Five patients with familial congenital deafness were screened for 15 deafness genes on the 454 Genome Sequencer in order to evaluate if next generation sequencing enables the detection of mutations. Three of these patients are members of families in which a region of interest has previously been characterized by linkage studies. Therefore, mutations in respectively CDH23
(for patients 1 and 2) and OTOF
(for patient 4) were expected. In patient 5, a known mutation (c.236
A) in TMC1
had to be confirmed. For patient 3, no mutation was known.
We chose the Roche 454 next generation sequencing platform because of the longer read lengths (up to 400
bp) of the pyrosequencing approach. We decided to screen the 15 most important autosomal recessive deafness genes, selected on their reported mutation frequency [4
] (Table ). By limiting the number of genes to 15, we estimated to analyze 10 to 15 patients within a single sequencing run [10
The enrichment protocol prior to the sequencing was based on an individual PCR amplification of all amplicons covering the coding region of the 15 genes. A prerequisite to use this PCR approach is that the reaction conditions for all the amplicons are uniform, enabling a semi-automated workflow by using 384-well plates and liquid handling robots. In a pilot experiment, we designed 162 forward and reversed primers for the coding regions of OTOF and CDH23, using our in-house developed primer design software package primerXL, in order to evaluate the uniformity of the reaction conditions. Amplification was performed on normal genomic DNA (Roche diagnostics) and all amplicons, except 3, could be amplified. An average Cq of 27 was obtained (Figure ). The 3 failed amplicons were redesigned. Subsequently, primers were designed for the coding regions of the remaining 13 genes. In the end, a set of 646 primer pairs covering all the coding sequences and most of the UTRs of the 15 deafness genes in our setup was developed (Figure ).
Cq values for the 15 genes. The Cq values for the amplicons covering all of the 15 genes are displayed. One amplicon of the TRIOBP gene failed during this PCR (red).
In the next step, the patients’ DNA samples were amplified. Initially, the PCRs for patients 1, 2 and 3 were performed. These patients were analyzed with the standard chemistry on a full PicoTiterPlate using 2 large regions (Table ). The total number of reads after filtering and mapping was 325.272, with an average read length of 250
bp. The resulting mean coverage was 221x (patient 1), 168x (patient 2) and 194x (patient 3). Approximately 95
% of the screened amplicons had a coverage above 38, which makes them available for variant analysis (Table ) [10
]. As such high average coverage is not necessary for reliable variant analysis; the next run was modified accordingly.
For patients 4 and 5, 1/5 of the capacity of a Standard run with 2-lane gasket was used for each patient. The average coverage was lower than expected and therefore an additional Titanium run (1/12) was performed to have extra reads (Table ), resulting in an average total coverage of 73 and 88 for respectively patient 4 and patient 5 (Table ).
A novel missense mutation in CDH23
could be identified for patients 1 and 2, a result that is in agreement with the convincing linkage data previously found (data not shown). Both are homozygous for c.5527
T resulting in an amino acid substitution from Asp to Tyr at position p.1843 in exon 43. The mutation was found in 241 reads out of 241 and in 101 reads out of 101 for patient 1 and 2 respectively. Subsequently, the mutation was confirmed with Sanger sequencing. The mutation is classified as probably damaging by Polyphen, is not tolerated by SIFT (Table ) and was not reported before, nor as disease causing mutation nor as a SNP. The mutation was analyzed in the whole family and showed full co-segregation with the phenotype.
New variants observed in patients 1, 2 and 4
In patient 3, we couldn’t identify a mutation that is undeniably disease causing. After filtering all the variants, obtained with data analysis and filter settings as described in ‘Methods’, we obtained a list of 171 variants. Ninety-seven of these are located in the introns (further than 10 basepairs away from the exonic regions). Another set of 22 variants is located in the UTRs and are all reported as polymorphisms. Of the 171 variants, 48 of them are previously reported as polymorphism or were identified in several patients in the lab. Eventually, 7 unique variants remained. Five of them are synonymous SNPs with no effect on splicing based on splice prediction programs. One deletion (c.3636_3637del) located in the OTOF
gene, results in a frame shift (p.Phe1212fs). This mutation could not be confirmed with Sanger sequencing (see Table ). The second mutation results in an amino acid substitution in the CDH23
gene and was found in 50 out of 96 reads. This variant is not known as a polymorphism. However, it’s unclear that this missense mutation is causal as the prediction tools are not conclusive (see Table ). No additional mutation was found in the CDH23
gene. Further evaluation of the known polymorphisms revealed a low population frequency for 4 of them. Although all were reported as non-pathogenic, we took a closer look at these. One variant occurs in the TMPRSS3
gene with a relative frequency in the reads of 1, previously reported as rs35227181 and considered to be non-pathogenic. Also the prediction tools didn’t classify the variant as disease causing and therefore we considered this base change as a rare variant. In OTOF
a compound heterozygous variant was found: c.2317
T (p.Arg773Cys) and c.4936
T (p.Pro1646Ser). Both variants were detected with a relative frequency of 0.5 and are predicted to be possibly damaging (Table ). The two variants were previously described in the literature, where they were considered as non-pathogenic [11
]. Another variant was found in PCDH15
in 380 out of 381 reads. The variant is known as rs4935502 and non-pathogenic according to dbSNP. Nonetheless, it is regarded as probably damaging by PolyPhen-2 and has a Grantham score of 126 (see Table ). All the homopolymer repeats of these deafness genes (Table ) were reanalyzed with Sanger sequencing since the 454 technology does not allow reliable sequencing of repeats greater than 6. No mutation could be detected in this analysis. It remains possible that the mutation is located in another deafness gene, not yet incorporated in our set, or that patient 3 is a non-genetic deafness case.
Variants observed in patient 3
The variant analysis of patient 4 revealed a new homozygous missense mutation in the OTOF
gene (exon 26: c.3263
C (p.Leu1088Pro)). This mutation is in agreement with linkage analysis, previously performed, suggesting a disease causing mutation in the OTOF
gene. Prediction programs revealed the mutation as possibly damaging (PolyPhen) and not tolerated (SIFT) and therefore can most likely be considered as disease causing. The mutation was investigated in all available family members and showed full co-segregation with the deafness. The prediction results of the novel mutations, with the different software approaches, are listed in Table .
In patient 5, included as a control, we could confirm the heterozygous mutation in the TMC1
gene previously found with Sanger sequencing, with a relative variant frequency of 0.5. The mutation NM_138691.2:c.236
A was found in 13 reads out of 26. This substitution is located in the donor splice site of intron 7.