|Home | About | Journals | Submit | Contact Us | Français|
Our understanding of hereditary hearing loss has greatly improved since the discovery of the first human deafness gene. These discoveries have only accelerated due to the great strides in DNA sequencing technology since the completion of the human genome project. Here, we review the immense impact that these developments have had in both deafness research and clinical arenas. We review commonly used genomic technologies as well as the application of these technologies to the genetic diagnosis of hereditary hearing loss and to the discovery of novel deafness genes.
Great progress has been made in our understanding of hereditary hearing loss since the discovery of the first human deafness gene in 1997 (Kelsell et al., 1997). To date, 57 non-syndromic deafness genes and more than 1,000 discrete deafness-causing mutations have been described (http://deafnessvariationdatabase.org). While this striking heterogeneity underscores the exquisite sensitivity of the human auditory system to perturbations, it also presents a major challenge to the development of a comprehensive genetic testing platform. This challenge, however, must be embraced. Deafness is the most common sensory deficit in humans. It occurs in 1 in 500 births and affects 278 million people worldwide (Morton & Nance, 2006; Smith, Bale & White, 2005). Establishing a robust platform for its evaluation will change the clinical management of deaf and hard-of-hearing patients, ultimately improving quality of life and saving healthcare dollars by making other investigations unnecessary.
As described in this review, great strides in DNA sequencing have been made since the completion of the human genome project. These developments are having an immense impact in both the research and clinical arenas. In the former, they have led to the discovery of new non-syndromic hearing loss (NSHL) and syndromic HL genes that are being more rapidly than ever translated to the clinical arena as part of comprehensive, first-line, inexpensive and highly sensitive clinical diagnostic platforms. It is clear that this advance in the evaluation of hereditary hearing loss positively impacts families affected by hearing loss (Robin et al., 2005; Withrow et al., 2009), but it will also be the foundation for novel gene or even mutation specific treatment options to slow progression or prevent hearing loss (Hildebrand et al., 2007).
First described in 1977, chain termination or Sanger sequencing has been the screening method of choice for genetic research and clinical genetic diagnostics for more than 30 years (Sanger, Nicklen & Coulson, 1977). The human genome project (HGP) was completed in 2001 primarily using automated capillary chain termination sequencers. However, since the completion of the HGP the output of these sequencers has been massively outpaced by the need for rapid and low-cost sequencing for clinical diagnostics and genetic research. Massively parallel sequencing (MPS), also called next-generation sequencing (NGS) and second-generation sequencing, has been developed over the past five years to meet this demand. Although there are now several MPS platforms available, they share several features including high output, high sequencing depth, and a relatively short sequence read-length.
The most advanced automated capillary chain terminator sequencers have a 24-hour output of 120,000 bp (basepairs) and cost $4,000 per megabase (million base pairs, Mb) sequenced (Metzker, 2010). Using a single sequencer it would take 73 years and cost $200,000 to sequence the 3.2 Gigabases (billion base pairs, Gb) in a single human genome. In contrast, the output of a single MPS machine is now greater than 30 Gb in 24 hours and costs less than $2 per Mb, meaning that a human genome can be sequenced in one day for a fraction of the cost. However, it is important to note that these calculations do not factor in bioinformatics analysis to map and annotate the large quantities of sequencing data generated. The increased output of MPS is associated with an increased error rate when compared with chain termination sequencing, although this rate varies by sequencer from 0.1% to 2% (Glenn, 2011). To overcome this increased error rate, template DNA strands are sequenced repeatedly to attain an adequate “coverage” or “read depth” to provide high confidence that a variant is truly present.
There are three commercial MPS platforms routinely used in the research arena − 454, Illumina and SOLiD – each with its own strengths and weaknesses (Figure 1). For technical details interested readers are referred elsewhere (Mardis, 2008; Metzker, 2010). The first MPS platform introduced was the 454 sequencer (Roche Life Sciences; (Margulies et al., 2005)). Available since 2005, its output is relatively low (1 Mb per run), its run time is short (less than 1 day) and its read lengths are long (generally >300 base pairs), three features that make data analysis relatively straightforward. Pyrosequencing is used, meaning that as DNA polymerase incorporates a nucleotide into the growing DNA strand and ATP hydrolysis occurs, pyrophosphate release is recorded. A limitation of this method is that base calling is difficult for sequence regions containing long stretches of the same nucleotide (homopolymer DNA segments) meaning these segments are prone to insertion-deletion errors. By contrast, substitution errors, where an incorrect nucleotide change is detected, are rarely encountered in 454 sequence reads.
