The identification of a RNA splice variant in TULP1 in two siblings with early‐onset photoreceptor dystrophy

Abstract Background Early‐onset photoreceptor dystrophies are a major cause of irreversible visual impairment in children and young adults. This clinically heterogeneous group of disorders can be caused by mutations in many genes. Nevertheless, to date, 30%–40% of cases remain genetically unexplained. In view of expanding therapeutic options, it is essential to obtain a molecular diagnosis in these patients as well. In this study, we aimed to identify the genetic cause in two siblings with genetically unexplained retinal disease. Methods Whole exome sequencing was performed to identify the causative variants in two siblings in whom a single pathogenic variant in TULP1 was found previously. Patients were clinically evaluated, including assessment of the medical history, slit‐lamp biomicroscopy, and ophthalmoscopy. In addition, a functional analysis of the putative splice variant in TULP1 was performed using a midigene assay. Results Clinical assessment showed a typical early‐onset photoreceptor dystrophy in both the patients. Whole exome sequencing identified two pathogenic variants in TULP1, a c.1445G>A (p.(Arg482Gln)) missense mutation and an intronic c.718+23G>A variant. Segregation analysis confirmed that both siblings were compound heterozygous for the TULP1 c.718+23G>A and c.1445G>A variants, while the unaffected parents were heterozygous. The midigene assay for the c.718+23G>A variant confirmed an elongation of exon 7 leading to a frameshift. Conclusion Here, we report the first near‐exon RNA splice variant that is not present in a consensus splice site sequence in TULP1, which was found in a compound heterozygous manner with a previously described pathogenic TULP1 variant in two patients with an early‐onset photoreceptor dystrophy. We provide proof of pathogenicity for this splice variant by performing an in vitro midigene splice assay, and highlight the importance of analysis of noncoding regions beyond the noncanonical splice sites in patients with inherited retinal diseases.


| INTRODUCTION
Inherited retinal diseases are characterized by the progressive degeneration of photoreceptor and/or retinal pigment epithelium cells, and are a major cause of irrecoverable visual impairment. Retinitis pigmentosa (RP) encompasses the most common group of inherited retinal diseases, with a worldwide prevalence of approximately 1 in 4,000 individuals (Pagon, 1988). RP is characterized by the rod photoreceptor degeneration that precedes cone photoreceptor degeneration. Patients generally present with night blindness, followed by a gradual constriction of the visual field. The visual acuity typically remains relatively preserved until the final stages of disease. Characteristic fundus features include bone spicule pigmentation, attenuation of retinal vessels, and a waxy pallor of the optic disc. RP can follow all Mendelian patterns of inheritance: autosomal dominant, autosomal recessive and X-linked. In contrast, Leber congenital amaurosis (LCA) is considered the most severe form of inherited retinal disease. It is characterized by severe loss of visual function, nystagmus, photophobia, amaurotic pupils (i.e., sluggish or near-absent pupillary responses), high hyperopia, oculo-digital signs such as poking, pressing, and rubbing the eyes, and severely reduced or absence of electroretinogram responses. LCA is generally inherited in an autosomal recessive manner (den Hollander, Roepman, Koenekoop, & Cremers, 2008;Weleber, Francis, Trzupek, & Beattie, 1993).
Early-onset RP and LCA represent a continuum of retinal dystrophies, and are generally differentiated based on the age of onset; patients with an onset after infancy (variably defined as age one or two) are diagnosed as having RP, while LCA generally presents in the first months of life (Kumaran, Moore, Weleber, & Michaelides, 2017). This arbitrary cut-off point gives rise to large clinical and genetic overlap between both phenotypes. In addition, both RP and LCA display large clinical and genetic heterogeneity. The genetic heterogeneity is illustrated by the 87 genes that have been associated with nonsyndromic RP, and the 25 genes that are associated with LCA (Retnet; available at https://sph.uth.edu/retnet/) (Kumaran et al., 2017;Verbakel et al., 2018). Ten genes have been associated with both RP and LCA, among which TULP1 (OMIM: 602280) (Verbakel et al., 2018) .
To date, a molecular diagnosis can be identified by whole exome sequencing (WES) in approximately 60%-70% of RP and LCA patients (Haer-Wigman et al., 2017;Kumaran et al., 2017;Tiwari et al., 2016;Zhao et al., 2015). In the remaining cases, the causative variants could be located in a gene that has not yet been associated with early-onset retinal dystrophies. Alternatively, the genetic defect could reside in a gene that has previously been associated with RP, but the mutation may not have been detected using WES because the variant resides in a GC-rich region, concerns a structural variant, or was not covered for another reason. Finally, the pathogenic variant could have been missed because of too stringent variant filtering procedures or because it resides outside of the coding regions and splice sites. With the advent of therapeutic options for inherited retinal disorders, it becomes essential to also obtain a molecular diagnosis in patients without a conclusive genetic diagnosis.
In recent years, various studies have shown the importance of searching for variants beyond the coding and splice site regions. Up to 15% of LCA patients carry a deep-intronic mutation (c.2991+1655A>G) in CEP290 (Coppieters et al., 2010;den Hollander et al., 2006;Perrault et al., 2007) and these patients may benefit from an upcoming treatment with antisense oligonucleotides (AONs) (Dulla et al., 2018) (ClinicalTrials.gov NCT03140969). Additionally, studies in Stargardt disease have identified numerous variants leading to an alternative splicing or pseudoexon inclusion in ABCA4 (Albert et al., 2018). In this study, we provide evidence for pathogenicity of the first intronic variant outside of the splice site consensus sequence in TULP1, which we coin a nearexon aberrant RNA (NEAR) splice variant, segregating with a previously described pathogenic missense variant in two siblings with early-onset retinal dystrophy.

