Early-Onset Stargardt Disease Caused by Homozygosity of a Complex ABCA4 Allele from Eastern Africa: Two Case Reports

Abstract

Introduction

Biallelic pathogenic variants in the ABCA4 gene are the leading cause of inherited retinal diseases. Over 1,200 pathogenic or likely pathogenic ABCA4 variants have been reported, resulting in a broad clinical spectrum of ABCA4-retinal dystrophies (ABCA4-RD), with Stargardt disease being the most common. Most patients with ABCA4-RD are compound heterozygotes, carrying two pathogenic ABCA4 variants in trans.

Case Presentation

We report 2 unrelated patients with early-onset (≤12 years) Stargardt disease, both found to be homozygous for a complex ABCA4 allele containing the hypomorphic c.5882G>A p.(Gly1961Glu) variant and the c.634C>T p.(Arg212Cys) variant. Both patients underwent detailed clinical assessment and genetic screening, including whole exome or genome sequencing. In vitro assays were performed to assess the individual and combined effect of these variants on the ABCA4 protein. The identified ABCA4 variants were expressed in HEK293FT and HeLa cells to assess their protein expression levels and intracellular localization compared to the wild type (WT) ABCA4 protein. Molecular analysis revealed that the Arg212Cys variant and the doubly mutated allele showed similarly reduced protein expression, while Gly1961Glu expressed close to WT level. Both variants, individually and combined, localized to intracellular vesicles similarly to WT ABCA4.

Conclusion

This study highlights the genetic complexity of ABCA4-RD and the significance of pathogenic variants in cis. It also emphasizes the challenge of accurately predicting the functional consequences of specific ABCA4 alleles with in vitro assays.

Introduction

Inherited retinal diseases (IRDs) constitute a genetically and clinically heterogeneous group characterized by retinal degeneration and progressive vision loss [1]. The most common cause of IRDs is pathogenic variation in the ABCA4 gene, with missense changes accounting for most disease-causing ABCA4 variants observed to date [2–4]. ABCA4 encodes the ATP-binding cassette transporter ABCA4, an ATP-dependent flippase primarily expressed within the photoreceptor cells of the retina, where it plays an important role in the visual cycle. Loss of ABCA4 activity results in the formation of toxic bisretinoids that injure the retinal pigment epithelium (RPE) and photoreceptors [4, 5]. ABCA4-retinal dystrophies (ABCA4-RD) follow an autosomal recessive inheritance pattern, but additional intronic cis-acting variants, such as c.769-784C>T, can complicate the pathogenic mechanism by modifying the penetrance of common hypomorphic variants, contributing to increased phenotypic severity [6, 7]. This results in a broad clinical spectrum, ranging from aggressive cone-rod dystrophy to milder Stargardt disease, fundus flavimaculatus, or late-onset geographic atrophy [3, 6, 8, 9]. Phenotypic and genotypic data from studies involving large patient cohorts have shown that two null alleles typically result in early disease onset and rapid progression to advanced stages. Conversely, genotypes containing more common hypomorphic variants such as c.5882G<A p.(Gly1961Glu) are generally associated with later-onset Stargardt disease [10, 11]. However, here we report 2 unrelated Stargardt patients of Somali origin with early disease onset, both homozygous for a complex ABCA4 allele containing the Gly1961Glu variant, thereby challenging this genotype-phenotype correlation matrix [11]. The early onset in these patients is likely attributed to homozygosity for their second variant, c.634C>T p.(Arg212Cys), previously associated with moderate to severe, early-onset disease. To investigate the impact of these variants, both individually and in combination, we assessed their effects on ABCA4 protein expression and intracellular localization in vitro, relative to wild type (WT) protein.

