HiPSCs were obtained from an RDH12-AD patient described by Sarkar et al. (2020) and an unrelated age- and gender-matched unaffected control (Méjécase et al., 2020; Sarkar et al., 2021). Three clones for each line were expanded, as previously reported (Méjécase et al., 2020; Sarkar et al., 2021). HiPSCs were differentiated into retinal organoids following a feeder-free xeno-free protocol (Reichman et al., 2017). Optic cups were picked between days 21 and 32 and cultured in 6-well plates with 10 ng/mL of human FGF2 for 7 days (Reichman et al., 2017). Each optic cup was selected and cultured in a 96-well plate, with the previously reported protocol to obtain laminated retinal organoids (Gonzalez-Cordero et al., 2017). Retinoic acid (0.5 µM) was maintained until the latest time point (week 44).

Retinal organoids were cryoembedded at two time points (4× week-18 retinal organoids per line obtained from two differentiations and 5× week-44 retinal organoids per clone obtained from up to four differentiations) for immunostaining as previously reported (Reichman et al., 2014). Immunofluorescence staining was performed on 12-µm thick sections. Samples were incubated in a permeabilizing solution (0.5% Triton X-100 and 0.1% Tween 20 with PBS) for 1 h, in a blocking solution (PBS, 0.2% gelatin, and 0.25% Triton X-100) for 1 h at room temperature, and then in primary antibodies solutions diluted in blocking solution (Supplementary Table S1) overnight at 4°C. After washes in PBS with 0.1% Tween 20 solution, sections were incubated with secondary antibodies (Supplementary Table S1) diluted in blocking solution for 1 h at room temperature. Sections were mounted with ProLong Diamond Antifade Mountant with DAPI (Thermo Fisher Scientific). CRX-positive cells with DAPI co-staining were manually counted in a stack of three slides using Fiji to determine the percentage representation. A t-test was performed with p ≤ 0.05 considered statistically significant.

A TUNEL assay was performed using the ApopTag Fluorescein In situ Apoptosis Detection Kit (Millipore) per the manufacturer’s instructions. The entire organoid section (19–28 images per stack) was imaged using an LSM710 confocal microscope at magnification ×40, and a maximum intensity projection Z stack was performed for each image (Lane et al., 2020). TUNEL-positive nuclei were counted manually on Fiji, and a t-test was performed, with p ≤ 0.05 considered statistically significant.

Week 44 retinal organoids (4× week-44 retinal organoids per line) were fixed in 4% paraformaldehyde overnight at 4°C, preserved in an increased sucrose gradient (10%–30%, in PBS1X) overnight for each glucose concentration at 4°C. Retinal organoids were then frozen in isopentane at −50°C. Samples and cryosections were kept at −80°C until use. RNAscope was performed to study RDH12 (1623951-C2), RBP4 (469781), RLBP1 (414221), and ARR3 (486461-C2) mRNA levels, following the manufacturer’s instructions (RNAscope Multiplex Fluorescent Reagent Kit v2), with a target retrieval step for 5 min and protease III steps for 30 min. The entire organoid section was imaged with a Zeiss Invert880 confocal microscope at magnification ×40, and a maximum intensity projection Z stack was performed for each image. Each dot corresponding to one mRNA molecule was counted with a 3D object counter plugin (Bolte and Cordelières, 2006) with Fiji. DAPI nuclei were counted manually to calculate the volume of dots per cell in each image. A t-test was performed with p ≤ 0.05 considered statistically significant.

Two retinal organoids per line (week 37) were fixed with Karnovsky’s fixative (2.5% glutaraldehyde, 1% PFA, 80 mM cacodylate buffer (pH 7.4), and 20 mM NaOH) for 1 h at room temperature in the dark. After washes in 0.2 M sodium cacodylate solution, samples were postfixed in 1% aqueous osmium tetroxide for 1–2 h at room temperature. Serial dehydration with increasing ethanol concentration (50%, 70%, 90%, and 100%) was performed, followed by a 100% propylene oxide bath. Samples were then embedded in 100% EPON resin. The samples were serially sectioned at the thickness of 70 nm using a UC7 ultramicrotome (Leica Microsystems), and sections were picked up on Formvar-coated 2-mm slot copper grids (Gilder Grids). The sections were post-stained with 2% UA and lead citrate, and sections containing the region of interest were imaged using a 120 kV JEM-1400Flash transmission electron microscope with a sCMOS Flash camera (JOEL, United Kingdom). For each organoid per line, three regions 60 µm apart were studied; photoreceptor segments, which could be small outer segment-like structures or inner segments, were counted and measured from the outer limiting membrane (Supplementary Figure S1). As the two structures can be difficult to distinguish due to the section and the angle of the organoid, they were not separated in our measurements. The measurement and counting were performed blind. As no significant difference was observed between the two researchers’ analyses for three blinded regions (Supplementary Figure S1), the remaining analysis (measurement and counting) was performed by only one researcher. A total of 2,097 and 1,238 segments were counted in the unaffected control and RDH12-AD samples. Observations can only be drawn from this as statistical analysis is not appropriate for n = 2 retinal organoids per group.

