed ionized calcium-binding adapter molecule 1 (IBA1)  staining to detect whether microglia were reactive in the absence of EC-derived p120. As expected, we observed abnormal proliferation of amoeboid-shaped microglia with enlarged cell bodies in the retinas of P8 Ctnnd1 iECKO/iECKO mice ( Figure 4, C and D), which are markers of reactive microglia (44). These results indicated reactive gliosis in the Ctnnd1 iECKO/iECKO mouse retinas.
A mature vasculature can be divided into arteries, veins, and capillaries. In P9 wild-type retinas, the major arteries and veins were segregated into alternating and radially arrayed territories ( Figure 4E); however, several crossings between arteries and veins were observed in the P9 Ctnnd1 iECKO/iECKO mouse retinas ( Figure 4F), which closely resembles the phenotype observed in Ndp KO retinas (36).
Loss of Ctnnd1 causes abnormal VEGF expression, cell proliferation, and vessel pruning. To further explore the mechanism of defective vascular development in Ctnnd1 iECKO/iECKO mouse retinas, we first asked whether the dense appearance of the distal vasculature resulted from filopodia-extending tip cell hyperproliferation. The results showed that tip cells at the leading edge were more abundant in the P6 Ctnnd1 iECKO/iECKO retinas than in those of the littermate controls (Supplemental Figure 3, A and B), suggesting abnormal proliferation of Ctnnd1-null ECs at the peripheral area of the retinal vasculature. We further applied EdU incorporation assay to confirm whether mitotic cell proliferation of the peripheral ECs increased. Consistently, the number of relative EdU + ECs drastically increased at the angiogenic front of the P6 Ctnnd1 iECKO/iECKO mouse retinas (Supplemental Figure 3, C and D).
Normally, astrocytes prominently secrete VEGF in the angiogenic front and form a gradient in the avascular areas ahead of the vessel plexus, regulating tip cell proliferation and extension during retinal vascular development (45). To explore the mechanism of hyperplasia and delayed radial outgrowth in p120-defective retinal vessels, we examined the abundance of Vegfa in the P6 Ctnnd1 iECKO/iECKO and control retinas. Fluorescence immunostaining of flat mounts showed strong upregulation of Vegfa expression in both the angiogenic front and the avascular areas upon Ctnnd1 depletion (Supplemental Figure 4, A-C), presumably in response to erythrocyte leakage and retinal hypoxia. However, the Vegfa level in the remodeling plexus was comparable between the control and Ctnnd1 iECKO/iECKO retinas (Supplemental Figure 4, B and C).
Collagen IV is a component of matrix deposits that regressing vascular ECs leave (46). Collagen IV + and isolectin B4matrix sleeves, which indicate vessel pruning, were highly abundant in P6 Ctnnd1 iECKO/iECKO mouse retinas (Supplemental Figure 4D), indicating that loss of p120 function might lead to abnormal vessel pruning. This may be responsible for decreased vessel density in the remodeling plexus ( Figure 3, A and C).

Depletion of CTNND1 in human retinal microvascular endothelial cells inhibits the cadherin and Wnt signaling pathways.
To explore the molecular mechanism through which p120 regulates EC function, we applied lentivirus-directed shRNA knockdown of CTNND1 to primary human retinal microvascular endothelial cells (HRECs). An unbiased proteomic analysis was performed on the control and CTNND1-knockdown HRECs. A large number of differentially expressed proteins were detected, as shown in the volcano plot, and the abundance of the p120 protein was confirmed to be substantially decreased ( Figure 5A and Supplemental Table 1). Of those involved, the cadherin and Wnt signaling pathways were the most affected upon knockdown of CTNND1 ( Figure 5B). We first focused on the cadherin signaling pathway and applied Western blot analysis to the HREC lysates. In the absence of p120, the protein levels of α-catenin, β-catenin, and VE-cadherin were considerably downregulated ( Figure 5, C-F), whereas RT-qPCR analysis revealed unchanged mRNA levels for these genes ( Figure 5, G-I), indicating that p120 regulates the expression of cadherin/catenin components at the protein level rather than at the mRNA level. We further applied immunofluorescence staining to evaluate the AJ integrity of p120-defective HRECs. The intensities of membrane-localized cadherin/catenin complex proteins in CTNND1-depleted HRECs were found to be remarkably decreased compared with those in control HRECs ( Figure 6, A-E), which might lead to vessel leakage in vivo. These results suggested that p120 might play a pivotal role in the regulation of AJ integrity and protect cadherin/catenin components from degradation.