Case Reports

Clinical Assessment and Genetic Screening of the Patients

Patient 1 (P1) experienced reduced vision from age 5 and was referred to ophthalmology testing at Haukeland University Hospital at age 12. The clinical assessment included measurement of best corrected visual acuity (BCVA), full-field electroretinography (ffERG), optical coherence tomography (OCT), ultra-widefield fundus autofluorescence (UWF-FAF) images, and Goldmann perimetry. Her BCVA was 20/135 (OD) and 20/100 (OS). Goldmann perimetry revealed normal peripheral visual fields (data not shown). A red-green color deficit was found. UWF-FAF images showed bilateral circular macular changes resembling bull’s eye maculopathy and circular hyperfluorescence around the atrophic foveal region but no pallor of the optic disc and normal peripheral retina (Fig. 1a, b). OCT revealed loss of the macular photoreceptor layer (Fig. 1c). Additionally, full-field electroretinography showed severely reduced cone function, but normal rod responses. P1 is of Somali origin with no family history of IRD. His family members (parents and five siblings) did not report any visual issues, and there was no known consanguinity.

Fig. 1.

Fundus images and OCT of Patient 1 (P1). a Fundus photography reveals a bull’s eye maculopathy. b Fundus autofluorescence imaging reveals circular hyperfluorescence around the atrophic parafoveal region. The patient had normal peripheral retina in both eyes. c OCT scan showing complete loss of the outer retina in the central 1 mm of the macula.

Patient 2 (P2), previously described in a large IRD cohort study [1], is also of Somali origin and was seen in the retina clinic at the University of Iowa at age 11. The clinical assessment included measurement of BCVA, intraocular pressure, slit lamp biomicroscopy, indirect ophthalmoscopy, Goldmann perimetry, UWF-FAF images, and OCT. Her symptoms began at age 8 when she noticed a reduction in her acuity when reading small print. She denied other visual symptoms including nyctalopia and photophobia. Upon examination, her BCVA was 20/70 (OD) and 20/80 (OS). Fundus imaging revealed circular macular changes resembling the bull’s eye maculopathy, normal discs and vessels with a cup-to-disc ratio of 0.2 for each eye (Fig. 2a, b). There was a 500-micron zone of outer retinal and RPE atrophy centered on fixation on both eyes. This atrophic area was surrounded by an elliptical annulus of small pisciform flecks that was 2.5 mm in vertical dimension and 3 mm in horizontal dimension (Fig. 2b). The patient had normal peripheral retina in both eyes. The choroidal vasculature for each eye was less visible than normal, suggesting a slight opacity at the level of the RPE. OCT revealed loss of outer retinal structures in the central 500 microns (Fig. 2c). Anterior to this, there was thickening of the external limiting membrane and thinning of the ellipsoid such that these two bands were roughly equal in reflectivity. Goldmann perimetry revealed normal I2e, I4e, and V4e isopters for each eye (Fig. 2d). There were small central scotomas to the I2e and I4e stimulus on both eyes. Her parents and four siblings (three sisters and one brother) had no visual complaints and there was no known consanguinity.

Fig. 2.

Fundus images, OCT, and Goldmann perimetry test of Patient 2 (P2). a Color fundus photographs revealed normal disks and vessels with a cup-to-disk ratio of 0.2 for each eye. The choroidal vasculature for each eye was less visible than normal, suggesting a slight opacity at the level of the RPE. b Infrared fundus photographs showed circular macular changes resembling the bull’s eye maculopathy. The patient had normal peripheral retina in both eyes. c OCT images revealed loss of outer retinal structures in the macula. There was thickening of the external limiting membrane and thinning of the ellipsoid such that these two bands were roughly equal in reflectivity. d Goldmann perimetry revealed normal I2e, I4e, and V4e isopters for each eye. There were small central scotomas to the I2e and I4e stimulus in both eyes.

Whole exome sequencing (WES) was obtained for P1, and whole genome sequencing (WGS) was obtained for P2 and their unaffected father. The Illumina sequence data were analyzed using standard GATK-based pipelines. Regions of homozygosity were called using bcftools/RoH [12] and Alissa (Agilent) tools. Genetic screening revealed homozygosity for two pathogenic missense variants in the ABCA4 gene (NM_000350.3) c.5882G>A p.(Gly1961Glu) and c.634C>T p.(Arg212Cys). Homozygosity was confirmed by genetic testing of the parents. All four parents were heterozygous for Arg212Cys, and three of the four parents were heterozygous for Gly1961Glu. The asymptomatic 32-year-old father of P2 was found to be homozygous for Gly1961Glu (Fig. 3). WES of P1 revealed 8.1 million base pairs of homozygosity surrounding the ABCA4 locus, while WGS of P2 revealed 1 million base pairs of homozygosity (Fig. 3). WGS of the father of P2 revealed about 50 thousand base pairs of homozygosity containing Gly1961Glu (Fig. 3).

