Introduction

The ATP-binding cassette sub-family A member 4 (ABCA4) (also known as Rim protein or ABCR) is an ABC transporter1,2. It consists of 2,273 amino acids and is expressed in the rim region of rod photoreceptor outer segment disc membranes and in foveal and peripheral cone photoreceptor outer segment disc membranes3. Unlike many ABC transporters, which function as efflux pumps, ABCA4 transporter is a unique import pump4. It transports all-trans-retinal (ATR) or N-retinylidene-phosphatidylethanolamine (PE) from the luminal side to the cytoplasmic side. Functional deficiency of this transporter leads to accumulation of ATR or N-retinylidene-PE in photoreceptors and retinal pigment epithelium (RPE) cells When N-retinylidene-PE accumulates, the production of N-retinylidene-N-retinylethanolamine (A2E) increases. Further, A2E has several detrimental effects on RPE cells; RPE cells deaths is induced through the generation of reactive oxygen species and dysfunction of lysosomal degradation. Ultimately, photoreceptors are lost5.

Many genetic variants of ABCA4 are associated with various forms of retinal dystrophy, including Stargardt disease, retinitis pigmentosa, and cone-rod dystrophy1,6,7,8,9,10,11,12. In patients with ABCA4-associated retinopathy, the characteristic clinical symptoms of the disease such as macular affection, fundus flecks, and peripapillary sparing can be observed. The severity of these clinical symptoms can vary depending on the stage of the disease13. ABCA4 variants have been extensively studied in diverse populations and disease-causing ABCA4 variants vary according to race or ethnicity13.

Functional studies of the ABCA4 missense variants have been reported. For example, the transport activities of p.Gly863Ala, p.Asn965Ser, p.Lys969Met, and p.Lys1978Met variants found in Stargardt disease were significantly reduced compared with that of the wild-type ABCA4 following overexpression in HEK-293T cells4.

Although the missense variants of ABCA4 account for the largest proportion of disease-causing variants, recent studies have indicated that variants in other areas of the gene also play an important role as disease-causing variants13. For example, 64 non-canonical splice site variants indicated splicing defects, such as exon skipping, exon elongation, or intron retention, and many of these variants resulted in 100% aberrantly spliced RNA in previous studies14,15,16,17. Furthermore, 20 variants among 35 deep-intronic or near-exon variants were reported to have deleterious or severe effects in ABCA4-associated retinopathy, based on in vitro splice assays and analysis of photoreceptor progenitor cells13. In addition, 12 variants indicated a moderate effect. Recently, Bauwens et al. found that two non-coding deep-intronic variants, c.768+3223C>T and c.2919–383C>T, caused significant downregulation in their in vitro reporter studies14.

Further, the transcription factors AIRS (-762 from the position of the putative transcription initiation site), AP-1/Nrl-like (-708), Mash (-655), Ret-4 (-489), RAR/RXR (-269), and CRX/Ta (-33) were predicted to bind to the ABCA4 promoter by computational analysis using the Transfac database. In particular, the -762, -708, and -655 regions are hypothesized to have an important effect on retina function as they correspond to the region responsible for retina-specific adjustment18. Although this study highlighted the clinical importance of the ABCA4 proximal promoter, no functional studies of its genetic variants have been reported. Recently, human eye-specific cis-regulatory elements (CREs) were identified and functionally characterized using a comprehensive analysis of tissue-specific CREs, transcription factor binding, and gene expression in the retina, macula, and retinal pigment epithelium/choroid19. This study revealed that the interaction between tissue-specific CREs and transcription factors is important for the expression of genes such as ABCA4.

Stargardt disease is an autosomal recessive disease caused by ABCA4 variants (STGD1)3. Recently, Kim et al. found several ABCA4 variants in six Korean patients with Stargardt disease using targeted exome sequencing20. This study examined the effect of ABCA4 variants on ABCA4 expression. In addition, functional characterization of the genetic variants of the ABCA4 proximal promoter region determined the mechanism by which these ABCA4 variants alter promoter activity.

