Gene Variant Spectrum in Probands With Familial Exudative Vitreoretinopathy Using an Expanded Panel

Abstract

Purpose

To investigate the gene variant spectrum in patients with familial exudative vitreoretinopathy (FEVR).

Methods

Probands clinically diagnosed with FEVR and their relatives were enrolled and clinical information and DNA collected. An expanded FEVR panel was used, including six recognized FEVR genes (FZD4, NDP, LRP5, TSPAN12, ZNF408, and CTNNB1) and 19 genes previously associated with ocular features overlapping FEVR (FEVR-associated genes). Variants identified using targeted next-generation sequencing and/or Sanger sequencing were analyzed and classified using the American College of Medical Genetics and Clinical Genome Resource Sequence Variant Interpretation (ClinGen SVI) working group recommendations to detect disease-causing variants (DCVs).

Results

Analyses of data from a cohort of 94 probands provided a molecular diagnosis for 39 (41.5%) probands: 34 (87.2%) had a single DCV, whereas 5 (12.8%) harbored more than 1 DCV. Of 41 total DCVs in solved probands, 33 (80.5%) were in 4 of the 6 recognized genes, LRP5, FZD4, TSPAN12, and NDP, whereas 8 were found in FEVR-associated genes, 6 in KIF11, and 2 (LAMA1 and DOCK6) each in association with a KIF11 DCV. Reanalyzing variants using the latest criteria impacted the variant classification in five probands (5.3%), changing variants that were once deemed likely pathogenic to variants of uncertain significance.

Conclusions

The expanded FEVR gene panel detected DCVs in nearly one-half of our cohort. Including the criteria used in classification will improve transparency of variant calls as more data become available. Four FEVR genes account for most cases, and the role of rare FEVR genes and candidate genes requires further study.

Familial exudative vitreoretinopathy (FEVR) is characterized by a developmental failure of complete retinal vascularization at birth.^1–3 Clinical manifestations range from asymptomatic peripheral nonperfusion detectable only with intravenous fluorescein angiography³ to blindness typically occurring in early childhood secondary to neovascularization with macular dragging, retinal tears, folds, or detachment in 20% to 40% of affected eyes.²

Initially described in 1969⁴ as an isolated ocular condition, FEVR is now recognized to be genetically heterogeneous with variable expressivity, interocular and intrafamilial variability, as well as associations with a variety of syndromes (Fig. 1).^5,6 FEVR genes listed in the Leiden open variation database (LOVD)⁷ include FZD4,¹ NDP,⁸ LRP5,⁹ TSPAN12,¹⁰ ZNF408,¹¹ and CTNNB1¹² (OMIM: 604579, 300658, 603506, 613138, 616454, and 116806, respectively). Of the recognized FEVR genes in the LOVD and OMIM databases, five participate in a shared Wnt signaling pathway. NDP encodes Norrin, which is the ligand for the FZD4 receptor. Norrin binding to FZD4 results in recruitment of the co-receptors LRP5 and TSPAN12 into a large protein complex that activates an intracellular signaling pathway and results in b-catenin (encoded by CTNNB1) stabilization and translocation into the nucleus and activation of genes involved in retinal vascular development.¹³ The role of ZNF408 in retinal vascular development remains unclear.

Figure 1.

Isolated and syndromic FEVR: genetic causes and inheritance. Genes in bold text designate the six FEVR genes, to differentiate them from genes that may cause a phenotype overlapping that of FEVR, that is, FEVR-associated genes. Owing to variable expressivity or effect of the variant on protein function, among other, 6 of the 25 genes have been reported with or without extraocular manifestations, highlighting the possibility that some cases diagnosed as FEVR may be caused by one of those six genes. NDP has been reported as a cause of FEVR or Norrie disease, including ocular manifestations in carriers that range from a forme fruste to severe disease (e.g., childhood onset retinal detachment), although typically less severe than what is seen in affected males. The predominant mode of inheritance of FEVR caused by LRP5 and FZD4 is autosomal dominant (AD), but AR forms, including syndromic forms, have been reported.

