EYS mutations and implementation of minigene assay for variant classification in EYS-associated retinitis pigmentosa in northern Sweden

Retinitis pigmentosa (RP) is a clinically and genetically heterogeneous group of inherited retinal degenerations. The ortholog of Drosophila eyes shut/spacemaker, EYS on chromosome 6q12 is a major genetic cause of recessive RP worldwide, with prevalence of 5 to 30%. In this study, by using targeted NGS, MLPA and Sanger sequencing we uncovered the EYS gene as one of the most common genetic cause of autosomal recessive RP in northern Sweden accounting for at least 16%. The most frequent pathogenic variant was c.8648_8655del that in some patients was identified in cis with c.1155T>A, indicating Finnish ancestry. We also showed that two novel EYS variants, c.2992_2992+6delinsTG and c.3877+1G>A caused exon skipping in human embryonic kidney cells, HEK293T and in retinal pigment epithelium cells, ARPE-19 demonstrating that in vitro minigene assay is a straightforward tool for the analysis of intronic variants. We conclude, that whenever it is possible, functional testing is of great value for classification of intronic EYS variants and the following molecular testing of family members, their genetic counselling, and inclusion of RP patients to future treatment studies.


Results
EYS mutations in the RP cohort from northern Sweden. 81 cases with clinical diagnosis of RP were available for the mutation analysis. The inheritance pattern was defined from the family history but remained uncertain in some cases. DNA from patients with arRP, RP18, RP26, RP74, RP103, RP154, RP160, RP165, RP6795 and VC101 (Table 1, Supplementary Figure S2) was analysed using a targeted analysis of 501 known variants in 16 arRP-associated genes, EYS not included. No mutations were identified by that time. Later, RP160 (Family 501, Supplementary Figure S1, Figure S2) analysed by NGS with a 56 arRP-associated gene panel, was found to be heterozygous for a novel sequence variant; an 8 bp deletion in exon 43 in the EYS gene c.8648_8655del. Two other affected family members (RP5436 and RP6795) were also heterozygous carriers of the same deletion. Sanger sequencing of exon 43 in 78 RP patients revealed five heterozygotes (RP15, RP18, RP74, RP103 and RP105) and five homozygotes (VC101, RP26, RP154, RP165 and RP6019) ( Table 1). As a next step, all coding sequences of the EYS gene were sequenced in the eight heterozygotes to identify a second pathogenic variant (RP15, RP18, RP74, RP103, RP105, RP160, RP5436 and RP6795). Three potential culprit variants were detected c.1155T>A (RP15, RP103, RP105), c.2992_2992+6delinsTG (RP15) and c.3877+1G>A (RP103 and RP105) ( Table 1, Supplementary Figure S3). Consequently, all 81 RP patients were sequenced regarding the three identified EYS variants. The c.1155T>A was found in cis with the 8 bp deletion (c.8648-8655del) in four patients from Family 012 (VC101, RP15) and Family 335 (RP103, RP105). All four patients were carriers of bi-allelic mutations; VC101 was homozygous for the cis allele, RP15 was a compound heterozygote with c.2992_2992+6delinsTG, RP103 and RP105 were compound heterozygotes with 3877+1G>A ( Table 1). The five remaining RP cases (RP18, RP74, RP160, RP5436 and RP6795) with only one EYS mutation were subjected to MLPA to detect copy number variants. By using a MLPA probe ligation site situated 133 bp before exon 42 we detected a ~ 41 kb duplication of exon 42 in three members of Family 501. The size of the duplication was narrowed to a ~ 13 kb area between genomic coordinates chr6:63,721,660 in exon 43 and chr6:63,734,700 in intron 41 (Genome Browser GRChr38/ hg38) by using of three custom designed MLPA probes (Supplementary Figure S4). The exact breakpoints of the duplication could not be identified. Three affected individuals in the Family 501 were shown to be compound heterozygotes inheriting the duplication and the 8-bp deletion from each parent. The second EYS mutation in Table 1. EYS variants in arRP patients in the northern part of Sweden. 1 Two variants are present on one allele and another variant is present on the second allele in this patient. 2 Pathogenic variants were present on both alleles in this patient. 3 Duplication breakpoints remain unknown. 4 Only one likely pathogenic variant was found. All sequence positions are denoted according to NM_001142800.1 (human reference genome GRCh38/ hg38). In vitro splice assay using EYS minigenes carrying c.2992_2992+6delinsTG variant. To analyse how the EYS c.2992_2992 + 6delinsTG variant affects splicing, wt and mt EYS minigenes covering exon 19 and part of the adjacent intron sequence were generated (Fig. 1). The minigenes were transfected into HEK293T and ARPE-19 cells and the splicing was analysed. The wt EYS minigene transfected into HEK293T cells resulted in one distinct product of 406 bp (Fig. 1). By sequencing, we could confirm that the 406 bp product represents correct splicing of exon 19. Transfection of the mt EYS c.2992_2992+6delinsTG minigene into HEK293T cells resulted in a 260 bp product, due to exon 19 skipping (Fig. 1). Transfection of the ARPE-19 cells with wt EYS minigene resulted in three products; 406 bps, ~ 320 bps and 260 bps ( Fig. 1), where correctly spliced transcript was the most prevalent. When the mt EYS c.2992_2992+6delinsTG minigene was transfected into ARPE-19 only the product with skipped exon 19 was observed (Fig. 1). Thus, we observed exon skipping in both human embryonic kidney HEK293T and retinal pigment epithelium ARPE-19 cells when the c.2992_2992+6delinsTG variant was introduced. Skipping of exon 19 is predicted to result in a premature stop codon and due to absence of exon 19, the amino acid nomenclature would be p.(Phe950Hisfs*4).
