Summary
Correct identification of the molecular consequences of pathogenic genetic variants is essential to the development of allele-specific therapies. However, such molecular effects may remain ambiguous following genetic sequence analysis alone. Here, we identify exonic codon-altering variants that are also predicted to disrupt normal RNA splicing in the context of inherited retinal disease. NR2E3 c.932G>A (p.Arg311Gln) is a variant commonly associated with enhanced S cone syndrome. Previous studies using mutagenized cDNA constructs have shown that the arginine to glutamine substitution at position 311 of NR2E3 does not meaningfully diminish function of the rod-specific transcription factor. Using retinal organoids, we explored the molecular consequences of NR2E3 c.932G>A when expressed endogenously during human rod photoreceptor cell development. Retinal organoids carrying the NR2E3 c.932G>A allele expressed a transcript containing a 186-nucleotide deletion of exon 6 within the ligand binding domain. This short transcript was not detected in control organoids or control human donor retina samples. A minigene containing exons 5 and 6 of NR2E3 showed sufficiency of the c.932G>A variant to cause the observed splicing defect. These results support the hypothesis that the pathogenic NR2E3 c.932G>A variant leads to photoreceptor disease by causing a splice defect and not through an amino acid substitution as previously supposed. They also explain the relatively mild effect of Arg311Gln on NR2E3 function in vitro. We also used in silico prediction tools to show that similar changes are likely to affect other inherited retinal disease variants in genes such as CEP290, ABCA4, and BEST1.
Using patient-derived retinal organoids, minigene assays, and in silico prediction, this study shows that nonsynonymous exonic variants can act through splice alteration in inherited retinal disease. The NR2E3 variant c.932G>A (R311Q) causes aberrant splicing when expressed in retinal cells, contrary to previous reports of this variant’s mechanism.
Introduction
Enhanced S cone syndrome (ESCS) (OMIM: 268100) is an inherited retinal disease characterized by developmental malformation of the outer neural retina.1,2,3,4,5 Classic features of this disease include night blindness and increased sensitivity to blue light.1 Loss of function of the rod-specific transcription factor nuclear receptor subfamily 2, group E, member 3 (NR2E3, HGNC: 7974) is the most common cause of ESCS.1 While for many patients ESCS is a relatively stationary disorder,6 NR2E3 variants can cause a more severe form of retinal degeneration known as Goldmann-Favre syndrome. Interestingly, a clear genotype-phenotype correlation has yet to be identified.7
More than 80 pathogenic variants in NR2E3 have been described to date.8 One of the more common ESCS-causing mutations in NR2E3 is a coding sequence variant at position 932 (NM_014249.4:c.932G>A), which converts codon 311 from CGG to CAG.9 This variant has been previously thought to exert its pathogenic effect through a resulting arginine to glutamine residue change at this position (p.Arg311Gln). Residue 311 is found within the ligand-binding domain of NR2E3, further supporting the potential for protein dysfunction following normal amino acid sequence disruption. However, conflicting evidence from in vitro experiments and in silico modeling exists for the p.Arg311Gln substitution being sufficient for loss of NR2E3 protein function.10,11,12 Notably, expression of mutangenized NR2E3 cDNAs to generate protein containing the Arg311Gln change have not shown markedly diminished NR2E3 behavior in terms of co-factor binding or the ability to act at known regulatory elements.10,11,12
To evaluate the pathobiology of genetic variants such as p.Arg311Gln, the use of disease-relevant tissue types that natively express the gene of interest are more likely to provide meaningful results. Unlike the skin, which can readily be biopsied for direct evaluation, surgical sampling of the neural retina would result in a loss of visual function in the sampled area and, as such, is not something that is done unless absolutely necessary (e.g., to identify a pathogen that would otherwise destroy the retina). Patient induced pluripotent stem cell (iPSC)-derived retinal organoids, which follow normal retinal developmental programs, express retina-specific genes,9,13,14 and replicate retina-specific isoform expression,9,15,16 are the ideal alternative to primary retinal tissue for evaluation of the molecular consequences of pathogenic variants. In this study, we investigated the molecular significance of the NR2E3 c.932G>A (p.R311Q) variant using retinal organoids. We found that, in human retinal cells, c.932G>A creates an alternative 3′ splice site within exon 6. The resulting transcript lacks 186 bases of the 5′ end of exon 6, presumably causing an in-frame deletion of 62 amino acids from NR2E3’s ligand-binding domain. We showed that this change was predictable based on in silico models of variant effect. Organoids carrying the NR2E3 c.932G>A variant are comparable to organoids carrying other pathogenic, loss-of-function NR2E3 variants in terms of NR2E3 expression, and cell type composition.
In a cohort of 1,000 inherited retinal disease families, only 4 other single nucleotide variants currently annotated as missense variants reach the same prediction confidence threshold for splice site alteration. Using cultured fibroblasts from a patient carrying one such variant in the gene encoding centrosomal protein 290 (CEP290, HGNC:29021; c.1711G>A (p.Gly571Arg)), we confirmed that this variant leads to the expression of a longer isoform containing a frame-shifting intronic inclusion. Together, these results shed new light on the pathogenic mechanism of a common pathogenic NR2E3 variant and highlight a potential gap in the annotation of exonic splice variants in inherited retinal disease genes.
Material and methods
Approval and participant consent
This study was approved by the Institutional Review Board of the University of Iowa (project approval 200202022) and adhered to the tenets set forth in the Declaration of Helsinki. Individuals provided written, informed consent.
