Non-coding single-nucleotide and structural variants affecting the EYS putative promoter cause autosomal recessive retinitis pigmentosa

Abstract

Purpose:

Variants in untranslated genomic regions are difficult to identify as pathogenic but are capable of causing disease by interfering with gene expression. This study aimed to characterize the effect of variants identified in the 5′-untranslated region of EYS in patients with autosomal recessive retinitis pigmentosa (RP).

Methods:

Variant screening included gene panels, Sanger, exome, and genome sequencing. Functional validation included an electrophoretic mobility shift assay and various luciferase assays.

Results:

Patients with RP from 6 EYS biallelic Arab-Muslim families harbored a 5′ noncoding EYS variant, c.−453G>T, and 4 harbored a structural variant affecting the 5′ noncoding exons. Electrophoretic mobility shift assay analysis revealed an effect on binding of transcription factors for c.−453G>T and a neighboring variant c.−454G>T. Dual luciferase assays using overexpression of various transcription factors showed distinct effects on expression. c.−453G>T was associated with higher luciferase expression with CRX overexpression and c.−454G>C with OTX2 overexpression. In addition, the 2 variants were found to influence translation by affecting upstream initiation codons. Interestingly, visual function of EYS RP patients who harbor c.−453G>T are better than those with biallelic null EYS variants.

Conclusion:

Our analysis revealed both single-nucleotide and structural variants in the EYS promoter as the cause of autosomal recessive RP. These variants may affect EYS expression via a dual mechanism by altering transcription factor binding affinity at the EYS promoter and by affecting upstream open reading frames.

Introduction

Inherited retinal diseases (IRDs) are a group of genetic disorders with a wide array of clinical presentations and genetic causes. Uncovering the cause of IRDs is often challenging, and even by using exome sequencing (ES), the diagnostic yield reaches approximately 50% to 60%.^1,2 The unsolved cases are due to several factors, including novel genes and elusive variants, such as regulatory variants that affect gene expression.^3–5 Additionally, some variants may fail to be identified by the current filtering methods of next-generation sequencing (NGS) data. Genome sequencing (GS) raises the yield slightly; however, many of the non-coding variants are classified as variants of uncertain significance, which often require additional functional studies to establish their pathogenic effect. A region of interest in this regard is the 5′-untranslated region (UTR), which includes internal ribosome entry sites, upstream open reading frames (uORFs), microRNA binding sites, and regulatory regions, such as promoter sites and enhancers that bind transcription factors (TFs) and transcription initiation complex proteins.⁶ These regulatory elements usually influence transcription but might also affect translation via uORFs.

In the majority of cases, ES kits include only partial coverage of 5′-UTR regions. In IRD genes, only 3% to 20% of the 5′-UTR regions are covered by the most commonly used ES enrichment kits.⁷ GS and specifically designed panels can provide coverage of these regions. Because of the limitations of the coverage and necessity of functional assays to determine pathogenicity, a relatively low number of such variants have been described as disease causative in IRDs^7–12 and variants in only 5 genes have been reported in IRD promoter regions in BBS10¹⁰ (HGNC:26291), MERTK⁷ (HGNC:7027), NMNAT1¹¹ (HGNC:17877), PRPF31⁹ (HGNC:15446), and TMEM216¹² (HGNC:25018).

The presence of uORFs in 5′-UTRs has proven to have relevance to inherited diseases. uORFs have an initiation codon, which is upstream to the primary ORF and in-frame with a stop codon. uORFs may overlap with the primary ORF (pORF),⁷ and these uORFs can decrease pORF protein expression by as much as 80%.¹³ Variants perturbing uORFs are an underrecognized disease-causing mechanism.¹⁴ In ocular diseases, variants affecting an existing uORF or introducing a novel uORF have been shown to cause aniridia and macular dystrophy through decreased translation of the pORF of PAX6⁸ (HGNC:8620) and RDH12⁷ (HGNC:19977), respectively.

