Molecular Characterization of FZD4, LRP5, and TSPAN12 in Familial Exudative Vitreoretinopathy

Abstract

Purpose: Familial exudative vitreoretinopathy (FEVR) is a rare hereditary disorder characterized by the failure of peripheral retinal vascularization. The genes FZD4, LRP5, and TSPAN12 are known to be associated with the autosomal inheritance form of FEVR. In this study, we performed mutation screening for FZD4, LRP5, and TSPAN12 in patients with clinical diagnosis of FEVR. In patients with no mutation detected, sequencing analyses for ZNF408, a novel gene potentially related to FEVR, and two other genes related to retinal development, LGR4 and ATOH7, were performed.

Methods: Mutational studies were done in 51 unrelated patients with diagnosis of FEVR during 2008 to 2012 at the Seoul National University Hospital. These patients were screened previously for NDP gene and confirmed to be negative for mutations. Diagnosis of FEVR was established by ophthalmic examinations. Data collected from medical records included sex, age at diagnosis, clinical presentation, and angiographic findings.

Results: In this study, we identified 3 known mutations, 10 novel variants with high possibility of pathogenicity, and a whole gene deletion in a total of 18 unrelated patients of 51, resulting in 35.3% of patients being genetically confirmed as having FEVR. Among the patients with pathogenic mutations detected, FZD4 mutations accounted for the largest proportion of autosomal inheritance FEVR cases (13/18 patients, 72.2%), followed by LRP5 (4/18 patients, 22.2%) and TSPAN12 (1/18 patients, 5.6%). No pathogenic mutations were identified in ZNF408, LGR4, and ATOH7. A significant difference in FEVR stage and visual acuity was observed according to the gene involved, showing that patients with FZD4 mutations had milder phenotype.

Conclusions: Mutations of FZD4 accounted for the largest proportion, which could be directly applied to the testing strategy to start with screening for FZD4 mutations. Panel sequencing consisting of related genes would be an alternative choice for the diagnosis of FEVR. Also, genotype–phenotype correlation suggested in this study could be helpful in genetic counseling of the probands and their family members as well.

Familial exudative vitreoretinopathy (FEVR), first described by Criswick and Schepens in 1969,¹ is a rare hereditary disorder characterized by failure of the peripheral retinal vascularization. Clinical manifestation of the disease can be variable, ranging from nonsymptomatic vascular anomalies in the peripheral retina to bilateral retinal detachments with blindness. The prevalence of this disorder is unknown, but it is likely to be underestimated due to high proportion (79%) of peripheral retinal vascular anomalies detected in asymptomatic family members.² Familial exudative vitreoretinopathy does not follow a predictable timeline of progression, occurring throughout childhood and adulthood, and thus, the accurate diagnosis of FEVR is important for long-term monitoring. The condition also is genetically heterogeneous, and found in various modes of inheritance. Autosomal dominant inheritance is the most common form in FEVR, and FZD4, LRP5, and TSPAN12 are the known genes associated with the disease. An additional FEVR-related genetic locus (EVR3) has been mapped to chromosome 11p12–p13, but the gene is not known yet.³ Genes LRP5 and TSPAN12 also are reported to be involved with autosomal recessive inheritance of the disease, usually resulting in more severe phenotype.^4,5 Possibility of autosomal recessive FZD4 mutations has been mentioned previously,^6,7 but to our knowledge there are no cases with two clear pathogenic mutations reported to date. Additionally, X-linked FEVR is caused by mutations in the NDP gene, which also is known as a causal gene of Norrie disease. Norrie disease is one of the NDP-related retinopathies, showing clinically overlapping phenotype with X-linked FEVR. 

The four genes known to be associated with FEVR (NDP, FZD4, LRP5, and TSPAN12) are the essential components of the wingless (Wnt) pathway in the retina. Norrin protein encoded by the NDP gene, which is not a typical Wnt pathway ligand, binds to the receptor complex of the Wnt pathway composed of the seven-pass transmembrane Frizzled (FZD) receptor along with its coreceptor, low-density lipoprotein receptor-related protein (LRP), and acts as a ligand in the retina.⁸ Another causative gene, TSPAN12, also is known to form a receptor complex with FZD4 and LRP5. Pathogenic mutations affecting the function of these genes results in abnormal retinal vascular formation. 