In silico variant effect prediction
Twenty-one NR2E3 variants were retrieved from ClinVar (https://www.ncbi.nlm.nih.gov/clinvar/, accessed December 2023). Only variants classified as having a clinical significance of “pathogenic” or “likely pathogenic” and a molecular consequence of “missense” were used. SpliceAI (Version 1.3.1) was run against the hg19 human reference genome assembly with default parameters. For AlphaMissense analysis, pathogenicity scores for variants of interest were retrieved from the genome-wide results previously generated by Google DeepMind (AlphaMissense_aa_substitutions.tsv, accessed December 2023). Results from these analyses were processed and plotted using R (version 4.3.0).
Patient-derived iPSCs
This study was approved by the Institutional Review Board of the University of Iowa (project approval #200202022) and adhered to the tenets set forth in the Declaration of Helsinki. Patient iPSCs were generated previously from two individuals with molecularly confirmed ESCS as described.17
Retinal organoid differentiation
Retinal differentiation was performed as described previously with minor modifications.17,18 Briefly, iPSCs were cultured on laminin 521 coated plates in E8 medium. Embryoid bodies (EBs) were lifted with ReLeSR (STEMCELL Technologies) and transitioned from E8 to neural induction medium (NIM; DMEM/F12 [1:1], 1% N2 supplement, 1% non-essential amino acids, 1% Glutamax [Thermo Fisher Scientific], 2 μg/mL heparin [Sigma-Aldrich], and Primocin [InvivoGen]) over a 4-day period. On day 6, NIM was supplemented with 1.5 nM BMP4 (R&D Systems). On day 7, EBs were adhered to Matrigel coated plates (Corning). BMP4 was gradually transitioned out of the NIM over 7 days. On day 16, the media was changed to retinal differentiation medium (RDM; DMEM/F12 [3:1], 2% B27 supplement [Thermo Fisher Scientific], 1% non-essential amino acids, 1% Glutamax. and 0.2% Primocin). On days 25–30, the entire EB outgrowth was mechanically lifted using a cell scraper and transferred to ultra-low attachment flasks in 3D-RDM (RDM plus 10% fetal bovine serum); Thermo Fisher Scientific), 100 μM taurine (Sigma-Aldrich), 1:1,000 chemically defined lipid concentrate (Thermo Fisher Scientific), and 1 μM all-trans retinoic acid (until day 100; Sigma-Aldrich). The cells were fed three times per week with 3D-RDM until the point of RNA isolation.
Retinal organoid immunohistochemistry
Fixation, sectioning, staining, and imaging of retinal organoids was performed at day 160 of differentiation, as described previously.5 Briefly, organoids were fixed with 4% paraformaldehyde, cryopreserved, and 15 μm sections were cut. Sections stained overnight with primary antibodies (OXT2; R&D AF1979, RCVRN; Millipore AB5585). Fluorophore conjugated secondary antibodies (Thermo Fisher Scientific) were incubated on sections for 1 h, and cell nuclei were counterstained using DAPI (Thermo Fisher Scientific, catalog 62248). Images were acquired using a Leica TCS SPE upright confocal microscope system (Leica Microsystems).
RNA isolation and cDNA synthesis
RNA was isolated using the NucleoSpin RNA Kit (Machery-Nagel) following the manufacturer’s protocol. In addition to the on-column DNaseI treatment, an additional DNA digestion was performed with RQ1 RNase-Free DNase following the manufacturer’s protocol for RT-PCR (Promega). cDNA synthesis was carried out using the Superscript IV VILO system (Invitrogen) following the manufacturer’s protocol. RNA was stored at −80°C in water and cDNA was stored at −20°C.
RT-PCR
Custom oligos were designed using the PrimerQuest online tool (Integrated DNA Technologies) and synthesized by Integrated DNA Technologies. Oligos were resuspended to a concentration of 100 μM in water and stored at −20°C. The PCRs were carried out using BIOLASE DNA polymerase (Bioline) with 35 cycles, an annealing temperature of 52°C for NR2E3 and 50°C for CEP290, and a 30-s extension at 72°C. PCR products were electrophoresed on 2% agarose gels with SYBR Safe stain (Invitrogen).
RT-PCR amplicon sequencing
Bands of interest were cut from the gels, and the amplicons were purified using the QIAquick Gel Extraction kit following manufacturer’s protocol (Qiagen). Sanger sequencing was performed by Functional Biosciences. The resulting chromatograms were visualized using SnapGene (version 6.1.2).
Long-read gDNA amplicon sequencing
The entire ∼8-kb NR2E3 locus was amplified using NR2E3 F and NR2E3 R primers (Table 1) with the Expand Long Template PCR System (Roche) buffer system 1. PCR was performed for 30 cycles with an annealing temperature of 58°C and an extension of 11 min with extension time increasing 5 s per cycle for the last 20 cycles, following the manufacturer’s protocol. PCR products were prepared for long-read sequencing using the Native Barcoding Kit 24 V14 (Oxford Nanopore Technologies) following the manufacturer’s protocol. Briefly, sample index barcodes and sequencing adapters were added to amplicons via ligation and libraries were pooled for sequencing on an R10.4.1 flow cell on the MinION Mk1B instrument (Oxford Nanopore Technologies). Base calling and demultiplexing was performed with Guppy (version 6.3.8) using the dna_r10.4.1_e8.2_260bps_sup.cfg configuration and reads were mapped to the hg38 reference genome with Minimap2 (version 2.24).19 Aligned reads were visualized with the Integrative Genomics Viewer desktop application (version 2.16.2).
Table 1.
