2.1 Clinical assessment

History and full ophthalmic examination including best-corrected visual acuity using the Early Treatment of Diabetic Retinopathy Study (ETDRS) letter chart and dilated fundus examination was undertaken. Multimodal retinal imaging included ultra-wide-field color fundus photography, green-light autofluorescence (AF) imaging (P200Tx and California, Optos plc, Dunfermline, UK), 30° and 55° scanning laser ophthalmoscopy including near-infrared reflectance (NIR), blue-light and near-infrared AF and spectral domain optical coherence tomography (SD-OCT, Spectralis OCT2 and Spectralis HRA2, Heidelberg Engineering, Heidelberg, Germany).

2.2 Genetic analysis and interpretation

Genetic analysis utilized genomic DNA extracted from peripheral blood samples collected and stored as detailed previously (Paterson et al., 2012). Analysis of proband genomic DNA targeted known variants in 28 genes associated with autosomal recessive retinitis pigmentosa (ARRP) using the Apex ARRP Array (version 5.3; performed by Asper Ophthalmics, Tartu, ESTONIA) coupled with haplotype analysis to exclude non-candidate genes (performed by the Australian Genome Research Facility (AGRF), Brisbane, Australia), as described previously (Paterson et al., 2012). Bidirectional Sanger sequencing of coding and flanking intronic regions of candidate genes CRB1, PDE6A, and RPE65 was performed on DNA of the proband and an affected sibling to detect novel candidate variants in these genes. The phase of CRB1 variants was established in parental DNA, and their presence in two siblings was ascertained by bidirectional Sanger sequencing (performed by ASPER Ophthalmics or AGRF). Variant nomenclature is in accordance with recommendations of the Human Genome Variation Society (den Dunnen et al., 2016unnen et al., 2016), with sequencing results aligned to the CRB1 reference sequence NM_201253.2 (OMIM#604210). Nucleotide 1 corresponds to the A of the ATG initiation codon.

Variant pathogenicity assessment encompassed in silico pathogenicity predictions, allele frequencies sourced from gnomAD (Karczewski et al., 2020), data from variation databases (LOVD/ClinVar/dbSNP) and the scientific literature, and the clinical diagnosis. In silico pathogenicity assessment utilized Mutation Taster (Schwarz, Rodelsperger, Schuelke, & Seelow, 2010), SIFT (Ng & Henikoff, 2001), PolyPhen2 (Adzhubei et al., 2010), Align GVGD (Tavtigian, Byrnes, Goldgar, & Thomas, 2008), REVEL (Ioannidis et al., 2016), M-CAP (Jagadeesh et al., 2016) and the Alamut Visual splicing prediction module (Version 2.11; Interactive Biosoftware). Pathogenicity interpretation was performed in accordance with ACMGG/AMP guidelines (Richards et al., 2015). Additionally, in silico splicing analyses were performed using Human Splicing Finder, Version 3.1 (Desmet et al., 2009) and EX-SKIP (Raponi et al., 2011).

2.3 Cell culture and differentiation

Patient iPSC (LEIi006-A) and the normal control iPSC line (Cat#A18945, Thermo Fisher Scientific) were cultured in feeder-free conditions, on geltrex (Thermo Fisher) coated culture plates in mTeSR1 medium (StemCell Technologies), as previously described (Zhang et al., 2018). The mutations found in peripheral blood DNA were also confirmed in these patient-derived iPSC lines. For directed differentiation into neural retinal organoids (NRO) we followed a previously published protocol (Mellough et al., 2015, 2019). Patient and control iPSC colonies were dissociated into small pieces by EDTA buffer (0.5 mM EDTA and 30 mM NaCl in DPBS (-Ca⁺⁺, -Mg⁺⁺) and then cultured in the DMEM/F-12 medium (Thermo Fisher) with decreasing concentrations of KnockOut Serum Replacement (KOSR, Thermo Fisher) (20% for the first 5 days, 15% until day 12, 10% until day 35) containing MEM Non-Essential Amino Acids Solution (NEAA; Thermo Fisher), B27 (Thermo Fisher) and IGF-1 (StemCell).

2.4 Transcript analysis

Total mRNA was isolated from early (day 35) NRO using TRIZOL and cDNA was synthesized using the RT² First Strand Kit (Qiagen, Hilden, Germany) according to the manufacturer's instructions. Partial cDNA sequences were amplified with Q5 High-Fidelity Polymerase (New England Biolabs, Ipswich, Massachusetts, USA) using primers targeting exons 3–7 or exons 7–8. Primer sequences are listed in Table S1. PCR products were resolved on a 2% agarose gel. The 50 bp DNA Ladder (New England BioLabs) was used as a molecular size marker. Gels were imaged using the ChemiDoc XRS+imaging system (BioRad). After purification with the Gel Purification and PCR Clean-up System (Promega), PCR fragments were assessed by Sanger sequencing (AGRF).

