By Julia Karow
This story was originally published March 17.
A new, comprehensive sequencing test to pinpoint the cause of genetic hearing loss promises to provide both precise diagnoses and clues to disease progression and therapeutic interventions.
The test, called OtoSCOPE for Otologic Sequence Capture of Pathogenic Exons, was developed by researchers in the Molecular Otolaryngology and Renal Research Labs, or MORL, at the University of Iowa. The non-profit lab performs both CLIA-certified genetic testing as well as research on deafness and renal disease. They plan to start offering the test clinically toward the end of summer.
MORL currently performs single-gene genetic testing on more than 1,200 deafness patients per year, receiving samples from both within and outside the US.
"We really think this is going to revolutionize patient care for individuals affected by hearing loss," said Eliot Shearer, an MD/PhD candidate at the University of Iowa who helped develop the sequencing-based test.
The test uses Agilent's SureSelect in-solution capture and Illumina's sequencing platform to screen nearly 60 known non-syndromic hearing loss genes for pathogenic mutations, which are confirmed by Sanger sequencing in the CLIA-certified portion of the lab before they are reported to doctors and patients.
At the end of last year, Shearer and his colleagues published a proof-of-concept study of OtoSCOPE in PNAS, showing that it is reproducible, and that its sensitivity and specificity are good enough for diagnostic use.
They are currently focused on running the test more efficiently, so they can offer it for around $1,000 to $2,000 this summer. The goal is to reduce the price further, to around $500, eventually. "We already have a lot of patients contacting us about this test," Shearer said.
Like other genetic tests currently offered by MORL, the researchers expect the new test to be covered by many insurance plans.
About 1 in 500 children suffer from severe profound hearing loss, and many children born deaf already undergo genetic testing, typically for mutations in a single gene, GJB2, which accounts for about half the cases of autosomal recessive non-syndromic hearing loss. At MORL, a test that sequences exon 2 of the GJB2 gene and screens for two deletions involving the GJB6 gene costs $324.
If the GJB2 test comes back negative, "most often, the testing is left there, although a few centers around the country go on to screen more genes, maybe one or two more," said Bob Eppsteiner, a resident physician in the department of otolaryngology at the University of Iowa. That can get expensive quickly, he said — up to $800 per gene.
If those tests are also negative, "then patients are kind of left in this void, where we have those 59 genes that it could be, but there is no good way to go about sequencing those," Shearer said. Due to a lack of studies, it is also unknown to what percentage of hearing loss each of these genes accounts for.
Beyond getting a precise diagnosis, there are other benefits for patients in learning what mutation causes their hearing loss. For example, some children whose only initial symptom is deafness may actually suffer from a broader syndrome that a genetic test could reveal, and other symptoms that develop later could be predicted or even delayed. In the case of Usher syndrome, for example, patients can start wearing protective eye glasses to delay the onset of blindness.
In the future, scientists may also be able to correlate specific mutations with the extent and severity of hearing loss, Shearer said, as well as whether patients can benefit from hearing aids or cochlear implants. He said that studies have already shown that patients with mutations in GJB2 benefit from cochlear implants, but similar correlations are not yet know for other types of mutations.
And longer term — a decade or more from now — gene therapy for deafness may become available, he said, the success of which will also depend on a patient's specific mutations.
[ pagebreak ]
Technology Assessment
When the Iowa researchers set out to develop their test about two years ago, they compared two technical approaches, one using NimbleGen capture arrays and 454 sequencing, the other Agilent's SureSelect in-solution capture and Illumina sequencing.
Both capture methods performed well, according to Michael Hildebrand, a postdoc at the University of Iowa who co-developed the test, "but there are other considerations beyond whether or not they work for clinical testing" — for example, how easily and efficiently the test can be run.
Performing the capture in solution is easier than on an array, he said, because it can be done with the help of an "ordinary PCR machine" rather than specialized hybridization equipment, so the scientists chose the SureSelect/Illumina method to run their test.
Also, newly discovered hearing loss genes can easily be included in the assay by adding new capture baits, which would be more difficult to do with an array. At least 140 genetic loci have been mapped that are associated with hearing loss, according to Hildebrand, but underlying genes have only been identified for about a third of them. Researchers at MORL are actually screening patients for mutations in almost 200 candidate genes for deafness, using the SureSelect/Illumina approach as well. If they discover new ones, they will transfer them to the diagnostic test.
In addition, it is easier to scale up the number of samples with an in-solution method. "We are hoping to do over 1,000 samples per year when this test is up and running as a clinical test," Hildebrand said. "You really need an efficient, reliable method, and definitely, SureSelect offers this."