We recently reported that inactivation of CTNNA1 (α-catenin) in ECs results in reduced membrane localization of the cadherin/catenin complex (20). However, to our knowledge it has not been determined whether depletion of α-catenin affects the total protein levels of cadherin/catenin components. Interestingly,  Western blot analysis of HRECs transfected with control or CTNNA1 shRNA exhibited identical total protein levels of other cadherin/catenin components (Supplemental Figure 5), further indicating that p120 might be a specific regulator protecting the cadherin/catenin complex from degradation.
Notably, in addition to functioning in cadherin signaling, β-catenin is more widely known as a core component of the Wnt signaling pathway. We next asked whether the Wnt signaling pathway was disturbed in CTNND1-depleted HRECs. The mRNA levels of several Wnt-targeted genes, including CyclinD1 (CCND1), c-Myc (MYC), c-Jun (JUN), DKK1, CD44, and claudin5 (CLDN5), were significantly reduced in the absence of p120 ( Figure 7A). Additionally, we observed an appreciable reduction of the CyclinD1, c-Myc, and glucose transporter type 1 (GLUT1) proteins in CTNND1-depleted HRECs (Figure 7, B-E).
Since defective Wnt signaling was previously reported to result in compromised in vitro vascularization (46-48), we applied a Matrigel tube formation assay to investigate in vitro vascularization in CTNND1depleted HRECs. As expected, we observed defective in vitro tube formation upon deletion of CTNND1 in the HRECs ( Figure 7F). Given that Wnt signaling regulates vascular cell proliferation (36), we unsurprisingly detected decreased mitotic cell proliferation in CTNND1-depleted HRECs (Figure 7, G and H), which is consistent with the consequence of defective Wnt signaling. These results confirmed the compromised activity of the endogenous Wnt/β-catenin signaling pathway in the absence of p120, which might cause defective proliferation and vascularization of HRECs.
We noticed a discrepancy between the compromised HREC proliferation and the increased EC proliferation of Ctnnd1 iECKO/iECKO in the angiogenic front. Since levels of Vegfa, a master regulator of EC proliferation, were increased in the angiogenic front and avascular areas of the Ctnnd1 iECKO/iECKO mouse retinas (Supplemental Figure 4, A-C), we hypothesized that excessive Vegfa secreted by astrocytes might be attributable to hyperplasia of the peripheral retinal vasculature. Therefore, we applied recombinant VEGF-165 to test VEGF signaling responses in control and CTNND1-depleted HRECs. Western blot analysis revealed that treatment with VEGF-165 led to elevated phosphorylation of VEGFR2 and ERK1/2 in both control and CTNND1-depleted HRECs (Supplemental Figure 6). Intriguingly, in the presence of VEGF-165, the phosphorylation levels of VEGFR2 and ERK1/2 were higher in the CTNND1-depleted HRECs than that in the control (Supplemental Figure 6), suggesting that HRECs were more sensitive to VEGF-165 upon CTNND1 depletion. Additionally, we also investigated the effect of VEGF-165 on Wnt signaling activity. However, we did not observe alterations of β-catenin and CyclinD1 protein levels upon VEGF-165 treatment, neither in control nor in CTNND1-depleted HRECs (Supplemental Figure 6).