Genomic sequencing of large FEVR cohorts has focused on the genes traditionally associated with FEVR, with detection rates ranging from 6% to 27% in FZD4, 9% to 41% in LRP5, 3% to 13% in TSPAN12, 4% to 11% in NDP, and 0.7% to 1% in ZNF408.^14–19 Studies have reported results using different gene panels assessing four to nine genes and using variable criteria to determine pathogenicity. Additionally, because most were published before 2022, recent Clinical Genome Resource Sequence Variant Interpretation (ClinGen SVI) working group recommendations were not applied and these studies may have artificially increased the determination of variant pathogenicity. A recent meta-analysis using data published before 2022 showed that variants in these genes account for 43.7% of cases, suggesting that additional FEVR genes remain unidentified.⁵ One recent study comparing disease caused by one of the four FEVR genes in the Norrin–Fzd4 complex vs those that did not harbor a variant in the four genes does use the most up-to-date criteria and found that the disease course and severity differed between the two groups.¹⁹ Importantly, developmental peripheral nonperfusion of the retina is not specific to FEVR. Several genes were reported recently in association with an ocular phenotype overlapping that of FEVR, that is, FEVR-like, with and without extraocular manifestations: KIF11,²⁰ ATOH7,²¹ ILK,²² JAG1,²³ LRP6,²⁴ CTNNA1,²⁵ CTNND1,²⁶ DLG1,²⁷ TGFBR2,²⁸ RCBTB1,²⁹ COL9A1,³⁰ DOCK6,³¹ ARHGAP31,³¹ NOTCH1,³² TUBGCP4,³³ TUBGCP6,³³ PLK4,³⁴ CDK19,³⁵ and LAMA1.³⁶ With the exception of KIF11, these genes are either rare causes of FEVR, or manifest FEVR-like disease in a single cohort, or are described as case reports, and their recent association in most cases underscores the possibility that they may account for a group of molecularly undiagnosed cases in previous cohort studies. We refer to these genes as “FEVR-associated genes” for the purposes of this study. Inheritance patterns among these and the known FEVR genes depend on the genetic cause (Fig. 1).

The purpose of this study was to assess the diagnostic value of an expanded FEVR gene panel encompassing 6 FEVR and 19 FEVR-associated genes. We investigate the proportion of previously reported and novel variants predicted to be damaging using guidelines of the American College of Medical Genetics (ACMG) and Genomics³⁷ and the ClinGen SVI recommendations³⁸ in a cohort of 94 FEVR probands.

Methods

Patients

Participants were recruited between 1998 and 2024. The study was approved by the Research Ethics Board of the IWK Health Centre, Halifax, Canada, and written consent was obtained in accordance with the Declaration of Helsinki. Patients manifesting retinal features compatible with a diagnosis of FEVR, affected relatives, and first-degree relatives of these patients were invited to participate. Patients with a history of premature birth at less than 36 weeks of gestation were excluded to avoid confusion with retinopathy of prematurity, which can mimic FEVR. Recruitment was performed by the referring physician. From 1998 to 2010, inclusion criteria for male patients required prior exclusion of NDP-related disease as a clinical test was available and research funding was prioritized to gene discovery.

Clinical Examinations

Clinical data included results from eye examinations and, whenever possible, intravenous fluorescein angiography and fundus photography. Ethnicity was provided at the discretion of the referring physician.

Molecular Analysis

DNA was extracted from peripheral blood or saliva samples using standard protocols. Early in the study, Sanger sequencing was used to identify rare variants in known genes, and this method alone identified a molecular diagnosis for 18 probands. Seventy-six, including two who were diagnosed with Sanger sequencing in prior publications, underwent whole exome sequencing using the Illumina NextSeq 550. Reads were assembled using the Burrows Wheeler Alignment³⁹ (Fig. 2). Genomic variants were called using the Genome Analysis Toolkit⁴⁰ and Picard.⁴¹ The Integrative Genomics Viewer⁴² was used to verify the phase in a case with two variants in the same gene. A virtual FEVR gene panel was used in those 76 probands to identify variants in FZD4 (NM_012193.4), LRP5 (NM_002335.4), TSPAN12 (NM_012338.4), NDP (NM_000266.4), ZNF408 (NM_024741.3), CTNNB1 (NM_001904.4), KIF11 (NM_004523.4), ATOH7 (NM_145178.4), ILK (NM_004517.4), JAG1 (NM_000214.3), TGFBR2 (NM_001024847.2), RCBTB1 (NM_018191.4), DOCK6 (NM_010812.4), ARHGAP31 (NM_020754.4), NOTCH1 (NM_017617.5), CTNNA1 (NM_001903.5), CTNND1 (NM_001085458.2), DLG1 (NM_004087.2), COL9A1 (NM_001851.6), LRP6 (NM_002336.3), TUBGCP4 (NM_001286414.3), TUBGCP6 (NM_020461.4), PLK4 (NM_014264.5), CDK19 (NM_015076.5), and LAMA1 (NM_005559.4). Variants were annotated with Annovar⁴³ and compared against dbSNP,⁴⁴ 1000 Genomes project,⁴⁵ and the genome aggregation dataset version 4⁴⁶ using a minor allele frequency of 0.5%. Pathogenicity predictions were determined using REVEL⁴⁷ and CADD.⁴⁸ The variants were evaluated according to ACMG guidelines,³⁷ including SVI recommendations,³⁸ and cross-referenced with the LOVD human variome FEVR databases.⁷ Each variant was analyzed independently by three raters. A minimum 20× coverage was used to report pathogenic and likely pathogenic variants collectively referred as disease-causing variants (DCVs). Segregation analyses were performed whenever possible. We also report variants of uncertain significance (VUS) as a reference (Supplementary Table).