Results

Effects of ABCA4 variants on its expression

In a previous study, five missense variants—p.Arg290Gln, p.Asp645Asn, p.Thr1117Ala, p.Cys1140Trp, and p.Asn1588Tyr—in six Korean patients with Stargardt disease were identified through genetic analysis20. In the present study, we constructed vectors containing a wild-type ABCA4 gene and its variants and investigated ABCA4 expression levels of the variants on the plasma membrane using cell surface biotinylation assays and immunofluorescence. We observed that four of five variants had significantly decreased ABCA4 expression on the plasma membrane, whereas the expression of p.Asp645Asn was comparable with that of the wild-type (Fig. 1). In particular, the ABCA4 expression of p.Thr1117Ala, p.Cys1140Trp, and p.Asn1588Tyr was remarkably decreased by 70.9%, 40.6%, and 87.9%, respectively, compared with that of the wild-type. Further, ABCA4 expression of p.Arg290Gln was decreased by 28.4%. Proteins with trafficking defects typically undergo intracellular degradation by proteasomes or lysosomes. The ABCA4 expression of p.Arg290Gln, p.Thr1117Ala, p.Cys1140Trp, and p.Asn1588Tyr recovered to 109.8%, 72.5%, 91.2%, and 95.0%, respectively, of naïve wild-type ABCA4 expression, after treatment with the proteasomal proteolysis inhibitor MG132 (Fig. 2a). In addition, the ABCA4 expression of these variants recovered to 100.5%, 88.9%, 108.1%, and 100.9%, respectively, of naïve wild-type ABCA4 expression, after treatment with the lysosomal degradation inhibitor bafilomycin A1 (Fig. 2b). Our data suggest that p.Arg290Gln, p.Thr1117Ala, p.Cys1140Trp, and p.Asn1588Tyr are susceptible to intracellular degradation and that the decreased ABCA4 expression of these variants can be a result of proteasomal or lysosomal degradation. We also confirmed the ABCA4 expression on the plasma membrane of cells expressing each variant using immunofluorescence and observed that the ABCA4 expression of three variants, p.Thr1117Ala, p.Cys1140Trp, and p.Asn1588Tyr on the plasma membrane was markedly decreased (Fig. 3). In the case of p.Arg290Gln, although ABCA4 expression on the plasma membrane was less decreased than that of the three variants, a large faction of ABCA4 was present in the endoplasmic reticulum.

Figure 1
figure 1

Effect of ABCA4 variants on ABCA4 surface expression. HEK-293T cells were transfected with ABCA4 wild-type or variant plasmids, and a surface biotinylation assay was conducted. Images were cropped, and full-length blots were presented in Supplementary Fig. 1. Data represent the mean ± SD from three independent experiments analyzed using one-way analysis of variance followed by Dunnett’s two-tailed test. **P < 0.01, ***P < 0.001 versus wild-type.

Full size image
Figure 2
figure 2

Effect of MG132 or bafilomycin A1 on ABCA4 expression. ABCA4 expression was examined after transfection with ABCA4 wild-type or variant plasmids. Immunoblotting was performed after treatment with MG132 (a) or bafilomycin A1 (b). Images were cropped, and full-length blots were presented in Supplementary Fig. 2. Data represent the mean ± SD from three independent experiments analyzed using Student’s t test. *P < 0.05, **P < 0.01, ***P < 0.001 versus ABCA4 expression without MG132 or bafilomycin A1 treatment.

Full size image
Figure 3
figure 3

Effect of ABCA4 variants on subcellular localization. After transfection with ABCA4 wild-type or variant plasmids, HEK-293T cells were fixed, permeabilized, and immunostained using primary antibodies for anti-FLAG M2, BiP endoplasmic reticulum marker, or giantin Golgi marker. Scale bars, 10 μm.

