Evaluation for Retinal Therapy for RPE65 Variation Assessed in hiPSC Retinal Pigment Epithelial Cells

Human induced pluripotent stem cells (hiPSCs) generated from patients and the derivative retinal cells enable the investigation of pathological and novel variants in relevant cell populations. Biallelic pathogenic variants in RPE65 cause early-onset severe retinal dystrophy (EOSRD) or Leber congenital amaurosis (LCA). Increasingly, regulatory-approved in vivo RPE65 retinal gene replacement therapy is available for patients with these clinical features, but only if they have biallelic pathological variants and sufficient viable retinal cells. In our cohort of patients, we identified siblings with early-onset severe retinal degeneration where genomic studies revealed compound heterozygous variants in RPE65, one a known pathogenic missense variant and the other a novel synonymous variant of uncertain significance. The synonymous variant was suspected to affect RNA splicing. Since RPE65 is very poorly expressed in all tissues except the retinal pigment epithelium (RPE), we generated hiPSC-derived RPE cells from the parental carrier of the synonymous variant. Sequencing of RNA obtained from hiPSC-RPE cells demonstrated heterozygous skipping of RPE65 exon 2 and the introduction of a premature stop codon in the mRNA. Minigene studies confirmed the splicing aberration. Results from this study led to reclassification of the synonymous variant to a pathogenic variant, providing the affected patients with access to RPE65 gene replacement therapy.


Introduction
Leber congenital amaurosis (LCA) and other early-onset severe retinal dystrophies (EOSRD) have a prevalence of approximately 1 : 50,000 worldwide [1]. Symptoms may include nystagmus, photophobia, and the oculodigital sign of eye rubbing or poking, and when these are present in the first year of life, a diagnosis of LCA may be suspected. Ophthalmic investigations demonstrate severely diminished or often undetectable full-field electoretinogram (ffERG) traces, indicating severe photoreceptor pathology [1]. There are currently more than 24 disease genes associated with LCA and other forms of EOSRD and a molecular diagnosis may be obtained in approximately 70-80% of families [2]. In view of the phenotypic and genetic heterogeneity of LCA and EOSRD, the relatively recent mainstream adoption of genomic technologies in diagnostic laboratories has enabled improved access for families to clinical molecular testing and diagnosis.
The RPE65 gene encodes the retinal-specific isomerohydrolase enzyme, a 533 amino acid, and 61 KDa protein. RPE65 is involved in the visual cycle recovery phase and the conversion of 11-trans retinyl ester to 11-cis retinal after pigmentary excitation from light [3]. Autosomal recessive homozygous and compound heterozygous disease-causing variants in RPE65 have been well described and account for~16% of individuals with LCA or EOSRD, and variants occur less frequently, at~2%, in individuals with retinitis pigmentosa (OMIM: #18600) [4,5]. Autosomal recessive LCA and EOSRD-associated RPE65 variants reported to date include missense, frameshift, premature stop, in-frame deletion, and splicing variants [6]. To date, there are no clear genotype-phenotype correlations reported, with the severity or the onset of disease presentation seemingly independent to the type or location of the causative variants detected [7].
There is a new era emerging in ophthalmology and precision medicine with the first ocular gene therapy receiving regulatory approval in several jurisdictions, namely, the RPE65 gene replacement therapy voretigene neparvovecrzyl [8]. To receive this, therapy patients must have clinical features of retinal dystrophy, viable retinal cells, and biallelic pathological RPE65 gene mutations (https://www.tga.gov.au/ apm-summary/luxturna).
Diagnostic genomics is rapidly expanding the ability to efficiently interrogate the human genome, and standardised approaches to variant classification have been adopted widely by diagnostic laboratories [9,10]. However, sometimes a variant is classified as a variant of uncertain significance (VUS), even if the clinical phenotype is in agreement with the disease. This lack of confirmed pathological diagnosis makes these patients ineligible to receive RPE65 gene therapy. Functional genomic studies regarding the deleterious effects of a VUS on RNA or protein function are currently considered strong evidence of pathogenicity; however, these studies are often hampered by other technical complexities such as the disease gene being exclusively expressed in clinically unobtainable tissue types.
Advances in stem cell methods and technologies have led to the application of patient-derived hiPSCs to generate ret-inal pigmented epithelial cells (hiPSC-RPE) and retinal organoids (hiPSC-RO) to model retinal disease mechanisms and therapies directly in target human retinal cells [11,12]. In this study, we generated a hiPSC-RPE model to examine the effects of a novel genomic VUS in two siblings from a family presenting with autosomal recessive EOSRD. The hiPSC-RPE model facilitated access to RPE65 mRNA to investigate the molecular consequence of this cryptic genomic variant and determine its pathological impact.