Considering that the existing proportion of mutation detection does not generally exceed 50%,² searching for novel causative genes of FEVR should be continued to increase the proportion of molecularly confirmed patients. In this study, we chose several other genes as FEVR-related candidates. One of them was ZNF408, which was suggested recently to be associated with autosomal dominant FEVR.⁹ Though it has not been explained how this gene acts along with other components of the Norrin pathway, mutants of this gene resulted in abnormal retinal vasculogenesis, suggesting that ZNF408 might account for a considerable proportion of FEVR. Another gene, LGR4, has been reported previously to act as a receptor for Norrin, with a high possibility of having a role in the Wnt signaling pathway related to retinal vascularization.^10,11 Additionally, ATOH7, mutations of which originally were reported in patients with persistent hyperplastic primary vitreous (PHPV), also was included due to the phenotype of PHPV resembling FEVR and also the role of ATOH7 in retinovascular disease.¹² 

Here, we show mutation spectrum of FZD4, LRP5, and TSPAN12 in patients with clinical diagnosis of FEVR, who had been confirmed previously as negative for NDP mutations. Mutation screening for ZNF408, LGR4, and ATOH7 genes also was performed in patients with no detected mutation in three autosomal inheritance FEVR-related genes. Then, we reviewed genotype–phenotype correlation of each gene in molecularly confirmed patients. 

Materials and Methods

Subjects

A retrospective chart review was performed in patients with a diagnosis for FEVR between January 2008 and December 2012 at the Seoul National University Children's Hospital. The diagnostic criteria for FEVR composed of three of the following: birth at full term or premature birth with no evidence of retinopathy of prematurity, a presence of peripheral retinal avascular area, and variable degree of nonperfusion, vascular leakage, or retinal neovascularization in fluorescein angiography. Patients were included if a final diagnosis of FEVR was given after the complete ophthalmic examination, including fundus examination and fluorescein angiography using Retcam (Clarity Medical Systems, Inc., Pleasanton, CA, USA). 

Among the patients diagnosed with FEVR by clinical diagnostic criteria, patients with NDP mutations were excluded and only those confirmed as negative for NDP mutation were included. Mutational studies were done in 51 unrelated patients who agreed on the genetic testing and, among them, three had another family member with the diagnosis of FEVR. When these patients were confirmed with a mutation, family members also were tested. Informed consent was obtained from all individuals after explaining the nature and possible consequences of the study. Experiments were performed according to the Declaration of Helsinki and were approved by the hospital's ethics committee (IRB No. 1503-036-654). 

Sequence Analysis

Whole blood was collected from patients in EDTA blood collection tubes and the DNA was extracted using Gentra PureGene blood kits (Qiagen, Valencia, CA, USA) according to the manufacturer's instructions. Polymerase chain reactions were done using primers specific for all coding exons (Supplementary Table S1). The amplified products were purified with ExoSAP-IT and sequenced on an ABI 3730 analyzer using a BigDye Terminator v3.1 Cycle sequencing kit (Life Technologies, Carlsbad, CA, USA). Sequences were analyzed using Sequencher 5.1 (Gene Codes Co., Ann Arbor, MI, USA). Three genes known to cause autosomal dominant FEVR (FZD4, LzRP5, and TSPAN12) were analyzed sequentially. In addition, cases with negative mutation in all three known FEVR-related genes were screened for mutations in ZNF408, LGR4, and ATOH7. 

Multiplex Ligation-Dependent Probe Amplification (MLPA)

Gross gene deletions and duplications for FZD4 and LRP5 were screened via MLPA using the SALSA P285-C1 LRP5-NDP-FZD4 (MRC-Holland, Amsterdam, The Netherlands). Products of PCR were analyzed on an ABI 3130 analyzer with Genemarker ver. 1.51 (Softgenetics, State College, PA, USA). 