Full size image

Frequency of ABCA4 promoter variants

Three common single nucleotide polymorphisms (SNPs) (minor allele frequency ≥ 5%)—c.-1086A>C (rs2151846), c.-900A>T (rs3789452) and c.-761C>A (rs3761906)—in the ABCA4 proximal promoter region were identified using SNP data from the database of single nucleotide polymorphisms (dbSNP) of the National Center for Biotechnology Information (NCBI) (https://www.ncbi.nlm.nih.gov/snp/). Using the rs number of each variant, we obtained the frequencies of these variants from the 1,000 Genomes Project (phase 3) (https://www.ensembl.org/) in four different ethnic groups: 661 Africans, 347 Americans, 504 East Asians, and 503 Europeans (Table 1). The frequencies of each minor allele in Koreans are similar to those of East Asians, except that the minor allele of the c.-1086A>C variant is C in all four ethnic groups, but A is the minor allele in Koreans [data obtained from 1,465 Koreans from the Korean Reference Genome (KRG) Database, http://152.99.75.168:9090/KRGDB/]. In addition, four haplotypes were assembled (Table 2). Haplotype 4 (H4) was used as the wild-type in this study according to the ABCA4 mRNA sequence (GenBank accession number; NM_000350.3). In addition, H1 consisted of a c.-761C>A variant, whereas H2 and H3 were composed of c.-1086A>C and c.-900A>T and c.-1086A>C, respectively (Table 2).

Table 1 Frequencies of ABCA4 proximal promoter variants.
Full size table
Table 2 Frequencies of ABCA4 proximal promoter haplotypes.
Full size table

Effects of ABCA4 variants on promoter activity

H1 and H3 haplotypes indicated significantly altered luciferase activities compared with that of the H4 wild-type (P < 0.01 and 0.001, respectively); H1 had 22.8% increased luciferase activity, while H3 had 37.0% decreased luciferase activity (Fig. 4). The luciferase activity of H2 was comparable with that of the wild-type. In addition, the variant in H2, c.-900A>T, indicated a 27.0% increase in luciferase activity compared with that of the wild-type (Fig. 4).

Figure 4
figure 4

Effect of ABCA4 promoter variants on luciferase activity. Luciferase activity was measured after transfection of wild-type ABCA4 reporter plasmid or reporter plasmids containing ABCA4 variants into HCT-116 cells. The luciferase activity of each variant was compared with that of wild-type ABCA4. The data (mean ± SD) represent triplicate measurements from a representative experiment. **P < 0.01, ***P < 0.001 versus wild-type.

Full size image

Transcription factors involved in regulating ABCA4 promoter activity

ConSite (http://consite.genereg.net/cgi-bin/consite) and MatInspector (Genomatrix Software GmbH, Munich, Germany) was used to predict transcription factors binding the ABCA4 promoter near the region harboring the three variants: c.-1086A>C, c.-900A>T, and c.-761C>A. Further, GATA binding protein 2 (GATA-2) and hepatic leukemia factor (HLF) were predicted to bind to the ABCA4 promoter region and exhibited a large difference in their binding affinity between the wild-type and variant sequences. This was validated using gel shift assays with unlabeled GATA-2 or HLF consensus oligonucleotides and antibodies against GATA-2 or HLF. GATA-2 more strongly bound to the wild-type ABCA4 promoter with c.-1086A, by 2.22-fold (SD = 0.36), than the one with c.-1086C variant (lanes 4 and 7, Fig. 5a). HLF bound to the ABCA4 promoter region near the c.-900A>T variant and more strongly bound to the c.-900T variant, by 1.92-fold (SD = 0.52) than to the c.-900A wild-type (lanes 4 and 7, Fig. 5b). Additionally, GATA-2 bound to the ABCA4 promoter region near the c.-761C>A variant and more strongly bound to the c.-761A variant, by 1.67-fold (SD = 0.23), than to the c.-761C wild-type (lanes 4 and 7, Fig. 5c).

