Comprehensive analysis of syndromic hearing loss patients in Japan

More than 400 syndromes associated with hearing loss and other symptoms have been described, corresponding to 30% of cases of hereditary hearing loss. In this study we aimed to clarify the mutation spectrum of syndromic hearing loss patients in Japan by using next-generation sequencing analysis with a multiple syndromic targeted resequencing panel (36 target genes). We analyzed single nucleotide variants, small insertions, deletions and copy number variations in the target genes. We enrolled 140 patients with any of 14 syndromes (BOR syndrome, Waardenburg syndrome, osteogenesis imperfecta, spondyloepiphyseal dysplasia congenita, Stickler syndrome, CHARGE syndrome, Jervell and Lange-Nielsen syndrome, Pendred syndrome, Klippel-Feil syndrome, Alport syndrome, Norrie disease, Treacher-Collins syndrome, Perrault syndrome and auditory neuropathy with optic atrophy) and identified the causative variants in 56% of the patients. This analysis could identify the causative variants in syndromic hearing loss patients in a short time with a high diagnostic rate. In addition, it was useful for the analysis of the cases who only partially fulfilled the diagnostic criteria.

Mutation spectrum and clinical features of BOR syndrome patients. We conducted genetic analysis of 59 probands with clinical findings of BO/BOR syndrome (16 typical cases, 43 atypical cases, Supplementary Table S1), and identified causative heterozygous variants in 19 probands (diagnostic rate 32%). We identified the causative variants in 12/16 typical cases, but the causative variants were identified in only 7/43 atypical cases. Table 3 summarizes the identified variants and clinical features of the probands and all affected family members for 18 families with EYA1 mutations, and one family with a SIX1 mutation. There were no BO/BOR cases caused by the SIX5 gene variant. Among the 12 EYA1 variants, 8 were truncating variants (five were nonsense, one was frameshift, two were splice site), three were missense variants and one was a copy number variation (one copy number loss). Four of them were novel variants and 8 of them were previously reported. JHLB4043 had one copy loss detected using NGS read depth data, which seemed to be deleted in all of the EYA1 gene, confirmed by array Comparative Genomic Hybridization (aCGH). The mutations identified in this study were located in exon 6 to exon 13, and frequently observed in exon 8 and exon 12. Two or more cases carried the same variants (p.R264X, p.R275X, c.867 + 5 G > A, p.R328X and p.R407Q). EYA1 variants were mainly identified from autosomal dominant families (10/18 cases); however, we also identified variants from 7 sporadic cases. Among them we confirmed de novo mutations in four families ( Supplementary Fig. S1). The case with a SIX1 mutation was also caused by de novo mutation ( Supplementary Fig. S1). In terms of the clinical features of all BO/BOR-affected patients with EYA1 and SIX1 gene variants (19 probands and their family members who carried the same variants; 34 patients in total), the most frequent symptom was hearing loss (31/ 32, 97%). Unilateral hearing loss was observed in 2 cases. The most frequent type of hearing loss was moderate mixed hearing loss. Middle and/ or inner ear anomalies were observed in 22 of 23 cases who underwent CT imaging (96%). Twenty-seven of 31 cases had preauricular pits (87%), and 14 of 25 cases for whom information was available had branchial anomalies (56%). Renal anomalies, on the other hand, were revealed in only one of 7 cases for whom kidney abnormalities were examined (14%). It is noteworthy that there were only a limited number of cases (7/34) with renal ultrasonographic information available in our cohort, thus the frequencies of renal anomalies may be underestimated. The presence of branchial or renal anomalies was not correlated with the severity of hearing loss. Furthermore, no relationship was found between genotype and clinical findings. As rare symptoms, one patient had hemifacial palsy, and 3 cases had eye symptoms.