2.5 Immunostaining analysis

Retinal organoids were fixed with 4% paraformaldehyde, washed, then permeabilized using phosphate-buffered saline (PBS) with 0.1% Triton X-100 for 15 min. The samples were then incubated in 5% BSA in PBS for 1 hr at room temperature. Primary antibodies were applied at 4°C overnight. Secondary antibodies were applied for 2 hours at room temperature. Nuclei were stained with DAPI. The following primary antibodies were used in this study: Mouse anti-Pax6 (1:100, Abcam, AB78545), Rabbit anti-Recoverin (1:200, Merck, AB5585). Secondary antibodies used in this study included Alexa Fluor 546 Goat anti-mouse IgG (1:500, Molecular Probes, A-11003), and Alexa Fluor 546 Goat anti-rabbit IgG (1:500, Molecular Probes, A-11035).

2.6 Transmission electron microscopy (TEM) analysis

Retinal organoids were fixed in 4% paraformaldehyde, 2% glutaraldehyde in 0.1 M phosphate buffer, pH 7.4. After 24 hours, they were post-fixed in 1% osmium tetroxide in the same buffer for 6 minutes (2 minutes on, 2 minutes off and 2 minutes on) with biowave (PELCO). Samples were then dehydrated in a graded series of ethanol and acetone followed by embedding in resin containing Procure 812, Araldite 502, DDSA and BDMA. Ultrathin (100 nm) sections were prepared and examined with a JEM2100 electron microscope (JEOL).

2.7 Quantitative PCR

Total mRNA was isolated from NRO cultures using TRIZOL and cDNA was synthesized using the RT² First Strand Kit (Qiagen). Quantitative PCR (qPCR) was performed using RT² SYBR Green qPCR Mastermix (Qiagen) and the CFX Connect Real-Time System (BioRad). Data were analyzed using the ΔΔCT method. Gene expression values were normalized to expression levels measured in LEIi005-A iPSC. Sequences for primers used in this study are listed in Table S1.

3.1 Clinical presentation

The patient was a 50-year-old female initially diagnosed with rod-cone dystrophy and macular cysts at the age of 21. Her presenting Snellen visual acuity was 20/30 in both eyes. Fundus examination in the early stages of the disease showed bilateral cystoid maculopathy with numerous white dots in the periphery resembling fundus albipunctatus. Over the 30 years of follow-up, her vision declined to perception of light in both eyes and the majority of the retina was replaced by heavy intraretinal pigmentation sparing the para-arteriolar regions in the periphery (Figure 1a-b). There was gross thickening of the retina due to disorganized retinal layering (Figure 1c). She also developed non-visually significant posterior subcapsular cataract at age 46.

The proband's brother presented at age 18 with a visual acuity of 20/40 in both eyes and peripheral retinal schisis that led to an initial clinical diagnosis of Goldman-Favre syndrome. He had bilateral cataract extraction at the age of 36, and 9 years later, his vision was reduced to perception of light. He also had heavy intraretinal pigmentation with para-arteriolar sparing (Figure 1d-e) and thickened retina (Figure 1f). Neither the proband nor the affected sibling manifested variable vascular abnormalities or retinal telangiectasia with exudation. The proband's parents underwent similar clinical examinations, with no retinal abnormalities detected (data not shown).

3.2 Genetic analysis and pathogenicity assessment

Proband DNA was analyzed by Apex ARRP Array, coupled with haplotype analysis to further identify candidate genes. Subsequent Sanger sequencing of PDE6A and RPE65 excluded novel candidate variants in these genes. Two variants in CRB1 were identified: c.1892A>G and c.2548G>A. Segregation analysis confirmed the paternal (former) and maternal (latter) origin of these variants and their segregation with disease (Figure 2a). These missense variants had predicted amino acid substitutions p.(Tyr631Cys) and p.(Gly850Ser), respectively, expected to affect conserved laminin A/G-like domains (Figure 2b).

In silico pathogenicity analysis provided conflicting results for c.1892A>G: despite aberrant splicing predictions, Mutation Taster returned a benign prediction, as did SIFT, PolyPhen 2 and REVEL. AlignGVGD and M-CAP returned a possibly pathogenic result, and EX-SKIP predicted the mutant allele was more likely to cause exon skipping than the wild-type allele. Analysis of this variant using Human Splicing Finder (HSF, version 3.1) predicted potential alteration of splicing due to the introduction of an exonic splice silencer (ESS) site and a cryptic donor splice site (DSS) in exon 6 (Figure S1). Comparison of normal and c.1892A>G CRB1 splice site sequences in exon 6 revealed the cryptic DSS present in the mutant sequence had a lower HSF score (67.55) than the natural DSS (94.75), suggesting it may not be a strong competitor for the natural DSS. The strength of the Fas-ESS hexamer sequence generated by the c.1892A>G variant could not be assessed by the HSF software, but may promote skipping of exon 6 during pre-mRNA splicing. Interrogation of variation databases and the literature indicated that this variant is rare (allele frequency = 0.000008), and is reported as causative for autosomal recessive retinal dystrophy or retinitis pigmentosa (Hettinga, van Genderen, Wieringa, Ossewaarde-van Norel, & de Boer, 2016; Talib et al., 2017). Recent reports have additionally provided in vitro evidence of mislocalized CRB1 protein and misplaced photoreceptors in patient-derived retinal organoids subsequent to this variant occurring in trans with a nonsense CRB1 variant (Quinn et al., 2019). The scientific literature, thus, suggests a pathogenic status for this variant.