It has been widely established that Gsk3β regulates phosphorylation of β-catenin at the Ser33, Ser37, and Thr41 residues, providing a docking site for E3 ubiquitin ligase β-TrCP, which mediates ubiquitination and subsequent proteasomal degradation of β-catenin (49,50). Thus, we asked whether p120 protects β-catenin from degradation by inhibiting this Gsk3β-mediated process. We applied treatment with LiCl or CHIR-99021, inhibitors of Gsk3β, to restore β-catenin signaling in p120-deficient HRECs. Western blot analysis showed that LiCl or CHIR-99021 considerably restored the endogenous protein levels of β-catenin, as well as downstream targets CyclinD1 and c-Myc (Figure 8, A-E). Not surprisingly, both LiCl and CHIR-99021 failed to restore the protein levels of VE-cadherin (Figure 8, A and E). Immunofluorescence staining also showed that treatment with LiCl or CHIR-99021 partially restored the levels of membrane-and nucleus-localized β-catenin in p120-deficient HRECs (Supplemental Figure  7, A and B). Additionally, treatment with LiCl or CHIR-99021 remarkably promoted cell proliferation in CTNND1-depleted HRECs (Supplemental Figure 7, C and D). These results suggested that p120 protects β-catenin from degradation and that loss of p120 might impair the proliferation of ECs through inactivation of Wnt signaling, which could be corrected by inhibition of Gsk3β-ubiquitin-guided β-catenin degradation.
To explore whether Ctnnd1 depletion affects Wnt signaling in vivo, a survey of several markers of the Wnt-regulated genes in the retinal vasculature was conducted to show a decrease of lymphoid enhancer binding factor 1 (LEF1), a transcriptional mediator of Norrin/β-catenin signaling (Supplemental Moreover, given that LiCl has previously been applied to normalizing ocular vessels in Lrp5 -/and heterozygous EC-specific Ctnnb1-knockout mice (17,38), we applied intraperitoneal injection of LiCl (10 mg/kg) to Ctnnd1 iECKO/iECKO mice to investigate whether overactivation of β-catenin signaling could rescue defective retinal angiogenesis. Interestingly, LiCl treatment considerably restored vascular density in the remodeling plexus of P7 Ctnnd1 iECKO/iECKO mouse retinas (Figure 8, F and G). However, LiCl failed to promote vascular progression and prevent leakage of erythrocytes in the Ctnnd1 iECKO/iECKO mouse retinal vasculature (Figure 8, F and H).
CTNND1 variants disrupt the cadherin/catenin complex and inactivate Wnt signaling activity. To confirm the effects of the CTNND1 variants R317C, K623*, and R700Q on the biological functions of p120, we applied site-directed mutagenesis to introduce these variants into CTNND1 expression plasmids, which are resistant to shRNA targeting without changing the amino acid sequence. Wild-type CTNND1, variant CTNND1, or vector plasmids were then transfected into CTNND1-knockdown HEK293T cells to evaluate the effects of variants on the function of the p120 protein. As predicted, Western blot analysis showed truncation of the K623* variant without affecting protein levels, while the R317C and R700Q variants produced identical bands to the wild-type CTNND1 expression plasmid (Figure 9, A-E). Notably, overexpression of wildtype protein remarkably increased the endogenous levels of β-catenin, which further triggered activation of Wnt-targeted genes, including CyclinD1 and c-Myc (Figure 9, A-E). However, the R317C, K623*, and R700Q variants failed to restore expression levels of β-catenin, α-catenin, and Wnt-targeted genes, indicating that these variants might disrupt the normal function of p120 (Figure 9, A-E).
