Patients with Usher syndrome (MIM 276900-2, MIM 276905, MIM 605472) have autosomal recessive retinitis pigmentosa (arRP) and sensorineural hearing impairment. In addition, vestibular dysfunction can be seen in a subset of patients. Usher syndrome occurs in ∼1/20 000 individuals and represents 50% of all cases with deafblindness.^1–3 Three clinical subtypes can be distinguished.⁴ Patients with Usher syndrome type I (USH1) show severe to profound congenital hearing loss, RP and vestibular areflexia. Patients with Usher syndrome type II (USH2) have moderate to severe hearing loss, RP and normal or variable vestibular function (Huygen et al, unpublished data). Usher syndrome type III (USH3) patients present with progressive hearing loss, RP and variable vestibular function. The eight genes known to be involved in Usher syndrome are CDH23, MYO7A, PCDH15, USH1C and USH1G for USH1, USH2A and VLGR1 for USH2, and USH3A for USH3.^{5,6,7,8,9,10,11,12,13,14,15} The USH2A gene is also implicated in arRP.^16–18 Likewise, mutations in CDH23, MYO7A, PCDH15 and USH1C have been reported in patients with non-syndromic hearing impairment.^7,12,19–22

Identification of the causal mutations is important for the early diagnosis of Usher syndrome, which is relevant for the decision whether or not to elect for a cochlear implant, for genetic counselling and for prenatal diagnosis. In addition, gene-specific treatments might become available in the near future, necessitating knowledge of the underlying gene defect. Comprehensive molecular diagnostics for Usher syndrome has been hampered both by the genetic heterogeneity and the large number of exons for six of the eight known Usher genes. The five USH1 genes collectively contain 179 protein-coding exons, the two USH2 genes comprise 162 protein-coding exons, and the USH3 gene USH3A has six protein-coding exons, some of which are alternatively spliced (UCSC Human Genome Browser at http://genome.cse.ucsc.edu). Thus far, large-scale mutation screening has been carried out using single strand conformation analysis and denaturing gradient gel electrophoresis, with subsequent sequence analysis of fragments with an aberrant migration pattern. For routine analysis, these techniques are both time consuming and expensive.

In this study, we used microarray technology that was previously used for efficient mutation analysis of the ABCA4 gene in patients with either autosomal recessive Stargardt disease or autosomal recessive cone-rod dystrophy,^23,24 as well as for seven genes implicated in Leber congenital amaurosis (LCA).^25,26 It is inherent to the arrayed primer extension (APEX) that it detects only known mutations and, therefore, its efficiency is highly dependent on the extent of earlier mutation analysis efforts. Here, we describe the first comprehensive Usher genotyping microarray which allows the identification of known Usher gene defects in a high-throughput and cost-effective manner.

METHODS

Patients

For validation purposes, we obtained genomic DNA from 158 patients with Usher syndrome in which the variants were originally identified. In order to test the efficiency of the Usher microarray (hereafter denoted “evaluation”), 370 patients were studied. Of these, 170 unrelated Usher syndrome patients were ascertained in seven European countries: Belgium (20 patients), Denmark (10 patients), UK (34 patients), Germany (27 patients), Italy (20 patients), Spain (5 patients), Switzerland (7 patients) and the Netherlands (47 patients). A further 200 patients from the USA (Boys Town, Omaha) were ascertained. In total, we tested DNAs from 140 patients with USH1, 189 patients with USH2, 21 patients with USH3 and 20 patients with atypical Usher syndrome (USHA). The patients, except for 104 USH2 patients from the USA who were previously shown not to carry the USH2A c.2299delG (p.E767fs) mutation, had not been previously tested for the presence of Usher mutations. Research procedures were in accordance with institutional guidelines and the Declaration of Helsinki. Informed consent was obtained at each centre from all patients after the nature of procedures to be performed was fully explained.

