Genotypes of USH2A-related disorders
We identified 14 biallelic variants in USH2A, either homozygous or compound heterozygous, in a trans configuration. The segregation of these variants was confirmed by Sanger sequencing. Overall, 18 mutant alleles were implicated in the diagnosis of USH2A-related phenotypes, including c.251G > A:p.Cys84Tyr, c.2209C > T:p.Arg737*, c.2802 T > G:p.Cys934Trp, c.4372C > T:p.Arg1578Cys, c.4858C > T:p.Gln1620*, c.7120 + 1475A > G, c.8232G > C:p.Trp2744Cys, c.85559-2A > G, c.10593del:p.Ile3532Phefs*18, c.10712C > T:p.Thr3571Met, c.10724G > T:p.Cys3575Phe, c.11156G > A:p.Arg3719His, c.12708 T > A:p.Cys4236*, c.13112_13115del:p.Gln4371Argfs*19, c.14134-3169A > G, c.13964 T > C:p.Leu4655Pro, c.14835del:p.Val4946Trpfs*4, and c.14911C > T:p.Arg4971* (Table 1 and Supplementary Fig. S1). The amino acid residues were highly conserved in orthologs from several species (Supplementary Fig. S1), a finding consistent with the corresponding high Genomic Evolutionary Rate Profiling (GERP + +) score. Two variants, c.2802 T > G:p.Cys934Trp and c.8559-2A > G variants, were present more than twice. Consistent with our findings, these two variants have been recurrently identified and are recognized as East Asian-specific founder and mutational hotspot alleles associated with USH2A gene23. Four variants were novel, including two missense variants, one frameshift variant, and one deep intronic variant. The frameshift variant p.Ile3532Phefs*18 had a premature termination codon, which resulted in nonsense-mediated mRNA decay. This variant was extremely rare in population databases, including the Korean Reference Genome Database, which is based on genetic data from South Koreans. Furthermore, co-segregation analyses confirmed the presence of this variant in trans with a known pathogenic variant (c.8559-2A > G). Based on the American College of Medical Genetics and the Association for Molecular Pathology (ACMG/AMP) guidelines24,25, this novel frameshift variant was classified as pathogenic (PVS1, PM2, PM3). The novel missense variants p.Cys3575Phe and p.Leu4655Pro, located in the fibronectin type III-20 and fibronectin type III-32 domains, respectively, were also extremely rare in population databases and occurred in trans with a pathogenic variant. In silico analysis, using Combined Annotation Dependent Depletion26, predicted that these variants would be pathogenic. Based on the ACMG/AMP guidelines24,25, the novel missense variants p.Cys3575Phe and p.Leu4655Pro were classified as likely pathogenic (PM2, PM3, PM5, PP4) and variants of uncertain significance (PM2, PM3, PP4), respectively. Furthermore, two deep intronic variants, c.14134-3169A > G and c.7120 + 1475A > G, were identified by trio-based whole-genome sequencing (WGS) and bioinformatics analyses. The previously reported c.14134-3169A > G variant in intron 64 led to pseudo-exon activation, which resulted in a truncated protein27. The novel deep intronic variant c.7120 + 1475A > G in intron 37 was predicted to introduce pseudo-exon activation. This newly inserted pseudo-exon had a premature termination codon, which probably resulted in a truncated translation product. We have reclassified the conflicting or variants of uncertain significance documented in the ClinVar database, using the ACMG/AMP guidelines24,25. Upon evaluation using either the ClinVar database or the ACMG/AMP guidelines, 16 of the 18 variants (with the exceptions of c.251G > A:p.Cys84Tyr and c.13964 T > C:p.Leu4655Pro) were determined as either pathogenic or likely pathogenic (Table 1). The genotype information of the disease-causing USH2A variants, mapped to both hg19 and hg38, is described in Supplementary Table 1.
To investigate the allelic hierarchy of USH2A-related disorders, we categorized patients with USH2A biallelic variants into three groups based on their truncated and nontruncated mutant alleles: group 1 consisted of individuals with two truncated mutant alleles; group 2 consisted of individuals with one truncated and one nontruncated mutant allele; and group 3 consisted of individuals with two nontruncated mutant alleles (Supplementary Table 2).