At week 44 (day 303), 12 replicates (two organoids per replicate) from two differentiations of two RDH12-AD clones and eight replicates from three differentiations from one unaffected control clone were snap frozen. RNA was isolated and quantified, and cDNA was generated from pooled organoids (Qiagen, United Kingdom). cDNA libraries were sequenced on a HiSeq 2500 with paired-end reads of 150 bp using Illumina Stranded Total RNA with Ribo-Zero Plus (Illumina, United Kingdom). Raw reads were quality- and adapter-trimmed using Cutadapt (version 3.4) (Martin, 2011) before alignment. Reads were mapped, and subsequent gene levels were counted using RSEM 1.3.1 (Li and Dewey, 2011) and STAR 2.7.10a (Dobin et al., 2013) against the human genome GRCh38 using annotation release 95, both from Ensembl. Normalization of raw count data and differential expression analysis was performed with the DESeq2 package (version 1.38.3) (Love et al., 2014) within the R programming environment (version 4.2.2) (R Development Core Team, 2008). The pairwise comparison was performed with the contrast function, from which genes differentially expressed (adjusted p-value less than 0.05) between different conditions were determined. Gene lists were used to look for pathways and molecular functions with over-representation analysis using Reactome (Jassal et al., 2020) and Gene Ontology (GO) (Mi et al., 2019). Genes involved in RDH12 pathways were collected from the PathCards database (https://pathcards.genecards.org/, accessed 29 September 2023).

All statistical analyses were performed using R for Statistics (https://www.r-project.org/, v4.4.3).

The RNA-seq data generated by this study have been deposited in the NCBI Gene Expression Omnibus under the access code GSE271751, including unprocessed FASTQ files and associated gene count matrices.

HiPSC derived from an RDH12-AD patient and an unaffected control were differentiated into retinal organoids and characterized after 18 weeks and 44 weeks of differentiation (Figure 1). At week 18, both unaffected and RDH12-AD retinal organoids developed rod and cone photoreceptors, and there was no significant difference in either cell death (p = 0.42) or in the number of photoreceptor cells (CRX-positive cells, p = 0.62, Figure 1A). In both mutant and unaffected control retinal organoids, recoverin was localized in photoreceptor segments, with some recoverin-positive cells detected in the inner nuclear layer, corresponding to bipolar cells. At week 44, there was no significant difference in cell viability (p = 0.13), and both photoreceptor cells grew inner (recoverin-expressing) and outer segment-like structures (with L/M-opsin and rhodopsin expression) (Figure 1B). However, the percentage of photoreceptors (CRX-positive cells) in RDH12-AD retinal organoids was significantly decreased compared to the unaffected control at week 44 (p = 0.04, Figure 1B). RDH12 was detected in the photoreceptor inner segments in both the unaffected control and RDH12-AD retinal organoids at week 18 and week 44.

Transmission electron microscopy of week 37 retinal organoids was undertaken to look for intracellular changes (Figure 2). Both unaffected control and RDH12-AD organoids developed segments around the organoid edges, with mitochondria and a clear, regular outer limiting membrane (Figure 2A). Small rudimentary outer segment-like structures and connecting cilia were observed in both WT and RDH12-AD organoids (Figure 2A). Blinded analyses of three regions per organoid showed unaffected control retinal organoids have an average number of 62 ± 5 photoreceptors per 100 µm whilst RDH12-AD have fewer photoreceptors (33 ± 4 photoreceptors per 100 µm) (Figure 2B). The average photoreceptor segment length was 10.09 ± 6.48 µm in unaffected controls (range between 0.25 µm and 44.02 µm) and reduced to 5.51 ± 3.42 µm in RDH12-AD retinal organoids (range between 0.36 µm and 26.77 µm) (Figure 2C).