| Ethical compliance
The study adhered to the tenets of the Declaration of Helsinki and was approved by the local ethics committee. Written informed consent was obtained from both patients and their parents prior to inclusion in this study.

| Clinical evaluation
A family with two siblings with an autosomal recessive early-onset retinal dystrophy was clinically examined at the Radboud university medical center in Nijmegen, the Netherlands. Clinical data were obtained from the medical records of the patients. Patients' medical history was registered with special attention for the age at onset, initial symptoms | 3 of 10 VERBAKEL Et AL. and the course of the disease. In addition, both patients were re-evaluated after the identification of the genetic cause of disease. We performed a detailed ophthalmic examination, which included best-corrected visual acuity, slit-lamp biomicroscopy, and ophthalmoscopy. Fundus photography, spectraldomain optical coherence tomography (SD-OCT; Spectralis HRA+OCT, Heidelberg Engineering, Heidelberg, Germany), and fundus autofluorescence (FAF; HRA+OCT, Heidelberg Engineering, Heidelberg, Germany) imaging were performed. The visual field was assessed using a Goldmann perimeter. Full-field electroretinography (ffERG) recordings were performed according to the International Society for Clinical Electrophysiology of Vision (ISCEV) guidelines and assessed applying local standard values (McCulloch et al., 2015).

| Genetic analysis
Genomic DNA was extracted from peripheral lymphocytes according to standard procedures. WES was performed in a certified DNA diagnostic laboratory in both siblings (Haer-Wigman et al., 2017). The exome was enriched using Agilent's SureSelectXT Human all Exon V5 (Agilent Technologies, Santa Clara, CA). Subsequently, next-generation sequencing using an Illumina HiSeq 4000 sequencer (Illumina, Inc. San Diego, CA), read alignment to the human reference genome (Genome Reference Consortium Human Reference 37/hg19) using Burrows-Wheeler Aligner, and variant calling with the Genome Analysis Toolkit were performed at BGI-Europe (Copenhagen, Denmark). Copy number variants were detected using CoNIFER 0.2.0, and variants were annotated using a custom designed in-house annotation strategy.

| Variant prioritizing
Prioritizing candidate variants for causality was based on their presence in both affected siblings, a minor allele frequency of < 0.5% in ExAC, dbSNP and the Nijmegen in-house database consisting of 15,576 individuals, their effect (i.e., nonsense, frameshift, canonical (donor +1 and +2, acceptor −1 and −2) and noncanonical (donor +3 to +6, acceptor −3 to −14) splice site variants), and the occurrence in a homozygous or compound heterozygous state. Moreover, for the remaining heterozygous variants in currently known retinal dystrophy-associated genes, we manually assessed the BAMfiles to verify if all exons (potentially harboring pathogenic variants) were covered.