The currently available structural and mutagenesis analyses exhibited direct interaction between p120 and E-cadherin via broad cadherin binding interfaces in the p120 armadillo repeat (ARM) domains (1-5 and 7-10 ARM domains) (52,53). The R700 residue is located in the eighth ARM domain of p120 ( Figure 9F), and substitution of Arg with Cys might disrupt potential p120/cadherin binding sites. The Lys623* variant was proved to result in a truncated protein that lacked the 7-10 ARM domains and the C-terminal domain ( Figure 9F), which is likelier than the R700Q variant to cause disrupted p120/cadherin binding surfaces. The R317 residue is located in the N-terminal domain, which is structurally adjacent to the first ARM domain ( Figure 9F). However, since R317 was not included in the structure currently available, the exact role of this residue in the binding of p120 and cadherin is unknown. To investigate the mechanism by which these variants affect the cadherin/catenin complex and Wnt signaling activity, we tested whether these variants disrupt interactions between p120 and VE-cadherin or β-catenin using co-immunoprecipitation analysis. The results revealed that wild-type p120 strongly interacted with overexpressed VE-cadherin and endogenous β-catenin ( Figure 9G). In contrast, the R317C, K623*, or R700Q variants showed remarkably compromised interactions with both VE-cadherin and β-catenin ( Figure 9G), indicating that these variants in CTNND1 caused a disrupted cadherin/catenin complex, which could result in degradation of β-catenin and compromised signaling transduction. p120 genetically interacts with β-catenin and α-catenin. We have previously reported that EC-specific knockout of Ctnna1 (α-catenin) in mice causes FEVR-like phenotypes through disruption of the cadherin/ catenin complex and abnormal overactivation of β-catenin signaling (20). However, in this paper, we observed disruption of the cadherin/catenin complex and inactivation of β-catenin signaling in EC-specific Ctnnd1-knockout mice and CTNND1-depleted HRECs. We therefore asked whether Ctnnd1 knockout in murine ECs could result in distinct phenotypes from those of Ctnna1-knockout mice. Aligned comparisons of these retinas showed that, despite both models exhibiting vascular hyperplasia and extensive leakage of erythrocytes in the peripheral areas of the superficial retinal vasculature, Ctnnd1 iECKO/iECKO mice exhibited sparsely distributed vasculature in the remodeling plexus compared with the extreme hyperplasia seen in the Ctnna1 iECKO/iECKO case (Supplemental Figure 9). This discrepancy might be partially due to the opposite roles that α-catenin and p120 play in the regulation of Wnt/β-catenin signaling. We constructed Ctnnb1 iECKO/iECKO mice, which exhibited retinal vascular phenotypes similar to those of Ctnnd1 iECKO/iECKO mice, except with less leakage. This indicated that disruption of AJs might play a pivotal role in the maintenance of vascular integrity (Supplemental Figure 9). We further bred double-heterozygous knockout mice that each lacked 1 allele of Ctnnd1, Ctnna1, or Ctnnb1 to elucidate whether p120 genetically interacts with α-catenin or β-catenin. We crossed Ctnnd1 iECKO/+ mice with either Ctnna1 fl/+ or Ctnnb1 fl/+ mice to generate wild-type, single-gene heterozygous, or doubleheterozygous offspring. Notably, the vasculature of Ctnnd1 iECKO/+ mouse retinas showed moderate FEVRlike phenotypes, including delayed superficial vascular progression, leakage of erythrocytes, and crossings of arteries and veins ( Figure 10). Further deleting 1 allele of the Ctnnb1 in Ctnnd1 iECKO/+ mouse ECs led to increased vessel leakage, artery/vein crossings, and peripheral vascular hyperplasia, as well as more severely compromised vascular progression compared with Ctnnd1 iECKO/+ mice ( Figure 10, A and C), which closely recapitulated the phenotypes observed in Ctnnd1 iECKO/iECKO mice. Deletion of 1 Ctnna1 allele from Ctnnd1 iECKO/+ mice caused severe vascular leakage and hyperplasia in the leakage areas, whereas the vascular progression was undisturbed (Figure 10, B and D).

Discussion
p120 is an indispensable component of the cadherin/catenin complex that participates in the normal function of several tissues and organ systems (54-56). However, the regulatory role of p120 in postnatal retinal vascular development remains unclear due to the embryonic lethality of Tie2-Cre-mediated knockout of  Ctnnd1 in the early embryonic stage (33). In this study, using WES on the genomic DNA of probands in 140 FEVR-associated families without known genetic variants, we identified 3 heterozygous variants in the CTNND1 gene, including 1 nonsense and 2 missense variants.
