Functional Validation of a Novel Deep Intronic IMPG2 Variant Causing Pseudoexon Activation in Retinitis Pigmentosa with Macular Involvement

Background

Inherited retinal dystrophies (IRDs) are a group of genetically heterogeneous disorders characterized by the progressive degeneration of retinal photoreceptors or retinal pigment epithelium (RPE). Clinically, IRDs typically present with night blindness, peripheral visual field loss, and progressive decline in visual acuity, which may ultimately lead to blindness.^1,2 Among them, retinitis pigmentosa (RP) is the most common subtype, with a global prevalence estimated at 1 in 3000 to 1 in 5000 individuals. To date, more than 100 genes have been identified as being associated with RP, reflecting its considerable clinical and genetic heterogeneity.^3,4

The IMPG2 (interphotoreceptor matrix proteoglycan 2) gene encodes a structural protein localized to the interphotoreceptor matrix between photoreceptors and the RPE, playing a crucial role in maintaining photoreceptor integrity and function. Biallelic variants in IMPG2 have been reported to cause autosomal recessive RP as well as certain subtypes of vitelliform macular dystrophy.^5,6 Patients with IMPG2-related retinopathy often present with bilateral night blindness, progressive peripheral visual field constriction, and characteristic bone-spicule pigmentation in the retina. Optical coherence tomography (OCT) commonly reveals disorganization of the macular architecture, and electrophysiological tests indicate severely impaired retinal function.

Despite major advances in gene discovery, 52–74% of IRD cases remain without a molecular diagnosis after exome- or panel-based testing.^7,8 This “missing heritability” is frequently attributed to pathogenic variants outside coding regions, including deep intronic and regulatory changes.⁹ Notably, an estimated 9–30% of pathogenic variants in Mendelian disease act by altering splicing rather than directly changing protein sequence.⁹ Whole-genome sequencing (WGS) helps bridge this gap by detecting intronic and structural variants overlooked by whole-exome sequencing (WES), with reported incremental diagnostic yields ranging from a few percentage points to ~24%, depending on cohort and methodology.¹⁰ As deep intronic and structural variants are often missed by WES, WGS has become an essential tool to improve diagnostic yield in IRDs with negative or inconclusive exome results.^10,11 Recent ACMG/ClinGen Splicing Variant Interpretation (SVI) recommendations also emphasize that RNA-based functional assays provide important evidence for interpreting splicing variants, particularly those located in noncanonical or deep intronic regions.¹² Emerging long-read cDNA/RNA sequencing technologies also provide a direct means to detect aberrant splicing events, including pseudoexon activation, and are increasingly applied in IRD diagnostics.¹³

Here, we report a non-consanguineous patient with bilateral night blindness who, by WGS, was found to harbor a novel homozygous deep intronic variant in IMPG2 (c.909-1659A>G). Using family-based segregation, in silico splice prediction, detailed retinal imaging and electrophysiology, and an in vitro minigene splicing assay, we demonstrate aberrant splicing attributable to this variant. In accordance with ACMG/AMP criteria,¹⁴ the variant was reclassified from VUS to likely pathogenic. These findings expand the IMPG2 mutational spectrum and highlight the critical role of functional assays in interpreting deep intronic variants in IRDs.

Materials and Methods

Patient and Clinical Evaluation

We enrolled a 54-year-old female proband clinically diagnosed with retinitis pigmentosa (RP). The patient exhibited congenital night blindness, progressive peripheral visual field constriction, and gradual decline in visual acuity in recent years. Comprehensive ophthalmic assessments included medical history review, best-corrected visual acuity measurement, slit-lamp biomicroscopy, fundus photography, optical coherence tomography (OCT), OCT angiography (OCTA), multifocal electroretinography (mfERG), and microperimetry (MAIA).

This study was approved by the Ethics Committee of Boai Hospital of Zhongshan (Approval No.: KY-2025-008-05). Institutional approval was required and obtained for both the study procedures and the publication of the case details. Written informed consent was obtained from all participants in accordance with the Declaration of Helsinki.