EYS (HGNC:21555) is specifically expressed in photoreceptors,^15,16 and no orthologs are present in the mouse genome and some other species.¹⁵ Pathogenic EYS variants cause mainly retinitis pigmentosa (RP) (OMIM 268000) and are reported to also cause cone-rod dystrophy and macular dystrophy constituting a relatively large proportion of IRD cases.^17,18 EYS is one of the largest human genes and the largest identified IRD-associated gene consisting of 43 exons of which the first 3 and part of the fourth are noncoding. Most of the variants identified in EYS are loss of function, including nonsense, frameshift, and most of the splice-site variants¹⁸ (https://databases.lovd.nl/shared/genes/EYS).

A cis-regulatory analysis of EYS showed that part of the 5′-UTR, including exons 1 and 2, contains the putative active EYS promoter as indicated by binding of photoreceptor TFs, such as OTX2 (HGNC:8522), CRX (HGNC:2383), and NRL (HGNC:8002).^19,20 In this study, we have identified novel noncoding variants, including single-nucleotide variants (SNVs) and copy-number variations (CNVs), and used functional assays to show their effect on gene expression by affecting the promoter site and upstream initiation codons in the 5′-UTR.

Materials and Methods

A more detailed description of the methods can be found in the Supplemental Methods section online.

Patient recruitment and clinical evaluation

This study adhered to ethical standards, receiving approval from the Hadassah Hospital Institutional Review Board. A full ophthalmological examination, including imaging and electroretinography, was performed.

Genetic analysis

DNA samples were screened using various methods, including the BluePrint IRD panel (351 genes), single-molecule molecular inversion probes IRD panel (113 genes), ES, and GS. Homozygosity mapping was performed on ES and GS data of 3 homozygous cases. The following NM (NM_001142800.2) was used to describe all EYS variants identified, and variant calling used the GRCh38/hg38 reference genome from the University of California, Santa Cruz genome browser.²¹

Electromobility shift assay

Intact nuclei were isolated from frozen bovine retina as previously reported.²² A region of 35 base pairs (Supplemental Table 1) of the 5′-UTR region of EYS was designed for the assay.

Luciferase assay for promoter region

Based on previous analyses,¹⁹ a chromatin accessible region was identified which included the variants NM_001142800.2:c.−453G>T and c.−454G>C and which is a binding region for several TFs including OTX2, CRX, and NRL. The PGL4.23 vector was used to test the effect of the EYS promoter construct inserted immediately upstream to the minimal promoter for Firefly luciferase (Supplemental Methods).

Luciferase assay for upstream initiation codons

A dual luciferase assay was performed to evaluate the impact of the variants on upstream initiation codons. The detailed protocol can be found in the Supplemental Methods section.

Results

EYS sequence pathogenic variants in the 5′-UTR

NGS analysis of IRD gene panels identified 3 cases that were homozygous for the EYS c.−453G>T variant (MOL0991–1, MOL1935–1, and TB328) and 3 cases that carried this variant in a heterozygous state and were also heterozygous for a second EYS pathogenic variant (MOL1547–1, MOL1673–1, and MOL2084–1) (Supplemental Figures 1A and 2). The second hits in the compound heterozygous cases were NM_0011428 00.2:c.1765A>G p.(Arg589Gly), c.4361_4362delinsAG p.(Ser1454*), and c.(2137+1_2138–1)_(2846+1_2847)del (including exons 14–18), respectively (Table 1^23–26). Although c.1765A>G is predicted to result in a missense change, we used a splice assay to show that it affects splicing and is likely to be null (Supplemental Figure 3 and Supplemental Results). All cases with the c.−453G>T variant were of Arab-Muslim origin, suggesting a founder effect. This was confirmed by identification of a shared homozygous region between 3 cases homozygous for the variant and a shared haplotype (Supplemental Figure 4).

Table 1.