Fig. 3.

Homozygous regions containing ABCA4 in two families. Both affected patients, P1 and P2, are homozygous over the entire ABCA4 locus including Arg212Cys and Gly1961Glu. The unaffected father of P2 has a much smaller region of homozygosity containing only Gly1961Glu.

Protein Expression and Intracellular Localization of ABCA4 Variants

ABCA4 constructs encoding either the individual variants or the doubly mutated allele were expressed in HEK293FT and HeLa cells to assess protein expression levels and intracellular localization, as previously described [13]. As shown in Figure 4a and b, Gly1961Glu showed a protein expression level comparable to WT ABCA4 (∼87%), while Arg212Cys and the doubly mutated allele showed reduced expression (∼45–46%) compared to WT. qPCR confirmed equal mRNA expression levels of the variants and WT ABCA4 (data not shown). Immunofluorescence in HeLa cells revealed vesicle-like localization for all variants, consistent with WT ABCA4. This was in line with the normal protein expression of Gly1961Glu but was somewhat unexpected for Arg212Cys and the doubly mutated allele, given their reduced protein expression.

Fig. 4.

Protein expression and intracellular localization of Arg212Cys, Gly1961Glu, and the doubly mutated allele in transfected HEK293FT and HeLa cells, respectively. a Image of stain-free gel used for total protein normalization and a representative Western blot labeled with the Rho-antibody targeting 1D4-tagged ABCA4. b Protein expression of variants relative to WT (set to 100%), shown as mean ± SD (n = 4). Each dot represents independent experiments. c Intracellular localization of variants and WT ABCA4. Anti-calnexin was used as an ER marker (green), Rho1D4 antibody was used to target ABCA4 (red), and DAPI for nuclear stain (blue).

Discussion

ABCA4 is very genetically heterogeneous, with most patients with ABCA4-RD being compound heterozygotes for pathogenic variants [6]. Homozygosity for Arg212Cys has previously been shown to cause early-onset (≤11 years) Stargardt disease, similar to that observed in our patients [14, 15]. In contrast, Gly1961Glu is a mild ABCA4 variant associated with later-onset Stargardt disease in the homozygous state [16]. However, its pathogenicity in homozygosity may not reach the pathogenic threshold unless a cis-acting modifier is present [6, 7]. More than 50 such complex ABCA4 alleles have been identified, consisting of multiple pathogenic variants in cis. When an additional pathogenic variant is present in cis, it can lead to an earlier and more severe disease phenotype. For example, both patients and the unaffected father of P2 are homozygous for Gly1961Glu, yet only those with an additional Arg212Cys variant exhibit macular disease. While this distinction is obvious in the present case because the additive variant is easily detectable and interpretable in the coding sequence, some phenotype altering variants lie in non-coding sequences, some lie thousands of base pairs from the nearest coding sequence, and many remain undiscovered. Therefore, as shown in Figure 3, it is more accurate to assert that some alleles that contain Gly1961Glu cause disease in the homozygous state than to assert that some patients who are homozygous for Gly1961Glu have macular disease. Still, the current status of P2’s father as clinically normal could be explained by the late disease onset for the Gly1961Glu allele. Altogether, the existence of both pathogenic complex alleles and deep-intronic variants, underlines the importance of establishing the correct haplotype phase of variants found during diagnostic work-up.

The population prevalence of Gly1961Glu is quite high in eastern Africa (8–11%), where it is believed to have originated [16, 17]. Although there is no known consanguinity in P1’s pedigree, the 8.1 million base pairs of homozygosity suggest that P1’s parents are only ten or twelve meioses apart in their pedigree (Fig. 3). In contrast, P2’s parents, despite carrying the regional Gly1961Glu/Arg212Cys allele, are more distantly related, reflected in P2’s 1-million-base-pair homozygous region. More striking still is P2’s father, who carries one of his Gly1961Glu variants on an allele without Arg212Cys. Gly1961Glu was fixed in the human population tens of thousands of years ago, long before the addition of Arg212Cys on the allele now prevalent in eastern Africa.