Audiological phenotypes and allelic hierarchy
The audiological characteristics of our cohort are shown in Table 2; the corresponding audiograms are shown in Fig. 1. Based on the latest audiograms, three patients (18.8%) had normal hearing (SH479, SH767, and SH707), indicating nonsyndromic RP. The remaining 13 patients (81.2%) had varying degrees of SNHL; this was typically mild-to-moderate at low frequencies and severe-to-profound at high frequencies. Among these 13 patients, six had moderate SNHL (46.2%), five had moderately severe SNHL (38.5%), and two had severe SNHL (15.4%). Subjective onset of hearing loss was congenital in six (46.2%), prelingual in two (15.4%), and postlingual in five (38.5%) patients. All but one patient (SH351) had symmetric bilateral hearing loss that deteriorated at higher frequencies (10 of 12, 83.3%), or they exhibited a horizontal trace pattern (2 of 12, 16.7%). Consistent with the findings in previous studies21, during the follow-up period, which ranged from 0.3 to 14.1 years with a median duration of 1.3 years, all 11 patients who underwent hearing evaluations had milder progressive hearing loss. One patient (SH351) with biallelic truncated variants underwent unilateral CI in their worse ear before the age of 13 years. Neither congenital cytomegalovirus infection nor inner ear anomalies underlying asymmetric hearing loss were identified in this patient (SH351). Most patients with confirmed hearing loss used hearing aids.
The mean hearing thresholds were 59.7 dB, 51.1 dB, and 29.1 dB in group 1 (two truncated alleles), group 2 (one truncated and one nontruncated allele), and group 3 (two nontruncated alleles), respectively (Fig. 2a). The thresholds were consistently highest across frequencies in group 1 and lowest in group 3. We recorded mean hearing thresholds for group 1 versus group 2 versus group 3 at 250 Hz (35.0 vs. 24.3 vs. 15.0, p = 0.015 by Kruskal–Wallis rank sum test), 500 Hz (46.3 vs. 36.9 vs. 20.0, p = 0.021 by Kruskal–Wallis rank sum test), 1000 Hz (62.5 vs. 50.6 vs. 27.5, p = 0.005 by Kruskal–Wallis rank sum test), 2000 Hz (65.0 vs. 56.9 vs. 32.5, p = 0.024 by Kruskal–Wallis rank sum test), 4000 Hz (65.0 vs. 60.0 vs. 36.3, p = 0.032 by Kruskal–Wallis rank sum test), and 8000 Hz (78.3 vs. 75.7 vs. 41.3, p = 0.013 by Kruskal–Wallis rank sum test). Patients with at least one truncated allele (groups 1 and 2) tended to have more severe hearing loss than patients with biallelic nontruncated alleles (group 3; Fig. 2b); patients with one truncated allele had higher hearing thresholds than patients with biallelic nontruncated alleles (Fig. 2c). Nonsyndromic RP was more common among patients in group 3 than among patients in the other groups. All patients in group 1 had congenital or prelingual onset of subjective hearing loss, whereas most patients in the other groups (except three patients in group 2) had postlingual onset or normal hearing. Moreover, patients in group 1 were younger at ascertainment than patients in groups 2 and 3, although this difference was not statistically significant (Fig. 2d, p = 0.132 by Kruskal–Wallis rank sum test). In summary, our audiological data suggest that truncated alleles are associated with more severe phenotypes for both the extent and severity of progressive hearing loss; these effects were greater for biallelic truncated alleles, implying that the USH2A gene exhibits an allelic hierarchy from an audiological perspective.
Retinal phenotypes and allelic hierarchy
As previously reported, age has a strong effect on retinal phenotypes. Patients under 16 years of age did not report ophthalmological symptoms prior to clinical evaluation. Previous report shows that the onset of syndromic RP, although earlier than non-syndromic RP, is still delayed after adolescence15,16,17. Eight of the patients were under 16 years old, and their phenotypes mimicked nonsyndromic hearing loss. Therefore, eight patients aged < 16 years were excluded from analyses of retinal phenotypes; this led to exclusion of all participants from group 1. The remaining patients in group 2 were 33, 47, 57, and 57 years old; the remaining patients in group 3 were 30, 37, 40, and 65 years old. Visual acuity, electroretinography (ERG), Goldmann visual field (GVF), optical coherence tomography (OCT), fundus photographs, and fundus autofluorescence (FAF) assessments were used to evaluate each patient’s retinal phenotype.