Primers used for RT-PCR, dPCR, and minigene construction
Short name | Name | Sequence (5′–3′) |
---|---|---|
– | IVS4_NdeI_Fwd | CATATGCAGCAGGGCTTGGCTTC |
– | IVS6_BamHI_Rev | GGATCCGCTGCAGACATGGACCT |
F2 | HBB_Exon1_Fwd | AGGAGAAGTCTGCCGTTACT |
F1 | NR2E3_Exon5_Fwd | ATTGATGTCACCAGCAATGAC |
R1 | NR2E3_Exon6_Rev | CCTTCATGCAGGCAAACTC |
F3 | HBB_Exon2_Fwd | GTCCACTCCTGATGCTGTTATG |
R2 | HBB_Exon3_Rev | GAGCGAGCTTAGTGATACTTGTG |
F4 | CEP290_Exon17_Fwd | CTTTGATTGTGCTTCACTCAT |
R3 | CEP290_Exon18_Rev | CGTCAAATGGCTCAAGAAAG |
18S F | 18S_Fwd | CGGCTACCACATCCAAGGAAG |
18S R | 28S_Rev | GCTGGAATTACCGCGGCTGCT |
NR2E3 F | NR2E3_gDNA_Fwd | ATGGTGAACAGCTGAGCACA |
NR2E3 R | NR2E3_gDNA_Rev | TGGGATGCCATCTGGAGAGA |
Long-read cDNA sequencing
Long-read sequencing of NR2E3 cDNA was carried out using the MinION sequencing platform (Oxford Nanopore Technologies). cDNA was generated using the Chromium Next GEM Single Cell 3′ Kit v3.1 (10X Genomics) from dissociated retinal organoids as described previously.5 Libraries were generated from this cDNA using the PCR-cDNA Sequencing Kit (SQK-PCS111) and sequenced with an R9.4.1 flow cell on the MinION Mk1B instrument (Oxford Nanopore Technologies). Reads were processed with Guppy and Minimap2 as described above.
Minigene construction
The region containing exon 6 of NR2E3 (i.e., IVS5-IVS6) was amplified from genomic DNA isolated from patient 1 (c.932G>A (p.R311Q)/c.1192-A>C) using Q5 High-Fidelity DNA Polymerase (New England Biolabs) with 35 cycles, an annealing temperature of 71°C, and a 1-min extension at 72°C. Primers containing NdeI and BamHI restriction sites were used (Table 1). The resulting amplicon was cloned into the pCR2.1-Blunt-II backbone using the Zero Blunt TOPO PCR Cloning Kit following manufacturer’s protocol (Invitrogen). Clones were screened by Sanger sequencing to identify clones containing either G or A at c.932. Plasmids from selected clones were digested with NdeI and BamHI (New England Biolabs) and the resulting insert gel purified as above and ligated into the modified pcAT7-Glo1 vector (a gift from Brenton Graveley, Addgene Plasmid # 160996) prepared with NdeI and BglII digest and gel purification. Minipreps of the resulting clones were prepared using the NucleoSpin Plasmid kit following the manufacturer’s protocol (Machery Nagel) and confirmed by diagnostic EcoRI digest and Sanger sequencing across the insert.
Minigene transfection
HEK293T cells (ATCC) were cultured in MEMα (Gibco) with 10% fetal bovine serum (Atlas Biologicals) and 100 μg/mL Primocin (InvivoGen). Cells were transfected at ∼60% confluence with Lipofectamine 3000 following the manufacturer’s protocol. We used 500 ng of minigene construct per well of a 24-well plate. As a transfection control, a copGFP-expressing plasmid was used. At 48 h after transfection, total RNA was isolated using the NucleoSpin RNA kit (Machery Nagel) as described above.
Single-cell gene expression dataset acquisition
Previously published single-cell gene expression data available on GEO was accessed for reanalysis.5,20 Specifically, from GSE266526, sample GSM8249205 (c.932G>A (p.R311Q)/c.1192-A>C, patient 1 from this paper) was used. From GSE236197, samples GSM8181640 (c.932G>A (p.R311Q)/c.219G>C (p.R73S), patient 2 from this paper), GSM8181650 (control 1 from this paper), GSM8181651 (c.1192-A>C/c.1192-A>C), and GSM8181652 (c.1192-A>C/corrected) were used.
Single-cell gene expression data analysis
Reads were mapped to the human reference genome build hg38 (GRCh38-2020-A reference assembled by 10X Genomics) and quantified using Cell Ranger v7.0.0 (10X Genomics). Cell Ranger was run using default settings including counting intron mapped reads. Cells with between 2,000 and 6,000 detected features (genes) and fewer than 10% mitochondrial reads were included in downstream analysis. Counts were normalized using the LogNormalize method in Seurat (v4.3) with a scale factor of 10,000. The 2,000 gene with greatest variability were identified using the vst method. Integration anchors were identified by canonical correlation analysis using 25 dimensions and integration was performed yielding an adjusted count matrix for the 2,000 most variable genes. This matrix was scaled and used for PCA. PCA dimensions 1–25 were then used as input for UMAP and cluster identification using the Louvain algorithm. Finally, cell types were annotated as previously described5 using reference mapping from a previously published human fetal retina scRNAseq dataset13 and manual refinement. Plots were generated using scCustomize.21
Results
Predicted molecular consequences of genetic variants in NR2E3
RNA splicing is initiated via conserved motifs that define splice junctions. Models built on these patterns offer a predictive tool for splice site prediction.22 Similarly, predictive models of the consequences of amino acid changes on protein structure and function have been constructed.23 We used SpliceAI and AlphaMissense to ask whether the NR2E3 c.932G>A (p.Arg311Gln) variant and other pathogenic missense variants within NR2E3 would be predicted to cause changes in RNA splicing or protein function (Figure 1). Twenty-one missense variants reported in ClinVar to be either pathogenic or likely pathogenic were used in this analysis. For each variant, the AlphaMissense score and SpliceAI Delta Scores were plotted. NR2E3 p.R311Q returned the lowest AlphaMissense score of this collection (0.093) as well as the highest Delta Score (for Acceptor Gain) from SpliceAI (0.69). No other pathogenic missense variant in NR2E3 was predicted as strongly to alter splicing.
Figure 1.