Bulk RNA sequencing was performed on mature RDH12 and unaffected control retinal organoids at week 44 to identify disease-causing pathways. In mature RDH12-AD retinal organoids, 2,446 genes were differentially expressed, with 1,424 upregulated and 1,022 downregulated, compared to the control expression (Figure 3A; Supplementary Table S2; Supplementary Figure S2). Gene ontology (GO) analysis identified GO overrepresentation of passive transmembrane transporter activity pathways (GO:0022803, GO:0005249, GO:0005261, GO:0046873, GO:0015103, GO:0015267, GO:0022836, GO:0005267, GO:0022843, GO:0005244, GO:0022832, GO:0015079, GO:0005216, GO:0022824, GO:0005230, GO:0022834, GO:0005231, GO:0005237, GO:0005245, GO:0005254, GO:0016917, GO:0004890, GO:0022835, GO:0015276, GO:0015085, GO:0005262, GO:0005253, GO:0015108, GO:1904315, GO:0099095, GO:0022851, and GO:0008331), neurotransmitter receptor activity (GO:0098960, GO:0099529, and GO:0008503), nucleoside-triphosphatase regulator activity (GO:0060589, GO:0030695, and GO:0005096), photoreceptor activity (GO:0009881 and GO:0008020), lipid phosphatase activity (GO:0042577 and GO:0008195), modified amino acid binding (GO:0072341 and GO:0001786), extracellular matrix structural constituent (GO:0005201 and GO:0030020), glycosaminoglycan binding (GO:0005539), transmembrane receptor protein tyrosine kinase activator activity (GO:0030297), ephrin receptor activity (GO:0005003), cGMP binding (GO:0030553), SNAP receptor activity (GO:0005484), cell–cell adhesion mediator activity (GO:0098632), coreceptor activity (GO:0015026), collagen binding (GO:0005518), actin binding (GO:0003779), cadherin binding (GO:0045296), integrin binding (GO:0005178), 1-phosphatidylinositol binding (GO:0005545), tubulin binding (GO:0015631 and GO:0043015), and microtubule binding (GO:0008017) in RDH12-AD (Figure 3B; Supplementary Table S3).

We examined different retinal cell markers (Kim et al., 2023) and found cone-specific markers were downregulated in RDH12-AD retinal organoids compared to the unaffected control: OPN1LW, TDRG1, GRK7, RAB41, MYL4, PDE6H, CNGB3, PDE6C, GNAT2, GUCA1C, GNGT2, ARR3, SLC24A2, KCNB2, and PEX5L (Supplementary Table S4; Figure 4A). However, only two stress and/or apoptosis markers were differentially expressed: MOAP1 (LFC −0.804, padj < 2.109 × 10⁻⁸) and AIFM2 (LFC 0.639, padj < 3.276 × 10⁻²) (Deng et al., 2018) (Supplementary Table S4).

We compared the expression of genes previously reported to be associated with inherited eye diseases (Supplementary Table S5; Figure 4). Interestingly, 49 genes associated with retinal dystrophies were significantly differentially expressed in RDH12-AD (21%, n = 229), with 38 downregulated and 11 upregulated (Figures 4B–D; Supplementary Table S5). These genes are associated with syndromic retinal disorders (ACBD5, ADAMTS18, ADGRV1, CAPN5, CC2D2A, CDH23, CLRN1, KIF11, LZTFL1, NPHP1, PEX7, POC1B, USH1C, and VCAN), retinitis pigmentosa (ADAMTS18, CC2D2A, CLRN1, CNGA1, DHDDS, IMPDH1, IMPG2, NEK2, PRCD, RDH12, RLBP1, and TULP1), other retinal dystrophies (CAPN5, LRP5, OPN1LW, OPN1MW, PLA2G5, RB1, RBP4, RGS9, RGS9BP, and RS1), Leber congenital amaurosis (GUCY2D, IMPDH1, RD3, RDH12, RPGRIP1, and TULP1), cone dystrophy/cone-rod dystrophy (ADAM9, CNNM4, GUCA1A, GUCY2D, KCNV2, PDE6C, POC1B, RAX2, RIMS1, RPGRIP1, and UNC119), achromatopsia (CNGA3, CNGB3, GNAT2, PDE6C, and PDE6H), congenital stationary night blindness (LRIT3, GRK1, and RLBP1), and macular dystrophy (RAX2). Differential gene expression was also observed for the different groups of ocular diseases (Figure 4D; Supplementary Table S5): 26 genes are associated with ocular malformations (16%, n = 163), with 18 upregulated and eight downregulated; 12 are associated with cataracts (13%, n = 93), with seven upregulated and five downregulated; 10 are associated with anterior segment developmental anomalies including glaucoma (20%, n = 49), with seven upregulated and three downregulated; six associated with microphthalmia, anophthalmia, and coloboma (17%, n = 36), with three upregulated and three downregulated; two are associated with eye movement disorders (15%, n = 13), with one upregulated and one downregulated; two are associated with albinism (13%, n = 15), with one upregulated and one downregulated; and one gene associated with optic atrophy was downregulated (8%, n = 12).

