Molecular diagnostic challenges for non‐retinal developmental eye disorders in the United Kingdom

Abstract Overall, approximately one‐quarter of patients with genetic eye diseases will receive a molecular diagnosis. Patients with developmental eye disorders face a number of diagnostic challenges including phenotypic heterogeneity with significant asymmetry, coexisting ocular and systemic disease, limited understanding of human eye development and the associated genetic repertoire, and lack of access to next generation sequencing as regarded not to impact on patient outcomes/management with cost implications. Herein, we report our real world experience from a pediatric ocular genetics service over a 12 month period with 72 consecutive patients from 62 families, and that from a cohort of 322 patients undergoing whole genome sequencing (WGS) through the Genomics England 100,000 Genomes Project; encompassing microphthalmia, anophthalmia, ocular coloboma (MAC), anterior segment dysgenesis anomalies (ASDA), primary congenital glaucoma, congenital cataract, infantile nystagmus, and albinism. Overall molecular diagnostic rates reached 24.9% for those recruited to the 100,000 Genomes Project (73/293 families were solved), but up to 33.9% in the clinic setting (20/59 families). WGS was able to improve genetic diagnosis for MAC patients (15.7%), but not for ASDA (15.0%) and congenital cataracts (44.7%). Increased sample sizes and accurate human phenotype ontology (HPO) terms are required to improve diagnostic accuracy. The significant mixed complex ocular phenotypes distort these rates and lead to missed variants if the correct gene panel is not applied. Increased molecular diagnoses will help to explain the genotype–phenotype relationships of these developmental eye disorders. In turn, this will lead to improved integrated care pathways, understanding of disease, and future therapeutic development.


| INTRODUCTION
Childhood visual impairment has a significant emotional, social and economic impact on the individual, their family, and society as a whole. An estimated 1.4 million children are blind worldwide (Gilbert, 2001), and in the United Kingdom, 1 in 2500 children under the age of 1 year are diagnosed as severely sight impaired with an estimated one-third having a genetic basis (Rahi and Cable, 2003).
Environmental factors, such as maternal alcohol intake or in utero infections, may cause some of these conditions (Busby, Dolk, & Armstrong, 2005;Chassaing et al., 2014;Givens, Lee, Jones, & Ilstrup, 1993), therefore a detailed prenatal history should be obtained. If an unremarkable pregnancy is reported a genetic basis should be considered.
MAC contributes up to 15% of childhood blindness and severe visual impairment worldwide (Hornby et al., 2000), with a cumulative incidence of 11.9 per 100,000 children (<16 years of age) in the United Kingdom (Shah et al., 2011). A prospective incidence study found that 2% of cases were due to environmental causes, and despite the assumption that the rest were genetic, only 6% of patients received a molecular diagnosis (Shah et al., 2011). MAC patients display significant phenotypic heterogeneity, forming part of a clinical spectrum and mixed phenotypes can often be seen in individuals, for example right microphthalmia with chorioretinal coloboma and left anophthalmia. Other ocular abnormalities, such as ASDA and cataract, can also be found in MAC patients causing a more complex presentation, with 60% having systemic associations (Richardson, Sowden, Gerth-Kahlert, Moore, & Moosajee, 2017). Over 90 genes linked to MAC have been identified with all forms of inheritance (de novo sporadic, autosomal dominant, autosomal recessive, X-linked dominant, and X-linked recessive), demonstrating genetic heterogeneity (Harding & Moosajee, 2019). In severe bilateral cases of microphthalmia and anophthalmia, a genetic cause can be found in up to 80% (Plaisancie, Calvas, & Chassaing, 2016), with heterozygous loss of function variants involving SOX2 and OTX2, and recessive biallelic changes in STRA6 being the most common (Williamson & FitzPatrick, 2014). However, this represents a small subset of patients with a prevalence of 2-3 per 100,000. For the majority, where asymmetry exists, particularly isolated unilateral cases, the diagnostic rates fall below 10%. Germline mosaicism, nonpenetrance and variable expressivity may be contributing factors (Chassaing et al., 2007;Faivre et al., 2006;Morrison et al., 2002).
The involvement of multiple ocular structures in patients with developmental eye disorders is due to the genes (a significant number being transcription factors) involved in early eye development having a spatiotemporal role in various ocular tissues, thus having a pleiotropic effect if defective. Current genetic testing practice in the United Kingdom for genetically heterogeneous eye conditions utilizes targeted gene panels (e.g., the Oculome; http://www.labs.gosh. nhs.uk/media/764794/oculome_v8.pdf) which encompass known disease-causing genes that cause both nonsyndromic and syndromic forms of disease. There are exceptions, for example, for children born with aniridia, an array-CGH is commonly used to detect a deletion involving the WT1 and PAX6 genes, if negative then Wilms tumor, aniridia, genitourinary anomalies, and mental retardation (WAGR) syndrome, can be ruled out and single gene PCR-based sequencing of PAX6 is undertaken to identify pathogenic variants causing isolated aniridia. It is important to consider that although targeted gene panels, such as the Oculome "Anterior Segment Dysgenesis" panel encompasses related conditions such as ASDA, corneal dystrophies and glaucoma related genes, if a patient also had cataract or coloboma, then the relevant panel may not be selected and the molecular cause missed. Hence accurate phenotypic descriptions using Human Phenotype Ontology (HPO) terms must be given so the clinical scientists can consider the differential genes that may be involved and apply multiple gene panels if necessary. Whole genome sequencing (WGS) remains a research-based test in the United Kingdom but is transitioning to clinical accreditation, however, similar principles will apply to selecting the correct panel of genes to be screened based on phenotype.
In contrast, inherited retinal disorders (IRDs), although considered phenotypically heterogenous, commonly have a symmetrical appearance with an onset after birth through to late adulthood which can be monitored closely with state-of-the-art retinal imaging. Over 250 disease-causing genes have been identified, mainly over the past two decades, and this had led to the first approved retinal gene therapy, voretigene neparovec, for autosomal recessive biallelic RPE65-retinopathy, with a multitude of gene/mutation-based clinical trials underway (Maguire et al., 2019;Miraldi Utz, Coussa, Antaki, & Traboulsi, 2018;Russell et al., 2017). Genetic diagnostic rates in IRDs vary according to the population being tested, but range between 50 and 70% (Audo et al., 2012;Bernardis et al., 2016;Consugar et al., 2015;Ellingford et al., 2016;Jiman et al., 2020;Tayebi et al., 2019). The progress seen in the IRD field is likely due to the consistent scientific investment made internationally. A PubMed search on 16th May 2020 of papers relating to IRDs with search terms "retinal dystrophy" came to 12,953, whereas those for "developmental eye disorders" was 7,689; "anterior segment dysgenesis" 2,402; "microphthalmia" 4,837; "anophthalmia" 1,612; coloboma 5,152; and "congenital cataract" 5,793.
In this study, we report our real-world clinical experience of genetic testing of pediatric patients with developmental eye disorders,