Mutation spectrum and clinical features of Waardenburg syndrome patients.
We conducted genetic analysis of 23 probands with hearing loss and one or more clinical findings typical of Waardenburg syndrome, and identified the causative heterozygous variants in 18 probands (diagnostic rate 78%). Table 4  of the identified variants were truncating variants (four were nonsense, 7 were frameshift, one was splice site) and only three cases had missense variants, one each in PAX3, MITF and SOX10. In addition, we also identified three cases with one copy number loss of the SOX10 gene identified from NGS read depth data and confirmed by aCGH. Thirteen variants were novel and five variants (three PAX3 mutations, and one each with MITF and SOX10 mutations) were previously reported. Computer prediction scores, allele frequency information and the pathogenicity classification for novel variants are listed in Supplementary Table S2. Autosomal dominant inherited cases were 2/4 in PAX3 cases, 4/5 in MITF cases, 1/8 in SOX10 cases, and 1/1 in EDNRB cases. The other 10 cases were sporadic cases, with confirmed de novo mutations in SOX10 in five cases (Table 4). In terms of the clinical features of the probands and all family members harboring the same causative gene variants (29 patients from 18 families in total), the most frequent symptom was hearing loss (27/29, 93%), followed by heterochromia iridis (23/28, 82%). The severity of hearing loss for each gene is shown in Fig. 1, with the frequency of the profound hearing loss higher in cases with MITF and SOX10 mutations. Two cases with PAX3 mutations had bilateral normal hearing and three cases with MITF mutations had unilateral hearing loss. Only a limited number of patients showed discoloration of the hair and skin: hair discoloration was seen in two cases (with SOX10 and MITF mutations), leukoderma in one case with a SOX10 mutation, and excessive freckles in three cases with MITF mutations. No abnormal musculoskeletal findings were observed in any case. Dystopia canthorum was seen in two cases with PAX3 mutations, and one each with MITF and SOX10 mutations. The other associated symptoms observed in SOX10 cases were ptosis (JHLB4270, JHLB4310), developmental delay (JHLB4310) and Asperger syndrome (JHLB3480). In addition, inner ear anomalies, including hypoplasia of the semicircular canal, cochlea, cochlear nerve, and saccular vestibule, were observed. It is suggested that there is no obvious correlation between the type of mutation and its location and the severity of the symptoms. Most of the clinical findings for cases associated with each gene were in agreement with previous reports; however, we identified phenotype-genotype disagreement in two Waardenburg syndrome 1 (WS1) cases (JHLB2469 with a MITF mutation and JHLB5132 with a SOX10 mutation).  www.nature.com/scientificreports www.nature.com/scientificreports/ The proband with spondyloepiphyseal dysplasia congenita had a novel truncating variant in COL2A1. She and her father, who harbored the same variant, had characteristic clinical features (cleft palate, short stature and short extremities). Their hearing level was severe to profound sensorineural hearing loss.

Mutation spectrum and clinical features of other syndromic hearing loss patients.
With regard to Stickler syndrome, we identified pathogenic variants in the COL11A1 (two cases) and COL11A2 (one case) genes. All identified variants were truncating (two were splice site, one was frameshift), with the two variants in COL11A1 being novel. One COL11A1 and one COL11A2 variant were identified from an autosomal dominant family, and one COL11A1 variant was identified from a sporadic case (de novo). As to the clinical features of the probands and all family members harboring the same causative gene variants (8 patients in total), hearing loss was observed in 75% of cases (3/3 with COL11A1 variants, 3/5 with COL11A2 variants), with the severity of hearing loss being mild to moderate. Two children of the proband with a COL11A2 variant (JHLB4181) carried the same variant but had normal hearing. Seventy-five percent of cases (6/8) had a cleft palate or uvula bifida (3/3 with COL11A1 variants, 3/5 with COL11A2 variants), and all three cases with COL11A1 variants had congenital myopia. One case harboring a COL11A2 variant, who was the son of the proband, had no symptoms.
A novel OPA1 variant was identified in one case who suffered auditory neuropathy with optic atrophy. Two other pathogenic amino acid substitutions have been previously identified in the same position. The proband had amblyopia since infancy, and bilateral moderate sensorineural hearing loss. OAE (Otoacoustic emission) presented a normal response, the ABR (Auditory Brainstem Response) threshold was out of scale, and MRI (magnetic resonance imaging) showed bilateral cochlear nerve hypoplasia. The proband's mother had similar symptoms (no DNA sample was available).
With regard to Pendred syndrome, we identified SLC26A4 variants in 32 probands with autosomal recessive inheritance or sporadic cases. No variants in KCNJ10 and FOXI1 were identified in cases with heterozygous SLC26A4 variants.