Molecular methods

Microarrays were designed and manufactured according to arrayed primer extension (APEX) technology.^27,28 A detailed description of the methodology can be found elsewhere (www.asperbio.com).^29,30 In brief, 5′-modified sequence-specific oligonucleotides are arrayed on a glass slide. In general, these oligonucleotides are designed with their 3′ end immediately adjacent to the variable site. PCR-amplified and fragmented target nucleic acids are annealed to oligonucleotides on the slide, followed by sequence-specific extension of the 3′ ends of primers with dye-labelled nucleotide analogues (ddNTPs) by DNA polymerase. The APEX reaction is, in essence, a sequencing reaction on a solid support.

A total of 120 amplicons from eight genes (table 1) were amplified as described previously.²³ Primer sequences are available on request. In the amplification mixture, 20% of the dTTP was substituted by dUTP.²⁹ The amplification products were concentrated and purified (GENErALL PCR kit; General Biosystems, Seoul, Korea). The fragmentation of amplification products was achieved by adding thermolabile uracil N-glycosylase (Epicenter Technologies, Madison, WI) and heat treatment.²⁹ One-sixth of every amplification product was used in the primer extension reaction on the Usher microarray. Each APEX reaction consisted of fragmented and pooled denatured PCR products, 4 U Thermo Sequenase DNA Polymerase (Amersham Biosciences, Little Chalfont, UK), 1× reaction buffer and 1.4 μM final concentration of each fluorescently labelled ddNTP: Texas Red-ddATP, fluorescein-ddGTP (Amersham Biosciences), Cy3-ddCTP and Cy5-ddUTP (NEN, Boston, Massachusetts, USA). The reaction mixture was applied to a microarray slide for 15 min at 58°C. The reaction was stopped by washing the slide at 95°C in Milli-Q water.³¹ The slides were imaged with the Genorama QuattroImager (Asper Biotech, Tartu, Estonia) and the sequence variants were identified by Genorama Genotyping software (Asper Biotech).^23,30

Array-identified variants were confirmed by direct sequencing with the ABI PRISM Big Dye Terminator Cycle Sequencing V2.0 Ready Reaction Kit and the ABI PRISM 3730 DNA analyser (Applied Biosystems, Foster City, California, USA).

RESULTS

Construction of the Usher genotyping microarray

We compiled a list of 298 previously published and communicated sequence variants from the coding region and adjacent intronic sequences of the eight Usher genes: CDH23, MYO7A, PCDH15, USH1C, USH1G, USH2A, USH3A and VLGR1 (see supplementary table 1 available at http://jmg.bmj.com/supplemental). Intronic sequence variants were included when they were predicted to affect splicing, that is, altering the strictly conserved splice site sequences (nucleotides −2A or −1G of 3′ splice sites; nucleotides +1G or +2T of 5′ splice sites), causing transversion of one or a few of the pyrimidines (−5 through −14) of a 3′ splice site, changing a −3C nucleotide of a 3′ splice site, changing a +3, +4, +5 or +6 nucleotide of a 5′ splice site, or exonic nucleotides being part of splice sites, to a less favourable nucleotide. In addition, we included the putatively benign sequence variants c.5156A>G (p.Y1719C) in MYO7A and c.688G>A (p.V230M), c.1434G>C (p.E478D), c.2052A>G (p.Q684Q, possible cryptic splice site), c.2137G>C (p.G713R) and c.2522C>A (p.S841Y) in USH2A. Each sequence variant was queried in duplicate by the software, from both the sense and the antisense strand (Genorama Genotyping software).

We also included four variants on the chip despite the fact that there are inconsistencies in their notations in the original publications or erroneous genomic sequence data: CDH23 c.3880>T (p.Q1294X),¹²MYO7A c.269G>C (p.R90P),³²MYO7A c.4039_4053del (p.R1347_F1351del),³³ and USH2A c.233T>G (p.F78V). The USH1C_ABCC8 122 815-bp deletion⁷ is not represented on the chip.