Figure 5
figure 5

Gel shift assays. (a) 32P-labeled oligonucleotides (lanes 1–3, GATA-2 consensus; lanes 4–6, wild-type c.-1086A; lanes 7–9, c.-1086C variant) were incubated with nuclear protein extracts. Competition and supershift assays were performed using 100-fold molar excess of unlabeled GATA-2 consensus oligonucleotides (lanes 2, 5, and 8) and GATA-2 antibody (lanes 3, 6, and 9), respectively. (b) 32P-labeled oligonucleotides (lanes 1–3, HLF consensus; lanes 4–6, wild-type c.-900A; lanes 7–9, c.-900T variant) were incubated with nuclear protein extracts. Competition and supershift assays were performed using 100-fold molar excess of unlabeled HLF consensus oligonucleotides (lanes 2, 5, and 8) and HLF antibody (lanes 3, 6, and 9), respectively. (c) 32P-labeled oligonucleotides (lanes 1–3, GATA-2 consensus; lanes 4–6, wild-type c.-761C; lanes 7–9, c.-761A variant) were incubated with nuclear protein extracts. Competition and supershift assays were performed using 100-fold molar excess of unlabeled GATA-2 consensus oligonucleotides (lanes 2, 5, and 8) and GATA-2 antibody (lanes 3, 6, and 9), respectively. Each arrow indicates the band of the DNA–protein complex. Full film data were presented in Supplementary Fig. 3.

Full size image

Effects of GATA-2 or HLF on ABCA4 promoter activity

The effect of GATA-2 or HLF transcription factors on ABCA4 promoter activity was examined by measuring luciferase activity after co-transfection of wild-type or variant ABCA4 promoter vectors and each transcription factor into HCT-116 cells. Both transcription factors increased ABCA4 luciferase activity in a dose-dependent manner (Fig. 6). These results suggest that GATA-2 and HLF function as gene activators of ABCA4 transcription.

Figure 6
figure 6

Effect of transcription factors on ABCA4 promoter activity. Luciferase activity was measured 48 h after co-transfection of wild-type ABCA4 or its variant reporter plasmids and varying amounts of GATA-2 (a and c) or HLF cDNA (b). The reporter activity of each construct was compared with naïve luciferase activity. The data (mean ± SD) represent triplicate measurements in a representative experiment. **P < 0.01, ***P < 0.001 versus naïve luciferase activity.

Full size image

Discussion

This study examined the effect of p.Arg290Gln, p.Asp645Asn, p.Thr1117Ala, p.Cys1140Trp, and p.Asn1588Tyr missense variants on ABCA4 expression. These variants were found in six Korean patients with Stargardt disease, and three of them (p.Arg290Gln, p.Thr1117Ala, and p.Asn1588Tyr) were novel in a previous study20. The p.Asp645Asn and p.Cys1140Trp variants were previously found in Korean or Chinese patients with Stargardt disease21,22. Our study indicated that the significantly decreased ABCA4 expression of p.Arg290Gln, p.Thr1117Ala, p.Cys1140Trp, and p.Asn1588Tyr variants on the plasma membrane could be due to intracellular degradation. In particular, p.Thr1117Ala, p.Cys1140Trp, and p.Asn1588Tyr indicated remarkably decreased ABCA4 expression, whereas ABCA4 expression of p.Asp645Asn was comparable with that of the wild-type. In a previous study, p.Asp645Asn indicated an increased affinity for ATP compared with that of the wild-type in ATP-labeling experiments23. Additional functional experiments including ATPase activity assays and molecular mechanism experiments are required to determine the effect of this variant on ABCA4 functions. Recently, Runhart et al. reported that a mild variant of ABCA4, p.Asn1868Ile, had incomplete penetrance and could result in late-onset Stargardt disease when it presents in trans with other severe ABCA4 variants24,25. Therefore, even if p.Asp645Asn is insufficient to cause loss-of-function of ABCA4, it can cause a disease phenotype when it is associated with another severe ABCA4 variant. In previous studies, all Korean patients with Stargardt disease were heterozygotes for two ABCA4 variants20,26. The patient with p.Asp645Asn had another variant, p.Lys2049ArgfsTer12. Genetic testing of the parents of this patient was not performed. However, the family tree indicated that the parents were not affected by the disease, and the patient did not have any other ABCA4 variant other than the two variants. In addition, p.Lys2049ArgfsTer12 was present with the p.Thr1117Ala variant in another Stargardt disease patient. In this family, the patient's parents were not affected by the disease as well, and the patient did not have any other ABCA4 variant other than the two variants26. Further, p.Lys2049ArgfsTer12 was reported in two Korean patients with Stargardt disease in another study21. Based on these findings, it can be inferred that p.Lys2049ArgfsTer12 exists in trans with p.Asp645Asn in a patient with Stargardt disease. In a previous in silico analysis, p.Lys2049ArgfsTer12 was categorized as a severe variant27.