Haplotype Analysis With SNPs

To assess whether commonly found FZD4 variant, c.313A>G, came from a common founder, four single nucleotide polymorphisms (SNPs) near the coding region of FZD4 with minor allele frequency threshold of 20% were assessed. Three SNPS were located downstream of FZD4 coding region (rs75727571, rs713065, rs3802892) and one located upstream (rs7925666). Haplotype reconstruction and frequency estimation were performed using PHASE ver 2.1.¹³ 

Clinical Significance of the Novel Unclassified Variant

In cases of novel missense variants, pathogenicity of each variant was assessed by considering allele frequencies in approximately 180 normal controls, conservation of a given amino acid among species using MUSCLE,¹⁴ and in silico prediction results. Three in silico prediction analyses were used in this study: SIFT,¹⁵ PolyPhen,¹⁶ and MutationTaster.¹⁷ 

Clinical Manifestation According to Genetic Mutations

Clinical characteristics, including mainly angiographic findings, the presence of macular ectopia, macular dragging, retinal folds, and patient demographics, including sex, age at presentation, and cycloplegic refractive errors (spherical equivalent) at last follow up visit were analyzed in patients with FEVR. 

Statistical Analysis

For the statistical analysis, visual acuity is converted from Snellen value to logMAR value. Statistical analyses were performed using SPSS software version 18.0 (SPSS, Inc., Chicago, IL, USA). P values less than 0.05 were considered to be statistically significant. 

Results

Demographics

We surveyed 51 unrelated patients with clinical diagnosis of FEVR. Eight of them were girls and 43 were boys. The reason for sex bias was that many of the patients who had requested NDP sequencing before this study were included, who were mostly boys. The mean age at the time of diagnosis was 18 months (range, 19 days through 7 years old). Among three patients who had another family member with the diagnosis of FEVR, two of them shared the same mutation with their original probands. 

Overall Mutation Spectrum

We identified three known mutations, 10 novel variants with high possibility of pathogenicity, and a whole gene deletion in total of 18 families, resulting in 35.3% of patients being genetically confirmed as having FEVR (Table 1). Among them, FZD4 mutations were most common (13/18, 72.2%), followed by mutation in LRP5 (4/18, 22.2%) and TSPAN12 (1/18, 5.6%). Sequence analyses of all coding exons of ZNF408, LGR4, and ATOH7 were done in 33 patients with no mutations detected in FZD4, LRP5, or TSPAN12, but no additional mutation was found. 

Table 1

 

Clinical Characteristics of 18 FEVR Patients and 2 Family Members With Pathogenic Mutation Detected

 

Pathogenic Mutations Detected

Mutations of FZD4 were found in 13 patients. The previously reported variant of c.313A>G was recurrently found in four patients and c.1282_1285delGACA in two. Since c.313A>G was frequently detected in this study, we checked for the occurrence in the normal control group, but none was found. Haplotype of five patients with FZD4 c.313A>G mutation (patient ID 2-1, 2-2, 3, 4, 5) also were assessed to check the founder effect. A total of 4 haplotypes were reconstructed from our data (Table 2). Haplotype 2 was found in 4 patients who carried FZD4 c.313A>G mutation, but not in one patient. It is less likely that this mutation was derived from a common founder. 

Table 2

 

Gene FZD4 Haplotypes Reconstructed in Patients With c.313A>G Mutation

 

Among the novel variants found in FZD4, one was a nonsense mutation (c.160C>T) and two were frameshift mutations (c.539_540delAG and c.1210_1211delTT). One patient was found to carry a whole gene deletion of FZD4, which was detected by MLPA. Novel missense variants detected in FZD4, c.456C>G, c.470T>C and c.676T>A, also were considered as likely pathogenic mutations due to following reasons. These variants were in well-conserved region of FZD4 protein across all species considered (Fig. 1), all predicted to be damaging by more than two in silico analyses, and were not found in normal control group. In the case of c.470T>C (p.Met157Thr), pathogenic mutations also have been reported previously in the same amino acid position (p.Met157Lys and p.Met157Val).^18,19 

Figure 1

 

Amino acid conservation of six novel missense sequence variations that are likely to be pathogenic. (A) p.Asn152Lys and p.Met157Thr in FZD4. (B) p.Trp226Arg in FZD4. (C) p.Thr244Arg in LRP5. (D) p.Asp1366Glu in LRP5. (E) p.Leu19Arg in TSPAN12.

