Usher syndrome type 1 (USH1) is the most severe form of Usher syndrome and is characterised by profound congenital hearing loss, vestibular dysfunction, and prepubertal onset of retinitis pigmentosa. To date six loci, USH1B–G, have been mapped and five causative Usher genes have been cloned: myosin VIIA (MYO7A), harmonin (USH1C), cadherin 23 (CDH23), protocadherin 15 (PCDH15), and SANS (USH1G) known to be involved in USH1B, USH1C, USH1D, USH1F, and USH1G, respectively.^1,2

It is not possible to distinguish clinically between the forms of USH1. To date mutations in MYO7A and CDH23 have been reported most often. Patients with USH1 typically have mutations in CDH23 and PCDH15, resulting in truncated proteins (nonsense, frameshift, or splice mutations).^2–6 This is, however, not true for the MYO7A gene, where many missense mutations have been reported.^7,8,9,10,11 Demonstrating that these are truly pathogenic changes and not non-functional polymorphic variants complicates the interpretation.

MYO7A has 49 exons, CDH23 has 69 exons, and PCDH15 has 33 exons, posing a considerable problem for direct sequencing. Once the extent of genetic heterogeneity became apparent, haplotype analysis was introduced before mutation screening to assist in identification of candidate genes. This enabled us to search for regions of homozygosity in families with known or suspected consanguinity and to check for co-segregation in the six families with more than one affected child (exclusion mapping). It would also have been valuable if there had been a founder gene mutation in the population, such as the PCDH15 Arg245X mutation in Ashkenazi Jews¹² or the CDH23 IVS4+1G→A and/or Arg1502X mutations in Swedes.⁵

Using this strategy we identified 46 different pathogenic mutations in four of these genes (of which 29 are novel) and unambiguously genotyped 31 of 34 patients. MYO7A accounted for 55% of cases and the cadherin genes, CDH23 and PCDH15, accounted for 19% each. The USH1C gene is rarely involved and no mutation was identified in the USH1G gene.

METHODS

Patients

Patients were referred from medical genetic clinics distributed all over France. The parents were available in almost all cases. All patients had audiograms and fundus examination or electroretinograms (ERGs), or both. USH1 was diagnosed on the basis of congenital profound sensorineural deafness, vestibular dysfunction, and retinal degeneration. The degree of retinitis pigmentosa was variable among the patients, whose ages varied from 3 to 40 years (see table in the supplemental material, which can be viewed on the journal website, http://www.jmedgenet.com/supplemental).

This study was approved by the local ethics committee and consent to genetic testing was obtained from adult probands or parents of minors. Ethnic origins included Turkish, North African, Senegalese, Guinean, French West Indian, and white (Europid). A questionnaire was completed in the clinic to evaluate the possibility of consanguinity.

Haplotype analyses

In all, 32 microsatellite markers were used to build haplotypes at each known USH1 locus. The majority of markers correspond to Genethon markers, their relative order was confirmed by the ensembl genome server www.ensembl.org, and the detailed list is available in the supplemental material.

PCR amplification of the USH1 genes and sequencing

Genomic DNA was extracted from peripheral blood using standard procedures.

The coding exons and flanking intronic sequences of MYO7A, PCDH15, CDH23, USH1C, and USH1G were analysed by direct sequencing of 189 amplicons. Most of the primer sequences were obtained from published reports but a few had to be modified to ensure specific amplification or avoid any risk of allele dropout owing to the presence of a single nucleotide polymorphism (SNP) (identified on the Ensembl Genome server). The list of primers used and the conditions used for polymerase chain reaction (PCR) are available in the supplemental material. We customised the SCAIP (single condition amplification/internal primer) method, initially described by Flanigan (2003)¹³ which allows a single amplification condition for all exons of a gene, followed by sequencing on a single 96-well plate.

The same PCR primers were used for sequencing unless previous sequence analyses had shown poor quality results, in which case internal primers were designed. Sequences were run on an ABI 3100 DNA analyser and assembled using the ABI Prism Seqscape 2.1 using reference sequences of the USH1 genes extracted for the NCBI databases. Setting of the basecaller was according to the manufacturer, and records any base with a second peak of >5% as mixed.

