A detailed clinical and molecular survey of subjects with nonsyndromic USH2A retinopathy reveals an allelic hierarchy of disease-causing variants

Defects in USH2A cause both isolated retinal disease and Usher syndrome (ie, retinal disease and deafness). To gain insights into isolated/nonsyndromic USH2A retinopathy, we screened USH2A in 186 probands with recessive retinal disease and no hearing complaint in childhood (discovery cohort) and in 84 probands with recessive retinal disease (replication cohort). Detailed phenotyping, including retinal imaging and audiological assessment, was performed in individuals with two likely disease-causing USH2A variants. Further genetic testing, including screening for a deep-intronic disease-causing variant and large deletions/duplications, was performed in those with one likely disease-causing change. Overall, 23 of 186 probands (discovery cohort) were found to harbour two likely disease-causing variants in USH2A. Some of these variants were predominantly associated with nonsyndromic retinal degeneration (‘retinal disease-specific'); these included the common c.2276 G>T, p.(Cys759Phe) mutation and five additional variants: c.2802 T>G, p.(Cys934Trp); c.10073 G>A, p.(Cys3358Tyr); c.11156 G>A, p.(Arg3719His); c.12295-3 T>A; and c.12575 G>A, p.(Arg4192His). An allelic hierarchy was observed in the discovery cohort and confirmed in the replication cohort. In nonsyndromic USH2A disease, retinopathy was consistent with retinitis pigmentosa and the audiological phenotype was variable. USH2A retinopathy is a common cause of nonsyndromic recessive retinal degeneration and has a different mutational spectrum to that observed in Usher syndrome. The following model is proposed: the presence of at least one ‘retinal disease-specific' USH2A allele in a patient with USH2A-related disease results in the preservation of normal hearing. Careful genotype–phenotype studies such as this will become increasingly important, especially now that high-throughput sequencing is widely used in the clinical setting.


INTRODUCTION
Retinitis pigmentosa is the most common inherited retinal degeneration and a major cause of visual impairment among individuals aged 20-64 years. 1,2 It is genetically heterogeneous (over 60 genes implicated so far) and associated with significant variability in age of onset, disease progression and retinal appearance (RetNet; http://www.sph. uth.tmc.edu/retnet/, accessed 31 December 2014). 3 Although retinitis pigmentosa is a disease confined to the eye, some 20-30% of patients have associated non-ocular disease; Usher syndrome, in which retinitis pigmentosa is combined with, typically prelingual, sensorineural hearing loss, is the most frequent syndromic form. 4 Disease-causing variants in the USH2A gene are the most common cause of Usher syndrome (29% of all cases) and one of the most common causes of nonsyndromic autosomal recessive retinitis pigmentosa (19-23% of all cases). 4,5 The USH2A gene is located on 1q41 and has two alternatively spliced transcripts: a short one consisting of 21 exons, and a longer one consisting of 51 additional 3′ exons; the latter encodes a 5202 amino-acid matrix protein expressed specifically in photoreceptors and developing cochlear hair cells. 6,7 It has been shown that the USH2A protein is required for long-term maintenance of retinal photoreceptors and the development of cochlear cells. 7 Over 2900 coding variants have been reported in the USH2A gene (1000 genomes project database, National Heart, Lung and Blood Institute Exome Sequencing Project or NHLBI ESP and LOVD-USHBase, accessed 15 September 2014). Over 470 of these changes are presumed to be pathogenic (HGMD, accessed 15 September 2014); most of these disease-causing variants are found in one or a few cases each, with the exception of c.2299delG, p.(Glu767Serfs*21) and c.2276 G4T, p.(Cys759Phe), which are more commonly found in patient cohorts. The c.2299delG variant causes a substantial proportion of cases of Usher syndrome, 8 while the c.2276 G4T change has been associated mainly with disease confined to the eye. 9,10 This suggests the existence of alleles that are likely to be specific to those with nonsyndromic disease ('retinal disease-specific'). However, this has not been studied in detail and to date no other such 'retinal disease-specific' alleles have been identified.
Here, we have performed a comprehensive screen of the USH2A gene in 186 patients with autosomal recessive retinal degeneration and 1 no complaint of childhood hearing loss to gain insights into nonsyndromic USH2A-related disease. Detailed phenotyping, including fundus autofluorescence imaging and audiological assessment, was performed in individuals found to harbour two likely disease-causing variants. The effect of three splice site changes on nasal mRNA was investigated and multiplex ligation-dependent probe amplification (MLPA) was performed in selected patients with the aim of detecting deletions and duplications in USH2A. Additionally, we have investigated whether nonsyndromic versus syndromic disease can be explained by the presence of an allelic hierarchy of USH2A diseasecausing variants, and have addressed this by studying USH2A allelic heterogeneity in the discovery (n = 186) and two additional (n = 84 and n = 187) cohorts.