| In vitro midigene splice assay
TULP1 (GenBank: NM_003322.5) is a photoreceptor-specific protein and not expressed in available somatic cells. Therefore, a functional analysis of the putative splice variant c.718+23G>A in TULP1 was performed using a midigene assay. We designed two midigene constructs with an insert of 6.3 kb, a wild-type and mutant multi-exon splice vector, using a modified protocol of the previously described method ( Figure 3). (Sangermano et al., 2016) In short, exon 4-11 of TULP1 of the genomic DNA from a control individual was amplified using forward primer 5′-GGAGATCCCTAGGGTGAGGA-3′ and reverse primer: 5′-ATCAAAGCGAGAGGCCCTA-3′. Both primers have an attB1 and attB2 tag at 5′ end to enable Gateway cloning. The wild-type construct served as a template to generate the mutant construct of c.718+23G>A by mutagenesis PCR. Subsequently, wild-type and mutant constructs were incorporated into the pCI-NEO-RHO Gateway-adapted vector as previously described. (Sangermano et al., 2016) This resulted in a wild-type midigene c.718+23G and a mutant midigene c.718+23A. Finally, we transfected HEK293T cells with the wild-type or mutant midigene and studied the transcripts with reverse transcription-polymerase chain reaction with primers in exons 4 and 11 (RT-PCR) (Forward primer: 5′-GTCTACGCCAGGTTCCTCAG-3′ and Reverse primer: 5′-TCCTCGGGACAGATTGGTAG-3′). Figure 1 shows the pedigree of the two siblings we studied, both were from Dutch ancestry and presented with a retinal dystrophy that we classified as very early-onset forms of RP. An overview of the clinical characteristics at the most recent examination is provided in Table 1. The eldest patient, individual II:1, presented with a fine horizontal nystagmus at the age of three. Subsequently, he developed night blindness that became apparent at the age of five. His visual acuity gradually deteriorated from 20/50 at age five to 20/100 when 4 of 10 | VERBAKEL Et AL. he was 15 years of age. His younger sister (patient II:2) presented with a subtle horizontal nystagmus at the age of one. Her visual acuity deteriorated from 20/50 at the age of three to 20/60 at age 13, and she also experienced impaired night vision in early childhood. At the most recent examination, both siblings complained of photophobia. In addition, patient II:1 reported photopsias, particularly after a sudden increase in light intensity. Perimetry showed constriction of the visual field from age 11 in patient II:1, with a central island measuring up to 10 degrees at the age of 15 years. In patient II:2, the constriction of the visual field started at age 10 and progressed to a constriction up to 20-30 degrees at the age of 13 years. High hyperopia was present in both siblings, with spherical equivalent refractions ranging from 6.13 to 7.25. Both siblings were in good general health, and no extra-ocular conditions were reported.

| Clinical findings
Ophthalmoscopy showed peripheral bone spicule pigmentation, attenuated retinal vessels, and small hyperemic optic discs (often found in high hyperopia) in both siblings ( Figure 2). Fundus autofluorescence imaging revealed the characteristic hyperautofluorescent ring that represents the transition zone between intact and degenerated photoreceptor outer segments. This was confirmed by SD-OCT images that showed an intact ellipsoid zone layer inside the ring, and loss of the outer retinal layers outside of the hyperautofluorescent ring area. In addition, the SD-OCT image revealed diffuse thickening of the retina in patient II:1, which proved refractory to treatment with 125 mg oral carbonic anhydrase inhibitors three times a day. Cystoid macular edema was identified in patient II:2 at age 11. However, in her case, the cystoid macular edema revolved after treatment with carbonic anhydrase inhibitors, and did not recur-or at least in a severely reduced fashion-after sustained treatment with 125 mg oral carbonic anhydrase inhibitors two times a day. Electrophysiological examination at the age of five (patient II:1) and six (patient II:2) demonstrated a generalized  photoreceptor dystrophy with severely affected photoreceptor responses in a rod-cone pattern.

| Genetic findings
Initial analysis of WES data in patient II:1 detected two heterozygous variants, a c.1445G>A (p.(Arg482Gln)) missense variant in TULP1 and a c.1567C>T (p.(Arg523*)) nonsense variant in FAM161A. Subsequently, WES was performed in patient II:2; she also carried the heterozygous missense variant in TULP1, as well as a heterozygous c.3683A>G (p.(Tyr1228Cys)) missense variant in RPGRIP1. All coding regions of TULP1 were covered in the WES data of both siblings, and uncovered exons from RPGRIP1 and FAM161A were Sanger sequenced but did not reveal additional putative pathogenic variants in either sibling. Moreover, no copy number variants were identified in FAM161A, RPGRIP1 or TULP1 in both individuals. As the FAM161A and RPGRIP1 pathogenic variant were not shared between the siblings, we pursued to study TULP1 in more detail. The previously described variant c.1445G>A (p.(Arg482Gln) has a CADD-Phred score of 23.2, a PhyloP score of 6.1, and a Grantham score of 43. (Ajmal et al., 2012) An expanded analysis beyond the coding regions and putative splice site regions resulted in the identification of a heterozygous intronic variant, c.718 + 23G>A (Figure 3), found in both siblings. Segregation analysis confirmed that both siblings were compound heterozygous for the c.718+23G>A and c.1445G>A variants, while the unaffected parents were heterozygous (Figure 1). Although the c.718+23G>A variant, based on splice score prediction algorithms, does not alter the nearby splice donor site, this deep-intronic variant increased the splice prediction scores at position c.718 + 20 when using programs SpliceSiteFinder-like (+4.4%), MaxEntScan (+33.33%), GeneSplice (+12.5%), and Human Splicing Finder (+1.1%) (Figure 3). Additionally, c.718+23G>A introduced a new exonic splice enhancer (ESE) motif recognized by the SC35 exonic splice enhancer at the c.718+21 to c.718+29 positions where a putative splice donor site is already located in the wild-type mRNA, and reduces the number and strength of exonic splice silencer (ESS) motifs present in the reference sequence.