URLs and GenBank accession number

http://www.ensembl.org/Homo_sapiens/;

MYO7A: NM_000260; CDH23: AY010111; PCDH15: NM_033056; USH1C: NM_153676; USH1G: NM_173477

Control DNAs

Guthrie cards were obtained from the neonatal screening centre GREPAM in Montpellier. All samples were anonymised and neither phenotypic nor ethnic origin data were available. DNA was extracted using standard procedures. The Ensembl server was also scanned for SNPs.

In silico studies

The SIFT (Sorting Intolerant From Tolerant) program developed by Ng¹⁴ was used to predict the consequences of the amino acid substitutions on the protein function (http://blocks.fhcrc.org/sift/SIFT.html).

Analysis in parents and siblings

Whenever substitutions were identified in a patient, segregation analysis was carried out on all available family members to ascertain parental origin, by sequencing, denaturing high performance liquid chromatography, or restriction fragment length polymorphism.

RESULTS

Haplotype analyses

Initially, molecular analyses were focused on MYO7A and CDH23 genes by direct sequencing. However, as some patients had no mutation in either of these genes, it became necessary to analyse further USH1 genes as they were identified. It appeared useful to undertake indirect studies with markers surrounding the USH1 genes (table 1). Even when there was no known consanguinity it helped to prioritise the order in which genes were screened. Examples are U402 and U332, where mutations were found in PCDH15 and CDH23, respectively. Consanguinity or a founder effect, as shown by homozygosity for the haplotype and the mutation, was found in 10 families, although only in five of the cases had it been suggested by the family history. It was useful in a further six of the 23 families screened where there was more than one sibling. In addition, in patient U153, apparent non-inheritance was noticed for one intragenic marker of PCDH15 (D10S2536 localised in intron 3). Analysis using exonic and intronic SNPs confirmed a deletion spanning at least exons 3 to 5 of the PCDH15 gene (tables 1 and 2). Further deletions of PCDH15 were identified using quantitative PCR. D10S2522, adjacent to PCDH15, failed to amplify in patient U382 and a homozygous deletion of exon 1 was confirmed.

Mutational analysis

Eight families had mutations in MYO7A or CDH23 (U94, U95, U98, U155, U178, U179, U310, and U20). For the remaining families, sequencing was carried out on the genes selected or at least not excluded by haplotyping, in the following order: MYO7A, CDH23, PCDH15, USH1G, and USH1C.

The 46 mutations identified which could confidently be called disease causing are presented in table 2. Family data are reported in table 1. Twenty nine mutations are novel and include three deletions, eight nonsense and eight missense mutations, seven frameshift alterations, and three splicing aberrations. Mutations found in CDH23, PCDH15, and USH1C were overwhelmingly truncation mutations. A G→A mutation in the last nucleotide of exon 45 of CDH23 is most likely to act by altering splicing rather than by making the predicted Gly2017Ser change.

Genes were deemed to be disease causing if both alleles were predicted to code for prematurely truncated products by the presence of either a nonsense mutation, a splice defect, or a deletion leading to a frame shift; or if one allele with a missense mutation was detected intrans with an allele containing a nonsense mutation or the equivalent (in this study or in published work, as referenced in table 2). Missense changes that were found on both alleles were assessed on an individual basis.

In addition none of the missense changes was identified in a minimum of 168 chromosomes.

Because only one parent was available for U297 and U402, quantitative PCR was carried out for the exons carrying the deleterious mutations. U402 was shown to have the Arg991X mutation on both alleles, but U297 was in fact hemizygous for Arg290X and carried a deletion, in trans, spanning exon 8. The apparent homozygosity of the linked markers was caused by a combination of hemizygosity and non-informativeness.

Likely non-pathogenic variations

Various exonic sequence variants, listed in table 3, predicted to be non-pathogenic were identified during the study. They were either also found in controls or additionally in a patient who already had two disease causing mutations identified. Table 3 includes three variants previously described as disease causing but for which our data suggest they are likely to be polymorphisms (see Discussion).

Several intronic variants have also been identified and can be communicated on request.

DISCUSSION

This study confirms that providing a diagnostic service for USH1 is a challenging but soluble problem, both because three genes are involved at a significant level and because a wide range of mutations is found in each of these genes, with few if any hotspots. However, as we were able to detect putative disease causing mutations in more than 90% of families referred it implies that there is no additional major gene for USH1. Using the rapid prescreening test based on linkage and haplotype analysis, we show that a diagnostic test based on the three genes MYO7A, CDH23, and PCDH15 is feasible, even with their complex pattern of exons. Homozygosity for markers around a single gene proved useful in 10 families, indicating that the disorder arose by consanguinity or resulted from a founder effect. This was subsequently confirmed by the demonstration of a homozygous change in the relevant gene. In only five cases had the close relation been reported in advance. In the other cases there was either no family history available or no known consanguinity.