Subjects and Methods
2.1. Human Participants. All human subjects in this study underwent written informed consent. This study was approved by the Human Research Ethics Committee (HREC) of the Sydney Children's Hospital Network, Sydney, Australia, and adhered to the tenets governed by the Declaration of Helsinki.

Ophthalmic
Investigations. Detailed ophthalmic review was undertaken and included fundal examination and imaging, ffERG performed according to International Society for Clinical Electrophysiology of Vision (ISCEV) to examine retinal function, wide-field fundus autofluorescence (WF-FAF), and spectral domain optical coherence tomography (SD-OCT) assessment of the retinal structures.

Genomics and Bioinformatics.
The proband II-1 (Figure 1(k)) underwent commercial targeted capture retinal dystrophy gene panel testing (Molecular Vision Laboratories, USA), examining 280 retinal disease genes. Variant confirmation and segregation studies of candidate variants were undertaken using Sanger sequencing performed by Sydney Genome Diagnostics (Sydney Children's Hospital Network-Westmead, Australia), with PCR products sequenced at the Australian Genome Research Facility (Westmead, Australia). Variants were classified in accordance with established ACMG guidelines [9,10].

Stem Cells International
Authenticated Cell Cultures (ECACC)) by the pluripotent stem cell core facility, StemCore (University of Queensland, Australian Institute for Bioengineering and Nanotechnology, Brisbane, Australia).

Characterisation of hiPSC Lines: In Vitro Trilineage
Differentiation and Assessment of Pluripotency. To characterise pluripotency, embryoid bodies (EBs) were formed by transferring iPSC aggregates to nontissue culture-treated dishes containing E8 medium and 10 μM Y-27632 (Sigma). The next day, the medium was replaced with spontaneous differentiation medium: KnockOut Dulbecco's Modified Eagle's Medium, 20% KnockOut serum replacement, 1% nonessential amino acids, 1x penicillin-streptomycin, 1% glutamax, and 0.1 mM β-mercaptoethanol (Life Technologies). On day 7, EBs were plated onto Geltrex-coated wells and cultured a further 7 days in the spontaneous differenti-ation medium prior to cell harvest for RT-qPCR and immunofluorescence assessment of pluripotency.
To determine gene expression levels, we first extracted total RNA using the RNeasy micro kit (Qiagen, USA). RNA was then converted to cDNA using the SuperScript IV 1 st -Strand Synthesis System kit (Invitrogen, USA). Relative gene expression was determined using 1x SensiMix SYBR (Bioline) and the Rotor-Gene 6000 Cycler system (Qiagen). Relative gene expression was calculated by 2 -ΔCt analysis of housekeeping genes POL2A and HPRT against the lineage markers: AFP, EN1, PAX6, CDH20, PHOX2B, FOXF1, and HAND2. Statistical analysis was performed by unpaired t-test. Primers were designed using PrimerBlast and synthesised by Sigma (see Supplementary  Table S1).
For immunofluorescence assessment of pluripotency, we used the StemLight™ iPS Cell Reprogramming Antibody Kit   . Primers were designed using Primer3 and had a forward primer specific for the 5 ′ UTR of RPE65, with reverse primers specific to exon 4 and exon 5, respectively. PCR was performed using standard cycling conditions. Primers were synthesised by Sigma (see Supplementary Table S1). PCR products were resolved by agarose gel electrophoresis, with resultant bands excised and purified using the Wizard® SV Gel and PCR Clean-Up System (Promega, USA) before Sanger sequencing (Australian Genome Research Facility, Westmead NSW, Australia). An orthogonal in vitro assay to confirm potential aberrant splicing resulting from the novel RPE65 genomic variant was undertaken using an ExonTrap vector minigene system (MoBiTech, GmbH, Germany) (see Supplementary File, Methods).
2.9. Western Blotting. Total protein was extracted from hiPSC-RPE cells using RIPA buffer (ThermoFisher) and 1x protease inhibitor (Roche). Protein concentration of cell lysates was determined using the Direct Detect Assay-free cards (MerkMillipore, Australia), and 30 μg of protein from each lysate was resolved by 4-20% Tris-glycine SDS-PAGE (ThermoFisher, Australia). Protein was transferred onto a nitrocellulose membrane, which was blocked with 5% skim milk in 1x TBS and then incubated with rabbit anti-RPE65 (PA5-78414, ThermoFisher, USA) and mouse anti-β-actin (Sigma) antibodies, followed by fluorescent dye conjugated secondary antibodies (anti-rabbit IRDye 800CW and antimouse -IRDye 680RD, Millennium Science, Australia). Fluorescence detection was made using the ChemiDoc MP Imaging System (Bio-Rad, Australia), and acquired images were exported to Image Lab 6.0 software for quantification of RPE65 65 KDa protein band intensities normalised to βactin levels (loading control).