Mutations of LRP5 were detected in four patients. A previously reported mutation of c.1330C>T was found in one patient, and a novel c.1833dupG mutation in another. Two novel missense variants (c.731C>G and c.4098C>G) in LRP5 also were considered as likely pathogenic mutations. In the case of c.731C>G (p.Thr244Arg), a pathogenic mutation in the same amino acid position (p.Thr244Met) has been reported previously in an osteoporosis-pseudoglioma syndrome patient.²⁰ Only one patient was confirmed to have a TSPAN12 missense variant of c.56T>G, which also was classified as a likely pathogenic variant. 

A patient with FZD4 c.313A>G mutation (patient ID 2-1 in Table 1) had a maternal cousin also diagnosed with FEVR (patient ID 2-2), who carried the same mutation but more severe phenotype. Considering the pedigree, the mothers of both probands are suspected to be patients with unnoticed mild phenotype (Fig. 2). Unfortunately, genetic testing or fundus examination of family members other than the probands were not performed. Another patient with novel FZD4 mutation of c.470T>C (patient ID 7-1 in Table 1) had a sister who carried the same mutation, but with more progressed phenotype (patient ID 7-2). 

Figure 2

 

Pedigree of the two probands who were maternal cousins (Patient ID 2-1 and 2-2). Both carried a mutation of c.313A>G in FZD4. Mothers of each proband are suspected to be asymptomatic FEVR individuals carrying the same mutation.

Variants of Unknown Significance Detected

Variant with weak evidence of pathogenicity detected in this study were categorized as variants of unknown significance (Table 3). Allele c.3361A>G of LRP5, which had been previously reported as pathogenic mutation in FEVR, was classified as likely benign variant due to its known minor allele frequency of 0.872% in East Asian population. A novel in-frame duplication variant (c.653_676dup24 of FZD4) did not show strong evidence for pathogenicity, thus, regarded as unknown significance. An in-frame deletion variant of c.55_60delCTGCTG, which removes 2 leucine residues in the poly-leucine repeat in the LRP5 signal peptide, was not found in the normal control of this study. In a previously reported functional assay, c.55_60del variant did not show a statistically significant reduction in activity of Norrin signal transduction in vitro.²¹ Allele c.194C>T (p.Pro65Leu) of TSPAN12 was detected in two unrelated patients and not in the normal control, but two in silico analysis predicted that this variant may be benign. Allele c.484G>A (p.Val162Ile) variant of TSPAN12 was predicted to be likely pathogenic by in silico analyses, but it did not segregate with the proband's symptomatic brother. Additionally, novel synonymous variants detected in LRP5 (c.2160C>T, c.4488G>A, and c.4650C>T) were all considered benign, though the possibilities of abnormal splicing were not excluded in this study. 

Table 3

 

Variants of Unknown Significance Identified in this Study

 

Clinical Manifestation

Among the 18 patients with genetic mutation and two other family members who carried the same mutation as the proband, six were girls and 14 were boys. There was a significant difference in FEVR stage (P = 0.001, Kruskal-Wallis test) or visual acuity (P = 0.019, Kruskal-Wallis test) at last follow up according to the gene involved in our patients. Post hoc analyses using Bonferroni's correction showed that FEVR stage was significantly low and visual acuity significantly better in eyes with FZD4 mutation group than in eyes with LRP5 mutation (P = 0.002, P = 0.006, respectively, Mann-Whitney U test). Of importance, visual acuity was low in the eyes (n = 24) with macula ectopia or retinal fold compared to the eyes (n = 16) without them (P < 0.001, Student's t-test). Visual acuity was poor in boys than in girls (P = 0.014, Student's t-test). The representative color fundus photographs in patients with novel FZD4 mutation and LRP5 mutation were shown in Figure 3, and the representative fundus fluorescein angiographic findings in patients with novel FZD4 mutation in Figure 4. 

Figure 3

 

Fundus photographs of patients with a novel FZD4 mutation and LRP5 mutation. (A, B) Patient ID 10 (FZD4, c.1210_1211delTT). (A) Peripheral avascular area is shown in the right eye. (B) Macular RPE atrophy is shown in the left eye. (C, D) Patient ID 16 (LRP5, c.1833dupG). (C) Radial retinal fold to inferotemporal quadrant cause macula dragging in the right eye. (D) Radial retinal fold to temporal area is concomitant with macular-involving retinal detachment.