Validation of the Usher genotyping microarray

We collected DNA samples from 158 patients with Usher syndrome and at least one known variant. All variants, except two, CDH23 c.3617C>G and MYO7A c.223delG, were reliably identified. For the two unidentified variants (false negatives), primers have been redesigned and will be retested (see supplementary table 1 available at http://jmg.bmj.com/supplemental). For the remaining 140 variants, no DNA samples from the original patients were available as positive controls. However, the wild-type sequences of these variants were robustly identified. Since 102 of these variants represent nucleotide substitutions, the identification of the wild-type sequence can be regarded as a positive control since the same 25 mers are employed for the identification of both the wild-type and mutant nucleotides. However, we cannot exclude the possibility of erroneous sequence variant annotations, which we observed for a number of mutations in the respective original publications. Thirty eight non-validated variants represent insertions/deletions for which in some instances oligonucleotides linked to the microarray for detection of the wild-type allele differed from that for detection of the mutant allele.

Evaluation of the Usher genotyping microarray

The efficiency of the Usher microarray was evaluated using a total of 370 patients: 140 USH1, 189 USH2, 21 USH3 and 20 USHA. The results are given both separately and combined for the novel Usher patients from Europe and the USA (table 2). The European patients originated from eight Western European countries, that is, the Netherlands (12.7%), the UK (8.1%), Germany (7.3%), Belgium (5.4%), Italy (5.4%), Denmark (2.7%), Switzerland (1.9%) and Spain (1.4%). The majority of the patients from the USA were of European descent. For USH1, genotyping using the microarray revealed the highest percentage of patients with either one or two mutations, 51% for the European patients, 30% for the patients from the USA and 46% when both groups are combined. When counting mutated alleles, these percentages were 34%, 23% and 31%, respectively. The best results were observed for the USH1 patients of Danish origin in whom a mutation was found in 80% (8/10) of cases or in 65% (13/20) of alleles.

For USH2 the percentages of patients with variants were 25%, 24% and 25%, respectively. The percentages for detected USH2 alleles were 20%, 14% and 18%, respectively. A total of 104 of the patients with USH2 from the USA were pre-screened for the p.E767fs mutation and patients heterozygous or homozygous for this mutation were not included. This pre-screening did not influence the genotyping detection rate for the microarray which was 23% of the patients and 15% of the alleles. The numbers of patients with USH3 or USHA were relatively low. Consequently, the sensitivity of the microarray for these patient groups, although rather good in the present cohorts of patients, cannot be reliably assessed.

The USH2A variant c.2299delG (p.E767fs) was previously described as a common mutation in Caucasian patients with USH2, comprising between 13% and 39% of USH2A alleles.^34–38 The genotyping presented here revealed a second common mutation, c.11864G>A (p.W3955X) in exon 61 of this gene. This nonsense mutation truncates the long isoform of the USH2A protein.³⁹ Three of 47 European patients with USH2 were heterozygous for this allele, twice in combination with the p.E767fs mutation. In USH2 patients from the USA, 15 alleles with this mutation were detected in 13 of 142 patients. In total, 18 of 61 mutant alleles (30%) identified in patients with USH2 were p.W3955X. However, since some of the USH2 patients were pre-screened for the common p.E767fs mutation, this percentage was probably higher than would be expected in an unselected patient cohort. For the European patients this percentage was only 16% (3/19). The p.W3955X mutation was also heterozygously present in a patient with USHA in whom the second mutation was not identified. Haplotype analysis has to be performed to determine whether the p.W3955X mutation is a founder mutation as has been shown for p.E767fs.⁴⁰

Of the mutations detected, 191 were re-evaluated by sequence analysis and only two could not be confirmed (false positives), which illustrates the robustness of the method. One mutation, c.496+1G>A in the USH1C gene, was not detected by the microarray but was detected afterwards by sequence analysis (false negative). Segregation analysis could be performed in relatives of 11 probands in whom at least two variants were detected by chip analysis. In one of these families, the analysis revealed that two CDH23 variants were present on the same allele. For the remaining 10 probands the results are compatible with both alleles each carrying one mutation.