In vitro splice assay using EYS minigenes with the c.3877+1G>A variant. To demonstrate how the EYS c.3877+1G>A variant affects splicing, wt and mt EYS minigenes were generated (Fig. 2). The minigenes were transfected into HEK293T and ARPE-19 cells and the splicing outcome was analysed. The wt EYS minigene transfected into HEK293T cells resulted in two distinct products of 572 bp and 453 bp respectively (Fig. 2). The 453 bp product corresponded to correct splicing of exon 25, while the 572 bp fragment was the result of using exon 25 donor site and pSPL3 alternative donor site. The sequencing revealed inclusion of 119 nucleotides that covered EYS intron 25 sequence from c.3877+253 to c.3877+266 (14 bp) plus the entire NotI restriction site (8 bp), and continuing 97 bp into the pSPL3 vector before splicing out at a cryptic donor site within the pSPL3, and thereafter joining pSPL3 acceptor site SAv. Transfection of the mt EYS c.3877+1G>A minigene into HEK293T resulted in three products: 582 bp, 463 bp and 260 bp (Fig. 2). The 582 bp fragment had additional 10 bp of intron 25 sequence (c.3877+10) and 119 nucleotides as seen in the wt construct. The 463 bp product was shown to include the same 10 bp of intron 25 Table 2. Characterization of EYS variants identified in arRP cohort from northern Sweden. All sequence positions are denoted according to NM_001142800.1) (GRCh38/hg38 human reference genome). dbSNP -database of single nucleotide variations at https:// www. ncbi. nlm. nih. gov/ snp/. gnomAD-genome aggregation database at https:// gnomad. broad insti tute.  www.nature.com/scientificreports/ Clinical findings. In this study we could only investigate clinical presentation of EYS-associated RP in three individuals from Family 501 (Table 3). Visual acuity was unaffected with no significant cataract or other ophthalmologic diseases present during the first two decades for all three patients. On fundoscopy, normal optic discs, discrete arteriolar attenuation and single pigmentary deposits in the periphery were seen in RP6795 at 15 yo, in RP5436 at 18 yo and in RP160 at 19 yo (Fig. 3). Upon visual field examinations, case RP6795 had restricted peripheral visual fields with relative paracentral scotoma (Fig. 3). Her younger female relative, RP5436, had less affected peripheral vision with arcuate relative scotomas in the mid-periphery. RP160 at 19 yo presented a large paracentral scotoma on both eyes (Fig. 3). OCT performed in RP5436 and RP160 demonstrated a general thinning of the macula with some foveal sparing (Supplementary Figure S5). The full-field standard dark-adapted ERGs are presented in Supplementary Figure S6. The amplitudes of rod, mixed rod-cone, cone and 30 Hz flicker responses were all subnormal to non-recordable in all cases at young age, meaning that the photoreceptor functions were seriously affected already in their teens. The young man, RP160, examined at age 19 and re-exam- www.nature.com/scientificreports/ ined at age 30, presented several indicators of disease progression during this period. On fundoscopy the optic discs had gained a waxy pallor, alongside severe arteriolar attenuation, and an indication of a bull-eye lesion in the macula (Fig. 3). The retina was generally atrophic with the typical bone spicule pigmentation. His VF had severely deteriorated with some preserved central and peripheral islets. There was a notable progression in the visual field loss over time, going from arcuate and ring-shaped scotomas to only mid-central preservation (Fig. 3). ERG showed a decline of response amplitudes over time for both rods and cones, and the rod, mixed rod/cone responses were extinguished at age 30 (Supplementary Figure S6).