Study subjects
A total of 186 unrelated individuals with recessive retinal degeneration and no hearing complaint in childhood were ascertained from the clinics of Moorfields Eye Hospital (London, UK; discovery cohort). Of these, 168 patients were affected with rod-cone dystrophy (ie, retinitis pigmentosa), 12 with cone-rod dystrophy and 6 with childhood-onset retinal dystrophy. Further to this discovery cohort, 84 additional probands with recessive retinal degeneration (62 with nonsyndromic disease and 22 with Usher syndrome) were recruited at the Hospital for Sick Children (Toronto, ON, Canada); both cohorts underwent full sequencing of USH2A (see below). A third cohort of 187 unrelated patients with nonsyndromic, adult-onset, retinitis pigmentosa was also ascertained from the Moorfields Eye Hospital. This cohort was only used to test for selected variants in exons 13, 51, 57 and 63. A complete eye examination was performed and a detailed clinical history was obtained in all patients. Only patients with a family history compatible with autosomal recessive inheritance (ie, no evidence of dominant inheritance) were studied. For the purposes of this study, we define the phenotype observed in retinitis pigmentosa patients without prelingual/childhood-onset hearing loss as nonsyndromic disease.
Subjects with Usher syndrome were not tested for common genetic causes of nonsyndromic hearing impairment.
After informed consent was obtained, blood samples were collected and genomic DNA was extracted from peripheral blood leucocytes. Control DNA and mRNA were obtained from consented unrelated healthy individuals. All investigations were conducted in accordance with the Declaration of Helsinki principles. Institutional Review Board (IRB)/Ethics Committee approval was obtained from the Moorfields Eye Hospital and the Hospital for Sick Children local ethics committees.
USH2A screening and method used to distinguish disease-causing variants The 186 probands with recessive retinal degeneration and no hearing complaint in childhood and the 84 probands with recessive retinal degeneration underwent bidirectional DNA sequencing of the 72 exons of the USH2A gene. Primers were designed for all exons and intron-exon boundaries of the transcript with accession number NM_206933.2 (a total of 105 primer pairs; genomic reference sequence NG_008212.3). 8 The additional panel of 187 unrelated individuals with nonsyndromic, adult-onset, recessive retinitis pigmentosa underwent bidirectional DNA sequencing of exons 13, 51, 57 and 63 of the USH2A gene; these exons were the location of presumed 'retinal disease-specific' variants. In 24 patients from the discovery cohort, only one likely disease-causing USH2A variant was identified. These patients were: (i) sequenced for the USH2A c.7595-2144 A4G, p.Lys2532Thrfs*56 change, which causes inclusion of a pseudoexon; 11 and (ii) screened using MLPA to detect deletions and duplications in the USH2A gene. The SALSA MLPA FAMlabelled reagent kit with probe mixes P361-A1/P362-A2 developed by MRC-Holland (MRC-Holland, Amsterdam, The Netherlands) was used and reactions were performed according to the manufacturer's instructions. Two subjects with previously identified heterozygous deletions were included to act as positive controls and confirm the validity of the method.
Sequence alterations were classified as 'likely disease-causing variants' if they (i) were either nonsynonymous (including missense) or coding insertions/ deletions or splice site (positions ± 3) or large duplications/deletions and (ii) have a minor allele frequency (MAF) of 0.15% or less in the NHLBI ESP data   ; this is the MAF of c.2276 G4T, the most common change identified in patients with recessive retinitis pigmentosa. 9 The effect of synonymous variants on splicing was not assessed in the present study. All relevant data (variants and associated phenotypes) were submitted to the appropriate LOVD-USHBase, which can be accessed in http://www.LOVD.nl/ USH2A (submission IDs: 0004401 − 0004452).