The Illumina platform (Illumina Inc.) is the most widely used MPS platform and relies on cyclic reversible termination (CRT) technology (Ju et al., 2006). In CRT, chain-terminating fluorescent nucleotides are incorporated into the growing strand, imaged by the sequencer, and then the terminator along with the fluorescence moiety are cleaved off the nucleotide allowing incorporation of the next nucleotide. The newest version of Illumina sequencers offer up to 150 bp sequencing reads and the highest sequence output per day of the currently available sequencers. Simple base miscalls are the most common error type.
The third option, the SOLiD platform (Applied Biosystems Inc.), uses unique sequence-by-ligation (SBL) technology in which the DNA template is sequenced via ligation of two-base fluorescent probes (Valouev et al., 2008). A fluorescently labelled probe hybridizes to the DNA template to be sequenced, is joined to the growing strand by DNA ligase, and the fluorescence is imaged to allow the sequence to be determined. Like CRT, the most common error in SBL is base miscalls.
As a complement to the incredible sequencing output of MPS, genomic enrichment technologies to isolate and amplify selected regions of the genome have been developed. Together these platforms offer the ability to attain sequence data at an unprecedented rate, and shift the challenge from sequence generation to data analysis and interpretation.
In the near future, whole genome sequencing will be an integral part of a patient’s medical record. However, this type of personalized medicine is currently not feasible for several reasons: (i) the cost of sequencing the entire 3.2 Gb human genome at a sufficient level to accurately call variants is still too high; (ii) there is no agreed upon format for storage of genomic data; and (iii) the interpretation of variants in the human genome is still difficult. Targeted genomic enrichment offers an attractive intermediate solution. This selective type of sequence interrogation, which is also called targeted sequence capture, refers to any of several methods whereby a portion of the genome is targeted for sequencing as opposed to the entire genome.
There are two appealing reasons to focus on a specific part of a patient’s genome (for example, their exome or even a smaller set of targeted genes of interest) for genetic diagnosis. Firstly, exon-targeting results are easier to interpret as ~85% of pathogenic mutations are predicted to lie in coding and/or splice site regions of genes. Secondly, costs are dramatically decreased both in data generation and analysis when the amount of sequence is low, decreasing the bioinformatics burden. Indeed, at the current time, the most successful MPS clinical platforms rely on targeted genomic enrichment (Ng et al., 2009b; Shearer et al., 2010; Walsh et al., 2010a).
Targeted genomic enrichment allows polymerase chain reaction (PCR) amplification on a massively parallel scale (Figure 3c). Instead of amplifying a single genomic region, hundreds, thousands, or millions of genomic regions are specifically isolated from a pool of sheared DNA and amplified to create a DNA library, which is then sequenced using one of the massively parallel sequencers described above. The first method for targeted enrichment was based on DNA microarray technology and is often called solid-phase targeted enrichment (SPTE) (Albert et al., 2007). In this method, long oligonucleotides complementary to the regions of interest are synthesized on a glass slide and hybridized against fragmented genomic patient DNA. Repeated washings remove non-hybridized DNA leaving the regions of interest, which are then eluted, amplified and sequenced. SPTE requires customized hybridization equipment and a long protocol, two constraints that led to the development of solution-based targeted enrichment (SBTE) (Gnirke et al., 2009). In SBTE, targeted capture is completed using RNA- or DNA-based complementary probes, with the hybridization reaction occurring in solution in a standard thermal cycler.