Identification of novel variants in Usher genes

Since all four nucleotides are available during primer extension, the APEX technology allows the detection of new nucleotide variants for positions that were already known to be variant and are represented by an assay on the microarray. In this way two novel variants were detected and confirmed by sequence analysis. In PCDH15, c.400C>T is predicted to cause the nonsense mutation p.R134X. At this position the substitution of a guanine for a cytosine was already described to cause the missense mutation p.R134G in a family with non-syndromic hearing loss.²¹ In the USH2A gene, a c.1606T>A was detected, which is predicted to cause an amino acid substitution p.C536S. The nucleotide substitution c.1606T>C (p.C536R) was reported by Dreyer et al³⁴ in one of 31 unrelated patients with USH2 who also carried the p.E767fs variant.

Variants in genes known to be associated with a different USH type

In a number of patients, variants were detected that did not correspond with the type of Usher syndrome in the patient (table 3). In five patients with USH1, one or two mutations were detected in the USH2A gene. In USH2 patients, variants were detected in the MYO7A gene and also in the USH3A gene, and in USH3 patients variants were seen in MYO7A and in USH2A. Whether the heterozygous mutations in the MYO7A gene are indeed causative such that mutations in MYO7A can also cause USH2 and USH3, remains to be determined. In many cases, the clinical diagnosis may be ambiguous. This is also true for the heterozygous USH2A mutations in patients with USH1 and USH3 and USH3A mutations in patients with USH2.

In patients with USHA, mutations were detected in MYO7A, USH2A and CDH23 (table 3).

Variants identified in two Usher genes

In nine patients, variants in two different genes were detected, that is, in four samples used for validation and in five patients tested for evaluation of the chip (table 4). In three of the nine cases, a variation in a second gene was present in addition to two mutations in one of the other Usher genes. As segregation analysis could not be performed for any of these cases, we cannot conclude that they represent true digenic inheritance, whether the variant in the second gene contributes to the severity of the phenotype or whether the results are merely a coincidence. However, in one of the cases with three sequence variants, two in USH2A and one in MYO7A, the USH2A mutations are protein truncating. These truncating mutations are most likely sufficient to cause the disease and thus only a modifying effect for the p.R1343S variant in MYO7A seems plausible. A second case had two truncating mutations in MYO7A in addition to p.P608R in USH1C. Also for this case, the third variant might modify the phenotype but is not likely to be primarily disease causing. The combination of variants in the MYO7A and USH2A genes were seen in six of the eight cases with variants in two Usher genes, which is not unexpected since these genes are causative in the majority of the patients with USH1 and USH2, respectively, and thus are expected to exhibit the highest carrier frequencies.

USH variants with questionable pathogenicity

Several sequence variants in Usher genes have previously been described for which pathogenicity was uncertain. A number of these were included in the microarray and, based on the results of the evaluation experiment, were regarded as polymorphisms. Nine alleles with the missense variant p.E478D^17,41 in the USH2A gene were detected in patients with all three types of Usher syndrome and in three patients, one of each with USH1, USH2 and USH3, this variant was present in addition to two pathogenic mutations. The frequency in controls was determined for the Jewish population (0/120 alleles).⁴¹ However, we detected this variant in three out of 180 control alleles from the Dutch population. The USH2A variant p.V230M was detected nine times in patients with all three types of Usher syndrome. In these patients, only one putative pathogenic USH2A mutation was detected in addition to the p.V230M variant. Dreyer et al³⁴ did not detect this variant in controls, although Jian Seyedahmadi et al¹⁷ detected similar frequencies in patients and controls. The USH2A variant p.G713R was described by Dreyer et al³⁴ as likely to be pathogenic since it was not detected in controls. However, Jian Seyedahmadi reported that the variant did not co-segregate with the disease in some families. In the present genotyping, the p.G713R variant was detected as the only variant heterozygously in two patients with USH1 and in one patient with USH3. In an USH2 patient it was detected in addition to a truncating mutation. This variant was not present in an affected sib of one of the USH1 patients. The USH2A variant c.2052A>G (p.Q684Q)⁴² was described as possibly creating a cryptic splice site. This nucleotide substitution was detected once in an USH2 patient as the only variant. The USH2A variant p.S841Y was previously described to be present in two of 186 control alleles.³⁸ In the present genotyping this variant was detected in one USH1 patient with one MYO7A mutation and in one USH1 patient with one mutation in CDH23.