Genetic Testing and Bioinformatic Analysis

Genomic DNA was extracted from peripheral blood using EDTA anticoagulation. Whole-genome sequencing (WGS) was performed on the BGI high-throughput platform using paired-end short-read sequencing, with an average depth of 58.95× and >98% of the genome covered at ≥10× (hg19/GRCh37 reference). Variants were analyzed using BGI’s standard clinical WGS pipeline, and genes with established disease associations listed in the OMIM database were assessed. Population databases used for variant interpretation included gnomAD (r2.1.1; accessed 20 March 2024), ExAC (r1), 1000 Genomes (Phase 3), and ESP6500 (V2). Disease databases included ClinVar (accessed 20 March 2024) and the BGI-Phoenix internal disease-variant database. HGMD Professional 2024.1 (accessed 20 March 2024) was additionally queried to confirm the absence of previously reported pathogenic variants at this locus. Candidate variants were validated by Sanger sequencing. In silico splicing predictions were generated using SpliceAI, MaxEntScan, and NNSplice. SpliceAI (version 1.3.1) was applied using a Δ score threshold of ≥0.20 for potential splice-altering effects. Segregation analysis was performed in available family members after written informed consent.

Minigene Splicing Assay

To assess the functional consequences of the intronic variant, minigene constructs were generated using both pcMINI and pcMINI-C vectors. Each construct contained IMPG2 exons 9 and 10 with the intervening intronic sequence carrying either the wild-type (WT) or mutant (Mut; c.909-1659A>G) allele. Detailed primer sequences used for amplification and cloning are provided in Supplementary Table 1. Additional details of the minigene construction strategy are provided in the Supplementary Methods. Recombinant plasmids were transiently transfected into HEK293T and HeLa cells using a liposome-based method. After 24–48 hours, total RNA was extracted and reverse-transcribed into cDNA. RT-PCR was performed with vector-specific primers, and the products were separated by agarose gel electrophoresis, purified, and sequenced by Sanger sequencing. All experiments were independently repeated at least three times in both cell lines to confirm reproducibility. The design and maps of pcMINI and pcMINI-C constructs are provided in Supplementary Figure 1.

Variant Interpretation and Pathogenicity Assessment

Variant classification was performed according to the ACMG/AMP 2015 guidelines and subsequent refinements.^12,14 Evidence categories included population frequency, computational predictions, segregation analysis, and functional assay results. The final pathogenicity classification was determined based on the integrated evidence.

Results

Clinical Phenotype and Ocular Imaging Findings

The patient exhibited progressive visual decline over the years. Fundus examination revealed widespread bone-spicule-like pigment deposits in the mid-peripheral retina, markedly attenuated retinal vessels, optic nerve atrophy, and waxy pallor of the optic discs.

In October 2019, ophthalmic examinations showed:

1. Microperimetry (MAIA): Poor fixation response in both eyes; central fixation in the left eye was unstable; macular sensitivity was markedly reduced, falling below the 5th percentile reference range.
2. Multifocal electroretinography (mfERG): Residual responses were limited to the central foveal region, with severe peripheral dysfunction.
3. Ultra-widefield fundus imaging (Figure 1A and C): Bone-spicule pigmentation in the mid-periphery, narrowed vessels, and waxy optic discs.
4. OCT (Figure 1B and D): Spectral-domain OCT revealed thinning of the outer retinal layers with generalized loss of the ellipsoid zone in both eyes.
5. OCTA: Demonstrated reduced capillary density in the superficial and deep capillary plexuses.

Figure 1 Ocular imaging findings of the proband. (A and C) Ultra-widefield fundus images obtained using the Optos Daytona P200T scanning laser ophthalmoscope (Optos PLC, Dunfermline, UK), showing scattered bone-spicule–like pigment deposits in the mid-peripheral retina, attenuated retinal vessels, optic nerve pallor, and waxy discs. (B and D) Spectral-domain optical coherence tomography (SD-OCT) images acquired using the Cirrus HD-OCT 5000 system (Carl Zeiss Meditec, Dublin, CA, USA) with the Macular Cube 512×128 protocol, demonstrating thinning of the outer retinal layers and generalized loss of the ellipsoid zone in both eyes.

Clinically, the proband exhibited night blindness from childhood, progressive decline in visual acuity beginning in her 40s, and no improvement in vision following cataract surgery. During the past decade, peripheral visual fields progressively constricted, resulting in increasing difficulty with reading.