Figure 2.

Diagnostic flow diagram. Summary of diagnostic work up and genetic test results.

Results

A total of 42 distinct DCVs were identified in 40 of the 94 probands (Tables 1 and 2), none of which were present in homozygous state: 23 DCVs were previously reported and 19 novel variants predicted to be disease-causing represent new potential causes of FEVR or FEVR-like disease. One proband possessing a single DCV in DOCK6, a gene known to cause autosomal recessive (AR) disease (Fig. 1), was deemed unsolved. Among the remaining 39 probands with a molecular diagnosis, 34 harbored a single DCV and 5 possessed 2 DCVs. No proband possessed more than 2 DCVs. Of the 41 DCVs in solved cases (i.e., probands with at least 1 DCV explaining their clinical diagnosis or molecularly solved cases), 33 (80.5%) affect 1 of the 6 FEVR genes, whereas 8 (19.5%) were found in a FEVR-associated gene (Tables 1 and 2; Fig. 2).

Table 1.

Probands With a Single Diagnostic DCV

Proband	Gene	Nucleotide Change	Amino Acid Change	Variant Type	REVEL	CADD	SpliceAI	gnomAD	Sex	Reference	ACMG Ranking	SVI Evidence Categories
1	FZD4	c.107G>A	p.Gly36Asp	Missense	0.31	21.7	n/a	n/a	M	Toomes et al, 2004	Likely pathogenic	PS1, PM1, PP5, PP4, PP2
2	FZD4	c. 197T>A	p.Leu66Gln	Missense	0.81	25.2	n/a	n/a	F	This study	Likely pathogenic	PS4_Mod, PM1, PM2, PP3_Mod, PP4, PP2
3	FZD4	c.313A>G	p.Met105Val	Missense	0.511	24.9	n/a	0.00002398	F	Toomes et al, 2004	Pathogenic	PS1, PS3, PM1, PP3_Sup, PP5, PP2
4	FZD4	c. 316T>G	p.Cys106Gly	Missense	0.952	28.3	n/a	n/a	F	This study	Pathogenic	PP3_Strong, PM1, PM2, PP5, PP4, PP2
5	FZD4	c.341T>C	p.Ile114Thr	Missense	0.851	26.1	n/a	n/a	M	Robitaille et al, 2009	Likely pathogenic	PM1, PM5, PP3_mod, PP4, PP2
6	FZD4	c. 470T>A	p.Met157Lys	Missense	0.268	18.73	n/a	n/a	F	Robitaille et al, 2011	Likely pathogenic	PM1, PM2, BP4_Sup, PP4, PP2
7	FZD4	c.633delC	p.Tyr211Terfs*	Frameshift	n/a	n/a	n/a	n/a	F	Robitaille et al, 2011	Pathogenic	PVS1, PM2, PP4
8	FZD4	c. 677G>A	p.Trp226*	Stopgain	n/a	n/a	n/a	n/a	F	Zhang et al, 2010	Pathogenic	PVS1, PS1, PP3_Mod
9	FZD4	c.1282_1285delGACA	p.Asp428Ser*2	Frameshift	n/a	n/a	n/a	0.000007953	F	Yang et al, 2012	Pathogenic	PVS1, PS1, PM2
10	FZD4	c. 1463G>T	p.Gly488Val	Missense	0.951	29.2	n/a	n/a	F	Robitaille et al, 2011	Likely pathogenic	PP3_Strong, PM2, PM1, PP2
11	FZD4	c.1479_1484delGTGGAT	p.Met493_W494del	Indel	n/a	n/a	n/a	n/a	F	Robitaille et al, 2002	Pathogenic	PS3, PP1_Strong, PM2, PM4
12	FZD4	c.1487G>A	p.Trp496*	Stopgain	n/a	n/a	n/a	n/a	M	Robitaille et al, 2011	Pathogenic	PVS1, PP3_Mod, PM2
13	FZD4	c.1508dupC	p.Trp504Val*31	Frameshift	n/a	n/a	n/a	n/a	F	This study	Pathogenic	PVS1, PM2, PP4