Discussion
Taken into account a relative homogeneity of the population in northern Sweden and presence of founder and family specific mutations in several retinal genes 5,21-27 we were looking for shared genetic causes in genetically unresolved cases of arRP in northern Sweden. In this study we investigated RP cohort including 65 arRP patients with unknown genetic cause and identified mutations in the EYS gene in 14 cases including 4 family members. Thus, the EYS mutations were present in 16% of arRP cases. Two of six EYS mutations detected in this study were reported previously, c.1155 T > A in exon 7 18,28 and c.8648_8655del in exon 43 12,19 . Both mutations present on the same allele had been found earlier in three Finnish patients with retinal dystrophy 29 . The most common EYS variant in our study, c.8648_8655del was present in 13 cases. We were not able to connect the families of all 13 carriers but considering that 6 patients originate from the same region of Tornedalen (Pajala and Karesuando) adjacent to the Finnish boarder, it is likely that they have a common ancestor. Four of the 13 patients carried c.8648_8655del and c.1155T>A variants in cis indicating that they share the same ancestor as the Finnish cases 29 . The Finnish origin of the cis allele might explain why we were not able to relate the Swedish families using the Swedish church records.
In two patients, RP18 and RP74 we found only one likely pathogenic variant c.8648_8655del. As we only investigated the intronic sequences adjacent to the EYS exons, a deep intronic variant as a second causal variant cannot be ruled out. The same could be true for RP85 where we found only one heterozygous missense variant, c.6284C>T reported once in a compound heterozygous arRP patient with second EYS variant c.2234A>G 30 31 . In our cohort we detected genomic rearrangements in only one family, compare to 4% EYS CNV in French cohort 31 .
Since the pathogenic mechanisms in EYS-associated IRD are still poorly understood, the interpretation of the pathogenicity of many EYS variants is a challenge. The implementation of numerous computational tools to predict functionality of potentially disease-causing variants underscores the need of functional assays especially for the variants outside coding regions. According to Messchaert et al. (2018) the large size of EYS' cDNA, the absence in the genome of mice and rats and the retina specific expression represent the factors limiting an experimental assessment of the pathogenicity of EYS variants 19 . In this study we assessed functionality of two intronic variants both predicted to be pathogenic due to the changes in exon-intron boundaries. By implementation of in vitro minigene assays we easily confirmed exon skipping caused by both EYS variants, c.2992_2992+6delinsTG and c.3877+1G>A. It is worth to mention that splicing outcome of in vitro minigene assays depends on the content of DNA sequences of the gene of interest and the cells chosen for the assay. We found that an inclusion of 10 bp of consecutive intron 25 occurred in HEK293T cells transfected with mutant minigene EYS c.3877+1G>A, but not in ARPE-19 cells. Computational tools revealed two alternative sites in intron 25, 10 and 14 bp from exon-intron junction site. In the patient's DNA from which the minigene was constructed a common deletion at position c.3877+18_3877+22del on the same allele as c.3877+1G>A was present. This 5 bp deletion was predicted to quench the stronger alternative donor site at position + 14, corroborating our result. Not only mutant but also wildtype minigene showed complex splicing by incorporating a part of the amplicon's intron sequence at position c.3877+253 and retaining 97 bp of the pSPL3. This aberrant splicing is more likely to be HEK293T specific because it was not seen in the ARPE-19 cells.
Mutations in the EYS gene cause a variety of phenotypes such as RP and cone-rod dystrophy characterised by inter-and intrafamilial phenotypic diversities 10,13,32,33 . We provide clinical evaluation of only three patients from one family available for this study.
The proposed molecular mechanism of RP caused by EYS mutations is loss of protein function. The role of EYS is not yet understood but presence of epidermal growth factor-like and laminin domains might be important in cell adhesion, migration and intracellular signalling 34,35 . Also, it is known that an ortholog of the Drosophila spacemaker (spam) protein plays a critical role in maintenance of the photoreceptor morphology 36 . Currently 343 EYS mutations are reported in The Human Gene Mutation Database (http:// www. hgmd. cf. ac. uk/) with almost 44% predicted to result in protein truncation 19 . Nonsense or protein truncating variants are usually easy to classify as pathogenic according to ACMG recommendations 20 however missense and intronic variants require additional factors to be classified as likely pathogenic or pathogenic. According to the last EYS update 19 missense and splice site EYS variants represent 48.2% of all mutations, but only 8.6% can be classified as likely pathogenic while the rest are variants of uncertain significance. Lack of functional studies prevent classification of the variants as pathogenic and, thus might dismiss these variants in future genetic testing and counselling.