| Genetic screening methods
Only families with nonsyndromic and syndromic microphthalmia, anophthalmia, ocular coloboma (MAC), anterior segment dysgenesis anomalies (ASDA) including primary congenital glaucoma, corneal dystrophies, and aniridia, congenital cataract, infantile nystagmus and albinism were included in this analysis. Consecutive patients presenting to the pediatric ocular genetics service at MEH between 1st October 2017 and 30th September 2018 were investigated. In addition, the cohort of patients with corresponding diagnoses recruited into the UK Genomics England 100,000 Genomes Project (Turnbull et al., 2018)  In total, 69 patients from 59 families (95.2%) proceeded with genetic testing following informed consent, two patients declined, and it was not possible to obtain a sample from one patient. Of the 59 families who consented, 13 had a targeted gene panel (22.0%), 45 had WGS through the 100,000 Genomes Project (76.3%) and one family had single gene test for PAX6. Most families opted for WGS due to the superiority of the test and its coverage despite it being on a research basis. Those that underwent a targeted gene panel did so as they either did not meet the 100,000 Genomes Project study eligibility criteria, did not want to partake in a research study, or were concerned about the length of time to get results for family planning purposes.

| Outcomes for developmental eye disorders through whole genome sequencing
For the disease subgroups the following molecular diagnostic rates were achieved; 15.7% MAC (11/70), 13.0% ASDA (6/46, but 12.8%  (Table 2). WGS seeks to screen both coding and noncoding regions of the gene, but only 8 out of 93 (8.6%) variants were found to be noncoding (all were splice-site mutations); these were found in one cataract patient (HSF4), two nystagmus patients (both FRMD7), and five albinism patients (OCA2 [two patients], GPR143, SLC38A8, and TYR). Noncoding variants were only found in splice regions due to the limitations of the UK Genomics England 100,000 Genomes Project diagnostic pipeline.
As with clinically accredited diagnostic targeted gene panels (e.g., the Oculome), the focus was on the detection of class 4 and 5 variants. So while WGS has the capacity to detect all noncoding variation, only those with a canonical splicing effect or those previously identified and/or functionally proven variants will be regarded as class 4 or 5. For unsolved cases, the discovery of novel noncoding variants is undertaken by further data mining in the research setting. For MAC conditions, the Oculome panel test was able to solve 8.2% of cases but through our clinics and the 100,000 Genomes Project (using WGS), our patients had at least a twofold increase with 21.4% (3/14) and 15.7% (11/70) diagnostic rates, respectively. This confirms that WGS has the capacity to increase diagnostic rates. The most prevalent gene was PRSS56 (OMIM #613858), which causes autosomal recessive microphthalmia, isolated 6 (OMIM #613517), found in two unrelated families (16899 and 25356). All other MAC families had their own individual disease-causing gene, which can make genotype-phenotype correlations hard to determine as many more cases need to be identified to strengthen associations and prognosis. As only 11 out of 70 cases were solved, this suggests that many deep intronic variants may be as yet undetected due to the diagnostic pipeline limitations. This low solve rate also implies that more variants and novel genes remain undiscovered, and/or possible alternate non-Mendelian disease aetiologies, for example epigenetic or complex genetic causes. In a recent paper using the wildtype zebrafish as a model of optic fissure morphogenesis, ocular tissue from the site of the unfused, fusing and post-fusion optic fissure was taken for comparative global transcriptomic profiling against dorsal retina at the same timepoint (Richardson et al., 2019 Missense cohorts such as those recruited to the 100,000 Genomes Project (Caulfield et al., 2019) and the NIHR Bioresource (NBR-RD, 2019).