14	LRP5	c.772delC	p.Arg258Alafs*18	Frameshift	n/a	n/a	n/a	n/a	F	This study	Pathogenic	PVS1, PM2, PP5
15	LRP5	c.955dupT	p.Tyr319Leufs*18	Frameshift	n/a	n/a	n/a	n/a	M	This study	Pathogenic	PVS1, PM2, PP4
16	LRP5	c.1067C>T	p.Ser356Leu	Missense	0.912	28.2	n/a	0.000003987	M	Ai et al, 2005	Pathogenic	PS1, PS4_Mod, PP3_Mod, PM1, PP2
17	LRP5	c.1067C>T	p.Ser356Leu	Missense	0.912	28.2	n/a	0.000003987	M	Ai et al, 2005	Pathogenic	PS1, PS4_Mod, PP3_Mod, PM1, PP2
18	LRP5	c.1199C>A	p.Ala400Glu	Missense	0.868	25.9	n/a	n/a	M	Salvo et al, 2015	Pathogenic	PS1, PP3_Mod, PM2, PP2
19	LRP5	c.1229T>A	p.Leu410Gln	Missense	0.927	25.4	n/a	n/a	M	This study	Likely pathogenic	PP3_Mod, PM2, PP4, PM1, PP2
20	LRP5	c.1507G>A	p.Gly503Arg	Missense	0.881	27.9	n/a	0.00000796	F	Liu et al, 2017	Pathogenic	PS1, PP3_Mod, PP1_Sup, PP2
21	LRP5	c.2820_2827+4del	n/a	Splicing	n/a	n/a	0.99	n/a	M	This study	Pathogenic	PVS1, PM2, PP3_Mod, PP4
22	LRP5	c.2251dupT	p.Arg752Profs*61	Frameshift	n/a	n/a	n/a	n/a	F	This study	Likely pathogenic	PVS1, PM2
23	LRP5	c.2737dupT	p.Cys913Leufs*73	Frameshift	n/a	n/a	n/a	0.00000401	F	Ai et al, 2005	Pathogenic	PVS1, PP4, PP5
24	LRP5	c.3561delG	p.Lys1187Asnfs*63	Frameshift	n/a	n/a	n/a	n/a	M	This study	Likely pathogenic	PVS1, PM2
25	LRP5	c.4488+2T>G	n/a	Splicing	n/a	n/a	0.98	n/a	F	Toomes et al, 2004	Pathogenic	PVS1, PS1, PS4_Mod, PP5, PP3_Mod
26	LRP5	c.4488+2T>G	n/a	Splicing	n/a	n/a	0.98	n/a	F	Toomes et al, 2004	Pathogenic	PVS1, PS1, PS4_Mod, PP5, PP3_Mod
27	NDP	c.337G>T	p.Gly113Cys	Missense	0.705	29.2	n/a	n/a	M	This study	Likely pathogenic	PM5, PP3_Sup, PM1, PM2, PP2
28	TSPAN12	c.149+3A>G	n/a	Splicing	n/a	n/a	0.88	0.00000249	F	Poulter et al, 2010	Pathogenic	PVS1, PS1, PP3_Mod
29	TSPAN12	c.172-175delACTT	p.Tyr59Phefs*8	Frameshift	n/a	n/a	n/a	n/a	F	This study	Likely pathogenic	PVS1, PP5, PM2
30	TSPAN12	c.566G>A	p.Cys189Tyr	Missense	0.885	31	n/a	0.00000398	F	Xu et al, 2014	Likely pathogenic	PS1, PP3_Mod, PP5, PP2
31	KIF11	c.139C>T	p.Arg47*	Stopgain	n/a	n/a	n/a	n/a	n/a	Robitaille et al, 2014	Pathogenic	PVS1, PM2, PP3_Mod, PP4
32	KIF11	c.790-1G>T	n/a	Splicing	n/a	n/a	0.99	n/a	M	Robitaille et al, 2014	Pathogenic	PVS1, PM2, PP4
33	KIF11	c.648dupG	p.Lys216Lys*16	Frameshift	n/a	n/a	n/a	n/a	M	This study	Pathogenic	PVS1, PM2, PP4
34	KIF11	c.2195_2196delGA	p.Arg732Ilefs*9	Frameshift	n/a	n/a	n/a	n/a	M	This study	Pathogenic	PVS1, PM2, PP4
35	DOCK6	C.5231G>A	p.Trp1744*	Stopgain	n/a	n/a	n/a	n/a	F	This study	Likely pathogenic	PM2, PM6, PP3_Mod, PP5

F, female; gnomAD, genome aggregation dataset version 4; M, male; N/A, not available.

Distribution of DCVs among 35 probands.

Table 2.