Many defective transcripts are degraded by nonsense-mediated mRNA decay (NMD) 37 that could also partially explain the phenotypic variation 38 www.nature.com/scientificreports/ c.1211dupA, c.4957dupA and c.8805C>A revealed barely detected, low and almost normal level of expression of the mutant allele which suggested that almost complete NMD, partial NMD and escape from NMD took place 40 .
Interestingly, transcripts carrying the c.8805C>A mutation, that result in a premature termination codon in the last exon 43, escaped degradation by NMD and demonstrated normal expression levels 40 . The most common mutation in northern Sweden, c.8648_8655del and a large duplication located in exon 42 set grounds for study of a genotype-phenotype correlation.
A debatable pitfall of our study is the cascade-targeted mutation analysis of the EYS gene. Not all arRP patients in our cohort were subjected to the analysis of the entire EYS gene making it possible that some mutations escaped detection. However, our aim was to find common EYS variants in the homogenous population of northern Sweden and the 8 bp deletion found in 13 of arRP cases proved the case. The additional screening of the entire coding regions of the EYS gene in 24 unresolved arRP resulted in only one finding of a variant of uncertain significance (c.6284C>T). This leaves us with 28 unresolved arRP cases that have only been screened for the variants found in this study. The probability of finding yet another common EYS variant among these patients seems to be low.
In conclusion, our study revealed one of the most common genetic cause of autosomal recessive retinitis pigmentosa in northern Sweden. EYS mutations accounted for 16% of arRP cases with the most frequent pathogenic variant c.8648_8655del. The same mutation in cis with c.1155T>A indicated presence of Finnish founder allele in Swedish patients. We also demonstrated that in vitro minigene assay is important applicable tool for the functional analysis of intronic variants. If the pathogenic mechanisms for these variants can be determined, it will be of a great value for variant classification, molecular testing, genetic counselling, and future treatment of patients with EYS-associated RP.

Methods
Patients. In this study we used DNA from 81 patients with clinical diagnosis of retinitis pigmentosa (RP). Supposedly, 9 patients had autosomal dominant RP (adRP), 65 had autosomal recessive RP (arRP) and 7 patients had either adRP or arRP. Among arRP patients, 7 belonged to three families (Families 012, 335 and 501, Supplementary Figure S1) and the rest were simplex. Peripheral blood samples were collected in EDTA tubes and DNA was extracted as described elsewhere. The assignment of inheritance pattern was based on a family history provided by the patients. However, some of the elderly patients did not have details about visual impairment of their siblings and parents and most of the cases were regarded as simplex. DNA from siblings or parents was available only in three families. The study was approved by the Swedish Ethical Review Authority (Etikprövningsmyndigheten) and conducted in accordance with ethical principles for medical research involving human subjects as stated in Declaration of Helsinki. An informed consent was obtained from all patients or their parents before inclusion to the study. During first examination of patient RP6795 an informed consent was obtained from her parents. Figure 2. Exon skipping because of EYS c.3877+1G>A variant. (a) A schematic illustration of the pSPL3-EYS c.3877+1G>A minigene. Exon 25 of EYS gene (blue) and flanking introns (purple) were cloned into the NotI and EcoRI sites of pSPL3 vector (black) with a wildtype (wt) or mutant (mt) c.3877+1G>A between two pSPL3 exons (yellow); splice donor vector (SDv) and splice acceptor vector (SAv). Positions of SD6 and SA2 primers for splice analysis are displayed. Red blocks show two alternative splice sites in intron 25 and one alternative splice site in pSPL3. In the patient's DNA used for minigene construction, a common deletion at position c.3877+18_3877+22del on the same allele as c.3877+1G>A was present. Solid and dashed lines show different outcomes of splicing. HEK293T cells and ARPE-19 cells were transfected with wt or mt EYS hybrid minigene or an empty pSPL3 vector. After RNA extraction and cDNA synthesis, splicing products were amplified by PCR using SD6 and SA2 vector specific primers and visualized by agarose gel electrophoresis. Resolution of splicing PCR products derived from HEK293T are shown in (b) and from ARPE-19 in (c). Sanger sequencing confirmed that in the control pSPL3, only SDv-SAv splicing occurred, leading to a 260 bp product. Expected wt splicing of the EYS minigene was a 