Best-corrected visual acuity recorded in 2013 was 0.2 in the right eye (OD) and 0.1 in the left eye (OS). During follow-up, a formal refraction in 2019 showed that best-corrected visual acuity (BCVA) measurement had declined to counting fingers at 20 cm in the right eye, while the left eye remained partially correctable to 0.3−, indicating preserved but progressively impaired central visual function. A summary of the longitudinal clinical findings is provided in Table 1.

Table 1 Clinical Timeline Summarizing Major Symptoms, BCVA, Functional Tests, and Imaging Findings

Peripheral blood samples were obtained from available family members, including the proband, her parents, three siblings, her husband, and her two children, following informed consent. Sanger sequencing was conducted to evaluate co-segregation of the identified variant with the disease phenotype, and a pedigree diagram was generated to illustrate the inheritance pattern.

Genetic Testing Results

Whole-genome sequencing identified a novel homozygous deep intronic variant in the IMPG2 gene (NM_016247.4:c.909-1659A>G), located between exons 9 and 10, approximately 1659 bp from the exon–intron boundary. Genome-wide runs of homozygosity analysis from the clinical WGS data showed no large homozygous segments. This variant was absent from gnomAD, ClinVar, HGMD, and dbSNP. Sanger sequencing was performed in nine relatives, including the proband, her parents, three siblings, her husband, and her two children. Segregation analysis demonstrated that both parents were heterozygous carriers, consistent with autosomal recessive inheritance, whereas unaffected siblings carried either the heterozygous or wild-type allele (Figure 2).

Figure 2 Pedigree and Sanger sequencing validation of the IMPG2 c.909-1659A>G variant. (A) Pedigree of the family showing the segregation of the variant; the proband (arrow) is homozygous, and all tested relatives were confirmed by Sanger sequencing. (B) Sanger sequencing chromatograms. (a) homozygous mutant (proband); (b) heterozygous carrier; (c) wild-type individual.

Bioinformatic Prediction and Minigene Splicing Assay Results

In silico analysis predicted a deleterious effect of the IMPG2 c.909-1659A>G variant. SpliceAI yielded a score of 0.55, exceeding the recommended 0.50 threshold for a strong predicted splicing impact. Consistently, both MaxEntScan and NNSplice indicated impaired recognition of the native splice site and activation of a cryptic donor site, supporting the likelihood of pseudoexon inclusion.

To experimentally validate these predictions, dual minigene constructs (pcMINI and pcMINI-C) carrying either the wild-type (WT) or mutant (Mut) sequence were transfected into HEK293T and HeLa cells (Figure 3A). In all four experimental systems (pcMINI-HEK293T, pcMINI-HeLa, pcMINI-C-HEK293T, pcMINI-C-HeLa), RT-PCR revealed that WT constructs generated a single band of the expected size, whereas Mut constructs consistently produced an aberrant band with altered mobility (Figure 3B; original uncropped gels shown in Supplementary Figure 2). Splicing diagrams illustrated canonical exon 9–10 junctions in WT transcripts, while Mut transcripts showed retention of a 132 bp pseudoexon, introducing a premature termination codon (PTC) (Figure 3C). Sanger sequencing further confirmed pseudoexon retention in Mut transcripts (Figure 3D).

Figure 3 Minigene splicing assay of the IMPG2 c.909-1659A>G variant (pcMINI system). (A) Schematic of the pcMINI-IMPG2 constructs. (B) RT-PCR products from HEK293T and HeLa cells. Expected sizes: 311 bp (WT) and 443 bp (Mut, with 132-bp pseudoexon inclusion); observed bands matched these predictions. (C) Splicing patterns of WT and Mut transcripts; the red segment represents the 132-bp pseudoexon activated by the variant, which introduces a premature stop codon. (D) Sanger sequencing confirming the exon9–pseudoexon–exon10 junction in the Mut transcript.

Collectively, these results demonstrate that the IMPG2 c.909-1659A>G variant induces a robust and reproducible splicing defect across different vector and cell systems. The aberrant inclusion of a 132 bp pseudoexon introduces a PTC, making nonsense-mediated decay the most likely outcome and reducing the likelihood of generating a stable truncated protein.