Missense variant effect prediction in NR2E3
For 21 missense variants classified as pathogenic or likely pathogenic, AlphaMissense and SpliceAI prediction scores are shown. Variants are shown based on position with the NR2E3 coding sequence. Each variant is represented by a vertical gray line and colored points along the line represent the magnitude of prediction strength for missense or splicing impact. The p.R311Q variant is labeled and exhibits a higher score for splice acceptor gain (blue circle) and a lower score for missense impact (red star).
Generation of retinal organoids from NR2E3 patient iPSC lines
To express different NR2E3 variants including the c.932G>A variant in their native cellular contexts, we generated retinal organoids from patient-derived iPSC lines. NR2E3 variant presence did not grossly disrupt maturation of organoids by 160 days of differentiation (Figures 2A–2C). RNA from these organoids was used to further investigate the expression and processing of NR2E3 transcript.
Figure 2.
Retinal organoids with pathogenic NR2E3 variants
(A) Differentiation D160 retinal organoids from the control iPSC line demonstrate expression of OTX2 and RCVRN indicating formation of photoreceptor cells in the outer layer.
(B and C) Representative organoids from the NR2E3 c.932G>A (R311Q) patient lines P1 and P2 also demonstrate proper differentiation and generate photoreceptor cells marked with OTX2 and RCVRN. Scale bar, 50 μm.
Truncated NR2E3 transcript expression in c.932G>A organoids
We next tested the effects of c.932G>A on NR2E3 splicing directly using patient-derived retinal cells carrying this variant. cDNA derived from ESCS patient retinal organoids was sequenced using a long reads and splice junctions were counted (Figures 3A and 3B). The patient carried a previously described compound heterozygous NR2E3 genotype composed of variants c.219G>C and c.932G>A (Figure 3).17 Out of a total of 58, 28 reads exhibited a splicing between the 3′ end of exon 5 and the middle of exon 6 at the location of the c.932G>A variant (Figure 3B). No other aberrant splicing was observed, implying that c.932G>A creates an alternative splice acceptor within exon 6.
Figure 3.
Retinal organoid expression of a mis-spliced transcript near the c.932G>A variant
(A) Canonical NR2E3 gene structure including 8 exons. The positions of two pathogenic variants in exons 2 and 6 (c.219G>C and c.932G>A, respectively) are shown.
(B) Sashimi plot representing long-read RNA sequencing reads mapping to NR2E3. The number of reads spanning exons is shown. Sixty-seven reads span from exon 6 to exon 7. Thirty reads span from exon 5 to the canonical splice acceptor at the 5′ end of exon 6. Twenty-eight reads span from exon 5 to the presumptive novel splice acceptor in exon 6.
(C) PCR primer design to span the truncated portion of exon 6 in mis-spliced transcripts. Amplicon originating from normal transcript is expected to be 391 bp, while that from mis-spliced transcript is expected to be 206 bp.
(D) RT-PCR using F1 and R1 primers shown in (C). In control organoids (Ctl), only full-length amplicon is observed (black arrowhead). In organoids from two unrelated patients with compound heterozygous NR2E3 genotypes (P1, P2), both full length and truncated amplicon (red arrowhead) is observed. Two unrelated human donor retina samples (D1, D2) express only properly spliced NR2E3 transcript.
(E) RT-PCR for 18S rRNA serves as a loading control.
(F) Sanger sequencing of the shorter amplicon from P1 and P2 (shown boxed in D) shows consistent mis-splicing of NR2E3 exon 5 into the predicted alternative splice acceptor within exon 6.
To further confirm allele-specific mis-splicing of the NR2E3 transcript, primers spanning exons 5 and 6 of NR2E3 were designed and used for RT-PCR of RNA from D80-D100 retinal organoids (Figure 3C; see Table 2 for sample information). Two lines from unrelated ESCS patients were used, as well as an unrelated control organoid samples carrying two intact NR2E3 alleles. RT-PCR from both patient lines yielded two amplicons; one of the expected size for canonically spliced NR2E3 and a smaller amplicon showing the expected size following splicing using an exonic splice site created at c.932G>A (Figures 3D and 3E). To confirm the identity of these amplicons, the smaller amplicon from each sample was isolated and sequenced. Both patient lines showed the same splice junction of the 3′ end of exon 5 being spliced to an alternative acceptor within exon 6 at the expected position based on RNA sequencing data (Figure 3F).
Table 2.
Cell lines used
Short name | Sample | Differentiation time point | NR2E3 genotype | CEP290 genotype |
---|---|---|---|---|
P1 | Patient 1 | D83 | c.932G>A (p.R311Q)/c.1192-A>C | N/A |
P2 | Patient 2 | D105 | c.932G>A (p.R311Q)/c.219G>C (p.R73S) | N/A |
P3 | Patient 3 | N/A | N/A | c.1711G>A (p.Gly571Arg)/c.7048C>T (p.Gln2350∗) |
Ctl | Control 1 | D161 | N/A | N/A |
D1 | Donor 1 | N/A | N/A | N/A |
D2 | Donor 2 | N/A | N/A | N/A |
D3 | Donor 3 | N/A | N/A | N/A |
iPS | iPSC control | N/A | N/A | N/A |
The presence of normal exonic and intronic genomic DNA sequence in the two patients was confirmed by long-read sequencing of an amplicon spanning the entire NR2E3 locus (Figures S1A and S1B). Both patients exhibit compound heterozygous alleles. Other unique benign variants were observed in both samples. No structural variants (such as deletion involving exon 6) were observed, further supporting c.932G>A as being causal for transcript mis-splicing.