Najera and co-workers questioned whether the variant p.Y1719C in MYO7A was disease-causing³³ since it was detected in 3/168 control alleles. We detected this variant in four patients with all three different types of Usher syndrome, two of whom did not have an additional variant in one of the Usher genes. This does not support a causative effect of the mutation.

DISCUSSION

A cost-effectiveness comparison of APEX mutation analysis with other medium and high-throughput mutation detection techniques

Routine mutation analysis of patients suffering from a genetically heterogeneous disease has been severely hampered by the high costs associated with application of conventional analysis techniques for comprehensive mutation screening. Despite the fact that most Usher syndrome patients can be reliably grouped into one of two main clinical groups, a systematic mutation analysis of either the USH1 or USH2 genes still requires the analysis of 179 and 162 protein coding exons, respectively. As depicted in table 5, the APEX-based mutation detection array is an affordable and flexible system as new alleles can be added without the need to design completely new screening platforms. The main disadvantage is its low efficiency in early versions since it relies on the extent of previous mutation analysis studies performed using other techniques. The Affymetrix resequencing technique has two disadvantages. First, heterozygous deletions larger than 1 bp and all insertions and duplications are missed because the complexity of the microarray would increase disproportionally if matching primers were added to the chip. These variants comprise 47/292 (16%) of all pathologic USH variants. Second, the screening costs, including development costs, are quite high, prohibiting the use of this technique in a routine laboratory setting.

A direct comparison of efficiency with other mutation detection techniques is not straightforward since no comprehensive mutation detection studies have yet been performed for more than one Usher gene. For USH1, the most commonly used technique is single strand conformation polymorphism (SSCP) analysis, which yielded between 19% and 48% of mutant alleles in MYO7A,^32,33,43 which is comparable to our overall efficiency (31% of alleles; table 2). For USH2, SSCP and sequence analysis of exons 1–21 of USH2A yielded 23% and 38% of mutant alleles, respectively,^17,38 which is higher than with the APEX technology (18%). The relatively low yield of USH2 alleles is due to the fact that the USH2A gene was expanded only recently from 21 to 73 exons³⁹ and because comprehensive mutation analysis of this gene lags behind MYO7A.

Thus, efficiencies for the Usher chip at the moment can be best compared to the recently developed APEX chip for LCA, which is comparable in complexity (contains ∼300 variants found in seven genes) and yields on average 20–28% of LCA alleles.^25,26 The Usher microarray chip is particularly useful for the analysis of patients with USHA, who have been shown to carry mutations in the MYO7A, CDH23 and USH2A genes.

The overall efficiency of the Usher chip in the future is expected to increase in proportion to the addition of new variants. We propose the use of this mutation chip as a first-pass screening tool for all Usher patients. In patients with one or no variant, capillary-based heteroduplex analysis of the larger genes (CDH23, MYO7A, PCDH15, USH1C, USH2A, VLGR1) or sequence analysis of the smaller genes (USH1G, USH3A) can be performed. Novel variants from these studies can subsequently be added to the chip, resulting in a gradual increase in its efficiency. The Usher syndrome mutation detection chip can be obtained from Asper Ophthalmics (Tartu, Estonia; http://www.asperophthalmics.com/UshersyndromeDNAtest.htm).

Deficiencies in the Usher chip and other conventionally used mutation detection techniques

Currently, none of the standard mutation analysis techniques is able to identify heterozygous deletions that span one or both amplicon primers. As a result, we have no knowledge whatsoever about the frequency of deletions of this kind. Therefore, a customised probe set needs to be designed that can identify copy-number changes, for example, multiplex ligation dependent probe amplification or quantitative PCR.⁴⁴ Likewise, deep intronic mutations affecting splicing are missed as RNA analyses are not part of standard procedures due to restricted expression of the Usher genes.