In contrast, the c.2548G>A variant was predicted to be pathogenic by all in silico algorithms utilized. Despite some predictions of altered splicing motifs, EX-SKIP predicted the mutant allele was unlikely to induce exon skipping. HSF predicted potential alterations of splicing due to the disruption of an SF2/ASF exonic splicing enhancer (ESE) motif in exon 7 (Figure S2). However, the impact of the predicted loss of this ESE element is unclear. This rare variant (allele frequency = 0.000025) has been reported as causative for ARRP with or without preservation of para-arteriolar RPE, and LCA8 (Clark et al., 2010; den Hollander et al., 2004; Henderson et al., 2011). Taken together, we considered both of these variants likely candidates for disease (Table S2), although the effects of these variants at the mRNA or protein level had not been demonstrated.

3.3 Differentiation of photoreceptor progenitor cells from patient and control iPSCs

Given the in silico predictions of pathogenicity, we hypothesized that one or both variants might lead to alternative splicing of the CRB1 gene, leading to a deficit in functional CRB1 protein expression. To test this hypothesis, we generated neural retinal organoids (NRO) from control iPSCs and the LEIi006-A iPSC line previously generated from proband dermal fibroblasts (Zhang et al., 2018). After 35 days of differentiation, both control (Figure 3a-d) and patient (Figure 3e-h) iPSCs produced laminated NRO resembling early optic cups. Immunostaining demonstrated the localization of retinal ganglion cells (RGCs) and amacrine cells expressing PAX6 in the center of the organoid surrounded by a dense layer of neural retinal progenitor cells (Figure 3b-c and f-g). Clusters of immature photoreceptor progenitor cells (PPCs) expressing recoverin were identified in the developing retinal progenitor cell layer (Figure 3d and h). Ultrastructural examination of NRO by transmission electron microscopy identified photoreceptor inner segments (IS) on the surface of both control (Figure 3i,k) and patient (Figure 3j,l) NRO. Junctional complexes were observed in the subapical region of PPCs from both genotypes (Figure 3i-l). Expression of PAX6, RCVRN, and CRB1 in NRO was confirmed by qPCR analysis of day 35 NRO cDNA, using adult human retinal cDNA as a positive control (Figure 3m). All three retinal markers were upregulated in NRO and human retina compared with undifferentiated iPSC. PAX6 was expressed at similar levels in NRO and adult human retina, while RCVRN expression was lower in NRO than in adult human retina, consistent with the low numbers of PPCs observed at this early time point. CRB1 mRNA expression was increased in NRO compared with human retina. Patient NRO expressed moderately lower levels of PAX6, RCVRN, and CRB1 than control NRO, likely reflecting a lower efficiency of NRO induction from the patient iPSC line. Together, these results demonstrate the successful generation of early-stage NRO containing an inner RGC/amacrine cell layer and an outer retinal progenitor cell layer containing clusters of immature PPCs.

3.4 The CRB1 c.1892A>G variant leads to exon skipping

To examine splicing of the CRB1 gene in patient and control PPCs, we amplified partial cDNA sequences from early NRO using primers targeting exons 3-7. We detected a 0.5 kb band in patient, but not control, NRO that corresponded to a novel CRB1 transcript lacking exon 6 (Figure 4a). Sanger sequencing of this fragment demonstrated the splicing of exon 5 to exon 7, producing an in-frame deletion in the CRB1 coding sequence. If translated, the transcript would produce a protein with a 319-amino acid internal deletion, p.(Gly391_Arg709del). Since full-length CRB1 cDNA sequence was not detected in this experiment due to the large size of the full-length product (1.7 kb), we performed additional PCR screening using primers targeting exon 6 and the exon 6-7 junction to determine whether CRB1 transcripts containing exon 6 were produced from the paternal c.1892A>G allele. A single band amplified from proband NRO displayed a single “A” peak at position c.1892, indicating the product was derived from mRNA transcribed from the maternal c.2548G>A allele (Figure S3). These results indicate that the c.1892A>G variant results in the genesis of an ESS, which leads to exon 6 skipping during pre-mRNA splicing. Based on these results, we propose a new ACMGG/AMP classification (Pathogenic (ia)) for this variant (Table S2).

3.5 The CRB1 c.2548G>A variant is a missense mutation

Using primers targeting exons 7-8, we detected CRB1 transcripts in both patient and control NRO. Purification and sequencing of the patient NRO PCR product revealed a heterozygous c.2548G>A signal demonstrating the presence of CRB1 mRNA transcribed from both alleles (Figure 4b). These results indicate that CRB1 mRNA transcripts carrying the c.2548G>A missense mutation are produced in patient retinal cells. If translated, these transcripts are predicted to produce mutant CRB1 proteins carrying the p.Gly850Ser amino acid substitution.