List of arRP cases with variants affecting EYS 5′-UTR

Patient Number	EYS Variant 1	EYS Variant 2	Zygosity	Origin	Detection Method
MOL0991–1	c.−453G>T g.65707140C>A p.?, (VUS – PM223)	c.−453G>T g.65707140C>A p.?	Homozygous	Arab-Muslim	smMIPs
MOL1547–1	c.−453G>T g.65707140C>A p.?	c.1765A>Ga g.65334981T>C p.(Arg589Gly), (LP – PM3, PM2)	Heterozygous	Arab-Muslim	Gene panel
MOL1673–1	c.−453G>T g.65707140C>A p.?	c.4361_4362delinsAG g.64591505_64591506delinsCT p.(Ser1454*), (P – PM3, PVS1, PM2)	Heterozygous	Arab-Muslim	Gene panel
MOL1935–1	c.−453G>T g.65707140C>A p.?	c.−453G>T g.65707140C>A p.?	Homozygous	Bedouin	smMIPs
MOL2084–1	c.−453G>T g.65707140C>A p.?	c.(2137+1_2138–1)_ (2846+1_2847–1)delb g.(64886843_64902112) _(64997704_65057613)del p.? (VUS:0.45 2E24)	Heterozygous	Arab-Muslim	Gene panel
TB328	c.−453G>T g.65707140C>A p.?	c.−453G>T g.65707140C>A p.?	Homozygous	Arab-Muslim	GS
MOL1760–1	g.(65639892_65707133) _(65707227_?)del p.?c, (VUS:0.3 – 2C)	g.(65639892_65707133) _(65707227_?)del p.?c	Homozygous	Moroccan Jewish	Gene panel
DNA09–03009	c.−448+11007_863–28730dup g.65434097_65696128dup p.?d, (VUS:0.3 – 2K)	NM_001292009.2:c.9403_9404insN [376] g.63720625_63720626insN [376] p.(Tyr3156Glyfs*40)e, (VUS:0.45 – 2H, 4L)	Compound heterozygous	Dutch	GS
DNA19–14872 (reported as arRP16 by Reurink et al25	g.65570161_66472034del p.?f, (VUS:0.4 – 2C, 4L)	g.65570161_66472034del p.?f	Homozygous	Dutch	GS
071486 (reported as arRP6 by Reurink et al25	g.65658171_65718920del p.?g (VUS:0.3 – 2C)	c.2527G>A g.64912598C>T p.(Gly843Arg), (LP – PM2, PM5, PP5)	Heterozygous	Swiss	GS
RP228B (reported by Barragán et al26)	c.−454G>Ch g.65707141C>G p.?, (VUS – PM2)	c.−454G>C g.65707141C>G p.?	Homozygous	Spanish	PCR-based direct genomic sequencing of EYS

All variants listed in the table correspond to NM_001142800.2 except where otherwise indicated and NC_000006.12.

arRP, autosomal recessive retinitis pigmentosa; Del, deletion; GS, genome sequencing; LP, likely pathogenic; P, pathogenic; PCR, polymerase chain reaction; SINE, short interspersed nuclear element; smMIP, single molecule molecular inversion probes; UTR, untranslated region; VUS, variant of unknown significance.

^aThis variant affects the splicing of exon 11 (see Supplemental Results for more information).

^bDeletion of exons 14 to 18.

^cDeletion of exon 1 with unknown borders.

^dTandem duplication of exons 2 to 5.

^eInsertion of 376 base pairs of mobile elements: 82.71% SINE and 14.10% simple repeat. Inserted sequence can be seen in Supplemental Figure 7.

^fPartial deletion of 5′-UTR (involving exons 1 and 2).

^gDeletion of exon 1.

^hReported as c.−462G>C.

Interestingly, a previously reported adjacent variant, c.−454G>C (originally named c.−462G>C), classified as a variant of uncertain significance was identified in a homozygous state in a Spanish patient and reported as a possible cause of RP.²⁶ This variant was included in our functional analyses.

In addition to the SNV, we also identified CNVs involving 5′-UTR exons in 4 patients with RP (Table 1,^23–26 Supplemental Figures 5-7). These include 2 patients with nonidentical deletions of exon 1 (one of which has known borders—patient 071486 in Table 1^23–26), a patient with a homozygous deletion involving exons 1 and 2 (Supplemental Figure 6) and a patient with a tandem duplication of exons 2 to 5 in trans with an additional EYS pathogenic variant NM_001142800.2:c.9403_9404insN [376] p.(Tyr3156Glyfs*40) (Supplemental Figure 7). In all of the heterozygous CNV cases, we identified a second previously reported pathogenic variant.