Minigene validation of c.932G>A sufficiency for exonic splice acceptor creation
To test the sufficiency of the c.932G>A variant for causing mis-splicing between exons 5 and 6 of NR2E3, we next constructed a minigene plasmid to express these exons in a non-retinal cell line. The sequence containing these exons (i.e., spanning NR2E3 introns 4–6) was amplified and cloned into the pcAT7-Glo124 vector between human hemoglobin subunit beta (HBB) exons 1 and 2 (Figure 4A). The resulting construct drives constitutive expression under the chimeric CAG promoter.24,25 Constructs containing both the reference (G) and alternate (A) base at c.932 were generated. Minigene constructs were transfected into HEK293T cells and RT-PCR using intron-spanning primers was used to amplify spliced template (Figure 4B) across various junctions. As in the organoid samples, a truncated cDNA was observed in cells transfected with the c.932G variant-containing minigene (Figures 4C and 4D), but not in the wildtype condition. No endogenous NR2E3 expression in HEK293T cells was detected, consistent with previous reports of the retinal specificity of the gene. Notably, a small amount of normally spliced NR2E3 fragment was observed in cells transfected with the variant-containing construct (Figures 4C and 4D, black arrowhead). Normal splicing of HBB exons 2 and 3 was observed, indicating normal splicing of the expressed pre-mRNA in HEK293T cells (Figure 4E). We used 18S as a loading control (Figure 4F). Together, the results of this minigene experiment show that that NR2E3 c.932G>A is sufficient to cause mis-splicing in a minimal in vitro system. They also demonstrate that such mis-splicing is not photoreceptor cell specific.
Figure 4.
An NR2E3 minigene construct shows sufficiency of c.932G>A to cause mis-splicing
(A) Plasmid design for testing splicing of NR2E3 exons 5 and 6 containing the c.932G>A variant. The NR2E3 locus was cloned into the pcAT7-Glo1 expression construct between HBB exons 1 and 2 and expressed in HEK 293T cells.
(B) Design of RT-PCR primers to assess splicing of minigene transcript. Splicing between HBB exons 2 and 3 serve as a control of normal splicing in this experimental system.
(C and D) Amplicons from samples transfected with the mutant (A) minigenes were 185 base pairs smaller than those from wildtype (G) constructs. Normal amplicons shown with black arrowheads, truncated amplicons shown with red arrowheads. Of note, a small amount of normal-length amplicon is seen in both mutant conditions (black arrowheads, “A” lanes).
(E) Normal splicing of HBB exons 2 and 3 is observed in both conditions.
(F) Expression of 18S RNA serves as a loading control.
Expression of NR2E3 c.932G>A in human retinal organoid-derived photoreceptors
We next compared the gene expression and composition of retinal organoids containing the splice-altering NR2E3 c.932G>A variant with stage-matched organoids containing other pathogenic NR2E3 variants or wildtype NR2E3 allele. We aggregated publicly available single-cell gene expression data from differentiation day 160 (D160) retinal organoids derived from patient lines carrying the NR2E3 c.932G>A variant as well as other NR2E3 alleles5,20 (Figures S2A and S2B). Expression of NR2E3 RNA within rod photoreceptors was comparable at D160 between several cell lines carrying different combinations of NR2E3 allele (Figure 5A). As expected, rod photoreceptors from all lines expressed high levels of NRL (Figure 5B), the rod-specific transcription factor genetically upstream of NR2E3. Finally, we asked whether different NR2E3 variants would shift the production of rod and cone photoreceptors in D160 organoids. We have previously shown that, while NR2E3 variants are known to cause an increase in S cone number in patients with ESCS, loss of NR2E3 function does not cause loss of rod photoreceptor specification in organoids of human retina samples.5 We find that organoids expressing the NR2E3 c.932G>A variant generate similar proportions of rod photoreceptors to control lines and a line carrying homozygous null IVS2-1 variants (Figure 5C).
Figure 5.
NR2E3 c.932G>A does not dimmish RNA abundance or impair cell type emergence in D160 human retinal organoids
(A) Normalized expression of NR2E3 within D160 organoid-derived rod photoreceptors as measured by single-cell RNA sequencing is shown. Organoids were derived from five lines containing different combinations of NR2E3 alleles. Lines expressing the NR2E3 c.932G>A allele (shown in blue and green) do not exhibit decreased NR2E3 expression compared with other variants or control lines.
(B) Normalized expression of NRL in organoid-derived rod photoreceptors is shown. All lines produce rod photoreceptors that express high levels of the rod-specific factor NRL.
(C) The proportion of major retinal cell types from D160 organoids is shown. Cells were classified into type based on expression profile measured by scRNAseq. NR2E3 mutant lines expressing the c.932G>A variant produce all major cell types.
Predicting splicing consequences of inherited retinal disease missense variants
In a recent cohort of 1,000 inherited retinal disease families,26 767 disease-causing alleles were identified. Of these, 529 (69%) involved a single nucleotide change. A high proportion of genes that cause inherited retinal disease are expressed only in the retina9 or are spliced in a tissue-specific manner.16 Following our observation that NR2E3 c.932G>A acts through an unexpected pathogenic mechanism, we next asked how common such a finding might be within this inherited retinal disease cohort. A total of 529 single-nucleotide variants were analyzed using SpliceAI and delta scores for gain or loss of splice sites in the alternate allele were returned (Figure 6A). Only 9 of the 529 variants (1.7%) had a delta score at least as high as that of NR2E3 c.932G>A. 14/529 variants (2.6%) had a score of at least 0.5, a threshold for confident cryptic splice site identification using SpliceAI.22 Of the top nine variants, five had the highest delta score for donor gain, three for donor loss, and one for acceptor gain (NR2E3 c.932G>A). The variants were distributed among seven genes, with three found in ABCA4 and one found in each of six additional genes: BEST1, CEP290, NPHP1, NR2E3, RHO, and USH2A. Non-synonymous, non-truncating variants from the set of 14 with a SpliceAI delta score of greater than 0.5 were next analyzed with AlphaMissense (Figure 6B) to determine the predicted impact of the annotated amino acid substitution. Four of the nine variants were predicted to be likely benign as missense variants, while the other five were ambiguous. None of the nine variants were predicted to be likely pathogenic as missense.