Interestingly, the proband and his affected mother in Family-506, who were heterozygous for the K623* variant, manifested distinct features of BCD syndrome phenotypes in addition to FEVR phenotypes   [29][30][31][32]. BCD syndrome is a rare disorder characterized by typical manifestations of CLP, eyelid malformations, and ectodermal defects, as well as other atypical manifestations of hypothyroidism, imperforate anus, neural tube defects, and syndactyly. However, the FEVR-associated phenotype has not yet been reported in patients with BCD syndrome to our knowledge (29)(30)(31)(32). CTNND1 variants have been associated with BCD syndrome. Notably, of this and the 15 previously reported variants in the CTNND1 gene that cause BCD syndrome, 13 were frameshift or nonsense variants, indicating that p120 participates in different aspects of human development. Frameshift variants with severe disruption of p120 function might cause systemic development defects, which is consistent with the observation that the K623* variant causes BCD syndrome phenotypes in addition to FEVR (29)(30)(31)(32).
In vitro functional studies indicated that all variants compromised the Wnt signaling activity through disruption of interactions between p120 and VE-cadherin or β-catenin ( Figure 9). Thus, we conclude that these variants are candidate pathogenic variants capable of causing FEVR through disruption of the cadherin/catenin complex and inactivation of Wnt/β-catenin signaling. Intriguingly, depletion of Ctnnd1 in the postnatal mouse ECs largely phenocopied typical manifestations of FEVR, confirming that CTNND1 could be a candidate gene for FEVR.
To investigate the mechanism by which p120 regulates EC function, we applied an unbiased proteomics analysis of CTNND1-knockdown HRECs and found prominent enrichment of differentially expressed proteins in the cadherin and Wnt signaling pathways ( Figure 5, A and B). It has been previously reported that knockdown of CTNND1 in human dermal microvascular ECs results in downregulation of cadherin/catenin components (22). Additionally, overexpression of CTNND1 in HUVECs significantly increases the levels of VE-cadherin (57). As expected, in the CTNND1-knockdown HRECs, we found disruption of the AJs and downregulation of cadherin/catenin components, including α-catenin, β-catenin, and VE-cadherin, which were regulated at the protein level, rather than at the mRNA level ( Figures 5 and 6).
The role of p120 in Wnt signaling in various cells or organs remains controversial. In prostate cancer cell lines, p120 has been reported to activate the Wnt/β-catenin signaling pathway (58). However, p120 has a negative role in regulating the Wnt/β-catenin signaling pathway in epithelia (59). Additionally, it has been reported to be redundant in regulating the Wnt/β-catenin signaling pathway in human corneal ECs (60). Here, we showed that p120 functioned as an activator of the Wnt/β-catenin signaling pathway; loss of p120 resulted in degradation of β-catenin and further downregulated genes involved downstream of the Wnt/β-catenin signaling pathway.
In mouse retinas, vasculature formed from vessel sprouting subsequently undergoes vessel pruning from P5~P18 to form a functionally mature network (61). The Wnt/β-catenin pathway has been widely studied in the regulation of vessel pruning. Loss of endothelial Ctnnb1 or Lef1 in mice has been reported to cause premature vessel regression (61,62). In contrast, transgenic gain of Ctnnb1 function in mouse ECs or global deletion of a negative Wnt/β-catenin regulator Apcdd1 resulted in activation of Wnt/β-catenin signaling and delayed vessel pruning (63,64). Likewise, loss of Semaphorin 3G or the endothelial transcription factor ERG led to decreased Wnt/β-catenin signaling and increased vessel pruning in the mouse retinas (41,46). Thus, it is reasonable to speculate that the abnormally increased vessel pruning in Ctnnd1 iECKO/iECKO mouse retinas was a consequence of compromised β-catenin signaling (Supplemental Figure 4D). Furthermore, it is worth noting that although the lack of p120 in ECs led to compromised Wnt signaling and increased vessel pruning, the cell proliferation and vascular density in the angiogenic fronts of Ctnnd1 iECKO/iECKO mouse retinas were abnormally elevated. We hypothesized and supported that, in the angiogenic front, deletion of endothelial Ctnnd1 leads to extensive vessel leakage, which stimulates immune responses and excessive Vegfa secretion, as well as increased sensitivity of ECs to Vegfa, which bypasses the negative effect of compromised Wnt signaling and collectively contributes to hyperplasia in the angiogenic front.