To shed light on the effect of the 2 SNVs (c.−453G>T and c.−454G>C) on the function of the EYS 5′-UTR, we performed 3 sets of experiments to examine the regulatory potential of this region as a promoter and as a region containing upstream initiation codons that affect protein expression.

Electromobility shift assay

Aiming to study if the variants c.−453G>T and c.−454G>C affect TF binding and thereby the function of the EYS promoter, we performed electromobility shift assay using nuclear extracts from bovine retina. The analysis showed that the 35-bp fragment used (containing both variant sites) binds retinal proteins. Additionally, for the 2 variants, a prominent high-molecular-weight band that maintained a significant level of specificity as it got competed out by the wild-type (WT) oligo failed to get competed out by either of the variant oligos (Figure 1). When comparing between the WT probe and c.−454G>C probe conditions, we observed that the WT oligo showed 4 bands (1 major band and 3 lighter ones), whereas the c.−454G>C variant oligo showed only a single prominent band indicating a stark difference in the ability to bind to proteins in the nuclear extract. Using excess cold oligo, we showed that this single band gets effectively competed out by the WT, c.−453G>T variant and c.−454G>C variant oligos. Overall, this indicates that the c.−454G>C variant severely compromises protein binding (Figure 1), whereas the c.−453G>T variant has a more minor effect in this assay.

Figure 1. Electromobility shift assay for EYS NM_001142800.2:c.−453G>T (A) and c.454G>C (B).

The 35-bp oligonucleotide probes used in the experiment were end labeled with P-32 and incubated with bovine retina nuclear extracts and run on 8% nondenaturing tris-borate-EDTA gels. Every labeled probe was made to compete against excess cold probe. (A) shows EMSA gel profile for the probes containing wild-type (WT) and c.−453G>T variants. The top shifted band (arrow in lane 2) is competed out with the WT oligo (lane 3) but fails to get competed out by the oligo with c.−453G>T or c.−454G>C variants (lanes 4–5) indicating the specificity of the DNA protein interaction there. For the EMSA with the c.−453G>T variant, the top prominent band (lane 7) gets competed out with all 3 variants (lanes 8–10). (B) shows the EMSA gel profile for the oligos containing WT and c.−454G>C variant. The c.−454G>C variant shows only a single prominent shifted band (lane 7), which gets competed out by itself, as well as with the WT and c.−453G>T variant oligos (lanes 8–10). EMSA, electrophoretic mobility shift assay.

TF binding analysis using luciferase assay

Previous studies have shown that the exon 1 putative promoter region binds OTX2, CRX, NRL, and other TFs¹⁹ (Supplemental Figure 1B). In addition, using Jaspar, we calculated the effect of the 2 variants on the frequency matrix score of OTX2. Both variants increase the score by 5% for c.−453G>T and 21% for c.−454G>C, suggesting an increase in affinity for OTX2 binding (Supplemental Figure 8). On the other hand, the binding sequences of CRX, NRL and NR2E3 (HGNC:7974) are not affected by these variants. Aiming to examine this experimentally, we performed a dual luciferase assay on HeLa cells cotransfected with a Firefly luciferase expression plasmid cloned with the EYS exon 1 region (including the entirety of exon 1 and 109 base pairs upstream and 345 base pairs downstream for the WT, c.−453G>T, and c.−454G>C sequences), a TF overexpression plasmid coding for OTX2, CRX, NRL, and NR2E3 transfected both separately and combined, as well as a Renilla expression plasmid used as a control for transfection efficiency. In the absence of TF overexpression, all variant types caused significantly higher luciferase expression (Figure 2, Supplemental Tables 2 and 3). This indicates that the endogenous TFs expressed in HeLa cells bind to the cloned EYS region that is likely to function as part of the EYS promoter. All cases of TF overexpression caused higher luciferase expression in the presence of the EYS construct compared with when there was no overexpression. However, there was a difference between variants in some cases. Among the 4 studied TFs, the most dramatic effect was seen with OTX2, with the WT region showing an approximately 20-fold increase in luciferase expression compared with the control, whereas CRX showed an increase of up to 3-fold only (Figure 2).

Figure 2. Dual luciferase reporter assay with and without TF overexpression.