Recently, p.Arg290Gln and p.Asn1588Tyr were categorized as mild/moderate variants whereas p.Cys1140Trp was categorized as a causative variant of unknown severity using in silico analysis27. In previous studies, patients carrying the p.Arg290Gln or p.Asn1588Tyr variants had another ABCA4 variant, p.Leu1157Ter. In another patient, the p.Cys1140Trp variant existed in trans with an in-frame ABCA4 variant, p.Ile1114del26. To date, the severity of these two variants, p.Leu1157Ter and p.Ile1114del, has not been reported. In a previous study, the researchers investigated a genotype–phenotype correlation in their patients26. They observed that patients with p.Asp645Asn, p.Thr1117Ala, p.Cys1140Trp, or p.Asn1588Tyr variant indicated typical features of Stargardt disease, flecks, bilateral macular atrophy, and blocked fluorescence in fluorescein angiography. The patient with p.Arg290Gln exhibited bull’s-eye maculopathy without flecks. The age of onset was 11–16 years, and the median best-corrected visual acuity was 20/320. To determine whether the ABCA4 variants examined in this study have a significant effect on the phenotype of Stargardt disease, analyzing the phenotype according to the genotype, with a larger number of patients, is necessary.

Allikmets et al. reported several putative important sites in the ABCA4 promoter using the Transfac transcription factor database18. Although photoreceptor-specific elements were not identified in the ABCA4 promoter region, similar sequences were found in the regulatory regions of other photoreceptor genes including rhodopsin and S-antigen/arrestin. However, the functional roles of these sites were not examined using molecular experimental tools. Recently, Cherry et al. identified and functionally characterized human eye-specific CREs19. In their study, a promoter variant (rs11802887) and an enhancer variant (rs752024867) found in patients with Stargardt disease indicated significantly increased or decreased reporter activity, respectively, although both were not within the readily recognizable transcription factor-binding motif.

This study investigated the effect of genetic variants in the ABCA4 proximal promoter on ABCA4 transcription using in vitro assays. The luciferase reporter assay indicated that c.-1086A>C, c.-900A>T, and c.-761C>A variants had significantly altered luciferase activities compared with that of the wild-type. We found that GATA-2 was bound to the ABCA4 promoter region near the c.-1086A>C or c.-761C>A variant site. GATA-2 plays an important role in the development of various tissue types. The GATA family is originally divided into the hematopoietic- (GATA-1/2/3) or cardiac (GATA-4/5/6) family of transcription factors and several variants in GATA2 are associated with various blood diseases28,29,30,31,32. Further, GATA-2 plays an important role in other tissues, including retina, central nervous system, and prostate33,34,35. In particular, a previous study revealed that GATA-2 is an essential transcription factor for the transcription of neuroglobin, which is mainly expressed in the retina and brain35. Regarding the effect of GATA-2 on the regulation of transporters, GATA-2 plays an important role in maintaining water homeostasis in the body by controlling aquaporin 2 expression in the collecting tubules36. This study found that the sequence of wild-type c.-1086A (ACTATCTCT) is similar to the consensus sequence (WGATAR) of GATA-2. Our prediction that the c.-1086C variant (ACTCTCTCT sequence) reduces GATA-2 binding affinity, because it is less consistent with the GATA-2 consensus sequence, was confirmed using a gel shift assay. The c.-761A variant also had a sequence similar to the consensus sequence of GATA-2, GGGATGGAA. The wild-type c.-761C contains GGGCTGGAA, which is hypothesized to lower the binding affinity of GATA-2, and this was confirmed using a gel shift assay. In addition, we observed that GATA-2 acts as an activator of the ABCA4 promoter, similar to the role in the transcription of neuroglobin. Our data suggest that the decreased luciferase activity of c.-1086A>C results from the reduced binding of GATA-2, while the increased luciferase activity of c.-761C>A results from the increased binding of GATA-2.