Nasal epithelial mRNA analysis
Nasal epithelial mRNA analysis was performed in individuals carrying the variants c.12295-3 T4A (subject D13), c.9056-2 A4G (subject D10) or c.5776 +1 G4A (subject D10). Nasal epithelial cell samples were obtained by gently brushing the lateral inferior turbinate with bronchial cytology brush (Diagmed Ltd, North Yorkshire, UK) and RNA was extracted from these samples using the NucleoSpin RNA II Extraction Kit (Macherey-Nagel, Duren, Germany) according to the manufacturer's guidelines. cDNA was reverse transcribed using a cDNA Synthesis Kit (BioLine, London, UK) with a random hexamer primer mix. For reverse transcriptase (RT)-PCR reactions, USH2A was amplified between exons 27 and 32, exons 45 and 49, exons 58 and 64 and exons 62 and 66. The housekeeping gene β-actin was amplified as a positive control. The identity of USH2A RT-PCR products was established by direct sequencing using standard procedures (primer sequences and conditions are available on request).

Ophthalmological and audiological assessment
Detailed ophthalmological evaluation was performed in 23 probands with two likely disease-causing variants in USH2A (all from the discovery cohort); 4 affected siblings of the 23 probands were also assessed. Ophthalmological examination included best-corrected visual acuity testing, dilated fundus examination, colour fundus photography (TRC-50IA; Topcon, Tokyo, Japan), spectral domain optical coherence tomography (OCT) and fundus autofluorescence imaging. The Spectralis HRA+OCT with viewing module version 5.1.2.0 (Heidelberg Engineering, Heidelberg, Germany) was used to acquire tomographs in 24 patients; fundus autofluorescence images were acquired in 24 patients using the HRA2 and Spectralis HRA+OCT (over a 30°× 30°and/or a 55°× 55°field; Heidelberg Engineering) instruments. Pure tone audiometry was conducted, in a sound-treated booth, using a calibrated GSI 61 audiometer with TDH 49 headphones to assess hearing thresholds 12 in 19 patients with no complaint of hearing loss, who were found to have two disease-associated variants in USH2A (all from the discovery cohort); the method has been described previously. 8 The audiology thresholds (0.25-8 kHz) were reviewed for right and left ear separately and compared with age-and gender-matched percentile bands of normative data. 13 Patients were classified into three groups: Group 1 with normal hearing thresholds falling in the interquartile normative range for age and gender across all frequencies (1 A if all thresholds o40th percentile; 1B if thresholds fell in the 50-60th percentile band); Group 2 with high-frequency hearing thresholds (3-8 kHz) 475th percentile and markedly greater than low-frequency threshold (0.25-2 kHz) percentiles (ie, 30-70 percentile bands difference); and Group 3 with atypical/ abnormal audiometric configurations 14 and without other aetiological explanation.

RESULTS
USH2A retinopathy is a major cause of adult-onset recessive retinal degeneration In the discovery cohort (n = 186), a total of 206 different sequence alterations were recorded in the exons and intron-exon boundaries of USH2A; 88 variants were missense, coding insertions/deletions or splice site changes. Of these, 52 were rare with an MAF o0.15% and thus were classified as likely disease-causing (21 were novel to this study and 31 were previously reported). Notably, 22 patients were found to harbour two of these likely disease-causing variants, whereas 24 were found to carry one likely disease-causing variant. In the latter group, one proband was found on MLPA testing to harbour a heterozygous duplication of exons 57-60. None of them was found to carry the c.7595-2144 A4G change, which causes inclusion of a pseudoexon. 11 All 23 patients with two likely disease-causing variants were affected with adult-onset retinitis pigmentosa (Table 1). Therefore, 14% of patients with adult-onset recessive retinitis pigmentosa (23/168) were found to harbour two likely disease-causing variants in USH2A.
In the replication cohort of 84 patients with recessive retinal degeneration (syndromic and nonsyndromic), 25 additional probands with two likely disease-causing USH2A variants were identified. All 25 had a retinal phenotype consistent with retinitis pigmentosa; 11 of these had Usher syndrome type II and the remaining 14 reported no hearing complaint in childhood (Table 2).
When Some USH2A alleles are only associated with nonsyndromic retinal disease We define as 'retinal disease-specific' variants or alleles that (i) were present in more than one patient with nonsyndromic retinal degeneration (in our discovery cohort and/or the literature) and (ii) have not been clearly associated with Usher syndrome type II to date (Table 3). On this basis, the following variants were categorised as likely 'retinal disease-specific': c.2802 T4G, p.(Cys934Trp); c.10073-G4A; c.11156 G4A; c.12295-3 T4A; and c.12575 G4A. The c.2276 G4T variant that has been previously associated mainly with disease confined to the eye 9,10 was also included in this group. Notably, the most prevalent c.2276 G4T and c.10073 G4A variants were statistically significantly enriched in nonsyndromic cases compared with Usher syndrome type II cases (P = 0.0060 and P = 0.047, respectively (Fisher's exact test); the data on Usher syndrome type II were obtained from the UK National Collaborative Usher Study 8 ). All 'retinal disease-specific' variants were located in laminin-type EGF-like domains or fibronectin type 3 domains (Figure 1).