Probands With at Least One Diagnostic DCV

Proband	Gene	Nucleotide Change	Amino Acid Change	Variant Type	REVEL	CADD	SpliceAI	gnomAD	Sex	Reference	ACMG Ranking	SVI Evidence Categories
37	LRP5	c.1453G>A	p.Glu485Lys	Missense	0.89	28.1	n/a	n/a	M	This study	Likely pathogenic	PP3, PM2, PP4, PP3_Mod, BP5
	LRP5	c.2643_2652dupCATCCAGGGC	p.Leu886Profs*23	Frameshift	n/a	n/a	n/a	n/a		This study	Pathogenic	PVS1, PM2
38	LRP5	c.889dupA	p.Thr297Asnfs*3	Frameshift	n/a	n/a	n/a	n/a	F	Welinder et al, 2014	Pathogenic	PVS1, PM2, PP1
	LRP5	c.2827+1G>A	n/a	Splicing	n/a	n/a	0.99	n/a		Welinder et al, 2014	Pathogenic	PVS1, PM2, PP1
39	FZD4	c. 197T>A	p.Leu66Gln	Missense	0.81	25.2	n/a	n/a	M	This study	Likely pathogenic	PS4_Mod, PM1, PM2, PP3_Mod, PP4
	FZD4	c.204_220delGCACGAGCTGCAGACGG	p.His69Argfs*55	Frameshift	n/a	n/a	n/a	n/a		This study	Pathogenic	PVS1, PM2, PM1
40	KIF11	c.1408G>T	p.Glu470*	Stopgain	n/a	n/a	n/a	n/a	F	Robitaille et al, 2014	Pathogenic	PVS1, PM2, PP4
	DOCK6	c.1104+2T>C	n/a	Splicing	n/a	28	0.48	5.7E-05	F	This study	Likely pathogenic	PSV1 and PP3_Mod
41	KIF11	c.157C>T	p.Arg53*	Stopgain	n/a	n/a	n/a	n/a	F	Ostergaard et al, 2012	Pathogenic	PVS1, PS1, PM2, PP4
	LAMA1	c.7736delG	p.Ser2579Metfs*29	Frameshift	n/a	n/a	n/a	n/a		This study	Likely pathogenic	PVS1, PM2

F, female; gnomAD, genome aggregation dataset version 4; M, male; N/A, not available.

Distribution of DCVs among 5 probands with each two DCVs in the same gene or different genes.

DCVs in the FEVR Genes

We observed reported and novel DCVs in 4 of the 6 FEVR genes, FZD4, LRP5, TSPAN12, and NDP in 33 probands; none were present in ZNF408 or CTNNB1 (Tables 1 and 2; Fig. 2).

We found 15 separate LRP5 DCVs in 15 probands accounting for 16.0% of all probands or 38.5% of the molecularly solved probands: 7 were previously reported and 8 are new (Tables 1 and 2).⁴⁹ In total, there were 5 missense (Fig. 3), 7 frameshift, and 3 splicing variants. Thirteen probands possessed a single DCV in LRP5, and 2 were compound heterozygous. The two LRP5 DCVs found in more than one proband were reported previously, including one missense (p.Ser356Leu)⁵⁰ observed twice and one splicing variant (c.4488+2T>G)⁹ also observed in two probands. Unlike the previously reported case with LRP5 p.Ser356Leu in a compound heterozygous state and a diagnosis of osteoporosis–pseudoglioma syndrome (OPPG) (OMIM: 259770),⁵⁰ the probands in our study presented with FEVR alone and did not possess a second DCV. Finally, LRP5 DCVs were found in a compound heterozygous state in two probands, both with OPPG. The proband with LRP5 p.Glu485Lys/p.Leu886Profs*23 presented at 5 months of age with bilateral retinal detachments and a skeletal survey compatible with OPPG; although the parents were each found to carry one of the variants, both were asymptomatic. Details of the clinical, family, and genetic testing history of the proband with LRP5 c.889dupA and c.2827+1G>A variants were published previously.⁵¹ Briefly, the proband and her sister, both with bilateral retinal detachments soon after birth and biallelic DCVs in LRP5, were diagnosed with OPPG, although both parents, who each carried each one of the variants, had normal eye examinations and reduced bone mass.

Figure 3.

(A) Distribution and location of missense variants among A. LRP5. All of the missense DCVs identified, both known and new, are in the extracellular domain of the LRP5 protein and are each contained in a b-propeller domain, with these domains being important during FZD4 receptor complex formation and intracellular signal generation. (B) FZD4. Of the four previously reported missense DCVs, three are located in the N-terminal cysteine-rich domain (p.Met105Val, p.Ile114Thr, p.Met157Lys), with two of the newly identified FZD4 DCVs also located in the N-terminal domain (p.Leu66Gln, p.Cys106Gly). All of these variants are in the Norrin-binding domain of the FZD4 protein. Disruptions in Norrin-FZD4 binding significantly decrease the activation of the Norrin/Frizzled4 signaling pathway and cause FEVR.⁵¹ The remaining missense DCVs present in the C-terminus was previously reported and is required for glycosylation of the FZD4 protein.⁵¹ (C) KIF11. (D) TSPAN12. (E) NDP. (F) DOCK6. (G) LAMA1.