| In vitro midigene splice assay
Reverse transcription polymerase chain reaction showed the expected 832-bp wild-type fragment (Figure 3). In contrast, the mutant midigene showed a product that corresponds to a larger fragment, and absence of the wild-type fragment. Sanger sequencing verified that the mutant mRNA product contained a 20-nucleotide elongation of exon 7, which can be explained by the use of a cryptic splice donor site 20 nucleotides downstream of exon 7 (Figure 3). This elongation causes a frameshift and results in early protein truncation (p. (Thr241Glyfs*23)).

| DISCUSSION
In the present study, we identified a near-exon aberrant RNA splice variant that we termed a "NEAR" splice variant in TULP1 in trans with a previously described exonic variant in two siblings with early-onset RP. Both siblings showed a nystagmus, night blindness, and a reduced visual acuity early in life. Fundus examination at the age of 15 (patient II:1) and 13 years (patients II:2) showed bone spicule pigmentation, attenuation of the retinal vessels and a small, hyperemic optic disc. High hyperopia (i.e., refractive error of more than +5.00 D) was present in both siblings. TULP1-associated disease has previously been associated with myopia (den Hollander, Lopez et al., 2007;den Hollander, van Lith-Verhoeven et al., 2007;Hendriks et al., 2017;Souzeau et al., 2018) and hyperopia (den Hollander, Lopez et al., 2007;Khan, Bergmann, Eisenberger, & Bolz, 2015). However, high hyperopia has only been described in patients with LCA, and is thought to result from impaired emmetropization caused by early-onset visual impairment (Khan et al., 2015;Weleber et al., 1993).
Both siblings were diagnosed with early-onset RP based on an onset after the age of one. However, they also show characteristic aspects of LCA such as a nystagmus and hyperopia. The difficulty in classifying such patients arises from the strict and rather arbitrary separation of both entities, when in fact they represent a continuum of retinal dystrophies.
Thus far, three noncanonical splice site variants in the TULP1 gene (i.e., c.999+5G>C, c.1224A>G, and c.1496-6C>A) have been reported in patients with TULP1associated disease (den Hollander, Lopez et al., 2007;Gu et al., 1998;Hagstrom, North, Nishina, Berson, & Dryja, 1998). We identified two pathogenic variants in TULP1, a c.1445G>A (p.(Arg482Gln)) missense mutation and an intronic c.718+23G>A variant. The TULP1 p.(Arg482Gln) mutation, previously described in patients with early-onset RP, alters the structure and function of the Tubby domain, and is expected to affect TULP1 function (Ajmal et al., 2012). To our knowledge, pathogenicity of mutations in TULP1 such as the c.718+23G>A variant, which we coin as a NEAR splice variant, has not been described before. Our definition of a NEAR splice variant is a variant located outside of the splice site consensus sequence leading to an alteration of the splicing of a nearby exon, whereas a deep-intronic variant often leads to the inclusion or alteration of a cryptic exon.
The consequence of a NEAR splice variant depends on the context of the variant, such as the strength of nearby splice acceptor and splice donor sites, the presence and size of flanking exons, and the effect on the appearance or removal of ESE, ESS, intronic splice enhancer, and intronic splice silencer motifs. To assess the effect of this variant on splicing, we generated a midigene assay which contained exon 4 to 11 of TULP1. This analysis showed the use of a cryptic splice donor site 20 bp downstream, which causes a shift of the reading frame resulting in early termination of protein synthesis. The severe nature of the c.718+23G>A NEAR splice variant is supported by the absence of a wild-type fragment in cells transfected with the mutant midigene and corresponds with the severe phenotype observed in the patients.