On examining our data, it was rare for nonsyndromic cases to have two 'null' variants (ie, variants that are not missense and predicted to cause nonsense-mediated decay and/or significant truncation of the protein if translated), whereas this was common in those with Usher syndrome. Of 47 probands with nonsyndromic USH2A-related retinal degeneration (all three cohorts), only 5 had apparent biallelic 'null' variants (this is the exception due to c.12295-3 T4C; see Table 1). This is significantly different to 39 out of 71 patients with USH2Aassociated Usher syndrome 8 (P = 0.0001 (Fisher's exact test)). These data support the model that Usher syndrome represents the null phenotype consequent upon USH2A defects, and that 'retinal diseasespecific' alleles are partly functional, allowing them to contribute to normal cochlea development.
Assuming our model of allelic hierarchy is correct, further rare 'retinal disease-specific' alleles can be sought. In those probands with nonsyndromic retinal disease, an allele is likely to be 'retinal diseasespecific' if either (i) it is homozygous or (ii) it is paired with an allele that has previously and consistently been reported to be associated with Usher syndrome or (iii) it is paired with an allele that has previously and consistently been reported to be associated with nonsyndromic retinitis pigmentosa. On examining our discovery and replication cohorts in this way, the following alleles are likely to be 'retinal disease-specific': c.2332 G4T, p.     Numbering of USH2A variants has been assigned in accordance with NCBI Reference Sequence NM_206933.2. The complete list of references can be found in LOVD-USHBase. a A single heterozygous variant in MYO7A was also reported in this patient; *a homozygous (hom) variant in MYO7A was also reported in this patient. b After reviewing the clinical data, the patient was categorised as atypical due to adult onset of hearing loss (45 years old) and normal speech. c After reviewing the clinical data, the patient was categorised as atypical due to very mild, progressive hearing loss.  hyperautofluorescent ring represents a border between relatively preserved and diseased retinal tissue (Figure 3), and future structure-function correlation studies are expected to provide important insights into the clinical utility of this imaging modality Audiological phenotype in USH2A retinopathy is variable The 23 patients with two likely disease-causing variants from the discovery cohort and their four affected relatives reported no hearing loss in childhood, with 9 (33%) of these reporting subjective adultonset hearing loss. Audiological assessment was conducted in 19 of these patients. In 14 (74%) patients, it was consistent with a Group 1 phenotype (ie, thresholds within normal limits); of those, 9 (47%) were classified in Group 1A and 5 (26%) in Group 1B. Three (16%) patients were classified in Group 2 and 2 (10%) in Group 3 (Supplementary Figure S2). Qualitative analysis revealed the c.12295-3T4A variant to be correlated with a more severe audiological phenotype (Groups 2 and 3). There seems to be no obvious correlation between the severity of visual and audiological phenotypes (Table 1, Figure 4 and Supplementary Figure S2).

DISCUSSION
In the present study, we confirm that recessive variants affecting USH2A function are a common cause of retinitis pigmentosa with disease-causing variants being spread throughout the gene. When allelic heterogeneity was studied and compared with that reported in Usher syndrome, the concept of 'retinal disease-specific' USH2A alleles (ie, alleles associated with retinal degeneration and no hearing complaint in childhood) became apparent. The presence of at least one such allele in a patient with USH2A-related retinal degeneration results in relative preservation of hearing ( Figure 5). Five likely 'retinal disease-specific' variants (c.2802T4G; c.10073G4A; c.11156G4A; c.12295-3T4A and c.12575G4A) that are novel to this study were identified in addition to c.2276G4T, a relatively common sequence alteration previously associated with retinitis pigmentosa without hearing impairment. 9,10 Defects in the USH2A gene have been previously reported to account for 12-25% of all retinitis pigmentosa cases, dominant, recessive or X-linked; syndromic or nonsyndromic. 5 McGee et al 5 reported that among 80 patients with nonsyndromic recessive retinitis pigmentosa, 23% had one or two likely disease-causing variants in USH2A. This result was consistent with our findings: 24.7% (46/186) of patients with recessive retinal degeneration and no reported hearing loss in childhood harboured one or two likely disease-causing changes. Many variants in USH2A can be overlooked when only coding regions and intron-exon boundaries are sequenced. In previous studies, it was found that screening for duplications, deletions and a common deep-intronic sequence alteration (c.7595-2144A4G) detected a second disease-causing variant in 35% of cases with Usher syndrome type II that had only one variant affecting function on conventional Sanger sequencing of all USH2A exons. 11,47 In the present cohort, Figure 4 Variability in the severity of ocular and audiological phenotype due to mutations in USH2A in three patients of similar age. Fundus autofluorescence (FAF) imaging and foveal OCT show a better preserved retina in subject D2 (group 3 corresponding to abnormal audiological assessment) and in a patient with Usher syndrome type II compared with subject D3 (middle panel), who has normal hearing.