Results were presented as the fold change of normalized luciferase expression compared with the control condition without transcription factor overexpression. In the no TF condition, the luciferase expression of each variant was compared with the control condition in which the EYS construct was not transfected. P values were calculated using the Mann-Whitney test (n = 9 for all panels except for no TFs for which n = 45). TF, transcription factor; WT, wild-type

In the presence of OTX2 overexpression, the c.−454G>C variant showed a significant 27% increase in normalized luciferase expression compared with WT (P = .0078), whereas no effect was measured for c.−453G>T, which showed a similar expression level to that of WT. CRX overexpression induced higher luciferase expression in the c.−453G>T variant compared with the WT and c.−454G>C (+25%, P = .019 and +23%, P = .011, respectively). Overall, the normalized luciferase expression levels seen with CRX overexpression were about 7 times lower than those seen in OTX2 overexpression. NRL and NR2E3 overexpression had a much lower effect on luciferase expression in the presence of the studied 5′-UTR region.

EYS variants effect on upstream initiation codons

Variant c.−453G>T is located between 2 upstream start codons: an uATG with a predicted uORF of 84 nucleotides and an uTTG with a predicted uORF of 138 nucleotides (Figure 3A) (uORFdb).²⁷ Both uORFs have a predicted adequate Kozak sequence and are nonoverlapping with the EYS pORF. The c.−453G>T variant alters the Kozak sequence of the uATG from adequate to weak as the presence of a G at position +4 in the Kozak sequence is optimal for translation initiation.^13,28 The previously reported c.−454G>C variant changes the uATG to an uATC, which is considered a less efficient start codon.²⁹ uORFs are known to regulate translation of the pORF.¹³ Therefore, we hypothesized that both variants may affect ribosomal recognition of these uORFs and consequently affect EYS expression (Figure 3A). To evaluate this, dual luciferase assays were performed in ARPE-19 cells using a luciferase construct that contained the 5′-UTR of EYS cloned immediately upstream of the Renilla luciferase gene, and a downstream Firefly gene for normalization (Figure 3B). We compared luciferase activity of the WT construct with 4 different mutant constructs. Two constructs contained the variants identified in patients with IRD (c.−453G>T and c.−454G>C), whereas 2 other constructs contained variants disrupting the upstream initiation codons (NM_001142800.2:c.−456A>T and c.−451T>C) (Figure 3A).

Figure 3. Upstream open reading frame (ORF) dual luciferase assay.

A. Diagram of part of EYS exon 1 showing the upstream start codons (ATG and TTG at positions chr6:65,707,143 and chr6:65,707,139, respectively, (GRCh38/hg38)) and the 4 variants evaluated through luciferase assays. These include 2 variants identified in patients with IRD (c.−453G>T and c.−454G>C in bold) and 2 variants affecting the upstream initiation codons (c.−456A>T and c.−451T>C) (NM_001142800.2). Below is a schematic showing how upstream initiation codons are predicted to be translated in upstream ORFs (uORFs) that, when recognized, can lead to reduced translation of the primary ORF (pORF). B. Structure of the dual luciferase reporter vector, including all 4 noncoding exons (=whole 5′-UTR) cloned upstream of the Renilla reporter gene, and Firefly for normalization. C. Fold change of luciferase activity for each variant compared with WT. All variants result in a significant increase in luminescence of up to 27% (P < .001). D. Fold change of luciferase mRNA levels as measured by qRT-PCR for each variant compared with WT. None of the variants affected mRNA levels, suggesting an effect at the translational level. IRD, inherited retinal disease; qRT-PCR, quantitative reverse transcription-polymerase chain reaction; UTR, untranslated region; WT, wild-type.

All variants demonstrated a significant (P < .001) increase in luciferase signal up to 27%, suggesting a similar mechanism of action (+27% for c.−456A>T, +25% for c.454G>C, +20% for c.453G>T, and +23% for c.451T>C) (Figure 3C). Of the 2 IRD-causing variants, the highest increase was observed for c.−454G>C.

Importantly, mRNA expression analysis of Renilla using real-time quantitative polymerase chain reaction did not reveal a statistically significant difference between the WT and any of the mutant constructs (Figure 3D). Together, these data suggest an effect at the translational level rather than at the transcriptional level, which fits the hypothesis that the variants affect the uORFs.