Exome sequencing – targeting every exon of every gene for enrichment and MPS – is an intermediary step between small targeted capture platforms and whole genome sequencing. Whole exome sequencing is more costly than targeting a small subset of genes but nevertheless offers the options of restricted data analysis – that is, examining only known disease-causing genes for pathogenic variants. Two current constraints, however, are: (i) exome sequencing kits are generally not made-to-order and so your regions of interest may be poorly represented; and, (ii) the depth of coverage can be marginal for clinical diagnostics. For these reasons, MPS platforms for clinical genetics typically focus on only the genes known to cause a given disease. NSHL, for example, is ideally suited for this technology. The OtoSCOPE® SBTE platform for NSHL (described below) represents 1% of the exome or 0.02% of the genome.
Prior to and through the HGP, the emphasis was on faster and better methods for sequence generation. The immense output of MPS platforms has now shifted the bottleneck to data analysis. Common to all sequencing platforms is the need to analyse the millions or billions of short sequencing reads generated (Figure 2). The interpretation of these reads to generate a list of variants from a patient DNA sample for further analysis generally follows the following steps: (i) read mapping and alignment to the reference human genome; (ii) filtering for quality, variant identification; and (iii) variant annotation. Each step is equally important to the overall success of the analysis.
Sequence reads are first aligned to the reference genome generated by the HGP using algorithms such as Bowtie, BWA and BFAST, amongst others (Heng & Homer, 2010). In simple terms, these software tools attempt to simultaneously align each of the millions of individual short sequencing reads to the human genome. Once the sequencing reads are aligned, any variations between the aligned sequencing reads and the human genome can be identified. Common variant calling tools include Samtools and GATK (Homer, Merriman & Nelson, 2009; McKenna et al., 2010). The final step, variant annotation, provides information on each variant identified. For example, this step would determine if the variant is known or unique by querying other large sequencing projects (for example, the 1000 Genomes Project, http://1000genomes.org), and, if it lies within a gene, whether the change results in a protein alteration. This step helps to prioritize variants to identify those variants most likely to be pathogenic.
When taken together, genomic technologies allow the unprecedented ability to interrogate millions or billions of base pairs of the genome simultaneously, revolutionizing genetic testing and gene discovery.
The ideal genetic test is highly sensitive, specific and accurate; is comprehensive; and can be run at low cost. To date, no single genetic test for deafness has achieved these goals. As described above, using new genomic technologies a large number of genes can be screened simultaneously making truly comprehensive genetic testing for deafness possible. The efforts to reach this goal are described below, highlighting the strengths and weaknesses of the available technologies (summarized in Table 1 and Figure 3).
Single nucleotide extension microarrays provide a fast and inexpensive method to resolve mutations (Figure 3a). Also called primer extension microarrays, these platforms detect a mutation by hybridizing specifically designed primers to patient DNA followed by single nucleotide extension at the site of the mutation. As such, this technology can only target specific known mutations and thus is not comparable to direct sequencing of entire genes. The Hereditary Hearing Loss Arrayed Primer Extension (HHL APEX) was the first diagnostic test to employ this method for hearing loss and is based on previous APEX designs (Kurg et al., 2000). The HHL APEX allows for the evaluation of 198 known pathogenic mutations within eight hearing loss genes. This method is cost and time effective: during an 8 hour day, one imager can process 60–70 patient assays requiring approximately 1.5 hours of hands-on work (Gardner et al., 2006). Initial testing of the HHL APEX showed 100% specificity in control subjects (Gardner et al., 2006). However, in unknown subjects causative mutations were only identified in 12/144 individuals (8.3%) and potential pathogenic mutations in 4/144 individuals (2.8%) (Rodriguez-Paris et al., 2010). Therefore, this test provides little additional diagnostic value for individuals negative for GJB2 and/or GJB6 mutations (Rodriguez-Paris et al., 2010). The major shortcoming associated with use of the HHL APEX as a clinical diagnostic test stems from its inability to cover all of the more than 1,000 known pathogenic deafness mutations. Expanding microarrays to cover all known mutations resulting in NSHL is impractical because it requires constant modification and updating due to the continual discovery of novel mutations.