Figure 3.

Continued.

Figure 3.

Continued.

We identified 14 distinct FZD4 DCVs across 14 probands accounting for 35.9% of solved cases (14.9% of the cohort); 10 DCVs were observed previously and 4 are novel (Tables 1 and 2).^52–54 Of these 14 DCVs, 7 were missense (Fig. 3), 2 stop-gain, 4 were frameshift, and 1 was a 2-amino acid deletion. One FZD4 DCV was found in more than one proband (c.197T>A, p.Leu66Gln), and one proband with inoperable retinal detachment in one eye in infancy harbored two FZD4 DCVs, one frameshift, and one missense. This proband was adopted and the only information available on the parents was a diagnosis of amblyopia and poor vision in one eye in the birth mother. Although both variants are pathogenic, they were located on the same allele. It is uncertain whether there is a compound effect of the two variants, although truncated variants of FZD4 are known to undergo nonsense-mediated decay.¹³

Within the reading frames of the other two subunits of the FZD4 receptor complex, we identified three DCVs in TSPAN12, two novel frameshift variants, a previously reported missense variant (Table 1)⁵⁵ (7.7% of the solved cases; 3.2% of the cohort), and one new missense DCV in NDP (2.6% of the solved cases; 1.1% of the cohort).

Identification of DCVs in FEVR-Associated Genes

Eight DCVs were found in 3 of the 19 FEVR-associated genes: KIF11, DOCK6, and LAMA1 (Tables 1 and 2; Fig. 2). KIF11 has been reported in association with autosomal dominant (AD) FEVR features as a part of the microcephaly with or without chorioretinopathy, lymphedema, or impaired intellectual development syndrome (MCLMR) (OMIM: 152950), whereas the FEVR phenotype associated with variants in LAMA1 and DOCK6 manifests as part of syndromes that are inherited in an AR manner (Fig. 1).

Six variants in KIF11 were identified across six probands accounting for 15.4% of solved cases (6.4% of the cohort). Each DCV was either a stop-gain, frameshift, or splicing variant (Tables 1 and 2), consistent with the previously predicted haploinsufficiency associated with KIF11-related disease.⁵⁶ Four KIF11 variants were present in isolation and two were found in probands with another FEVR-associated gene DCV, one in LAMA1 and the other in DOCK6. Multigenic inheritance was only detected in these two probands.

A heterozygous LAMA1 frameshift variant (p.Ser2579Metfs*29) was found in one proband who also possessed a KIF11 stop-gain variant, as mentioned elsewhere in this article. The proband presented at 8 months of age with bilateral retinal folds and severe retinal dysplasia with no light perception vision in one eye. She was noted to have postnatal microcephaly (below the third percentile) and normal development at 13 years of age. LAMA1 variants can cause Poretti–Boltshauser syndrome (OMIM: 615960) in an AR fashion.³⁶ There were no signs of FEVR in either parent, and the influence of the additional LAMA1 DCV is unclear, especially in the context of disease variability in MCLMR alone and without DNA samples from the parents. Similarly, a variant affecting splicing of DOCK6 was identified in a patient who also harbored a KIF11 DCV. This patient had microcephaly, retinal falciform folds, and atrophic retinochoroidal changes. An affected sibling carrying both DCVs manifested microcephaly and atrophic changes without folds or detachment. No information was available from the parents. As with the previous case with KIF11 and LAMA1 DCVs, the influence of the additional DOCK6 DCV was unclear in these siblings.

A novel stop-gain variant was identified in DOCK6 in one proband. One parent with the same variant had a normal eye examination and was unable to complete an intravenous fluorescein angiography. Because the proband did not have other DCVs, and the variant was found in a heterozygous state, that proband did not receive a molecular diagnosis and remained unsolved.

Variants of Uncertain Significance

We identified 44 variants across 46 of the 76 probands with whole exome sequencing that were predicted to be of uncertain significance (VUS) (Supplementary Table).^57,58 Of these variants, 16 were in known FEVR genes and the remaining 27 were in FEVR-associated genes. Importantly, five variants (5.3% of probands) once deemed to be likely pathogenic changed to VUS after reanalyzing the data using updated criteria and recommendations (Supplementary Table). Two of the five variants would have suggested the condition could be AR LRP5 disease and another, digenic inheritance (FZD4 and CTNND1), with implications for genetic counselling, requiring confirmation of phase and disease status in the parents, in addition to prognostic implications in recessive and possibly digenic FEVR. An additional 2 of the 94 probands (2.1%) would have been inappropriately labelled as solved. The proband with a pathogenic DOCK6 stop-gain variant also harbored two other DOCK6 VUS and two TUBGCP6 VUS.