Clinical diagnosis

Aiming to compare the severity of retinal disease in patients with 2 null EYS variants (n = 41) with those who are homozygous or compound heterozygous for c.−453G>T (n = 6), we collected data regarding disease onset (in years), visual acuity, and electroretinogram (ERG) amplitudes (Figure 4). On average, patients with RP who harbored c.−453G>T had milder diseases with later onset, better visual acuity, and higher ERG amplitudes. Because of the relatively small number of patients with c.−453G>T, only the ERG amplitude comparisons showed a statistically significant difference (Figure 4, Supplemental Table 4).

Figure 4. Comparison of clinical parameters.

Data were collected from 2 groups of patients with RP because of EYS variants—those harboring only null pathogenic variants (blue bar) and those harboring at least 1 c.−453G>T allele (green bar). RP, retinitis pigmentosa.

Retinal imaging was collected from biallelic EYS cases (Supplemental Figure 9, Supplemental Table 5). The 2 c.−453G>T homozygous patients (Supplemental Figure 9A-L) presented overall bilateral and symmetric retinal atrophy encroaching the macula, marked by bone spicule pigmentation in the mid/far periphery combined with narrow and attenuated retinal vessels. Fundus autofluorescence images showed heterogeneous autofluorescence surrounding the hyper-fluorescent patch on the macula. Horizontal cross-sectional optical coherence tomography scans showed a preserved island of ellipsoid zone delimited with a thin atrophic retina compatible with the viable retina. No significant differences were observed in the retinal imaging findings between patients with the c.−453G>T variant and those with null EYS variants.

Discussion

In this study, we have identified a founder variant in the EYS 5′-UTR promoter region that affects the binding of TFs, may also affect the dynamics of upstream initiation codons, and is associated with a relatively mild form of RP. The discovery of pathogenic variants in noncoding regions, and particularly in promoter regions, is becoming more prevalent as NGS methods become more advanced.³⁰ Such variants account for some of the numerous genetic pathologies for which the cause is unknown. As mentioned above, the diagnostic yield in IRDs usually ranges from 50% to 60%,^1,2 and we predict that many of the unsolved cases are due to noncoding variants affecting gene regulation. To the best of our knowledge, 8 sequence variants (including EYS: c.−453G>T and c.−454G>C) in promoter regions of IRD-associated genes have been reported thus far as pathogenic out of 10,000 to 15,000 reported variants (0.05%−0.07% of all variants). This estimated value is much lower than the fraction of reported regulatory variants in other diseases. Based on the human gene mutation database,³¹ for all human inherited diseases, 6757 regulatory variants out of 504,008 total variants were reported, leading to an estimated fraction of 1.3% of all variants. Therefore, we predict that many pathogenic regulatory variants in IRD genes are still unknown. Indeed, a recent study reported 2 TMEM216 5′-UTR variants, c.−69G>T and c.−69G>A that are common in the African and South Asian populations, respectively.¹² It is therefore highly recommended to screen 5′-UTR exons of EYS in RP cases who are heterozygous for a single EYS pathogenic allele. The 5′-UTR can also be interrupted by CNVs affecting at least a single noncoding exon, as we showed here. The effect of 5′-UTR-affecting CNVs on protein expression might be different from SNVs because larger sequences are involved; therefore, a more dramatic effect on protein level is expected, leading to a more severe retinal phenotype, compared with the effect of 5′-UTR SNVs.

To date, only a few pathogenic variants have been reported in the 5′-UTR of IRD-associated genes. Two PRPF31 promoter CNVs eliminate exon 1, which contains the promoter region.⁹ Four SNVs disrupting promoters have been reported in 3 IRD genes ((BBS10) NM_024685.4: c.−80dup, (MERTK) NM_006343.3:c.−125G>A, and (NMNAT1) NM_022787.4:c.−69C>T and c.−70A>T). In all of the cases, these variants caused reduced expression levels as detected by either luciferase mRNA and/or luminescence levels and/or mRNA quantification from patient-derived cell lines. A luciferase assay examining a variant in the promoter of BBS10 (c.−80dup) found 70% reduced expression.¹⁰ A variant in MERTK (c.−125G>A) was found to reduce luciferase mRNA and expression levels.⁷ The variants c.−70A>T and c.−69C>T in NMNAT1 were found to cause decreased NMNAT1 mRNA levels and decreased luciferase expression in human retinal pigment epithelial RPE-1 cells.¹¹ An (RDH12) NM_152443.3:c.−123C>T variant (in cis with the hypomorphic variant c.701G>A) (NM_152443.3) was found to introduce an uORF resulting in reduced RDH12 protein but unaltered mRNA levels.⁷