Like single nucleotide extension microarrays, resequencing microarrays are very inexpensive and time efficient; however the detection method is more complicated (Figure 3b). The major difference is that a set of four probes are used simultaneously to sequence one base. Thus, there are four versions of each probe to test whether an A, G, C or T is found at a specific nucleotide position. In theory this means that any variant in the interrogated genes should be detectable.
This technology was used in the design of OtoChip™, which was developed at Harvard University. OtoChip™ includes 13 deafness genes totalling 27,000 bp and offers results in as few as 3–4 days, with one technician being able to complete 100 assays per month (Waldmuller et al., 2008). In the initial evaluation of this platform, seventy-four persons were tested for an overall mutation call rate of 99.6% and an accuracy of 99.88%. Of the non-control samples, a potentially causative mutation was identified in 27 of 61 (44%) (Prachi, Cox & Rehm, 2011).
Resequencing microarrays are unable to reliably detect insertions and deletions; however their greatest limitation is the number of nucleotides that can be investigated, which is restricted by the physical size of the microarray. At present, 19 of 57 known NSHL genes are tested on an OtoChip™ (http://pcpgm.partners.org/lmm/tests/hearing-loss/OtoChip). For persons with causative mutations in other NSHL genes, this platform is not helpful, thus limiting its overall usefulness.
OtoSCOPE® was developed at the University of Iowa to provide direct sequencing of all 57 known deafness genes simultaneously at a relatively low cost (http://www.morl-otoscope.org; (Shearer et al., 2010)). This platform makes use of SBTE to ‘capture’ all of the exons of the genes implicated in NSHL (Figure 3c). Coupled with MPS, it becomes possible to provide a relatively inexpensive yet comprehensive genetic test for deafness.
The first version of OtoSCOPE® targeted 54 deafness genes (421,741 bp of the human genome), including the Usher syndrome genes because in young children they mimic NSHL (Shearer et al., 2010). In a proof-of-principle study, 97.7% of targeted coding nucleotides were sequenced with a mean per-base coverage of 903 sequencing reads. Causative mutations were identified in both positive controls but not in the negative control sample, and in five of six persons with idiopathic hearing loss, causative mutations were identified (Shearer et al., 2010). While ‘failure’ to identify a cause for hearing loss in the undiagnosed person could represent a failure of the OtoSCOPE® platform, based on the number of unresolved loci it is more likely that this person segregates a novel genetic cause of NSHL not yet represented on OtoSCOPE®.
To increase throughput and make SBTE and MPS more economical, small oligonucleotide tags called ‘barcodes’ can be added to patient DNA fragments during the library preparation to allow fragments to be traced back to their unique source after multiple DNA samples are pooled and sequenced together (Cummings et al., 2010). Using SBTE, MPS and barcoding, Bell and colleagues have designed a platform to screen 437 genes implicated in severe, recessive diseases of childhood for $378 per sample (Bell et al., 2011). With the incorporation of molecular barcoding, OtoSCOPE® will become a routine clinical test.
Monogenic and complex genetic diseases have been traditionally studied using linkage mapping or association studies followed by Sanger sequencing-based screening to identify disease-relevant genes. These approaches suffer from low throughput and lack of functional insight. An example is the study of the molecular genetics of auditory impairment. Since 1997, 122 genetic loci have been associated with NSHL, and 39 recessive (DFNB), 23 dominant (DFNA), and 2 X-linked (DFN) genes have been cloned (http://hereditaryhearingloss.org). Thus, despite intense efforts by dozens of laboratories over nearly two decades, less than half of the genes at these loci have been identified. Moreover, a wide spectrum of polymorphisms in modifier genes is expected to influence disease onset and severity. For example, recent studies show that progressive and age-related hearing loss depend on genetic predisposition, where disease onset and progression are determined by complex interactions between genetic and environmental factors, further underscoring our lack of understanding of the full spectrum of genetic variants leading to deafness (Friedman et al., 2009; Huyghe et al., 2008; Van Eyken et al., 2007; Van Eyken, Van Camp & Van Laer, 2007; Van Laer et al., 2008).