Ophthalmic Clinical Presentations.
The two affected siblings in this study presented with nyctalopia in early childhood. Patient II-1 (Figure 1(k)), a female, presented at 4 years of age when her parents recognised her difficulty with night vision and adaptation to changing lighting conditions. Systemic assessment noted normal developmental milestones, normal hearing, normal bone development, and no family history of visual problems. Her best corrected visual acuity was 6/6 bilaterally. Her fundal appearance showed discrete RPE mottling in the midperiphery illustrating outer retinal atrophy and small white intraretinal spots in the perimacular region (Figures 1(a) and 1(b)). The international society for clinical visual electrophysiology (ISCEV) standard ffERG performed at 5 years of age showed undetectable scotopic responses in each eye, and the photopic responses were defined but reduced in each eye consistent with a rod-cone dystrophy (Figure 1(j) top row). A diagnosis of 5 Stem Cells International early-onset severe retinal dystrophy (EOSRD) was made. Repeat ffERG at 12 years of age identified further reduction in rod-cone function. A pattern electroretinogram (pERG) showed a normal response to the 15-degree field and reduced response to the 30-degree field in the right eye and reduced responses to both the 15-and 30-degree fields in the left eye. Wide-field fundus photographs attained at age 12 showed widespread distribution of retinal atrophy with associated white fleck-like lesions in the peripheral fundus. The WF-FAF was almost absent. SD-OCT assessment showed preservation of the ellipsoid zone and outer-retinal structures. (Figure 1(c)).
Patient II-2, a female, was noticed to have difficulty with night vision from the age of 2 years. There were no systemic medical associations. At age 5 years, she underwent a ffERG which showed an undetectable scotopic ERG and residual severely attenuated photopic ERG, consistent with an early onset severe retinal dystrophy in the (EOSRD) (Figure 1(j) middle row). Wide-field pseudocolor fundus imaging attained at age 5 showed widespread distribution of retinal atrophy with associated white fleck-like lesions in the peripheral fundus (Figure 1(d)). The WF-FAF was virtually absent (Figure 1(e)). SD-OCT showed preservation of the ellipsoid zone across the macula with slight macular thinning (Figure 1(f)).
Parental ophthalmic assessments including electrophysiology testing were entirely normal including pERG and ffERG WF-FAF and SD-OCT.

Genomics, Segregation Analysis, Bioinformatics, and
Review of the Clinical Phenotype. To establish the genotypic mutation of the EOSRD diagnosis, targeted capture gene panel testing (Molecular Vision Laboratories, USA) was undertaken. Analysis of the proband II-1 was unable to identify a clear cut clinically significant genetic cause. Interestingly, the results showed two variants in the RPE65 gene (NM_000329.2):c.93A>G;271C>T p.(Thr31Thr);(Arg91Trp) (Figure 1(k)). The presence of these individual variants was confirmed using Sanger sequencing in the proband and her affected sister. Segregation studies were consistent with autosomal recessive inheritance, with one of each of the variants present in each of the parents (Figure 1(k)). The RPE65:c.271C>T p.(Arg91Trp) variant was classified as pathogenic as it has been reported as disease causing in both compound heterozygous and homozygous states in patients with autosomal recessive LCA [14], EOSRD [15], and RP [16]. Functional studies using isomerohydrolase and protein stability assays of this variant showed a significant decrease in RPE65 enzymatic activity and protein level [17,18]. However, the second heterozygous variant RPE65:c.93A>G p.(Thr31Thr) was a synonymous change which was novel and absent from population databases including gnomAD (version 2.1.1). We noted that this RPE65:c.93A>G variant affected the penultimate nucleotide to the canonical donor splice site of exon 2. Predictive in silico splicing tools using Alamut Visual v2.13 (Interactive-Biosoftware, France) and SpliceAI [19] indicated reduction in the splicing efficiency of the natural canonical splice donor site of exon 2. These findings prompted review of the retinal phenotype in both siblings, and it was noted that peripheral retinal white flecks and reduced fundus autofluorescence would be consistent with RPE65-associated retinopathy, providing supporting evidence to further investigate the functional impact of the RPE65:c.93A>G p.(Thr31Thr) synonymous variant [20,21].