Interestingly, treatment with LiCl or CHIR-99021, which are inhibitors of Gsk3β, considerably restored Wnt/β-catenin signaling activity and promoted the in vitro proliferation of p120-depleted HRECs (Supplemental Figure 7). This indicates that p120 might protect β-catenin from Gsk3β-mediated ubiquitination degradation. It is notable that although LiCl treatment of Ctnnd1 iECKO/iECKO mice considerably increased the vascular density of their retinas (possibly due to enhanced β-catenin signaling and inhibited pruning), LiCl failed to rescue compromised vascular progression and BRB integrity, suggesting that disrupted cadherin/ catenin complexes play nonnegligible roles in the progress of FEVR-like phenotypes in Ctnnd1 iECKO/iECKO mice. These results further highlighted the role of normal AJ integrity in retinal vascularization, which might be a therapeutic target for FEVR.
Genetic and functional analyses revealed that heterozygous variants in the CTNND1 gene caused FEVR in a presumed autosomal dominant manner of inheritance, consistent with the fact that most FZD4, LRP5, and TSPAN12 variants are inherited as autosomal dominant FEVR (10,34,65). However, mice subjected to heterozygous global knockout of Fzd4, Lrp5, or Tspan12 exhibit normal vascularization (10,34,65). In contrast, heterozygous knockout of Ctnnd1 in mouse ECs manifested moderate FEVR-like phenotypes ( Figure 10), indicating that angiogenesis is more sensitive to the Ctnnd1 dosage than to the dosages of the abovementioned Wnt-related genes (10,34,65). We hypothesized that the disruption of the cadherin/catenin complex, in addition to compromised Wnt/β-catenin signaling, might be the mechanism of the haploinsufficiency of CTNND1 in the pathogenesis of FEVR.
Using structural and functional analyses, we revealed that variants in CTNND1 compromised the p120/ VE-cadherin and p120/β-catenin interactions (Figure 9, F and G). It is worth noting that p120 has been reported to indirectly interact with β-catenin or α-catenin (66); thus, we speculate that the compromised interactions between p120 variants and β-catenin were largely due to the compromised p120/cadherin interaction. Moreover, although it is sufficient to conclude that the K623* and R700Q variants compromised p120/VE-cadherin interaction by disrupting the binding surfaces in the ARM domains of p120, further investigation of the mechanism by which the R317C variant affects interactions between p120 and VEcadherin is limited by the lack of a 3-dimensional structure of the full-length p120/VE-cadherin complex.
In our previous research, we showed that conditional knockout of Ctnna1 in ECs resulted in FEVR-like vascular defects and a radically distinct vascular density compared with that of Fzd4-or Lrp5-knockout mice (20). Here, Ctnna1 iECKO/iECKO mice exhibited extensive hyperplasia compared with the sparsely distributed vasculature in the remodeling plexus of Ctnnd1 iECKO/iECKO mouse retinas (Supplemental Figure 9) (20). These observations indicate that, although p120 and α-catenin are indispensable components of the cadherin/ catenin complex associated with FEVR, they may act through distinct mechanisms.