The genomic technologies described above are being translated into high-throughout pipelines to accelerate the pace of gene discovery, functional annotation, and analysis of disease mechanisms. The first example of targeted genomic capture and MPS to identify a Mendelian disease gene was the application of exome sequencing to reveal a mutation in the gene DHODH as the cause of Miller syndrome (Ng et al., 2009a). Exome sequencing has now been validated as an efficient strategy for gene discovery because, as evidenced by positional cloning experiments, most Mendelian disorders are due to protein-coding variants and a significant proportion of rare nonsynonymous variants in the human genome are predicted to be deleterious (Chen, Ferec & Cooper, 2010; Ng et al., 2009a; Ng et al., 2009b; Teer & Mullikin, 2010).
To date, eight syndromic or nonsyndromic deafness genes have been identified using targeted genomic enrichment and MPS (Table 2). The first gene was TPRN, mutations in which cause nonsyndromic hearing loss at the DFNB79 locus (Rehman et al., 2010). This locus was mapped in a consanguineous Pakistani family to a 2.9Mb region containing 108 protein-coding genes at chromosome position 9q34.3. Using a custom set of Roche NimbleGen probes designed for enrichment and 454 sequencing of this region, after mapping and filtering, 2,419 total variants were identified of which eight homozygous variants were validated by Sanger sequencing. Only one variant – a nonsense mutation in the TPRN gene – remained after screening of ethnically-matched controls and segregation analysis. Sequencing this gene in three additional DFNB79 Pakistani families revealed pathogenic indel mutations, providing further confirmation of TPRN’s role in NSHL (Rehman et al., 2010). This gene encodes the taperin protein that has similarity to phostensin, an actin filament pointed-end-capping protein known to modulate the actin cytoskeleton. Immunohistochemical analysis revealed expression of taperin at the base of the stereocilia of the mouse inner and outer hair cells in a region known as the taper (Rehman et al., 2010). In this region the stereocilia diameter narrows and weakness of these tapers when TPRN is mutated may underlie DFNB79 deafness.
This landmark report was shortly followed by two studies that used target enrichment and MPS of the exome to identify novel nonsyndromic deafness genes. In the first study, homozygosity mapping of a consanguineous Palestinian family to the DFNB82 locus at 1p13.3 was combined with exome sequencing (Walsh et al., 2010b). The Agilent SureSelect 38Mb All Exon Kit was used to capture the exome of the proband and sequencing of 93% of protein-coding sequence in the DFNB82 interval at 10-fold or higher coverage was achieved using the Illumina Genome Analyzer IIx (GAIIx). Bioinformatics analysis revealed 80 high quality homozygous coding variants in the DFNB82 interval, of which seven were novel and only two altered amino acid sequence. One variant was shown to be a polymorphism by screening controls, and the other – a nonsense variant in the GPSM2 gene – was absent in controls and segregated with the hearing loss in the family. GPSM2 encodes a G-protein signalling modulator that functions in guanine nucleotide exchange and is expressed in the hair and supporting cells of the developing mouse cochlea, utricle, saccule and cristae, and the pillar cells of the mature cochlea (Walsh et al., 2010b). Based on regulation of apical-base spindle orientation and asymmetric cell division by its Drosophila homolog Pins, GPSM2 may be essential for cell polarity and the asymmetric organization of the apical-basal cellular axis of the organ of Corti.
In the second study, exome sequencing was used to confirm a deafness-causing mutation in the originally mapped DFNA4 family (Chen et al., 1995). Mutations in MYH14 had been excluded and a potential pathogenic variant was identified in the candidate gene, CEACAM16 (Zheng et al., 2011). The variant segregated with the hearing loss in the family but to exclude other possible variants, whole exome sequencing and MPS were performed using the Agilent SureSelect 50 Mb All Exon Kit and SOLiDv4 sequencing system. 98.2% of the DFNA4 interval was targeted and 70.2% of the protein-coding regions were covered by at least 10 high quality sequence reads. In addition to the CEACAM16 variant, 30 other variants were identified, all of which were polymorphisms, false positives, or did not segregate with the hearing loss in the family. CEACAM16 encodes an adhesion molecule expressed at the tips of the outer hair cell stereocilia and may be involved in their attachment to the tectorial membrane through interaction with α-tectorin (Zheng et al., 2011).