The interpretation of nucleotide changes in MYO7A is particularly difficult because of the large number of missense mutations. The Ala198Thr variation caused by a G to A substitution at position c.592 of MYO7A is predicted to alter the normal splicing of exon 6 (the Shapiro and Senepathy score¹⁵ is reduced from 83.6 to 71.2), as it lies in the relatively highly conserved last position of the exon donor site. First, we have interpreted the missense changes as disease causing if the missense alteration was identified in trans of a deleterious mutation in either this study or previous ones. None of the changes was found in at least 352 control chromosomes. This applied to Gly163Arg, Thr165Met, Thr204Ala, Glu1170Lys, and Arg1240Gln.^11,16 Second, we have interpreted them as disease causing if they are in trans to one of the above. This applies to Leu1858Pro, Arg1873Trp, and the in-frame deletion Phe1962del. Three patients remain where the pathogenic nature of the changes is less firmly based. U94 is a compound heterozygote for two missense mutations, His133Asp and Gly519Asp. Neither mutation was found in 300 control chromosomes. Their deleterious effect remains to be proven, but it is unlikely that these anomalies, localised in trans, are not related to the disease. Patient U314, whose parents were known to be consanguineous, was homozygous for Gly163Arg. Similarly U142 was homozygous for a Lys164Arg mutation. These fall in a cluster of mutations involving conserved residues GESGAGKTE (position 158 to 166) which form part of the ATP binding site.¹⁷ In particular residues GKT (163–165) are not only conserved among more than 10 species (including pig, drosophila, and yeast) but also among numerous myosins (not shown, ClustalX¹⁸). We consider both anomalies likely to be deleterious mutations.

Tyr1719Cys and Thr1566Met (MYO7A) have both previously been reported as mutations, but we have found them in controls (2/982 and 8/1104, respectively) and consider they should in future be considered non-pathogenic variants. Arg1060Trp (CDH23) was identified as a heterozygous mutation in a patient presenting with non-syndromic hearing loss.⁵ We found this variant in patient U177 who carries two truncating mutations in PCDH15. Neither of the parents was recorded as having hearing loss, although audiograms were not done.

Two missense changes were identified in CDH23. The c.6049G→A change (Gly2017Ser) corresponds to the last nucleotide of exon 45 and reduces the Shapiro and Senepathy score from 84 to 71.5. We consider it is most likely to affect splicing. There remains the Glu247Lys anomaly identified in trans of a truncating mutation in patient U335. No other pathogenic effect was found in the four other USH1 genes which were completely sequenced, but we cannot exclude the possibility of a missed mutation, allelic to Glu247Lys. Neither can we exclude the possibility of a deleterious genotype in an unknown USH1 gene. This patient had a particularly severe retinitis pigmentosa phenotype and we also excluded the two most common mutations, Cys759Phe and c.2299delG, found in patients with non-syndromic retinitis pigmentosa linked to USH2A. It is possible that the missense change causes an usually severe phenotype by a dominant negative mechanism.

Two patients were found with mutations in USH1C. One was homozygous for 238_239insC frameshift which has been reported previously.^19,20 The other patient was a compound heterozygote for a missense Arg103His in trans of a c.522-2A→T anomaly. This was the only potentially pathogenic anomaly identified after sequencing CDH23, PCDH15, USH1G, and USH1C. MYO7A was ruled out by linkage exclusion analysis in the patient and two normal siblings.

Relative involvement of the USH1 genes

Among the 34 families genotyped in this study, MYO7A remains the most frequently involved gene, but only accounts for 55% of the cases. CDH23 and PCDH15 are also common with 19% each. In our cohort, USH1C was only involved in 6% of cases and no mutation was identified in USH1G. Our results suggest a higher level of mutations in cadherin genes, particularly PCDH15, than in previous reports.^2,21

No mutational hotspot was identified apart from a clustering of mutations in the ATP binding site, nor any recurrent mutation (no allele frequency was above 6%). Twenty nine of 46 mutations are new, reinforcing the fact that most mutations are “private” or restricted to populations not studied so far.