Allelic hierarchy in USH2A retinopathy E Lenassi et al despite performing MLPA analysis and testing for the c.7595-2144A4G change, a significant number of cases with only one likely disease-causing in USH2A variant remained (23/186, 12.4%). This observation can be attributed to (i) a number of variants affecting function still being overlooked and/or (ii) some changes defined here as likely disease-causing being rare benign polymorphisms.
The first report of an USH2A change being associated with recessive retinitis pigmentosa without hearing impairment was by Rivolta et al 9 in 2000; this change was a G to T transversion in exon 13 (c.2276G4T). 9 Since then many studies have confirmed this finding (Table 1) and the c.2276G4T variant is often considered to be the most common disease-causing variant in patients with nonsyndromic retinitis pigmentosa. 9,10,24 In the present study, c.2276G4T was found in 6.5% (12/186) of cases; in all study subjects, it was in cis with a previously reported polymorphism c.2256T4C, p.(His752His), suggesting a common ancestral haplotype. However, previous reports have shown that the c.2276G4T can be a recurrent sequence alteration. 17 We have found five additional changes that are also associated with retinitis pigmentosa and no hearing complaint in childhood (presumed 'retinal disease-specific' variants). These include four previously reported missense variants (c.10073G4A, 5,8,23,26,41 c.2802T4G, 44 c.11156G4A 5 and c.12575G4A 5,8,23,26,41 ) and a splice site change (c.12295-3T4A 8 ). We have shown that the latter results in exclusion of exon 63 and an out-of-frame deletion. Three out of four patients harbouring this variant had no auditory complaint, despite the presence of an abnormal auditory phenotype, which would suggest a long-standing or very slowly progressive neurosensory hearing impairment ( Table 1).
Analysis of allelic heterogeneity in USH2A in our discovery cohort (23 probands with presumed USH2A-related retinitis pigmentosa and no hearing complaint in childhood) revealed that all but one (patient D9) patient harboured at least one 'retinal disease-specific' or novel (presumed 'retinal disease-specific') USH2A allele (Table 1). Importantly, a similar pattern ( Figure 5) was observed in our replication cohort (16 probands with presumed USH2A-related retinitis pigmentosa and no hearing complaint in childhood; Table 3). Only in one case (patient D9; Table 1) there appeared to be discordance: a c.2332G4T change was identified in homozygous state in a Somali patient with retinitis pigmentosa and normal audiometric testing; previously, this change has been reported in heterozygous state in a patient with Usher syndrome type II. 35 One explanation for this could be that this is a rare polymorphism. Recently, a similar allelic hierarchy has been reported for change affecting CDH23 function: a 'nonsyndromic deafness' CDH23 allele in trans configuration with a 'syndromic/Usher syndrome type I' CDH23 allele preserves vision and balance in deaf individuals. 48 Audiometric findings were within normal limits for most patients with two likely disease-causing USH2A variants and no hearing complaint in childhood (14 of 19 tested; Table 1). Notably, the severity of the retinal phenotype did not obviously correlate with the severity of the hearing impairment ( Figure 4); this is in keeping with previous reports. 24 Interestingly, the eldest subject D23 in the present study reported adult-onset hearing loss and had a hearing defect consistent with Usher syndrome type II at age 75 years. This is in keeping with the notion that recessive variants in USH2A cause a spectrum of hearing defects that range from an early-onset phenotype consistent with Usher syndrome type II to completely normal hearing.
We have shown that an allelic hierarchy of variants affecting USH2A function is likely with 'retinal disease-specific' alleles being phenotypically dominant to 'Usher syndrome type II' alleles. Although this finding has implications for counselling, the fact that USH2A diseasecausing variants are often private makes prediction of the fully evolved phenotype challenging. In any case, the audiological phenotype in USH2A-related disease is highly variable and a multidisciplinary approach is often relevant even to cases without hearing complaints in childhood.