453 bp product. Additional splicing took place in the wt minigene transfected into HEK293T cells where incorporation of 14 bp of the 3′ end of the EYS amplicon occurred, starting at c.3877+253 (red block) and continuing 97 bp into the pSPL3 vector before splicing to SAv. This aberrant splicing resulted in a 572 bp product. The same incorporation occurred in the mt minigene in HEK293T, with an additional 10 bp of consecutive intron 25 sequence in the 5′ end, leading to a 582 bp product. In HEK293T, mt minigene also produced an mRNA with expected splicing at the SDv-exon 25 junction, but with the incorporation of 10 bp consecutive sequence of intron 25, before joining SAv. This was not seen in ARPE-cells. In both HEK293T and ARPE-19 cells, the mt hybrid minigene resulted in exon skipping (d). Full-length gels are presented in Supplementary Figure S7. www.nature.com/scientificreports/ Next generation sequencing (NGS) of 56 arRP-associated genes was performed at Asper Biogene. Samples were analysed using the BWA Enrichment app in BaseSpace (Illumina). Alignment was performed with BWA and variants were called with GATK 41 . Mean coverage was 87 reads for the whole gene panel and mean coverage of EYS gene was 99 reads. Variants that had an alternative variant frequency less than 20% and/or had a call quality score lower than 20 were automatically filtered out. Confirmation of NGS findings and subsequent screening of EYS gene was done with Sanger sequencing using BigDye Terminator v3.1 Cycle Sequencing Kit (Applied Biosystems, Foster City, CA, USA). Primer sequences for the screening of the EYS gene are available upon request. The products of sequencing reactions were analysed on ABI 3500xL Dx Genetic Analyser (Applied Biosystems). Sequences were aligned and evaluated using the Sequencher software version 4.9 (Gene Codes Corporation, Ann Arbor, MI, USA). All changes were assigned according to the GenBank Reference Sequence Version FJ416331; GI: 212,675,237; Transcript Reference Sequence: NM_001142800.1 and described according the HGVS recommendations 42 .
For detection of copy number changes in exonic sequences of the EYS gene we used Multiplex Dependent Probe Amplification (MLPA) with SALSA MLPA Probe mix P328-A1-0811 or P328-A2-0217 lacking probes for exon 9 or exons 2, 7, 9 and 27, respectively (MRC Holland, Amsterdam, Netherlands). Additionally, three MLPA www.nature.com/scientificreports/ probes were designed aiming to detect breakpoints of the duplication detected in Family 501. The sequences of these MLPA probes are available in Supplementary Figure S4.  Table S1) and 1.25U of Taq-DNA polymerase with 5′-3′ exonuclease activity (Ampliqon A/S, Odense, Denmark). Ligation of the amplicons was performed according to manufacturer´s instructions by adding 50 ng of pGEM-T Easy vector (Promega, Fitchburg, WI, USA), 3 μl of PCR product, 3U of T4 DNA ligase and 1X Rapid Ligation buffer to a final reaction volume of 10 μl. For propagation of the pGEM-T Easy vector, XL10-Gold ultracompetent cells (Stratagene, La Jolla, CA, USA) were transformed by adding the pGEM-T ligation reaction to 50 μl of XL10-Gold suspension as described elsewhere. Plasmid DNA was extracted from the overnight cultures (1 × LB with 100 µg/ml carbenicillin) using standard protocol. The DNA concentration was measured on a NanoDrop 1000 v.3.8.1 (Thermo Fisher Scientific, Waltham, MA, USA). To identify clones with correct inserts in the pGEM-T vector, 0.5-1 μg of each plasmid DNA was treated with 3U of EcoRI and NotI (New England Biolabs, Ipswich, MA, USA) for 1-2 h at 37 °C and visualized on a 1% agarose gel. Plasmids with correct size of insert were analysed by Sanger sequencing to identify wild type (wt) and mutant (mt) constructs and to rule out any sequence deviations generated by PCR. Sanger sequencing was performed using BigDye Terminator v3.1 Cycle Sequencing Kit (Applied Biosystems). For minigene construction, pSPL3 was digested with EcoRI and NotI and the wildtype and mutant inserts excised from pGEM-T Easy vector were ligated into the pSPL3 vector using the same ligation protocol as for pGEM-T Easy system followed by transformation into XL10-Gold ultracompetent cells. Ratios for pSPL3 versus inserts were calculated using an online tool at www. prome ga. com/a/ apps/ bioma th/? calc= ratio. Sequence confirmation was done by Sanger sequencing with pSPL3 specific primers: forward SD6-5′-TCT GAG TCA CCT GGA CAA CC and reverse SA2-5′-ATC TCA GTG GTA TTT GTG AGC (Supplementary Information). Upon identification of one mutant and one wildtype minigene construct in the pSPL3 vector, the clones were propagated, and plasmid DNA was purified by midi-preparation using NucleoBond Xtra Midi kit (Macherey-Nagel, Düren, Germany).