The donor splice site of exon 7 of the TULP1 gene contains a fairly weak splice site as indicated by the score of 77.4 for Human Slicing Finder. A natural stronger cryptic splice donor site is present at the c.718+20 position in the wild-type mRNA with a score of 83.4 for the Human Splicing Finder prediction (Tang, Prosser, & Love, 2016). The presence of ESSs located in and near the c.718+20 position likely explain why the cryptic donor site is not utilized by the splice machinery in wild-type cells. The c.718+23G>A variant, however, shows a decreased number of ESSs in this region, and consequently enables recognition of the putative splice donor site by the spliceosome (http://www.umd.be/HSF3/). In addition, according to ESE predictions, this variant also creates a new binding site for exonic splice enhancer SC35 at the c.718+21 to c.718+29 positions that strengthens this putative splice donor site and generates preference for the c.718+20 donor instead of the canonical splice donor site. This supports the role of ESEs and ESSs in the splicing process, and highlights the importance of these factors when analyzing the pathogenicity of variants within or outside of the coding and splice site regions. F I G U R E 3 Molecular genetic characterization of the splice effect of the c.718+23G>A variant in TULP1. (a) Schematic representation of the TULP1 gene and enlargement of the wild-type and mutant DNA sequences at the exon-intron boundary of exon 7 of TULP1. The SpliceSiteFinder-like (SSFL, range 0-100), MaxEntSCan (MES, range 0-12), GeneSplicer (GS, range 0-24) and Human Splicing Finder (HSF, range 0-100) scores for the splice donor site are indicated above the gene. The red "A" highlights the variant c.718+23G>A identified in both siblings. The red numbers represent altered scores compared to the wild-type. The green circle implies a newly recognized SC35 motif. Dotted circles indicate exonic splice silencers no longer present by prediction tools (b) Schematic representation of the mutant pCI-NEO-RHO vector, containing exon 4-11 of the TULP1 gene used to transfect HEK293T cells with a wild-type or mutant midigene. (c) RT-PCR products of the wild-type and mutant midigene showing the expected 832-bp wild-type fragment and a 852-bp fragment of the mutant midigene corresponding to a 20-nucleotide elongation of the mRNA encoded by exon 7. The wild-type fragment was absent in the cells transfected with the mutant midigene. RT-PCR analysis of RHO exon 5 was performed as a control for efficient transfection. (d) Sanger sequence analysis of the RT-PCR fragments confirmed the wild-type and the 20-bp elongation of exon 7 in the mutant While this is the first intronic variant outside of the splice site consensus sequences deemed pathogenic in TULP1, the causality of deep-intronic variants has also been described in nine other retinal dystrophy genes: ABCA4, CEP290, CHM, OA1, OAT, OFD1, PROM1, PRPF31, and USH2A (Bax et al., 2015;Braun et al., 2013;Carss et al., 2017;den Hollander et al., 2006;Liquori et al., 2016;Mayer et al., 2016;Naruto et al., 2015;Rio Frio et al., 2009;Vache et al., 2012;van den Hurk et al., 2003;Webb et al., 2012) (www.dbass.soton. ac.uk). Intronic variants are likely to explain a substantial portion of the current genetically unexplained or monoallelic retinal dystrophy cases, and underscore the importance of genetic tests uncovering those regions, such as whole genome sequencing. Future studies with whole genome sequencing will likely increase the number of genetically solved patients. However, the increase in the use of whole genome sequencing will be accompanied by the detection of a large number of variants of unknown significance, and determining the functional role of these variants will remain a challenge.
The identification of biallelic variants in patients with a retinal dystrophy is essential for eligibility for upcoming genetic therapies. Besides gene augmentation therapy, patients with a TULP1 NEAR splice variant may benefit from AONs treatment, which can suppress the aberrant splicing effect by binding to the mutated region in the pre-mRNA. To date, proof-of-concept of AONs has been shown in both cell-based models and animal models for four retinal dystrophy genes: CEP290 (Collin et al., 2012;Dulla et al., 2018;Garanto et al., 2016;Parfitt et al., 2016), CHM (Garanto, van der Velde-Visser, Cremers, & Collin, 2018), RHO (Murray et al., 2015), and USH2A (Slijkerman et al., 2016), and represents a promising therapy for retinal dystrophies (Collin & Garanto, 2017).
In conclusion, we identified a pathogenic NEAR splice variant in TULP1 in trans with a known pathogenic missense variant in two siblings with an early-onset photoreceptor dystrophy in whom analysis of the exonic and consensus splice site regions did not identify the cause of disease. This highlights the importance of investigating noncoding regions in order to obtain a conclusive molecular diagnosis in patients with a hereditary retinal dystrophy.