Regarding the sequencing platform, he said that the long reads of the 454 platform are not necessary for this test. Since the average length of a human exon is only 150 base pairs, it can be covered with paired-end 2x100-base reads on the Illumina platform, which the lab currently uses. "For the small regions that we are targeting, the short reads are more than adequate," he said.
Shearer acknowledged that next-gen sequencing platforms have a higher raw error rate than Sanger sequencing, but "as long as you know that from the start, you can deal with that in your analysis," he said. "The key is filtering." For their paper, the scientists required a variant to be covered by 40 high-quality reads to be considered "good quality," and it had to be seen in at least 30 percent of those reads. They validated more than 600 variants by Sanger sequencing, and found a "very low false-positive and false-negative rate."
Also, some 5 percent of the targets are currently not covered by enough reads to call variants, so those regions still have to be sequenced by the Sanger method. "That will be done in the foreseeable future unless the methods improve," Hildebrand said. To improve evenness of coverage, and reduce the amount of Sanger sequencing required, the scientists are also currently "rebalancing" the capture probes.
The test is capable of detecting all kinds of mutations, including single-nucleotide variants, small deletions, and large deletions, the latter being confirmed by CGH arrays. "If there is a mutation, we have a pretty good shot at finding it," Shearer said.
All variants are run through an annotation pipeline, developed by collaborators at the University of Iowa, and compared to data from the 1000 Genomes Project and the HapMap Project to filter out common variants. After that, the researchers are usually left with 15 to 20 variants in the target genes per individual. The scientists then run these through several different pathogenicity programs, and "when all of these are concordant, then you have the highest positive predictive value for pathogenicity," Shearer said, leaving "just a handful" of variants. They then check if any of these are known deafness mutations — more than 1,000 have been reported so far — and whether they segregate in the patient's family. If they are novel, they also screen controls.
[ pagebreak ]
Having data from the 1000 Genomes Project and other human genome sequencing projects has been extremely valuable for filtering variants, he said. "It's one of the most powerful tools we have." However, he cautioned that many deafness-causing mutations are contained in dbSNP, so they have to filter based on allele frequencies rather than presence in the database.
Every potentially pathogenic variant is currently verified by Sanger sequencing. "The key right now is that next-generation sequencing is not CLIA-approved," Shearer said, making it necessary to report only Sanger-based results to patients, and this will likely not change "until there is a change in the certification."
Other, microarray-based diagnostic tests that screen a panel of hearing loss genes already exist, but they are less comprehensive and only test for known mutations. One test, for example, that uses primer extension arrays, is "pretty cheap and works pretty well; we use it for other things in our lab," Shearer said. "But you don't have the ability to find new mutations — you can only target the mutations that you put on the array."
Since their publication, the researchers have been making the test more efficient, as well as improving the evenness of coverage. "We are not a huge lab, so anything we can do to increase the efficiency is really helpful," said Shearer, adding that the CLIA-certified side of the lab has four research technicians.
As part of a collaboration with Agilent, for example, the lab will be getting a Bravo liquid handling robot in the near future, allowing them to do target captures in up to 96 samples at a time.
In addition, they have experimented with multiplexing on the Illumina HiSeq, pooling several samples per sequencing lane, which will help decrease the cost of the test. Prior to offering it clinically this summer, they plan to run several large multiplexing studies to determine how many samples they can run per lane, Hildebrand said.
While the sequence capture is done at MORL, all Illumina sequencing is currently performed at Baylor College of Medicine, but Shearer pointed out that the samples they currently run are only for research purposes.
For the clinical test, the Illumina sequencing will most likely be outsourced to another institution initially, while the validation by Sanger sequencing will be done at MORL's own CLIA lab in Iowa. Long term, MORL plans to get its own HiSeq, giving it better control over the timing of runs.
Shearer said that he and his colleagues are keeping an eye on new sequencing platforms, like the Illumina MiSeq and the Life Tech Ion Torrent, whose vendors have touted their promise for clinical use. But "we are really happy with the technology we are using right now," he said.
The Iowa team is not the only developing diagnostic gene panels based on next-gen sequencing — Emory Genetics Laboratory, GeneDx, Ambry Genetics, Arup Laboratories, the Mayo Clinic, Sistemas Genómicos, and others are doing so as well, for conditions such as hypertrophic cardiomyopathy, X-linked mental retardation, congenital muscular dystrophy, and hereditary colorectal cancer (IS 7/13/2011, IS 5/25/2010, and IS 2/1/2011). Also, another part of MORL is working on a test for kidney disease, using a similar platform, Shearer said.
"We like to think that we are on the forefront of this type of research, and moving this towards the clinic," he said. "But we think that in a couple of years, it's going to be really broadly adopted."
Have topics you'd like to see covered in Clinical Sequencing News? E-mail the editor at jkarow [at] genomeweb [.] com.