Genetic Variants Associated with Anterior Segment Dysgenesis in South Florida


 Background Anterior segment dysgenesis (ASD) comprises a wide spectrum of developmental conditions affecting the cornea, iris, and lens, which may be associated with abnormalities of other organs. Mutations in over 20 genes have been shown to cause ASD. To identify disease-causing variants, we performed exome sequencing in 24 South Florida families with ASD. Results We identified 12 likely causative variants in 10 families (42%), including single nucleotide or small insertion-deletion variants in B3GLCT, BMP4, CYP1B1, FOXC1, FOXE3, GJA1, PXDN, and TP63, and a large copy number variant involving PAX6 . Four variants were novel. Each variant was detected only in one family. Conclusions Presence of rare variants in multiple genes leading to ASD explains its genetic basis in over 40% of patients from South Florida. Exome sequencing for ASD allows us to identify causative variants, thus improving our ability to explain the underlying etiology in more families and to assist them with genetic counseling.


Introduction
Anterior segment dysgenesis (ASD) is a heterogeneous group of eye disorders affecting the cornea, iris, lens, zonule, trabecular meshwork, Schlemm canal, and ciliary body. Primary defects in migration or differentiation of the mesenchymal cells may cause ASD, and in turn impede aqueous humor outflow and elevate intraocular pressure [1]. Increased intraocular pressure is a major risk factor for glaucoma [2]. About 50% of individuals with ASD develop glaucoma, which often manifests before the age of 40 years [3,4]. ASD can present with ophthalmic findings only, such as in Peters anomaly and isolated aniridia, or as part of a multisystemic condition, such as Axenfeld-Rieger (AR; MIM 601542, 601090, 601499), Peters plus (MIM 261540), or SHORT (MIM 269880; Stature-Hyperextensibility of joints or hernia or both-Ocular depression-Rieger anomaly-Teething delay) syndromes. Primary congenital glaucoma (PCG) is included in the ASD spectrum due to the presumed abnomal trabecular meshwork and Schlemm canal development [1,2].
In large families with multiple affected members, ASD is usually inherited as an autosomal dominant trait, though autosomal recessive inheritance has been reported [1]. Well-known ASD genes are CYP1B1 (MIM 601771), FOXC1 (MIM 601090), FOXE3 (MIM 601094), PAX6 (MIM 607108), and PITX2 (MIM 601542) [1]. Mutations in PAX6, PITX2, and FOXC1 do not always correlate with specific ASD phenotypes. Patients with AR syndrome and PCG may have FOXC1 mutations. PAX6 mutations can occur in both Peters anomaly and aniridia, and CYP1B1 mutations may be the cause of Peters anomaly and PCG. Phenotype or genotype alone is insufficiently precise to classify or diagnose ASD [2]. In the present study, we have investigated the genetic origin of isolated or syndromic ASD in the diverse population of South Florida.
Affected or unaffected family members were available for the study in 13 families.
Families were recruited through the Bascom Palmer Eye Institute at the University of Miami, Miami, Florida. Probands were consecutive patients seen by an ophthalmologist for clinical diagnosis and management of ASD. Clinical evaluation of all affected individuals by an ophthalmologist included a slit lamp examination and dilated fundus exam. Further imaging and laboratory tests were performed when needed. DNA was extracted from peripheral leukocytes of each proband by standard protocols.

Genetic Screening
We performed exome sequencing (ES) in the Hussman Institute for Human Genomics at the University of Miami. We used Agilent SureSelect Human All Exon 60 Mb V6 for insolution enrichment of coding exons and flanking intronic sequences following the manufacturer's standard protocol (Agilent). A HiSeq 3000 instrument (Illumina) was used for sequencing and Genome Analysis Toolkit software package used for variant calling [5,6]. During the analysis, we focused on specific genes with putative pathogenic variants previously found in individuals with ASD (Additional file 1: Table S1).
We used Enlis genome software (https://www.enlis.com/) for annotation and variant filtering. As recommended, we filtered variants based on minor allele frequency of < 0.005 in gnomAD (www.gnomad.broadinstitute.org) when considering a recessive mode of inheritance and < 0.0005 when considering a dominant mode of inheritance [7,8].
Combined Annotation Dependent Depletion (http://cadd.gs.washington.edu/) [9], MutationTaster (http://www.mutationtaster.org/) [10], and Sorting Intolerant from Tolerant (http://sift.bii.a-star.edu.sg/) [11] in silico analysis tools were used for the pathogenicity prediction. Conservation of the variant was evaluated by using Genomic Evolutionary Rate Profiling (http://mendel.stanford.edu/SidowLab/downloads/gerp/) [12]. We used Copy Number Inference From Exome Reads to detect Copy Number Variants [13,14]. Sanger sequencing was performed to confirm the variants, and when other family members were available only those that showed complete segregation with the phenotype in the entire family were considered pathogenic. We used the American College of Medical Genetics guidelines to interpret variant pathogenicity [15,16].