Discussion

We assessed a FEVR gene panel encompassing a total of 25 known FEVR genes and FEVR-associated genes in our cohort of 94 FEVR probands. Using ACMG guidelines and applying SVI recommendations, we determined that 42.6% of probands (40/94) possessed a pathogenic or likely pathogenic variant, and a molecular diagnosis was possible in 41.5% (39/94). Thirty-three probands (84.6% solved cases) carried DCVs in 4 of the 6 known FEVR genes, and the remaining 6 (15.4%) were in FEVR-associated gene KIF11. Digenic inheritance was rare and, in each case, the influence of an additional, single LAMA1 or DOCK6 variant on MCLMR could not be verified, but remains possible. Applying current ACMG and SVI recommendations and reanalyzing the variants in our cohort impacted the molecular diagnosis reached over the course of our study in five cases, none previously published, among the 94 probands (5.3%). Variant classifications in the literature and in clinical and research databases, including our database, often do not specify the criteria that were applied in their classification. The clinical and population databases and in silico tools change over time. Knowledge from functional analyses may also contribute to changing a variant classification. However, which variant, if any, have contributed to the differences cannot be determined. By including the criteria used in our classification, future users of these data who may arrive at a different classification will be able to determine where their assessment differs, which is a valuable aspect of variant interpretation.

The DCV detection rate using our FEVR panel in the 94 probands, excluding single DCVs in genes known to cause AR disease exclusively, were as follows: 16.0% LRP5, 14.9% FZD4, 6.4% KIF11, 3.2% TSPAN12, and 1.1% NDP. The overall and gene-specific variant detection rates are comparable with the reported range, despite using the most extensive panel to date. Our data shows that this is due to the rarity of finding DCVs outside of the five most common genes, a finding also observed in other studies that include some of the rare genes in their panels. Kondo et al.,¹⁹ using similar criteria to determine pathogenicity as in our study found 13.9% DCVs in FZD4, 13.5% in LRP5, 4.6% in TSPAN12, 4.6% in NDP, and 1.8% digenic in various combinations. A 2022 meta-analysis⁵ calculated variant frequencies among 5 FEVR genes plus KIF11 using 3257 patients from 32 studies: a FEVR gene variant was found in 43.7%, of which 13.6% arose in LRP5, 11.5% in FZD4, 6.7% in TSPAN12, 5.7% in KIF11, 4.6% in NDP, and 1.6% in ZNF408. No criteria were described to verify the pathogenicity of the previously reported variants. Tao et al.¹⁸ identified variants in 67.5% of 120 probands using a panel of six FEVR genes (LRP5, FZD4, TSPAN12, NDP, KIF11, and ZNF408). Rare variants categorized as VUS or benign using ACMG guidelines were included in this count, and when those are excluded, the detection rate reduces to 45%. Qu et al.¹⁵ found rare variants in 40.54% of their cohort of 74 probands using a 9-gene panel that included RCBTB1, JAG1, and CTNNA1 in addition to the 6 genes in the Tao panel, and only 33.8% if VUS are excluded. Both the Tao and the Qu papers found most variants in LRP5 followed closely by FZD4, with rates of fewer than 6% in each of the remaining genes in the Tao report, and fewer than 11.1% in the Qu paper. No pathogenic or likely pathogenic variant was identified in ZNF408 in the Tao cohort and none in JAG1, ZNF408, or CTNNA1 in the Qu study.

Two factors may have decreased some of the DCV detection rates in our study compared with previous reports. The first is a confounder that explains the lower rate of FEVR caused by NDP, because probands were excluded if NDP prescreening was positive during the first one-half of our study. Our rate of 1.1% is much lower than previous studies using different methodologies reporting 4.6% to 9.7% of FEVR cases caused by NDP variants.^14–18 We also found slightly fewer cases caused by TSPAN12 DCVs. The reason is unclear, but may be due to the low numbers of affected probands among those with rarer genetic causes of FEVR. Second, our present application of the classification criteria led to a more conservative interpretation, decreasing the number of solved cases by 2.1%. Reanalyzing variants from previous literature describing rare FEVR candidate genes using the same criteria as in our study, along with a review of the functional evidence, could influence the variant detection rates and the inclusion of genes in a diagnostic FEVR panel.