HLF (originally known as E2A-HLF) in childhood acute lymphocytic leukemia is a chimeric transcription factor produced by the translocation of E2A on chromosome 19 and HLF on chromosome 1737. Although the t (19:17) E2A-HLF leukemia subtype is rare, it has a high mortality rate because it is resistant to chemotherapy38. HLF is mainly expressed in the liver, lungs, kidneys, and neurons, but not in normal blood cells39. The role of HLF in the body is not clearly elucidated although it is involved in activating coagulation factor VII and factor IX genes and in synapse formation in the central nervous system in a mouse model39,40. In a previous study, we revealed that HLF binds to the promoter region of lactosylceramide α-2,3-sialytransferase 5 (ST3GalV) and regulates the transcription of this gene41. ST3GalV is an enzyme involved in the production of ganglioside in the body. It is distributed throughout the body and is particularly highly expressed in the central nervous system42,43. In this study, we observed that the sequence of the c.-900T variant (GATAACACA) was similar to the consensus sequence of HLF (RTTACRYAAT). In addition, it was predicted that the binding affinity for HLF is weaker in the wild-type c.-900A because the wild-type sequence (GAGAACACA) is not as similar to the consensus sequence of HLF; this prediction was confirmed using a gel shift assay. Furthermore, HLF activated the ABCA4 promoter. These results suggest that the increased luciferase activity of c.-900A>T is a consequence of the stronger binding between the c.-900T variant and HLF.

This study revealed that several ABCA4 variants found in patients with Stargardt disease could affect ABCA4 expression, and common variants of the ABCA4 proximal promoter alter the ABCA4 transcriptional activity, which is regulated by GATA-2 and HLF. In particular, the H3 haplotype reduced promoter activity; therefore, investigating whether this haplotype is associated with the ABCA4-associated retinopathy, including Stargardt disease, through presenting with other ABCA4 variants, is necessary.

Methods

Plasmid construction

A reporter plasmid containing the ABCA4 wild-type promoter sequences (− 1,116 to + 50 bp from the translational start site of ABCA4) was amplified and inserted into the pGL4.11b[luc2] vector (Promega Corporation, Madison, WI, USA). A plasmid containing the wild-type ABCA4 cDNA (Horizon Discovery, Cambridge, UK) was subcloned into the p3XFLAG-CMV vector. Variant-bearing plasmids were generated using the QuikChange® II site-directed mutagenesis kit (Agilent Technologies, Santa Clara, CA, USA). Primers used for the construction of a reporter plasmid or variant-bearing plasmids are listed in Supplementary Table 1. All the DNA sequences were confirmed using direct sequencing.

Surface biotinylation assay

After transfection of the ABCA4 wild-type or variant-bearing plasmids into HEK-293T (human embryonic kidney) cells using Lipofectamine LTX and Plus reagents (Thermo Fisher Scientific, Waltham, MA, USA), biotinylation assays were conducted using a cell surface protein isolation kit (Thermo Fisher Scientific) according to the manufacturer's protocol. Protein samples from biotinylation assays were subjected to immunoblotting. Rabbit polyclonal anti-Na+/K+ ATPase α-1 antibody (Merck, Kenilworth, NJ, USA) was used as the internal standard. The signal was acquired using an Amersham ImageQuant 800 biomolecular imager (Cytiva, Marlborough, MA, USA) and the intensity of each band was measured using ImageJ software (National Institute of Health, Bethesda, MD, USA).