Discussion
In this study, we conducted a comprehensive analysis of Japanese syndromic hearing loss patients to clarify mutation spectrums and clinical features by using NGS analysis with a multiple syndromic targeted resequencing panel. This analysis had a high diagnostic rate (56%) and was suitable for comprehensive analysis. Further, it allowed us to clarify the types and frequency of causative genes in Japanese syndromic hearing loss patients. In addition, it was particularly useful in cases that only partially fulfilled the respective diagnostic criteria. To the best of our knowledge, this is the first study using targeted resequencing panel analysis for multiple syndromic hearing loss patients.
With regard to BO/BOR syndrome, the causative variants were identified in 32% (19/59) of probands (16 typical, 43 atypical). The diagnostic rate was increased to 75% when we restricted the analysis to typical BO/ BOR cases (12/16). Krug et al. reported the results of genetic analysis for a large number of BO/BOR patients and identified the causative variants in 36% of cases. Similar to this study, the diagnostic rate was increased to 76% when they restricted subjects to typical BO/BOR cases 28 . Unzaki et al. analyzed 36 Japanese families with clinically diagnosed BO/BOR syndrome and identified causative genes in 72% of them 29 . Thus, the diagnostic rate in this study was similar to the rates in these previous reports. EYA1 variants account for 95% of the causative gene variants identified in this study. Similarly, EYA1 was commonly identified in BO/BOR cases in previous studies; 85% in Japanese patients 29 and 93% in French patients 28   www.nature.com/scientificreports www.nature.com/scientificreports/ possibility that variants in other genes (such as SALL1) or genomic rearrangement (inversion or translocation in chromosome 8) may contribute to these cases. In this study, we also identified one copy number loss with a 2.8 Mb deletion of 8q13.2-q13.3 including the EYA1 gene in one familial case. The frequency of one copy number loss of the EYA1 gene was 6% (1/18) in this study. In other reports, copy number loss of the EYA1 gene was also involved in BO/BOR syndrome, with 7% to 10% or more of cases caused by EYA1 copy number loss 28,29 . The most frequent clinical feature was hearing loss, which was observed in 97% of cases (31/32), followed by preauricular pits in 88% (29/33). In other reports, the most frequent clinical feature was also hearing loss; however, the frequencies of other symptoms varied, with the frequency of renal symptoms higher in some reports [28][29][30] . Chen et al. reported renal anomalies in 67% of affected individuals 31 , with about 6% of them progressing to renal failure 32 . Some of them were asymptomatic in the first decade but required dialysis or renal transplantation in adulthood 33,34 . In this study, only one case showed congenital renal anomalies. One plausible reason for this lower rate of renal anomalies was that we enrolled BO/BOR candidate patients, and information regarding renal anormalies was available for only a limited number of patients (renal ultrasonographic information was available for only 7/34 cases). Therefore, more cases may have had renal symptoms. In cases in which BO/BOR syndrome is suspected clinically or genetically, even in the absence of renal dysfunction in early childhood, renal examination may be important.
It is noteworthy that three cases from two unrelated families with EYA1 variants presented visual symptoms (progressive disturbance of vision, amblyopia and hypermetropia), but visual symptoms are not typically associated with BOR syndrome. EYA1 is needed for the formation of the anterior portion of the eye 35 . Azuma et al. reported one case who presented with congenital cataracts with a BOR phenotype (cervical fistula, unilateral multicystic kidney and conductive hearing loss due to ossicular malformations), and others have also reported cases with visual symptoms (dysopia, cataract, micrognathia, and iris coloboma) 28,29,[36][37][38] . The frequency of amblyopia is reported to be 3.0% to 3.2% in the general population 39,40 , but the frequency of visual symptoms in the EYA1-related BO/BOR patients in this study was a little higher (9%). There is a possibility that visual symptoms actually represent a rare clinical feature of BO/BOR syndrome.