Previous studies have shown that the 5′-UTR region of EYS and noncoding exons 1 and 2 in particular, are chromatin accessible regions that bind a number of TFs^19,20 (Supplemental Figure 1B). Apart from the variant c.−453G>T (chr6:65,707,140C>A) reported in this study, 2 variants have been reported in the EYS noncoding exon 1 in patients with RP. One (c.−454G>C) was identified in a Spanish patient²⁶ and lies adjacent to the variant identified here. Both variants are relatively well conserved and have genomic evolutionary rate profiling scores of 4.19 (for c.−454G>C) and 1.22 (for c.−453G>T). An additional variant (c.−448+5G>A) with a genomic evolutionary rate profiling score of 3.22 is expected to affect the splicing of exon 1, which as an important part of the EYS promoter makes this a likely cause of disease.^32,33 In this study we provide evidence that the 2 SNVs identified in exon 1 affect TF binding and uORF efficiency. In light of our and previous findings, we encourage other researchers to ensure that the 5′-UTR region of EYS is not overlooked in diagnostic analyses for both SNVs and CNVs, especially because it is covered by many gene panels.

The TFs that were analyzed in this study were either known retinal TFs that were detected previously by ChIP-seq to bind the EYS exon 1 region (OTX2, CRX, and NRL)¹⁹ or known to be involved in photoreceptor differentiation (NR2E3).³⁴ OTX2 is a homeodomain TF that is expressed from the beginning of and throughout retinal development³⁵ and is also crucial to photoreceptor maintenance.³⁶ CRX, another homeodomain TF is involved in photoreceptor maintenance by regulating development and gene expression.³⁷ NRL (neural retina leucine zipper) interacts with CRX and maintains photoreceptor function through regulation of genes including rhodopsin^38,39 and plays a major role in photoreceptor differentiation and maintenance.⁴⁰ NRL and NR2E3 are rod-determining factors,⁴⁰ and NR2E3 is involved in photoreceptor development and phototranduction.⁴¹

Supporting previous studies, our luciferase assay indicates that EYS exon 1 serves as a binding region for TFs and can therefore be defined as part of the promoter. In addition, we show that all 4 TFs that we studied and their combination increase luciferase expression, presumably through binding to the promoter site with varying affinities. It should be noted that even in the absence of overexpression of exogenous TFs, there was an increase in luciferase expression due to the presence of the EYS promoter region in the HeLa cells. This indicates that there are probably endogenously expressed factors that increase transcriptional activity when they associate with the cloned EYS promoter region. It also indicates that even without overexpression of TFs, it serves as a promoter region and encourages higher expression. The level of luciferase expression measured with OTX2 overexpression was notably increased compared with all other TFs. This indicates that OTX2 has a more marked effect on EYS expression, which is aligned with the central role it plays in retinal function from early in development.³⁵

Our results indicate a complex effect of the studied TFs on the EYS promoter region in which OTX2 overexpression results in higher expression in the c.−454G>C condition and CRX results in higher expression of the c.−453G>T condition. When NRL and NR2E3 were overexpressed, there was no significant difference between the 3 EYS promoter sequences, indicating that the specific sequence is not significant to their binding. This was also true when all the TFs were overexpressed, suggesting that the interaction of the TFs with each other can override the effect of individual TFs. This study included a small selection of the many TFs that may be at play in regulating this promoter region and therefore the ultimate difference between the variants cannot be fully ascertained.