The median best-corrected visual acuity values, determined using the logarithm of the minimum angle of resolution (logMAR) scale, showed that there was no significant difference in retinal function among the groups (group 2: median = 0.699, range = 0–1.699; group 3: median = 0.134, range = 0–2; Fig. 3a; p = 0.597 by Wilcoxon rank sum test). All patients included in this study did not have other ophthalmological diseases, which may affect visual acuity, such as cataract or preretinal fibrosis, at the time of evaluation. Additionally, we evaluated ERG responses for each patient. The ERG response of many patients with RP is undetectable, which indicates reduced or no function28. We measured residual responses using three different ERG settings: 0.01 scotopic ERG for a rod response, 3.0 scotopic ERG for a combined rod and cone response, and a photopic test for a cone response. The response of 0.01 scotopic ERG in all patients were undetectable. These results are consistent with the findings in previous reports, in which the progression of RP initially affected rod function13,14. Only one-fourth of the patients in group 2 exhibited residual cone function, whereas half of the patients in group 3 exhibited a cone response. Representative graphs for 3.0 scotopic ERG are shown in Fig. 3b. The 30-Hz flicker test evaluates cone function and was used to investigate RP in a previous study29. Group 3 patients exhibited higher response amplitudes compared with group 2 patients, indicating that cone function had been more effectively preserved in group 3 patients, but it was statistically insignificant (p = 0.215 by Wilcoxon rank sum test; Fig. 3c). As ERG is often altered in RP patients before other signs or symptoms appear, the ERG of excluded patients under age 16 were also analyzed. With the exception of one patient in group 2, 0.01 scotopic ERG response was undetectable. The photopic ERG response was found in all excluded patients due to age, indicating the preservation of cone function in younger ages. The visual fields of all patients in group 2 and two patients in group 3 were restricted to a central area. However, the other two patients in group 3 had relatively preserved visual fields (Fig. 3d). These ERG and GVF results reveal that patients with USH2A-related RP who have truncated alleles tend to exhibit more severe functional retinal degeneration.
For evaluation of retinal structure, OCT was used to examine the integrity of the junctional complex, which consists of an interdigitation zone, an ellipsoid zone, and an external limiting membrane30. Only one-fourth of the patients in group 2 exhibited conserved junctional complexes, compared with three-fourths of the patients in group 3 (Fig. 4a). Additionally, we compared retinas from patients in groups 2 and 3 with respect to bony spicules and waxy optic discs, which indicate retinal degeneration31. Fundus photographs revealed that group 2 patients had more bony spicules, and these were distributed more widely. Although this difference may occur from the difference of duration of the disease, the number and distribution, combined with overall retinal intactness, supports that group 2 patients show more severe phenotypes in a qualitative manner. In addition, the spicules in group 2 patients were thicker and wider in size and more apparent overall. Furthermore, all four patients in group 2 exhibited waxy optic discs, indicating progressing pallor; only one-fourth of the patients in group 3 exhibited similarly degenerate optic discs (Fig. 4b). Patients in group 2 also exhibited more paramacular mottling in FAF images than patients in group 3. Patients from both groups also underwent evaluation of hyper-autofluorescent rings (i.e., Robson–Holder rings), which are characteristic of RP32,33. The ring structures present in group 3 patients were less obvious than the structures present in group 2 patients or were completely absent. In addition, although being fainter, the integrity of the ring structure was more intact in group 2 patients, indicating the absence of the hyper-autofluorescent ring is not due to further advance of degeneration. Indicating less severe phenotypes in group 3 (Fig. 4c). Overall, these OCT, fundus, and FAF results indicate that patients with truncated alleles tend to exhibit more severe structural disruption in their retinas.
Correlations between audiological and ophthalmological phenotypes
In this study, we analyzed the correlation between audiological and ophthalmological phenotypes in eight suitable patients (groups 2 and 3; Fig. 5). Although mean hearing threshold was not correlated with logMAR visual acuity (Spearman correlation coefficient = 0.137, p = 0.613), 30 Hz flicker ERG amplitude (Spearman correlation coefficient = − 0.456, p = 0.076), or the maximum isopter on GVF evaluation (Spearman correlation coefficient = 0.050, p = 0.853), there was a notable correlation between mean hearing threshold and 30-Hz flicker ERG implicit time (Spearman correlation coefficient = − 0.525, p = 0.037). In particular, the correlation had distinct effects on the presence of a truncated allele. Group 2 patients exhibited a negative correlation (Spearman correlation coefficient = − 0.775, p = 0.024), whereas group 3 patients exhibited no discernible correlation (Spearman correlation coefficient = − 0.271, p = 0.515). In the same vein, subgroup analysis of Group 2 revealed correlations between hearing threshold and visual acuity (Spearman correlation coefficient = 0.859, p = 0.006) and between hearing threshold and 30 Hz flicker ERG amplitude (Spearman correlation coefficient = − 0.827, p = 0.011).