It has been reported that α-catenin inhibits Wnt signaling by segregating the available pool of β-catenin in cytosol, and that variants of α-catenin disrupt the cadherin/catenin complex that releases β-catenin from cytosol for nuclear translocation and activation of β-catenin signaling, which may contribute to the pathogenesis of FEVR (20,(67)(68)(69)(70). Notably, we observed genetic interactions between p120 and β-catenin or α-catenin in mice (Figure 10), albeit with distinguishing manifestations. This discrepancy could be explained by the hypothesis that double-heterozygous loss of Ctnnd1 and Ctnna1 prominently contributes to compromised AJ integrity, leading to severe vascular leakage and associated immune responses that induce hyperplasia without affecting vascular progression, whereas double-heterozygous deletion of Ctnnd1 and Ctnnb1 impairs vascular development prominently through inactivation of Wnt signaling, thus leading to more typical FEVR-like phenotypes.
Taken together, we demonstrate that variants in the CTNND1 gene cause FEVR through the combined contributions of AJ disruption and Wnt signaling inactivation ( Figure 10E). This offers potentially novel pathogenic clues concerning the development of FEVR and other diseases involving the vascular system.

Methods
A full description of the methods is presented in the online-only Supplemental Methods.
Mouse models. Ctnnd1 conditional knockout mice (Ctnnd1 tm1Lfr /Mmucd) (71) were purchased from the Mutant Mouse Resource & Research Centers and bred into a background of Pdgfb-iCre-ER mice (72). The Ctnna1 iECKO/iECKO model was constructed as previously described (20). The Ctnnb1 loxP/loxP mice were obtained from The Jackson Laboratory (stock number 004152). Clinical data. Patients and their family members were enrolled and evaluated by ophthalmologists at Zhongshan Ophthalmic Center, Sun Yat-sen University, and Xinhua Hospital, Shanghai Jiaotong University.
Plasmids. Wild-type CTNND1 coding sequence was subcloned into a mammalian expression vector with N-terminal FLAG-tagged pCDNA3.1 vector (both from Youbio Biotechnology). CTNND1 variants were generated by site-directed mutagenesis using a Q5 Site-Directed Mutagenesis Kit (E0554S, New England Biolabs) and confirmed by Sanger sequencing.
Image acquisition. The microscope (LSM 900, Zeiss) was used to capture confocal images. The Spline Contour2 and line tools in Zen 2.1 software (Zeiss) were used to evaluate intensity mean value of cell junction proteins and vascular progression of whole mounts, respectively. Angiotool software was used to measure the percentage of vessel area as previously described (73). The immunoblot signals were detected by ChemiDoc (Bio-Rad) and analyzed by ImageJ software (NIH).
Data and materials availability. All the data and materials are available within the article and the supplemental materials. Supplemental data include 11 figures and 6 tables.
Statistics. Statistical analysis was performed with GraphPad Prism 8.0. To assess statistical significance with a 95% confidence level, unpaired 2-tailed Student's t test with Welch's correction was applied to comparisons between 2 experimental groups, while 1-way ANOVA or 2-way ANOVA with Dunnett's or Tukey's multiple comparison tests were performed in multiple comparisons among more than 2 experimental groups. All the experiments were conducted at least 3 times independently. P < 0.05 was considered statistically significant.
Study approval. The clinical study was conducted in accordance with the declaration of Helsinki and approved by the ethical oversight committee of Sichuan Provincial People's Hospital, Zhongshan Ophthalmic Center, Sun Yat-sen University, and Xinhua Hospital, Shanghai Jiaotong University. Written informed consent was obtained from all participants or legal guardians for minors. The Institutional Animal Care and Use Committee at the Sichuan Provincial People's Hospital approved all animal study protocols. All experimental procedures were carried out according to the approved protocols.

Author contributions
XZ, ZY, and XD conceived the research. MY, SL, and RZ performed the animal model study and cell biology, immunohistochemistry, and gene expression studies. LH, XJ, YH, JL, LP, ED, WL, ZZ, DJ, YZ, ZJ, and YY performed the construction of plasmids and animal breeding. XD, LH, and PZ recruited the participants and performed the WES study and Sanger sequencing analysis. MY, SL, XD, and XZ analyzed data. MY, SL, and XZ wrote the manuscript, with input from other authors.