The use of human induced pluripotent stem cells (hiPSCs) has been manipulated to generate models of human eye development, and could be used to identify more candidate genes by revealing underlying molecular mechanisms (Hung, Khan, Lo, Hewitt, & Wong, 2017;Llonch, Carido, & Ader, 2018). Optic vesicle-like structures derived from a microphthalmia patient with a VSX2 null variant (p.Arg200Gln) showed upregulation of WNT pathway components and misexpression of retinal pigment epithelium (RPE) markers at the expense of the neural retina (NR) lineage, which was rescued by pharmacological inhibition of WNT signaling (Capowski et al., 2016). This supports an important role for VSX2 in WNT signaling and maintenance of the NR through WNT pathway suppression. One potential problem of in vitro modeling is the lack of surrounding embryological tissue and external stimuli that may provide cues for in vivo development, this is particularly important given the involvement of multiple ocular structures in patients. Utilization of data from both cell-based systems and animal models will continue to provide a more complete and accurate representation. can guide the ongoing care and management of the patient. As more genes are identified, it is likely that we will begin to see more extensive spectrums of disease rather than distinct disease entities. Classifications of disease using eponymous names must also be superseded by a gene-based system where we can build our knowledge of genotypephenotype relationships.
Patient access to genetic testing has been a barrier to accurate, timely diagnoses and appropriate management. In countries with insurance-based systems, agencies are reluctant to fund genetic testing unless there is clear evidence the results will accurately determine the clinical status of the patient and directly influence management (Capasso, 2014). In the United Kingdom, the 100,000 Genomes Project was established to provide an infrastructure that allows NHS patients with rare disease to access WGS through a genomic medicine service with centralized funding (Patch & Middleton, 2019;Turnbull et al., 2018). This transition is underway, with designated laboratories undertaking ophthalmic genetic testing and oversight from Genomic Hubs permitting equity of genomics services (Royal College of Ophthalmologists, 2020). The National Genomic Test Directory for rare and inherited diseases has been formed detailing which test is available for each clinical indication, with the Genomics England Panel app listing the genes included in each panel, reviewed by experts to ensure there is evidence for its inclusion (Genomics England Panel App, n.d.). Such advances will accelerate time to diagnosis, improve diagnostic rates, permit precision management and cost saving on prolonged, multidisciplinary team assessments and investigations (Gillespie et al., 2014;Musleh et al., 2016). For patients and families with developmental eye disorders, an accurate molecular diagnosis earlier in the patient pathway potentially provides several benefits by addressing uncertainty, improving decision-making and elucidation of any systemic manifestations, a number of which are can be potentially life threatening. Research into developmental eye disorders will also greatly benefit from larger solved patient datasets.

| CONCLUSION
Understanding the etiology of developmental eye disorders remains challenging given their diverse phenotypes and genetic heterogeneity.
Diagnostic rates remain variable and relatively lower than the progress made with IRDs. NGS technologies allow variants to be screened in parallel and at relatively low cost. Expanding WGS from the research setting to an accredited clinical service will allow for more accurate diagnosis and improved management of patients.
Ensuring precise human phenotype ontology is used to document each clinical feature (not just the primary diagnosis) will enable clinical scientists to best apply the relevant diagnostic gene panel. Being able to gain a molecular diagnosis will further our understanding of the natural history of gene/variant-specific cohorts, reveal potential therapeutic targets and establish outcome measures for prospective future treatments.