Clinical features of the newly identified and eight previously reported patients are summarized in Table 1.

The patient (II:3) was a 28-year-old woman, who was the third child born to healthy non-consanguineous Korean parents with no family history of IDPOGSA (Fig. 1A). She was born at full term without complications; her length and weight were 51 cm and 3500 g, respectively. She walked before 1 year of age and spoke without delay. She showed normal motor development and growth until she entered school at 5 years of age.

Biallelic novel ABCA2 variants in a family and reported ABCA2 variants. A Family pedigree and biallelic novel ABCA2 variants. WT, wild-type. B Sanger sequencing confirmed these variants occurred trans in the patient (II:3). The intronic variant (c.5390-17C>A) was inherited from the father (I:1) and nonsense variant (c.6379C>T) from the mother (I:2). Red arrowheads indicate the variants. Intronic and exonic DNA sequence is denoted in lowercase and uppercase letter, respectively. C Schematic presentation of the ABCA2 gene (NM_001606.5) and novel reported variants: c.5390-17C>A located in intron 33 and c.6379 C>T located in exon 42. The detected variants are in red

At 6 years, she had a seizure episode with ocular deviation and unconsciousness, and was hospitalized for three days. The origin of the seizure was unknown. Brain magnetic resonance imaging (MRI) and electroencephalogram at 7 years of age were normal. After entering school, her learning difficulties, especially in mathematics, were recognized, requiring additional teaching support at primary to high school. Around 20 years of age, her parents noticed that her speaking was slow and fragmented.

She currently shows mild intellectual disability (ID) (Table S1) and ataxia. Although she can hear and understand, she cannot count well or handle money and sometimes cannot distinguish right or left sides. She can write, but is generally inactive and her movements are slow with poor fine coordination. Her latest neurological examination suggests non-progressive cerebellar ataxia with unknown age of onset. Brain MRI was normal with no cerebellar abnormalities. She also shows dysarthria, with a hoarse deep voice and short bursts of speech. Her height, weight, and head circumference are 168 cm, 55 kg, and 55 cm, respectively.

WES of the patient revealed two novel ABCA2 variants. Sanger sequencing confirmed a paternally inherited intronic variant and a maternally inherited nonsense variant, NM_001606.5:c.[5300–17C>A];[6379C>T] p.[?];[(Gln2127*)] (Fig. 1). These variants are absent from the Genome Aggregation Database (gnomAD v2.1.1, https://gnomad.broadinstitute.org/). SpliceAI [8] (https://spliceailookup.broadinstitute.org/) predicted an acceptor gain for the intronic variant. The nonsense variant is predicted as deleterious by multiple in-silico tools (Table S2).

We performed RT-PCR using total RNA extracted from a patient-derived lymphoblastoid cell line. Gel electrophoresis of RT-PCR products yielded multiple bands suggesting aberrant splicing by the intronic variant (Fig. 2A). The upper bands were degraded by T7E1 treatment, indicating formation of a wild-type (WT)/mutant heteroduplex (asterisks in Fig. 2A, B). To distinguish multiple transcripts, we performed TA-cloning using RT-PCR products treated with cycloheximide (CHX). Apart from WT transcript, we detected two types of aberrant transcripts, referred to as Mutant transcript A (MTA) and B (MTB) (Fig. 2C–E). In 26 sequenced colonies, counts of WT, MTA, and MTB were detected at 22, 2, and 2, respectively, suggesting that both mutant transcripts are unstable.

RT-PCR analysis and aberrant splicing. A Agarose gel electrophoresis (3%) of RT-PCR products using lymphoblastoid cell lines (LCLs) derived from the patient and a control (Ctrl). LCLs were cultured in medium with or without dimethyl sulfoxide (DMSO) or cycloheximide (CHX). ACTB was used as an internal control. Red and blue dots indicate bands representing Mutant transcript A (MTA) and Mutant transcript B (MTB), respectively. Asterisks indicate bands representing the wild-type/mutant transcript heteroduplex. RT, reverse transcription; NC, negative control. B Gel electrophoresis of products digested with or without T7-endonulease I (T7EI) using RT-PCR products from the patient and Ctrl. C–E Schematic presentation of the normally spliced transcript (wild-type) and two types of aberrantly spliced transcripts from the variant c.5390-17C>A (red arrows). C Wild-type transcript. D MTA: The “ag” sequence of c.5390-17C>A was recognized as a cryptic splice acceptor site, resulting in insertion of 15-bp and leading to a 5 amino acid in-frame insertion, p.Tyr1796_Gly1797insAlaProProCysAla. E MTB: The “AG” sequence located in exon 35 is recognized as a cryptic splice acceptor site, resulting in skipping of exon 34 and the first 7-bp from the 5′ end of exon 35, leading to deletion of 68-bp of cDNA. Intronic and exonic DNA sequence is shown as lowercase and uppercase letters, respectively. The canonical splice site is colored in blue, and recognized splice sites are underlined

In MTA, the paternally inherited intronic variant c.5300–17C>A creates a cryptic acceptor site 17-bp upstream of the 5′ end of exon 34. This results in insertion of 15-bp of intronic DNA into the cDNA, leading to a five amino acid insertion, p.Tyr1766_Gly1767insAlaProProCysAla (Fig. 2D). In MTB, the “AG” sequence in the middle of exon 35 is recognized as a cryptic acceptor site, resulting in skipping of exon 34 and the first 7-bp from the 5′ end of exon 35. This causes deletion of 68-bp of cDNA and induces a premature stop codon (Fig. 2E). We confirmed that MTB is a frameshift variant, p.Gly1768Ilefs188*. MTA and MTB are consistent with the upper and lower bands marked with red and blue dots, respectively (Fig. 2A, B). The fluorescence intensity of the bands with and without CHX treatment suggest that MTB undergoes nonsense-mediated mRNA decay (NMD). We predict that the protein changes due to c.5300–17C>A are p.[Tyr1766_Gly1767insAlaProProCysAla,Gly1768Ilefs188*]. The prediction result of SpliceAI [8] corresponded to MTA but not MTB, which was only experimentally verified.