Variant Pathogenicity Assessment

Based on the ACMG/AMP guidelines, the IMPG2 c.909-1659A>G variant was evaluated for pathogenicity. Its extremely low allele frequency in population databases such as gnomAD and dbSNP provided supporting evidence of rarity (PM2_Supporting). Segregation analysis demonstrated that both parents were heterozygous carriers, consistent with autosomal recessive inheritance, providing additional supporting evidence (PM3_Supporting). Functional validation using dual minigene splicing assays revealed a reproducible splicing defect with pseudoexon retention and generation of a premature termination codon, meeting the PVS1_Moderate criterion. In addition, the proband’s clinical presentation—typical of IMPG2-related retinal dystrophy—provided phenotype specificity supporting PP4. Taken together, these criteria support the classification of the IMPG2 c.909-1659A>G variant as Likely pathogenic (Table 2). The final ACMG/AMP evidence codes applied were PVS1_Moderate, PM2_Supporting, PM3_Supporting, and PP4, in accordance with the ClinGen SVI framework.^12,14

Table 2 Summary of the IMPG2 c.909-1659A>G Variant and ACMG Classification

Discussion

This study reports, for the first time, a functionally validated deep intronic variant of IMPG2 (c.909-1659A>G) associated with retinitis pigmentosa (RP) and macular involvement. IMPG2 is a known causative gene, with biallelic variants leading to autosomal recessive RP (RP56), and both monoallelic and biallelic variants implicated in vitelliform macular dystrophy.¹⁵ Previous studies have demonstrated that pathogenic IMPG2 variants are predominantly located in exons or canonical splice sites, with a substantial proportion resulting in truncated proteins.^5,16,17 In contrast, the variant identified in this study activates a cryptic splice site, leading to retention of a 132 bp pseudoexon and the introduction of a premature termination codon. Because the premature termination codon lies upstream of the final exon–exon junction, nonsense-mediated decay is highly likely, and heterologous minigene assays cannot fully recapitulate endogenous retinal nonsense-mediated decay efficiency. This expands the mutational spectrum of IMPG2 and highlights the pathogenic potential of deep intronic variants, which are often overlooked in routine genetic testing.

The clinical features of our proband—including lifelong night blindness, progressive visual decline, peripheral visual field constriction, and macular dysfunction—are consistent with previously described phenotypes of IMPG2-associated retinopathy.^5,15,16 Importantly, while earlier reports indicated that macular involvement is common and often occurs early in IMPG2-related RP, our findings further demonstrate that a noncoding deep intronic variant can drive this phenotype. This underscores the importance of considering noncoding regions in patients with characteristic phenotypes but no identified exonic pathogenic variants.

A major contribution of this study is the functional validation of the intronic variant using dual minigene splicing assays. For deep intronic variants and other variants of uncertain significance (VUS), bioinformatic predictions alone are often insufficient for reliable interpretation.^18–20 By consistently confirming aberrant splicing across two vector systems and two cell lines, we provided functional evidence to support reclassification of this variant from “uncertain significance” to “likely pathogenic” under ACMG/AMP guidelines. This highlights the clinical value of functional assays in variant interpretation, directly improving molecular diagnostic yield and providing robust evidence for patient management and genetic counseling.

Nevertheless, this study has certain limitations. First, the splice defect was demonstrated in vitro rather than in patient-derived RNA or protein. This is largely due to the tissue-specific expression of IMPG2 in the retina/photoreceptors and the practical difficulty of obtaining retinal tissue.²¹ Second, this is a single-case report, limiting the generalizability of the conclusions. Validation in larger cohorts, including transcriptome (RNA-seq) and protein-level studies or the use of patient-derived iPSC models, will be important to further elucidate the pathogenic mechanism of the c.909–1659A>G variant.

Looking ahead, combining functional assays with emerging technologies such as RNA-seq and long-read sequencing will be instrumental in clarifying the pathogenicity of noncoding variants. Capasso et al¹³ recently demonstrated that targeted long-read cDNA sequencing can identify previously undetected splice-altering events, providing molecular diagnoses for patients with unsolved inherited retinal dystrophies (IRDs). Integrating such molecular evidence with detailed phenotypic data is expected to move variant interpretation from “suspected” to “confirmed,” significantly increasing the molecular diagnostic yield and advancing precision medicine and genetic counseling in IRDs.