Characterisation of hiPSC Lines and Differentiation to hiPSC-RPE.
To investigate the predicted splicing defect, we generated two hiPSC lines, one from a healthy control, and the other from the heterozygous parental carrier of the novel RPE65:c.93A>G variant. Immunocytochemistry of both hiPSC lines demonstrated presence of OCT4, NANOG, and SOX2/SSEA-4 pluripotency markers (Figures 2(a) and  2(b)). Pluripotency was also demonstrated by increased transcript expression of genes for each lineage (ectoderm, endoderm, and mesoderm) after 2 weeks of in vitro trilineage differentiation (Figures 2(c) and 2(d)). Digital karyotyping using SNP chromosome microarray confirmed that both control and c.93A>G carrier hiPSCs had normal karyotypes (46,XY and 46,XX, respectively) ( Figure 2(e)).
We next sought to produce retinal-specific tissue and perform functional evaluation by differentiating both hiPSC lines into RPE cells. Monolayers of day 90-120 hiPSC-RPE cells showed clear pigmentation and typical cobblestone-like morphology characteristic of human RPE cells (Figures 3(a)-3(c)). ZO-1 staining identified tight junctions expressed throughout hiPSC-RPE monolayers (Figure 3(d)) together with the expression of RPE markers CRX and MITF (Figures 3(e) and 3(f)). Gene expression analysis of other RPE-specific markers BEST1, MERTK, MITF, PAX6, PMEL17, and RPE65 showed increased expression in both the control and carrier hiPSC-RPE cells (Figures 3(g) and 3(h)).

RNA Studies of Synonymous Variant in hiPSC-RPE Cells
Identify Exon Skipping Leading to Reduced RPE65 Protein Expression. We next used the differentiated hiPSC-RPE cells to test whether the RPE65:c.93A>G variant was interfering with donor splice site activity at the exon 2-intron 2 boundary. The removal of exon 2 from the NM_000329.2 RPE65 transcript is predicted to result in the introduction of a premature termination codon resulting in a mutant RPE65 allele of only 22 amino acids in length (Figure 4(a)). RNA from hiPSC-RPE cells from the control and the carrier parent was extracted and RT-PCR performed. To ensure the PCR amplicons generated encompassed the splice site predicted to be affected, we designed RPE65 5 ′ UTR primers and reverse primers specific to exons 4 and 5 (Figure 4(b), denoted as A and B, respectively). Wild-type and mutant bands were visible on agarose gel electrophoresis (Figure 4(b)), and cDNA sequencing of these products showed skipping of exon 2 in the mutant transcript (Figure 4(c)).
Further confirmation of exon 2 skipping of the novel synonymous variant was investigated using a minigene vector-based approach (see Supplementary File, Figure S1).
To investigate whether the aberrant splicing and predicted premature termination codon resulted in differences in RPE65 protein levels, we examined hiPSC-RPE cell lysates   Variant details were submitted to the ClinVar database with accession details: SCV001905505.