Most recently for NSHL genes, MPS and a selective targeted approach were used to identify SMPX as the causative gene at the DFNX4 locus in a large Dutch family with progressive, X-linked nonsyndromic hearing loss (Schraders et al., 2011). After excluding three candidate genes by Sanger sequencing, targeted enrichment of the entire X chromosome was completed in one affected family member using the Agilent SureSelect Human X Chromosome Kit followed by single-read sequencing on the Illumina GAIIX sequencer. Based on at least 10-fold coverage of 95.1% of targeted bases and bioinformatics analysis, only two novel gene variants were identified. One variant, a nonsense mutation in the SMPX gene, was located in the critical region that segregated with hearing loss in the DFNX4 family and was absent from controls. Screening probands from 26 additional X-linked families with hearing loss identified another family with a single base pair deletion in SMPX, confirming mutation of this gene as causative of DFNX4 deafness. SMPX encodes a proline-rich protein likely to be part of an actin-associated complex, regulated by IGF-1 and involved in cellular organization. Since the SMPX protein is expressed during development in various cochlear cell types including sensory hair cells, root cells, pillar cells and interdental cells, its mutation may disrupt the development and/or maintenance of these cell types and their associated pathways over time (Huebner et al., 2011; Schraders et al., 2011).
The remaining four genes thus far implicated in deafness by high-throughput sequencing strategies were identified in families segregating rare hearing loss syndromes (Table 2). Two families segregate ovarian dysgenesis and sensorineural hearing loss (with or without ataxia) of Perrault syndrome due to mutations in different genes. In one of these two families, whole exome sequencing was performed on an outbred American family too small for linkage mapping (Pierce et al., 2010). The exome of one affected individual from the family was enriched using an Agilent SureSelect All Exon Kit and sequenced on an Illumina GAIIx yielding 93.1% coverage of the entire exome by at least 10 high quality reads. Based on the inheritance pattern the list of candidates was narrowed to 207 rare nonsense, missense, frameshift or splice variants. The only gene containing two variants (one nonsense and one missense) was HSD17B4, and Sanger sequencing confirmed that the affected family members were compound heterozygotes for both mutations. This gene encodes 17β-hydroxysteroid dehydrogenase type 4 (HSD17B4), a peroxisomal enzyme involved in fatty acid β-oxidation and steroid metabolism that has also been associated with D-bifunctional protein (DBP) deficiency. By immunoblotting of patient lymphoblasts, the authors confirmed deficiency in HSD17B4 protein that likely explains the symptoms of Perrault syndrome observed in these patients, although the mechanisms involved have not been elucidated (Pierce et al., 2010).
In the other Perrault family a different strategy was taken. In this case the American family was larger and genome-wide linkage analysis yielded a significant locus at 5q31 (Pierce et al., 2011). Sanger sequencing of all 58 genes in this region revealed only one gene – HARS2 – with two predicted pathogenic rare variants that co-segregated with the Perrault syndrome phenotype and were absent in matched controls. Similar to the CEACAM16-DFNA4 study, an exclusionary approach was used to rule out all other variants in the region. The entire linked region was targeted for enrichment using Agilent SureSelect technology and sequenced on an Illumina GAIIX to at least 20-fold coverage for 97% of targeted bases. As expected HARS was again the only gene identified with two variants of functional effect. Both were missense variants that affected conserved residues, however, one variant also introduced an alternative splice site leading to an in-frame deletion. The HARS2 gene encodes a histidyl tRNA synthetase that catalyzes the linkage of histidine to its cognate tRNA and was shown to function in mitochondria (Pierce et al., 2011). Studies in C. elegans mutant for the HARS2 homologue hars-1 confirmed that downregulation of this aminoacyl tRNA synthetase reduces fertility, linking its mutation with the symptoms of Perrault syndrome.