Figure 6.
Splice variant prediction in an inherited retinal disease cohort
(A) We analyzed 529 variants with SpliceAI, returning 4 delta scores per variant. Variants are shown in order of rank of maximum delta score. Nine variants (including NR2E3 c.932G>A) had a score at least as high as NR2E3 c.932G>A. Five variants had a score lower than NR2E3 c.932G>A but higher than 0.5, a common threshold for SpliceAI prediction. Most variants (515) had maximum delta scores lower than 0.5.
(B) Only non-synonymous and non-truncating variants with a maximum delta score of at least 0.5 from A are shown (9/14 variants from A). The gene symbol and predicted amino acid change is shown for each variant. On the x axis, the AlphaMissense score is shown.
(C) RT-PCR was performed between CEP290 exons 17 and 18 to assess mis-splicing caused by the c.1711G>A (p.Gly521Arg) variant, as predicted by the analysis in (B). A larger product consistent with intron inclusion (red arrowhead) was observed in a compound heterozygous patient line (P3) but not a control line (iPS).
To further validate the above in silico predictions, we performed RT-PCR using RNA from a compound heterozygous patient fibroblast line carrying the CEP290 c.1711G>A (p.Gly571Arg) variant (Figure 6C). Of the collection analyzed by SpliceAI and AlphaMissense, this variant was most highly predicted to alter splicing and was predicted to be likely benign as an amino acid substitution (Figure 6B). Indeed, dermal fibroblasts from a patient carrying this CEP290 variant expressed a longer isoform of the transcript, presumably due to partial intron inclusion between exons 17 and 18. The CEP290 c.1711G>A variant is predicted by SpliceAI to abolish the canonical splice donor of exon 17 and lead to preferential usage of a cryptic donor within the following intronic sequence. Together, this causes a frame-shifting intronic inclusion, as observed by RT-PCR (Figure 6C).
Discussion
Many variants that have been reported to cause inherited retinal disease are located within genes expressed exclusively in the retina. As it is difficult to obtain affected retinal tissue, the molecular consequences of pathogenic variants are often hard to evaluate. For this reason, the protein-level consequence of a genetic variant is often predicted based on codon disruption or known splice sites. Here, we show that in vitro models of inherited retinal disease such as patient iPSC-derived retinal organoids can be used to investigate variant effects and demonstrate that presumed missense variants actually cause mis-splicing of transcript when expressed.
Previous studies of the NR2E3 c.932G>A (p.R311Q) variant have largely utilized mutangenized wildtype cDNA to compare variants using cell culture-based systems.10,11,12 While these strategies allow for highly controlled expression of different variants, they do not allow one to observe changes to splicing since only exonic sequences are present. In light of the findings presented here, the inability of c.932G>A to cause significant dysfunction of NR2E3 in previous studies can be explained by the splice modulating mechanism, indicating that in fact, an arginine to glutamine change at position 311 may not be pathogenic. Together, these results indicate that a modulation of splicing to express a full-length NR2E3 cDNA containing the c.932G>A variant may be sufficient to restore normal NR2E3 function. Further experiments must be performed to confirm this hypothesis. Additionally, since NR2E3 is a transcription factor active in early retinal development, it remains an open question whether restoring normal NR2E3 expression in the adult retina would restore any visual function or prevent progressive retinal degeneration.
While nonsense or missense variants may require viral vector-mediated gene augmentation, splice variants may be amenable to modulation by antisense oligonucleotides. Previous work has shown that such treatment approaches can be successful in inherited retinal disease caused by errors in splicing of photoreceptor cell genes.27 Expanding the catalog of known splice variants to include variants currently annotated as missense variants (such as those in Figure 6) may permit new treatment strategies for several inherited retinal diseases.
In this study the in silico splice prediction software SpliceAI was used to predict if known pathogenic missense mutations were likely to alter transcript splicing. In addition to c.932G>A (p.R311Q), missense variants in 13 genes previously reported26 to be disease causing were identified. Using patient derived fibroblasts, we were able to demonstrate that the C.1711C>A (p.G571R) variant in CEP290 caused mis-splicing of the CEP290 transcript, resulting in inclusion of a portion of intron 17 that was absent in the normal control (Figure 6). While these findings demonstrate the utility of SpliceAI, they also highlight the importance of using patient derived cells to validate in silico predictions. Unlike CEP290, which is expressed in dermal fibroblasts, the remaining 12 variants identified are expressed in retina specific genes and as such require generation of patient iPSCs and derivation of retinal organoids. As this strategy is both labor and resource intensive it is most suitable for validation of variants whose pathogenicity are suspected but unknown. For instance, using patient derived retinal organoids we recently demonstrated that a deep intronic variant in the gene ABCA4 creates a cryptic splice site and mis-splicing in photoreceptor and RPE cells.28 In the latter study, data pertaining to the prevalence of the novel intronic variant in cases vs. controls allowed us to narrow the list of disease-causing candidates. One could envision using SpliceAI in a similar fashion; that is, as a hypothesis generating tool. Regardless, to support the molecular diagnosis of an unknown genetic variant, validation of in silico predictions using relevant model systems such as patient-derived retinal organoids is required.
In conclusion, this work shows that correct identification of a variant’s effect on a protein—in this case, deletion of a large portion of its ligand-binding domain—can be ascertained through studies based on disease modeling in human cells and assays at the transcript-level. Future work will be required to understand how the deletion caused by c.932G>A alters NR2E3’s ability to interact with co-factors and carry out its role in photoreceptor development.