Discussion
In this study, we demonstrate the value of parental stem cell-derived retinal tissue for genetic variant classification. The siblings studied here had a severe retinal dystrophy phenotype and review of genotypic findings and segregation studies showed compound heterozygous variants in RPE65, with one classified as a pathogenic variant and the other a novel synonymous VUS, RPE65:c.93A>G p.(Thr31Thr). Ocular phenotypic review indicated features including peripheral retinal white flecks and reduced fundus autofluorescence which could be consistent with RPE65-related disease. Bioinformatics analysis of the synonymous variant suggested a splicing abnormality, requiring functional studies of RNA expression to confirm this. As RPE65 expression is limited to the clinically inaccessible RPE cells in the eye, we generated a parental-derived hiPSC line which in turn was differentiated to generate hiPSC-RPE cells carrying the synonymous variant. We demonstrated the skipping of exon 2 and change in reading frame in the RPE65 coding sequence, which resulted in a predicted severely truncated mutant allele of only 22 amino acids. Confirming the predicted protein truncation, examination of the parental carrier hiPSC-RPE cells showed a reduction in RPE65 protein expression. These findings were validated by an orthogonal ExonTrap minigene assay also demonstrating skipping of exon 2. The RPE65:c.93A>G p.(Thr31Thr) variant is a novel synonymous variant in RPE65 we have shown to be of pathological significance. These results indicate the presence of biallelic pathological variants in these patients, enabling eligibility for RPE65 gene replacement therapy.
The RPE65 gene encodes the catalytic enzyme retinal isomerase, which plays a critical role in maintaining the presence of 11-cis retinol in the visual cycle [3]. There are >120 disease-causing RPE65 variants reported in public online databases which have been associated with photoreceptor dysfunction and structural abnormalities involving the retina {https://www.ncbi.nlm.nih.gov/clinvar/RPE65; https://databases.lovd.nl/shared/genes/RPE65}. The majority of reported RPE65 disease-causing alleles are missense variants (49%), followed by truncating variants (36%), splice site variants (12%), and in-frame deletions (3%) [4]. Investigation using hiPSC technology and derivative tissues such as RPE cells could delineate the molecular aetiology for other known RPE65 variants. Interestingly, to date, there are no disease-associated splicing variants described resulting in the skipping of exon 2. There is, however, a recent report of a disease-associated nucleotide substitution implicating the final coding base of exon 2, namely RPE65:c.94G>T p.(Gly32Cys) [7,22]. From the current literature, the deleterious effect of this particular variant is unclear as to whether it is due to the amino acid substitution (which has a codon across the splice site junction), or is reflective of a defect in natural canonical splicing. Given that both the c.93A>G described in this study and the c.94G>T variant are near the 5 ′ donor splice site, it is likely that the c.94G>T variant may also influence the efficiency of natural canonical splicing at this location. Studies such as the one described here would help clarify this question. Future experiments using our hiPSC-RPE model could also be effective in measuring retinoid isomerohydrolase activity to demonstrate the effect of a 50% reduction in functional RPE65 protein, compared to a wild-type control.
Previous work involving another retinal-specific disease gene, BEST1, highlighted the value of patient-derived   [23]. These results suggested there may be differences in gene splicing of transcripts between RPE and non-RPE cells, as has been identified for some transcripts in photoreceptor cells compared with nonretinal tissues [24]. To ensure that splice assay results in this project were specific to RPE cells, we proceeded with hiPSC production and RPE cell derivation from a sample available from the carrier parent in this fam-ily. This demonstrated a splicing abnormality, which was also able to be confirmed in our minigene assay undertaken in a HEK293 model system.  9 Stem Cells International genomic and precision medical advancement. A key requirement for access to this therapy is monogenic disease with pathological, biallelic mutations as determined by diagnostic genomic testing [8]. Diagnostic laboratories currently follow strict guidelines for classifying genomic variants [9,10].
Using this systematic approach, curation of novel variants is often challenging, especially with patients needing to meet strict inclusion criteria for RPE65 gene augmentation therapy. The valuable contributions of modelling variants in iPSC-derived retinal cells is evident in emerging studies of 10 Stem Cells International frequently described IRD disease genes such as RPGR [11] and USH2A [25]. The hiPSC-derived cells in this study were critical for determining the pathological significance of a synonymous variant in RPE65 and clearly demonstrated the value of a stem cell-based approach.

Conclusions
We generated a carrier parent-derived hiPSC-RPE model to establish the molecular pathogenesis of a novel synonymous genomic variant RPE65:c.93A>G p.(Thr31Thr), originally classified as a VUS on diagnostic genomic testing. RNAbased studies confirmed an aberration in gene splicing and generation of a loss-of-function allele with reduction of RPE65 protein expression. This study highlights the power of hiPSC technologies to establish the pathogenicity of genomic variants in disease genes with tissue-specific expression and showcases the collaborative opportunities between diagnostic laboratories and stem cell medical research to ensure positive outcomes for patients in the era of gene therapies.

Data Availability
The data supporting the findings of this study are available upon request to the corresponding author. The genomic variant has also been submitted to the ClinVar repository.

Conflicts of Interest
R Jamieson and J Grigg have consulted for Novartis. All the other authors have no conflicts of interest to declare.