The third gene to be linked to syndromic deafness by targeted genomic capture and MPS was MASP1. In two small families with distinctive syndromic facial, umbilical, coccygeal and auditory features of, homozygosity mapping identified an autozygous region at chromosome 3q27 (Sirmaci et al., 2010). Whole exome sequencing of the proband from one family using the Agilent SureSelect All Exon Kit and Illumina GAIIx sequencing achieved 50X median coverage of the autozygous region. Of four rare variants identified in the region, only one, a missense change in the MASP1 gene, was absent in controls and segregated with the syndrome in the family. Sanger sequencing of the MASP1 gene in the second family revealed a pathogenic nonsense mutation, confirming it as the causative gene. MASP1 encodes a mannan-binding lectin serine protease known to activate the complement pathway by binding lectin, and its catalytic activity is predicted to be affected by both mutations. The syndromic features of these families could be explained by the role of one of the MASP isoforms, MASP-3, in the IGF-signalling pathway (Sirmaci et al., 2010). However, the mechanisms involved, including in the auditory dysfunction observed, remain to be resolved.
In the most recent study, the genetic cause of one form of syndromic deafness, sensory neuropathy and dementia, was resolved using exome sequencing (Klein et al., 2011). Four kindreds from the United States, Japan and Australia segregating sensorineural hearing loss, early-onset dementia and loss of sensation were phenotyped and using the largest kindred, linkage analysis identified a significant linkage peak at 19p13.2. To sequence variants in this interval at a high depth of coverage, exome sequencing was performed using two independent methodologies - Agilent SureSelect 38Mb All Exon Kit capture and Illumina GAIIX sequencing, and NimbleGen 2.1M Human Exome capture and Roche454 sequencing. Filtering and bioinformatics analysis revealed a novel, non-synonymous heterozygous missense mutation in the DNMT1 gene that segregated with disease status in the family. Sanger sequencing of all 41 exons of the DNMT1 gene identified heterozygous missense mutations in the remaining three kindreds that segregated with disease and were absent from controls. These mutations were shown to reduce the stability of the DNMT1 protein leading to degradation, reduced methyltransferase activity, impaired heterochromatin binding after S phase of the cell cycle, global hypomethylation and local hypermethylation (Klein et al., 2011). These defects manifest as neurodegenerative changes because DNMT1 is highly expressed in postmitotic neurons and the central nervous system. Another methyltransferase, LRTOMT, has been implicated in nonsyndromic deafness DFNB63, suggesting that regulation of methylation is critical in the auditory system (Ahmed et al., 2011; Du et al., 2008; Klein et al., 2011).
It is worth noting several of the important characteristics of these gene discovery studies (Table 2). First, it is not surprising that five of these six studies focused on recessively inherited deafness, since homozygous or compound heterozygous mutations are ‘easier’ to detect using MPS because this requirement adds another filtering criterion. Second, both simple nonsyndromic and complex syndromic forms of hearing loss can be resolved efficiently using this technology. Third, genes with both restricted (e.g. CEACAM16) and broad (e.g. HARS2) expression patterns are candidates in these types of studies. Fourth, small families with distinct and interesting phenotypes that are too small to map can be studied using these approaches. Fifth, both whole exome and smaller targeted capture platforms are suitable technologies. And finally, all three major massively parallel sequencers (454, Illumina, SOLiD) have been utilized for deafness gene discovery.
The impact of genomic technologies on the practice of Otolaryngology will be far reaching. It is already clear that comprehensive sequencing platforms are critical for genetic hearing loss based on the prevalence of the condition and its high genetic heterogeneity. Genomic technologies facilitate the discovery of new deafness genes at an unprecedented rate, thereby translating to improved patient care through more rapid diagnosis while also advancing our understanding of the molecular physiology of hearing and deafness. At the same time, the direct benefit of these genomic technologies to genetic testing for deafness will be the availability of comprehensive genetic diagnosis. Based on these improvements it is anticipated that within the next ten years genetic testing will become a standard of care in the clinic after history, physical exam, and audiometric profiling for deaf and hard of hearing patients.
NIDCD RO1s DC003544 and DC002842 to RJHS, NIDCD 1F30DC011674-01A1 to AES, and NHMRC Overseas Biomedical Postdoctoral Training Fellowship to MSH. We would like to thank the Medical Scientist Training Program at the University of Iowa for support to AES and CMS.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.