The combination of the high carrier frequency of Gly1961Glu in eastern Africa and the underexplored significance of the Gly1961Glu/Arg212Cys allele may also have implications for genetic counseling. A generalized counseling approach based on the premise that Stargardt disease is caused by genotypes consisting of two pathogenic alleles is inadequate, due to the considerable genotypic and phenotypic variability. However, the calculation of recurrence risks for carriers of the Gly1961Glu/Arg212Cys allele is challenging, due to the low number of affected individuals carrying the allele so far. This is also underlined by the father of P2 who is currently healthy in spite of being compound heterozygous for Gly1961Glu/Arg212Cys and Gly1961Glu.

Protein expression analysis showed that Gly1961Glu had similar expression to WT, whereas Arg212Cys had reduced expression, consistent with previous studies [4]. The doubly mutated allele showed equally low-expression level as Arg212Cys. Immunofluorescence revealed vesicle-like localization for all variants, challenging previous studies showing that low-expression variants are misfolded and retained in the endoplasmic reticulum (ER) [4, 18]. Understanding the functional impact of ABCA4 variants is also important as protein expression does not always correlate with remaining protein function. As ABCA4 is predominantly localized to the outer segment disc membranes of the photoreceptors, and associates with vesicles and ER in heterologous expression systems, ABCA4 WT and variant proteins need to be solubilized in detergent in order to carry out functional characterizations. Although this approach has proven to be useful for most ABCA4 variants, the current ATPase activity assay for ABCA4 is unreliable for Gly1961Glu due to its instability in CHAPS detergent, rendering the assay unreliable for the doubly mutated allele as well. Accordingly, it is important to develop additional assays to measure ABCA4 activity that does not require detergent solubilization. However, Arg212Cys has previously exhibited severely reduced ATP activity (50% of the WT), suggesting a similar effect for the doubly mutated allele [4].

In conclusion, this study underscores the genetic complexity of ABCA4-RD and emphasizes the impact of pathogenic variants occurring in cis. It demonstrates the challenge of accurately predicting the functional behavior of specific alleles with in vitro assays, especially for variants in cis. We suspect that these issues are valid for not only ABCA4 and IRD but also a wide range of other genes involved in Mendelian disorders.

The CARE Checklist has been completed by the authors for this case report, attached as online supplementary material (for all online suppl. material, see https://doi.org/10.1159/000547387).

Acknowledgment

We thank Karen Marie Hagen at the Department of Clinical Science, University of Bergen, for her technical assistance.

Statement of Ethics

The investigations involving P1 and P2 were approved by the Norwegian Regional Ethics Committee (Case No. 2017/2487/REC West) and the Institutional Review Board of the University of Iowa, respectively. The entire study adhered to the tenets outlined in the Declaration of Helsinki. Written informed consent was obtained from the patients for publication of the details of their medical case and any accompanying images.

Conflict of Interest Statement

The authors have no conflicts of interest to declare.

Funding Sources

This study was supported by grants from the Norwegian RP Association and the Norwegian Association of the Blind and Partially Sighted.

Author Contributions

S.A. and M.H.B. conducted the in vitro assays. S.A. and E.M.S. wrote the manuscript with input from M.H.B., E.B., I.A., C.B., M.I.F., and P.M.K. C.B. and E.M.S. performed the clinical examination. E.B., I.A., M.I.F., E.M.S., and P.M.K. performed the genetic analyses. A.P.D. and J.L.A. contributed to the study analyses. All authors reviewed and approved the final manuscript.

Funding Statement

This study was supported by grants from the Norwegian RP Association and the Norwegian Association of the Blind and Partially Sighted.

Data Availability Statement

All clinical data and analyses generated in this study are included in this article and its supplementary materials. Further inquiries may be directed to the corresponding author.