Pathogenicity of variants in Usher genes

When a known mutation is detected in both copies of one of the Usher genes, one can confidently conclude that these mutations are disease-causing. Two alleles were found in 24/140 (17%) USH1 patients, 9/85 (11%) USH2 patients, 2/21 (10%) USH3 patients and 2/20 (10%) USHA patients. However, when one mutated allele is observed, the chance that a false conclusion is made regarding causality of this particular variant for the phenotype in that patient, differs for each of the Usher syndrome genes. For the MYO7A gene in USH1 patients, the percentage of patients with a false interpretation based on the presence of one mutated allele for this gene was calculated to be 2.4; for the USH2A gene in USH2 patients, this percentage was calculated to be 0.8.⁴⁵ For diagnoses based on one allele with a missense mutation, this percentage is likely to be higher due to the fact that the pathogenicity of this type of mutation is difficult to prove. Due to the small sizes of the families with Usher syndrome and the fact that samples of other family members are often not available, it is mostly not possible to perform segregation analysis to prove the pathogenicity of a specific sequence variant.

We also included on the microarray a number of sequence variants that we assume are not pathogenic. The results in the present study support this assumption. However, it cannot be excluded that these variants can modulate the disease severity in a way comparable to that of a sequence variant in Bardet-Biedl syndrome. An exonic nucleotide substitution that affects splicing of the MGC1203 RNA modulates the phenotype of patients with Bardet-Biedl syndrome with mutations in the BBS1 gene because the BBS1 and MGC1203 proteins interact.⁴⁶ Since both the USH1 and USH2 proteins are known be part of an Usher interactome,^47–52 phenotypic modulation might well be possible and was already suggested in a family with homozygous USH3A mutations in the affected members and a mutated MYO7A gene in one of them.^53,54

Digenic inheritance in Usher syndrome?

Since Usher proteins interact with each other and form dynamic protein complexes, digenic or oligogenic inheritance of Usher syndrome would not be surprising. Indications for digenic inheritance involving mutations in the PCDH15 and CDH23 genes have been obtained for two USH1 patients and a patient with USHA,⁵⁵ but no elaborate segregation analysis was reported to strengthen this observation. The large number of patient DNAs screened in the present study is specifically suitable for detection of rare cases exhibiting digenic inheritance of the syndrome. From the nine cases with sequence variants in two different genes, one is unlikely to follow digenic inheritance since two truncating mutations in one of the genes were present. PCDH15 and CDH23, previously implicated in digenicity, were not involved in any of the remaining eight cases. Segregation analysis could not be performed for any of the cases and therefore, we cannot confirm or exclude digenic inheritance.

Use of the Usher microarray for patients with non-syndromic arRP and hearing impairment

Mutations in the USH2A gene have been described to be a common cause of RP and the p.C759F mutation is almost exclusively detected in patients with non-syndromic RP. Based on their results of screening exons 1–21 of the gene, Jian Seyedahmadi et al¹⁷ estimated that USH2A mutations are present in 7% of all RP cases in the USA and screening of only one third of exons 22–73 increased this to 16% for arRP patients.⁵⁶ Since all USH2A mutations detected in RP patients are represented on the Usher microarray, this is also suitable for this patient population. However, a mutation array with known mutations in a large number of RP genes, including USH2A, has been developed⁵⁷ and is more suitable for mutation detection in patients with RP.

Mutations in the USH1 genes CDH23 (DFNB12), MYO7A (DFNA11, DFNB2), PCDH15 (DFNB23) and USH1C (DFNB18) can cause non-syndromic hearing loss.⁵¹ Among these genes, CDH23 seems to be the most frequently involved in non-syndromic hearing loss. The hearing loss in DFNB12 patients ranges from moderate to profound, but also patients with hearing loss that started as mild in childhood have been described to have mutations in CDH23.²⁰ So far, there are no indications that mutations in the USH1 genes are a common cause of non-syndromic hearing loss. However, more extensive mutation analysis in this patient category is necessary to determine the value of the present microarray for this group of patients.

ELECTRONIC DATABASE INFORMATION

The UCSC Human Genome Browser is at http://genome.cse.ucsc.edu, the Asper Biotech website is at http://www.asperbio.com and the Asper Ophthalmics web site is at http://www.asperophthalmics.com/UshersyndromeDNAtest.htm