We identified 44 VUS in FEVR and FEVR-associated genes. Interestingly, 33% of the solved probands possessed at least one VUS vs 37% of probands without a diagnostic DCV, with several probands harboring combinations of VUS in more than one gene, mainly in the FEVR-associated genes. The classification process of variants in the common FEVR genes is easier owing to the greater availability of data compared with the rare genes. Many VUS in those rare genes may be reclassified eventually as DCVs in the future as more data become available. Consideration of multigenic inheritance may improve our ability to predict disease outcomes in the future. In support of this statement, a recent survey of protein-altering variants (i.e., not necessarily DCV) among seven FEVR genes (KIF11 was included as a FEVR gene in this paper) has identified digenic and trigenic inheritance in 34% of FEVR cases compared with 3.6% in the general population, suggesting a potential contribution to disease severity.⁵⁹ Li et al.⁶⁰ found digenic inheritance in 2.7% of probands (13/487), most of which consisted of a combination of LRP5 and FZD4 pathogenic, likely pathogenic, variants as well as VUS, and found that those with digenic variants in known FEVR genes tended to have more severe disease. In our study, proband #11 with the FZD4 DCV, a CTNND1 VUS, as well as an LRP6 VUS is part of a large pedigree with more than 33 affected individuals and manifested the most severe disease: she was born with retinal detachment in one eye and, in the other eye, severe retinal dysplasia and fold that eventually evolved to a complete detachment in the second decade of life. Our results, along with the findings from these publications, supports the value of performing a similar analysis in our cohort that would include all affected patients, not only probands, to optimize the prediction value of the presence of VUS on disease course and outcomes and possibly a combined effect to explain some of the unsolved cases. Such association studies will require functional analyses to demonstrate mechanisms of enhanced effects from gene variant combinations on disease severity.

The proband with a single heterozygous DOCK6 likely pathogenic variant was not included in the solved cases; single heterozygous variants in this gene have not been reported previously in patients with FEVR. This proband also had two DOCK6 VUS that may eventually be found to be pathogenic. A milder phenotype or reduced penetrance is also possible and has been reported in other conditions that can cause FEVR or FEVR-like disease, such as heterozygous carriers of OPPG LRP5 variants who were recognized with an increased risk of osteoporosis-related fractures (as was the case for one of our probands) and milder, asymptomatic FEVR.^50,61 In support of the need for careful eye examination of carriers of variants that cause AR disease is our discovery of two FEVR probands in our cohort with a single LRP5 p.Ser356Leu variant, a variant that has been reported in a compound heterozygous state in a patient with OPPG.⁵⁰ The increased risk of severe disease, including ocular and extraocular manifestations, for genes that cause syndromic and nonsyndromic forms of XLR or AR FEVR may be related to the effects on protein function caused by biallelic variants, variant location and type, and for carriers of NDP DCVs, the effects of lyonization.^13,62,63 Systematic eye examination of parents of children who manifest FEVR-like features and are diagnosed with Adams-Oliver syndrome caused by DCVs in DOCK6 (OMIM: 614194), NOTCH1 (OMIM: 190198), or ARHGAP31 (OMIM: 610911) and Poretti–Boltshauser syndrome (OMIM: 615960) caused by DCVs in LAMA1, for example, could reveal undiagnosed peripheral nonperfusion among carriers and confirm a role for these FEVR-associated genes in biallelic and heterozygous states. This is particularly true as peripheral non perfusion can escape detection on routine fundoscopy and may require intravenous fluorescein angiography.

A role for RCBTB1 in FEVR has been debated,^29,64–66 and many do not consider this gene to be associated with FEVR. However, a recent report¹⁵ identified a rare likely pathogenic missense variant in this gene in a parent–child pedigree, and we wanted to interrogate our cohort: we did not identify a DCV in this gene. One proband possessed a single variant that was reported in a homozygous state together with a second homozygous variant in the same gene in a patient with a progressive retinal dystrophy.⁶⁷

Our results support a stepwise, panel-based approach to analyses of genomic data in patients manifesting signs of FEVR, starting with the four most common FEVR genes, LRP5, FZD4, TSPAN12, and NDP, in addition to KIF11, mostly owing to the variable expressivity of MCLMR, followed by a second round of analyses using a broader, more inclusive panel if the first round fails to yield a molecular diagnosis. In the presence of extraocular manifestations suggesting a specific syndrome, targeted single gene testing remains appropriate. The remaining 2 FEVR genes and 18 FEVR-associated genes, as well as VUS in FEVR and FEVR-associated genes, will require further study to advance our understanding of their role as causes and modifiers of FEVR or a FEVR-like phenotype. Further genotyping and phenotyping of FEVR patients and relatives at risk is required to determine the full spectrum of disease-causing genes and will increase our understanding of the molecular processes necessary for retinal angiogenesis.