Waardenburg syndrome was subdivided into four types based on the clinical findings, and each causative gene was identified. We successfully identified the genetic causes in 80% of WS1 probands (4/5), 71% of WS2 probands (10/14), and 100% of WS4 probands (2/2). Hoth et al. reported that point mutations in PAX3 have been identified in more than 90% of affected individuals with WS1 or WS3 41,42 . In this study, we identified one case each with MITF and SOX10 variants from WS1. MITF and SOX10 variants were generally identified from WS2 or WS4 patients. Similarly, MITF, EDNRB, and SOX10 variants were identified from WS1 patients in previous reports [42][43][44] . The cause of this inconsistency between phenotype and genotype may be 1) a new genotype-phenotype correlation or 2) the wider distance between the inner canthus in the Japanese population. In the Japanese literature, Motomura reported the inter-inner canthal, inter-outer canthal and inter-pupillary distance for each age group among Japanese (published in Japanese) 45 . It appears that the W-index calculated from these data may exceed 1.95 in many age groups (Supplementary Fig. S5). In future, it may be necessary to consider ethnic differences when evaluating the W-index. Among the WS2 cases, we identified the causative variants in 29% of cases with MITF, in 36% with SOX10, and in 7% with EDNRB. Pingault 48 . Taken together, these results indicate that SOX10 variants may be more frequently identified in East Asian WS2 cases.
It is worth noting that we also identified one copy number loss of the SOX10 gene using NGS read depth data and confirmed by aCGH in three cases. Two probands had a large deletion within the chromosome 22q13.1, a proband had the whole SOX10 gene deletion, and the other proband in a familial case had a partial deletion of SOX10. To date, more than 20 cases caused by copy number variation in PAX3 or SOX10 have been reported 42,47,[49][50][51][52][53][54] . We identified one SOX10-assiociated WS case with developmental delay and one with Asperger syndrome. Both of these cases carried truncation variants; however, no cases were observed with developmental www.nature.com/scientificreports www.nature.com/scientificreports/ delay among the SOX10 CNV cases. Thus, the association between genotype and developmental delay phenotype remains unclear.
In addition, we also identified one familial WS case with variations in phenotype among family members. In the MITF family (JHLB1623), the father had only unilateral hearing loss and excessive freckles, and her younger brother had only bilateral severe sensorineural hearing loss, but both had the same variant. It is usually difficult to   Direct sequencing. After the filtering process, described previously 55 , we performed confirmation of the identified variant and family segregation analysis by Sanger sequencing.

CNV (Copy Number Variation) analysis.
CNV analysis was performed with NGS analysis read depth data according to the method described in a previous report 58 .

aCGH (Array Comparative Genomic Hybridization).
To confirm the CNVs identified from NGS read depth data, we performed array CGH analysis with the Agilent 8 × 60 K whole genome array (Agilent Technologies, Santa Clara, CA). We used the same DNA samples as for the amplicon re-sequencing, and quality assessment was also carried out. Ten microliters of genomic DNA solution (0.5ug of DNA) were fragmented, labeled with cyanine-3 for reference DNA samples and cyanine-5 for subjects, and then hybridized. Scanning of the array was carried out according to the manufacturer's recommended protocols. Scanned aCGH data were analyzed using CytoGenomics software version 3.0.6.6 (Agilent Technologies).