Another mechanism through which the 5′-UTR can modulate gene expression is through uORF dynamics. uORFs are present in approximately 50% of human transcripts.^13,28,42 Upstream initiation codons can be recognized by translational machinery, which may reduce translation of the primary ORF by 80% through various mechanisms.^7,13 Although the most frequently (90%−95%) used initiation codon is ATG, there is evidence of alternative initiation codons being recognized such as TTG, CTG, and GTG. In EYS, the 5′-UTR region contains several naturally occurring upstream initiation codons that are predicted to result in uORFs of 84 and 138 nucleotides, potentially producing short peptides. Both uORFs have adequate Kozak sequences and do not overlap with the primary reading frame. The variant identified in this article (c.−453G>T) affects an important nucleotide of the Kozak sequence of the uATG, whereas the previously reported variant (c.−454G>C) disrupts the uATG itself. We suggest that the effect of the variants on luciferase activity may be modulated by such a mechanism of distortion of uORF dynamics, thus influencing translation.

Although we performed distinct experiments to examine each of the potential mechanisms at play in the EYS 5′-UTR region (promoter and uORF), it is not possible to entirely separate the 2. In the experiments without TF overexpression, we cannot account for the effect of endogenous TFs that activate the promoter region. In the TF overexpression experiments, the cloned EYS region includes the upstream start codons, which may also affect the results. It is possible that each disrupted mechanism has a different effect on protein expression, meaning that the same variant can have a complex and multidirectional effect on the expression of a gene, either at the transcriptional or translational level, or both. In all experiments, an increase in expression was seen, but it is unclear whether this was an additive result of both mechanisms working together or if the directionality of one mechanism overrode that of the other.

In this study we transfected HeLa cells with the EYS region of interest together with vectors for TF overexpression. The factors chosen were based on published bioinformatic data indicating binding to this region in the human retina.¹⁹ However, we presume that these factors are only part of a more complex mechanism involving many other factors and interactions unique to the retina. It is possible that the trends seen in both luciferase experiments would change in degree or in direction when done in an environment that resembles the retina more closely. Studying protein expression levels in retinal organoids generated from iPSCs of patients might allow better understanding of the effect of these variants on EYS expression.

Based on the scientific literature and Leiden open variation database (as of August 1, 2024),¹⁸ 534 unique EYS pathogenic variants were reported, the vast majority (79%) are likely null (including frameshift, nonsense, and splicing variants); therefore, no EYS protein expression is expected in the retina of most affected individuals. Previous studies showed that the vast majority (~90%) of pathogenic EYS alleles cause RP which is more severe compared with other autosomal recessive) RP forms.⁴³ Additional autosomal recessive-IRDs, including macular dystrophy and cone-rod dystrophy, were also associated to EYS.¹⁸ Because the Arab-Muslim population in which the c.−453G>T variant was identified is not included in web-based population databases, it is not currently possible to assess the minor allele frequency of this variant in this population. However, the lack of heterozygotes among approximately 500 analyzed cases from our cohort indicates that c.−453G>T is a rare variant. The c.−453G>T variant reported here is likely to affect the expression level of EYS in the retina, unlike most reported EYS pathogenic variants. Such an effect on expression level, without affecting the protein sequence itself, is likely to result in a milder RP phenotype compared with patients with biallelic null variants. Our clinical analysis, indeed, verified that clinical parameters of patients who are either homozygous or compound heterozygous for c.−453G>T are compatible with RP and are better compared with RP patients with biallelic null pathogenic variants. Our data expands the clinical severity range of EYS-related RP and therefore might allow more accurate prognosis of patients with milder diseases-causing variants with a wider window for a possible intervention, once available.

The development of a comprehensive system for analyzing promoter and other regulatory regions would be hugely beneficial to increasing the diagnostic yield of human disease. A recent study introduced a novel prioritization strategy for 5′-UTR variants using bioinformatic predictions tools that assess the variant effect on features including uORFs, transcription start sites (TSSs), and internal ribosome entry sites.⁷ However, a systematic approach is still needed to functionally validate the effect on gene expression and to confirm pathogenicity.

Supplementary Material

Additional Information

The online version of this article (https://doi.org/10.1016/j.gim.2025.101427) contains supplemental material, which is available to authorized users.

Data Availability

The data presented in this study are presented within the article and supplemental material. The authors are willing to share materials, data sets, and protocols presented in this publication with other researchers upon request.