Results
Based on the clinical evaluations, seven probands were considered to have syndromes associated with ASD [AR, Peters plus, and oculo-dento-digital syndromes (MIM 164200)], and 17 were diagnosed with isolated eye anomalies (Additional file 1: Table S2). On average, each exome had 99.2%, 95.3%, and 89.4% of mappable bases of the Gencode defined exome represented by coverage of 1X, 5X, and 10X reads for ES, respectively. The average read depth was 71.9X and the coverage and average read depth are considered adequate for exome sequencing [17,18]. We have detected nine pathogenic or likely pathogenic variants and three variants of uncertain significance (VUS) that potentially explain the observed phenotypes in 10 probands out of 24 (42%) (Fig. 1, Tables 1 and 2; Additional file 1: Fig S1 and Table S3 show phenotypic features of unsolved probands).

Discussion
In this study, we detected potentially causative variants in 42% of probands with ASD, which is higher than the reported proportion, which ranges from 10-25% [27]. Table 3 summarizes the characteristics of different genetic studies on ASD. Potential explanations for a higher detection rate of causative variants in our cohort are ethnicities studied, differences in case selection, the number of genes analyzed, and sample size. Our cohort consisted of a unique demographic from South Florida, including large Hispanic and Caribbean populations. Earlier studies focused on European, Asian, African, and Middle Eastern populations [27,28]. We did not identify recurrent variants enriched in our cohort; the difference between ethnicities of our cohort and those of earlier studies does not appear to explain our higher detection rate. In our cohort, families with Hispanic ancestry appear to have a higher detection rate (Hispanic vs. Non-Hispanic is 7/12 vs 3/12).
Additionally, the difference between whites and blacks is noticable: only 1 out of 7 black families is solved while 9 of 17 white families studied are found to have potentailly causative variants. Majority of our black families were from the Caribbean, suggesting that the underlying genetic factors of ASD in the Caribbean remain largely unknown. Another important difference between our study and previous studies is the spectrum of ASD being analyzed. We looked at a wide range of ASD conditions, such as Peters anomaly, aniridia, AR syndrome, and PCG. Some other studies focused on a specific phenotype, such as Peters anomaly [23] or primary open-angle glaucoma/primary angle-closure glaucoma [27]. Recognized gene variants for some focused phenotypes are present in smaller portions of affected individauls, which likely contributes to higher detection rate in our study. We used ES to cover all genes previously associated with ASD and some previous studies used gene panels, which may not include all associated genes. While targeted next-generation sequencing gene panels potentially provide higher coverage for individual genes and lower cost, ES as a research tool reduces the need of development and validation of custom panels. Finally, our cohort is smaller in size compared to previous cohorts and may have a higher detection rate by chance.  The second variant shows a change from serine to leucine in position 759. This missense variant is also predicted to make an impact on protein function with a REVEL score of 0.8399 Each parent is heterozygous for one variant suggesting that these two variants are in trans (Fig. 1). Biallelic PXDN variants have been reported with various eye anomalies including microphthalmia, congenital cataracts, microcornea, sclerocornea, and glaucoma [30,31]. Therefore, it is possible that the identified variants are the cause of Peters anomaly in our patient. Similarly, one proband was homozygous for the FOXE3 p.I97M variant, which is a VUS. The allele frequency of this variant on gnomAD is 0.00002015.
Multiple in-silico prediction tools show a damaging effect. This variant has been previously reported in a case with ASD [26]. Therefore we consider the FOXE3 variant a likely cause of the eye phenotype in our proband.    Table S1: List of the genes used for the filtering by using ES in our cohort. Table S2: Syndromic and isolated subjects in both solved and unsolved probands.