Immunoblotting

Immunoblotting was performed using mouse anti-FLAG M2 primary antibody (Merck) or goat anti-β-actin antibody (Santa Cruz Biotechnology, Dallas, TX, USA) and the corresponding secondary antibodies. Fifteen micrograms of protein was loaded on mini-protein TGX gels (Bio-Rad, Hercules, California, USA) and transferred using a trans-blot turbo RTA transfer kit and trans-blot turbo transfer system (Bio-Rad). Cells were treated with 10 µM MG132 (Sigma-Aldrich, Burlington, MA, USA) or 10 nM bafilomycin A1 (MedChemExpress, Monmouth Junction, NJ, USA) 24 h after transfection to examine their effects on ABCA4 variants. Membranes were developed with ECL detection reagents (Thermo Fisher Scientific). The signal was acquired using an Amersham ImageQuant 800 biomolecular imager and the intensity of each band was measured using ImageJ software.

Immunofluorescence

HEK-293T cells were seeded into 4-well chamber slides (Thermo Fisher Scientific) and the ABCA4 wild-type or variant-bearing plasmids were transfected into those cells. After transfection, the cells were fixed and permeabilized with 100% methanol (prechilled at − 20 °C) at room temperature for 5 min and blocked with 1% bovine serum albumin for 30 min. To examine the intracellular localization of ABCA4, cells were incubated with anti-FLAG M2, anti-BiP (Abcam, Waltham, MA, USA), or anti-giantin (Abcam) antibodies. The cells were then washed thrice with phosphate-buffered saline, and incubated with Alexa Fluor® 488 rabbit anti-mouse IgG or Alexa Fluor® 594 goat anti-rabbit IgG (Life Technologies) secondary antibodies. Nucleic acids were stained with DAPI (4’,6-diamidino-2-phenylindole, Vector Laboratories, Burlingame, CA, USA). Images were obtained using a confocal laser scanning microscope and analyzed using an LSM image examiner (Carl Zeiss, Oberkochen, Germany).

Genetic analysis of the ABCA4 proximal promoter

To identify common SNPs in the proximal region of ABCA4 promoter, SNP data were used from the dbSNP of the NCBI. Next, the frequency data from the 1,000 Genomes Project (phase 3) in four different ethnic groups—661 Africans, 347 Americans, 504 East Asians, and 503 Europeans—were used to identify the frequencies of these variants. The frequencies of the minor alleles in each variant in Koreans were obtained from the KRG database. The haplotype frequencies were determined using Haploview software version 4.3 (Broad Institute, Cambridge, MA, USA). Nucleotide location numbers were assigned from the translational start site based on the ABCA4 mRNA sequence (GenBank accession number NM_000350.3).

Promoter activity

GATA-2 cDNA or HLF cDNA (Horizon Discovery) was subcloned into the pcDNA3.1 ( +) vector (Life Technologies, Carlsbad, CA, USA). Varying amounts of GATA-2 or HLF plasmids were co-transfected with the ABCA4 reporter plasmid (wild-type or variants) into HCT-116 (human colon carcinoma) cells using Lipofectamine LTX and Plus reagents. After transfection, luciferase activity was measured using the Dual Luciferase® Reporter Assay System (Promega Corporation) according to the manufacturer’s protocol and quantified using a luminometer (Promega Corporation).

Gel shift assay

Gel shift assays were performed as previously described44. In this study, 10–20 μg of a nuclear extract from HCT-116 cells was incubated with 32P-labeled (1 × 105 or 2 × 105 counts/min) oligonucleotides. Each sample was then subjected to electrophoresis for 90–150 min at 80 V, and the CP-BU film was exposed to the dried gel (Agfa, Mortsel, Belgium) at − 80 °C for 16 h. Competition assays involved adding 100-fold molar excess of unlabeled GATA-2 or HLF consensus oligonucleotides prior to the binding reaction. Antibodies (3–6 µg) against GATA-2 (Santa Cruz Biotechnology) or HLF (Merck) were used for the supershift assay. Band intensities were measured using ImageJ software. Oligonucleotides used in the gel shift assays are listed in Supplementary Table 1.

Statistical analyses

Statistical analyses were performed using GraphPad Prism 8.0 (GraphPad Software Inc., San Diego, CA, USA). P values of the results obtained before- and after MG132 or bafilomycin A1 treatment were calculated using Student’s t test for comparison. Other P values were calculated using one-way analysis of variance, followed by Dunnett’s two-tailed test. P < 0.05 was considered statistically significant.