RDH12 expression is downregulated in RDH12-AD retinal organoids compared to the unaffected control (LFC −1.158, padj < 4.633 × 10⁻³), and this was supported by RNA in situ hybridization (Supplementary Table S6; Supplementary Figure S3). RDH12 dysregulation may affect several pathways involving RDH12: the visual cycle in rod and cone photoreceptors, vitamin A and carotenoid metabolism, and retinol biosynthesis and metabolism (Sarkar and Moosajee, 2019). Interestingly, 45 unique genes involved in these pathways were dysregulated in RDH12-AD retinal organoids compared to the unaffected control (Figure 5A; Supplementary Table S6): of 18 genes involved in visual signal transduction in cones (78%, n = 23), 16 are downregulated, and two are upregulated; of 27 genes involved in visual phototransduction (25%, n = 110), 17 are downregulated, and 10 are upregulated; of 22 genes involved in the visual cycle in retinal rod (23%, n = 94), 17 are downregulated, and five are upregulated; of nine genes involved in vitamin A and carotenoid metabolism (21%, n = 42), seven are downregulated, and two are upregulated; two genes involved in retinol metabolism (5%, n = 38) are downregulated; and four genes involved in retinol biosynthesis are downregulated (31%, n = 13). Cone function appeared to be highly affected in RDH12-AD retinal organoids (Supplementary Table S6): GRK7 (LFC −1.974, padj < 2.061 × 10⁻¹⁰), OPN1MW3 (LFC −1.746, padj < 4.244 × 10⁻³), PDE6H (LFC −1.541, padj < 2.014 × 10⁻¹⁰), CNGB3 (LFC −1.514, padj < 4.937 × 10⁻⁵), PDE6C (LFC −1.509, padj < 1.042 × 10⁻¹²), GNAT2 (LFC −1.499, padj < 8.770 × 10⁻⁹), GNGT2 (LFC −1.447, padj < 1.042 × 10⁻⁵), ARR3 (LFC −1.438, padj < 1.528 × 10⁻⁶), SLC24A2 (LFC −1.403, padj < 1.079 × 10⁻³), GNB3 (LFC −1.202, padj < 6.145 × 10⁻⁴), RDH12 (LFC −1.158, padj < 4.633 × 10⁻³), GRK1 (LFC −1.132, padj < 6.192 × 10⁻⁵), RGS9BP (LFC −0.886, padj < 2.274 × 10⁻³), CNGA3 (LFC −0.824, padj < 2.127 × 10⁻²), GUCY2D (LFC −0.788, padj < 2.388 × 10⁻³), RGS9 (LFC −0.589, padj < 3.590 × 10⁻²); and OPN1MW (LFC 10.42, padj < 3.188 × 10⁻³⁹), and OPN1MW2 (LFC 10.658, padj < 2.172 × 10⁻²⁵). RNA in situ hybridization confirmed that RLBP1, a marker of the vitamin A pathway, was downregulated in the RDH12-AD retinal organoids compared to unaffected controls (p = 0.009, Supplementary Figure S3B). A trend toward reduction was seen with RBP4, which is involved in retinol biosynthesis (p = 0.121, Supplementary Figure S3C), and with ARR3, involved in cone phototransduction (p = 0.275, Supplementary Figure S3). The vitamin A pathway was downregulated in RDH12-AD retinal organoids (Figures 5B,C).