Figure 4

 

Fundus angiographic findings of patients with novel FZD4 mutations. (A, B) Patient ID 1 (c.160C>T). (A) Vascular leakage is shown at temporal avascular border. (B) Abnormal vascular branches are shown at the temporal retina with vascular leakage. (C, D) Patient ID 8 (c.539_540delAG). (C) Fibrovascular traction at inferotemporal area causes extramacular retinal detachment in the right eye. (D) Vascular leakage is shown at the inferotemporal avascular border in the left eye. (E, F) Patient ID 13 (whole gene deletion). (E) Sea fan–shaped abnormal vascular branching and leakage are observed at the temporal avascular border in the right eye. (F) Almost full retinal vascularization is achieved in the left eye.

Discussion

The existing proportion of FEVR patients who are genetically diagnosed by the detection of pathogenic mutations is approximately 40% to 50%.² Although the contribution of each gene to this disease differs between study populations, it is known that 4% to 40% is attributed to FZD4, 12% to 25% to LRP5, and 3% to 10% to TSPAN12.^22–28 In this study, pathogenic mutations were detected in 18 patients, resulting in 35.3% of patients being genetically confirmed as having FEVR. Two recurrent mutations, c.313A>G and c.1282_1285delGACA in FZD4 were detected in our patient group. Allele c.313A>G in FZD4, which was detected in four unrelated patients in this study, has been reported previously in Japanese and Chinese FEVR patients, suggesting that this mutation occurs frequently in the Asian population. Haplotype analysis indicated that this mutation was less likely to be derived from a common founder. Among 18 patients with pathogenic mutation, 13 had FZD4 mutations, showing that mutations in this gene comprise the highest proportion of autosomal inheritance FEVR in Korean population (13/51 patients, 25.5%). This proportion was similar to the previously reported 20%.^6,19 The proportion of LRP5 mutations contributing to FEVR in our study group seems lower than reported proportion of other studies.^22,26,29 

Interestingly, our study showed that carrying a mutation in FZD4 seemed to result in milder phenotype than LRP5. Despite the limitation of small number of molecularly confirmed patients, those with FZD4 mutations showed lower FEVR stages and better visual acuity compared to those with LRP5 mutations. Though the phenotype of FEVR has long been considered to be indistinguishable by gene involved, analyzing a larger population of molecularly confirmed patients actually might be able to suggest the genotype–phenotype correlation, to the level further than currently known phenotypes of osteopenia or osteoporosis in the patients with LRP5 mutations. 

Additionally, there was a suspected case of a gross gene deletion in TSPAN12, which was not confirmed in the present study. A symptomatic subject with a homozygous variant of c.484G>A in TSPAN12 did not segregate with the proband's symptomatic brother who had homozygous G. The homozygous state of this rare variant may imply gross deletion of the gene, but a clearer explanation will be provided when gene dosage analysis for TSPAN12 is performed. 

Overall, pathogenic mutations were not detected in more than half of the patients in this study. No significant mutations were detected in the novel causative gene ZNF408, or in the possibly related LGR4 and ATOH7 genes. In cases where no mutation was found, further analysis to determine the novel gene contributing to FEVR still is needed. Targeting Wnt pathway-related genes would be an effective strategy. Additionally, mutations in the yet to be defined gene EVR3 also may contribute to the disease. 

In conclusion, we showed the mutation spectrum of three genes: FZD4, LRP5, and TSPAN12 in Korean FEVR patients. Mutations of FZD4 were the most common in cases of autosomal inheritance FEVR. Thus, a testing strategy for FEVR starting with screening for FZD4 mutations should be applied clinically. Panel sequencing consisting of the genes mentioned above is an alternative choice. Genetic counseling of a proband and asymptomatic family members also would be helpful in further management and prevention of the disease. Further analyses of novel causative genes and genotype–phenotype correlations may contribute to a better understanding of the pathophysiological consequences of FEVR. 

Acknowledgments

The authors alone are responsible for the content and writing of the paper. 

Disclosure: S.H. Seo, None; Y.S. Yu, None; S.W. Park, None; J.H. Kim, None; H.K. Kim, None; S.I. Cho, None; H. Park, None; S.J. Lee, None; M.-W. Seong, None; S.S. Park, None; J.Y. Kim, None