Data and code availability
This study did not generate datasets. The sequencing data analyzed in this study is publicly available on GEO under accession numbers GSE266526 and GSE236197. The code generated during this study are available at Github (https://github.com/nkmullin/r311q_splice_paper).
Acknowledgments
This work was funded in part by NIH grants T32GM139776 and F30EY034009, and by the University of Iowa Institute for Vision Research. We acknowledge the extraordinary generosity of the patients whose donation of material made this study possible. We also thank the members of the Institute for Vision Research for their helpful conversation and advice. Data presented herein were obtained at the Genomics Division of the Iowa Institute of Human Genetics, which is supported, in part, by the University of Iowa Carver College of Medicine. We acknowledge specifically the advice and support of M. Boes, G. Hauser, K. Knudtson, A. Sheehan, and E. Snir.
Author contributions
N.K.M. and E.M.S. conceptualized the study. N.K.M., L.R.B., and K.R.A. performed experiments. K.R.A. generated cell lines. J.L.A. and E.M.S. provided clinical genetic information. N.K.M. carried out data analysis and generated figures. R.F.M., B.A.T., and E.M.S. supervised the study. N.K.M. wrote the first draft. L.R.B., R.F.M., B.A.T., and E.M.S. edited the manuscript. N.K.M., R.F.M., B.A.T., and E.M.S. acquired funding. All authors approved of the final manuscript.
Declaration of interests
The authors declare no competing interests.
Footnotes
Supplemental information can be found online at https://doi.org/10.1016/j.xhgg.2024.100357.
Web resources
https://www.ncbi.nlm.nih.gov/clinvar/.
https://github.com/Illumina/SpliceAI.
https://console.cloud.google.com/storage/browser/dm_alphamissense.
Supplemental information
References
- 1.Haider N.B., Jacobson S.G., Cideciyan A.V., Swiderski R., Streb L.M., Searby C., Beck G., Hockey R., Hanna D.B., Gorman S., et al. Mutation of a nuclear receptor gene, NR2E3, causes enhanced S cone syndrome, a disorder of retinal cell fate. Nat. Genet. 2000;24:127–131. doi: 10.1038/72777. [DOI] [PubMed] [Google Scholar]
- 2.Milam A.H., Rose L., Cideciyan A.V., Barakat M.R., Tang W.X., Gupta N., Aleman T.S., Wright A.F., Stone E.M., Sheffield V.C., Jacobson S.G. The nuclear receptor NR2E3 plays a role in human retinal photoreceptor differentiation and degeneration. Proc. Natl. Acad. Sci. USA. 2002;99:473–478. doi: 10.1073/pnas.022533099. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Cheng H., Khanna H., Oh E.C.T., Hicks D., Mitton K.P., Swaroop A. Photoreceptor-specific nuclear receptor NR2E3 functions as a transcriptional activator in rod photoreceptors. Hum. Mol. Genet. 2004;13:1563–1575. doi: 10.1093/hmg/ddh173. [DOI] [PubMed] [Google Scholar]
- 4.Chen J., Rattner A., Nathans J. The rod photoreceptor-specific nuclear receptor Nr2e3 represses transcription of multiple cone-specific genes. J. Neurosci. 2005;25:118–129. doi: 10.1523/JNEUROSCI.3571-04.2005. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Mullin N.K., Bohrer L.R., Voigt A.P., Lozano L.P., Wright A.T., Bonilha V.L., Mullins R.F., Stone E.M., Tucker B.A. NR2E3 loss disrupts photoreceptor cell maturation and fate in human organoid models of retinal development. J. Clin. Invest. 2024;134 doi: 10.1172/JCI173892. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Garafalo A.V., Calzetti G., Cideciyan A.V., Roman A.J., Saxena S., Sumaroka A., Choi W., Wright A.F., Jacobson S.G. Cone Vision Changes in the Enhanced S-Cone Syndrome Caused by NR2E3 Gene Mutations. Invest. Ophthalmol. Vis. Sci. 2018;59:3209–3219. doi: 10.1167/iovs.18-24518. [DOI] [PubMed] [Google Scholar]
- 7.Xiao S., Yi Z., Xiao X., Li S., Jia X., Lian P., Sun W., Wang P., Lu L., Zhang Q. Clinical and Genetic Features of NR2E3-Associated Retinopathy: A Report of Eight Families with a Longitudinal Study and Literature Review. Genes. 2023;14 doi: 10.3390/genes14081525. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Toms M., Ward N., Moosajee M. Nuclear Receptor Subfamily 2 Group E Member 3 (NR2E3): Role in Retinal Development and Disease. Genes. 2023;14 doi: 10.3390/genes14071325. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Mullin N.K., Voigt A.P., Cooke J.A., Bohrer L.R., Burnight E.R., Stone E.M., Mullins R.F., Tucker B.A. Patient derived stem cells for discovery and validation of novel pathogenic variants in inherited retinal disease. Prog. Retin. Eye Res. 2021;83 doi: 10.1016/j.preteyeres.2020.100918. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Escher P., Gouras P., Roduit R., Tiab L., Bolay S., Delarive T., Chen S., Tsai C.C., Hayashi M., Zernant J., et al. Mutations in NR2E3 can cause dominant or recessive retinal degenerations in the same family. Hum. Mutat. 2009;30:342–351. doi: 10.1002/humu.20858. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.von Alpen D., Tran H.V., Guex N., Venturini G., Munier F.L., Schorderet D.F., Haider N.B., Escher P. Differential dimerization of variants linked to enhanced S-cone sensitivity syndrome (ESCS) located in the NR2E3 ligand-binding domain. Hum. Mutat. 2015;36:599–610. doi: 10.1002/humu.22775. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Roduit R., Escher P., Schorderet D.F. Mutations in the DNA-binding domain of NR2E3 affect in vivo dimerization and interaction with CRX. PLoS One. 2009;4 doi: 10.1371/journal.pone.0007379. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Sridhar A., Hoshino A., Finkbeiner C.R., Chitsazan A., Dai L., Haugan A.K., Eschenbacher K.M., Jackson D.L., Trapnell C., Bermingham-McDonogh O., et al. Single-Cell Transcriptomic Comparison of Human Fetal Retina, hPSC-Derived Retinal Organoids, and Long-Term Retinal Cultures. Cell Rep. 2020;30:1644–1659.e4. doi: 10.1016/j.celrep.2020.01.007. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Cowan C.S., Renner M., De Gennaro M., Gross-Scherf B., Goldblum D., Hou Y., Munz M., Rodrigues T.M., Krol J., Szikra T., et al. Cell Types of the Human Retina and Its Organoids at Single-Cell Resolution. Cell. 2020;182:1623–1640.e34. doi: 10.1016/j.cell.2020.08.013. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Afanasyeva T.A.V., Corral-Serrano J.C., Garanto A., Roepman R., Cheetham M.E., Collin R.W.J. A look into retinal organoids: methods, analytical techniques, and applications. Cell. Mol. Life Sci. 2021;78:6505–6532. doi: 10.1007/s00018-021-03917-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Keuthan C.J., Karma S., Zack D.J. Alternative RNA Splicing in the Retina: Insights and Perspectives. Cold Spring Harb. Perspect. Med. 2023;13 doi: 10.1101/cshperspect.a041313. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Bohrer L.R., Wiley L.A., Burnight E.R., Cooke J.A., Giacalone J.C., Anfinson K.R., Andorf J.L., Mullins R.F., Stone E.M., Tucker B.A. Correction of NR2E3 Associated Enhanced S-cone Syndrome Patient-specific iPSCs using CRISPR-Cas9. Genes. 2019;10 doi: 10.3390/genes10040278. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Ohlemacher S.K., Iglesias C.L., Sridhar A., Gamm D.M., Meyer J.S. Generation of highly enriched populations of optic vesicle-like retinal cells from human pluripotent stem cells. Curr. Protoc. Stem Cell Biol. 2015;32:1H.8.1–1H.8.20. doi: 10.1002/9780470151808.sc01h08s32. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Li H. Minimap2: pairwise alignment for nucleotide sequences. Bioinformatics. 2018;34:3094–3100. doi: 10.1093/bioinformatics/bty191. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Marsh S.E. scCustomize: Custom Visualizations & Functions for Streamlined Analyses of Single Cell Sequencing. 2021. Zenodo. [DOI]
- 21.Jaganathan K., Kyriazopoulou Panagiotopoulou S., McRae J.F., Darbandi S.F., Knowles D., Li Y.I., Kosmicki J.A., Arbelaez J., Cui W., Schwartz G.B., et al. Predicting Splicing from Primary Sequence with Deep Learning. Cell. 2019;176:535–548.e24. doi: 10.1016/j.cell.2018.12.015. [DOI] [PubMed] [Google Scholar]
- 22.Cheng J., Novati G., Pan J., Bycroft C., Žemgulytė A., Applebaum T., Pritzel A., Wong L.H., Zielinski M., Sargeant T., et al. Accurate proteome-wide missense variant effect prediction with AlphaMissense. Science. 2023;381 doi: 10.1126/science.adg7492. [DOI] [PubMed] [Google Scholar]
- 23.Smith S.A., Lynch K.W. Cell-based splicing of minigenes. Methods Mol. Biol. 2014;1126:243–255. doi: 10.1007/978-1-62703-980-2_18. [DOI] [PubMed] [Google Scholar]
- 24.Adamson S.I., Zhan L., Graveley B.R. Vex-seq: high-throughput identification of the impact of genetic variation on pre-mRNA splicing efficiency. Genome Biol. 2018;19:71. doi: 10.1186/s13059-018-1437-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Stone N.E., Bohrer L.R., Mullin N.K., Cooke J., Wright A.T., Stone E.M., Mullins R.F., Tucker B.A. Device-free isolation of photoreceptor cells from patient iPSC-derived retinal organoids. bioRxiv. 2024 doi: 10.1101/2024.05.02.592255. Preprint at. [DOI] [Google Scholar]
- 26.Stone E.M., Andorf J.L., Whitmore S.S., DeLuca A.P., Giacalone J.C., Streb L.M., Braun T.A., Mullins R.F., Scheetz T.E., Sheffield V.C., Tucker B.A. Clinically Focused Molecular Investigation of 1000 Consecutive Families with Inherited Retinal Disease. Ophthalmology. 2017;124:1314–1331. doi: 10.1016/j.ophtha.2017.04.008. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Russell S.R., Drack A.V., Cideciyan A.V., Jacobson S.G., Leroy B.P., Van Cauwenbergh C., Ho A.C., Dumitrescu A.V., Han I.C., Martin M., et al. Intravitreal antisense oligonucleotide sepofarsen in Leber congenital amaurosis type 10: a phase 1b/2 trial. Nat. Med. 2022;28:1014–1021. doi: 10.1038/s41591-022-01755-w. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Burnight E.R., Fenner B.J., Han I.C., DeLuca A.P., Whitmore S.S., Bohrer L.R., Andorf J.L., Sohn E.H., Mullins R.F., Tucker B.A., Stone E.M. Demonstration of the pathogenicity of a common non-exomic mutation in ABCA4 using iPSC-derived retinal organoids and retrospective clinical data. Hum. Mol. Genet. 2024;33:1379–1390. doi: 10.1093/hmg/ddad176. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
This study did not generate datasets. The sequencing data analyzed in this study is publicly available on GEO under accession numbers GSE266526 and GSE236197. The code generated during this study are available at Github (https://github.com/nkmullin/r311q_splice_paper).