By Julia Karow
Next-generation sequencing has been hailed as a tool that will end the "diagnostic odyssey" of patients with rare inherited disorders, but it has been unclear how often the technology is actually successful in pinpointing causative mutations and yielding a molecular diagnosis.
A recent study by researchers at the Broad Institute and the Murdoch Childrens Research Institute in Australia suggests that for infants with mitochondrial disease, a comprehensive sequencing-based test involving more than 1,000 genes currently only leads to a firm diagnosis in about half of patients and identifies candidate genes in another 20 percent.
Although this success rate is an "enormous win" for patients compared to traditional diagnostic methods, "it's a cautionary tale that it is not going to be a perfect test," said Sarah Calvo, a senior computational biologist at the Broad Institute and one of the lead authors of the study. Even with technical improvements and a larger target region, clinicians should not expect a next-gen sequencing test like this to provide a diagnosis for every patient.
In their study, published online in Science Translational Medicine last month, the researchers focused on inherited mitochondrial oxidative phosphorylation disease, which affects about 1 in 5,000 live births and is genetically and phenotypically heterogeneous. So far, more than 100 causal genes are known for the disease, which can start in infants or adults, has a variety of symptoms, and differs in severity.
The scientists sequenced the "MitoExome," consisting of the 16.6-kilobase mitochondrial DNA as well as the exons of more than 1,000 nuclear genes that encode mitochondrial proteins, in 42 infants who had clinical and biochemical evidence of mitochondrial disease. Existing tests of candidate genes suggested by the phenotype had been unsuccessful in pinpointing a causal mutation in these patients.
The targets were enriched using Agilent Technologies' in-solution capture method and sequenced on the Illumina GAII platform. To interpret the data, the scientists focused on variants that were rare and predicted to be protein-modifying, as well as those that had previously been associated with disease. They prioritized those variants that were consistent with a recessive mode of inheritance.
Their analysis showed that only 10 patients, or about a quarter, had mutations in genes known to cause mitochondrial disease, and they were able to provide a firm genetic diagnosis for nine of them.
Another third of the patients had rare, protein-modifying, and recessive mutations in candidate genes not previously linked to the disease, and for two of those genes, the researchers could show through other experiments that they were indeed disease-causing. "I think many more on the list are going to be actually the causal genes, but we need to do a lot more experiments to get there," Calvo said.
Overall, the test discovered known or likely causal variants for only about half the patients, which the scientists attributed to a variety of reasons.
First of all, they might have missed causal variants in the targeted genes due to a lack of sensitivity of the technology, in particular for indels and large deletions, classes of variants that are known to be difficult to pick up with current technology. "I think that's a call for caution for people who are about to launch these large next-gen sequencing projects for diagnosis," Calvo said. "Lots of variants will be missed right now."
Another possibility — that causal mutations may have been located in genes that were not targeted — is unlikely in this particular case because the overwhelming majority of genes known to cause the disease are mitochondrial genes, which were comprehensively covered in the assay. However, causal variants could have been located in introns or regulatory regions of the targeted genes, which were not captured.
Further, the stringent filters used in the interpretation may have thrown out true causal variants. For example, some of them might not be rare, or they might change splicing but look as though they do not alter the protein.
Finally, the disease might not always be monogenic, as assumed in the analysis, but several genes acting in combination might be involved. That could especially be true for adult-onset forms of the disease. "Our paper focused on infants because genetically, they are going to be the most tractable to solve," said Calvo. "I think the diagnosis of adults is going to turn out to be much more difficult."
Even if technical limitations did not exist, "current NGS methods would still fail to find the genetic basis of disease if it resulted from epigenetic changes or the additive effects of many changes," agreed Lee-Jun Wong, director of the mitochondrial diagnostic laboratory at Baylor College of Medicine, who was not involved in the study.
The Broad-led study did show that the diagnostic success rate of MitoExome sequencing was higher in unselected infants, rather than the infant cohort the researchers studied where existing tests had failed. In a cohort of about 300 unselected infant patients, it yielded a diagnosis in almost half the cases and provided candidates in another 20 percent.
Calvo expects that the percentage of successful molecular diagnoses will improve in the next few years, but that the test, even if expanded, will never be able to yield an answer for all patients. "I think we are going to be able to get much better at finding [causal mutations], but we're not going to be fully successful," she said. "Even with some improvements, I think many patients are still going to remain mysteries, even if you sequence the whole genome."
She and her collaborators are currently applying MitoExome sequencing to a cohort of about 100 adult patients, some of whom might actually not have mitochondrial disease but might have been diagnosed with it because there was no better explanation. The disease is "so difficult to diagnose as adults that if you had a genetic test, it would really help patients," Calvo said.
However, the researchers are not planning to develop MitoExome sequencing into a CLIA-certified test that could be applied routinely in the clinic because whole-exome sequencing will soon be cheap enough to replace such a test. An exome sequencing test would be especially useful for "mystery cases" where a diagnosis is not possible in any other way. "It's going to have a huge utility once this test is rapid enough," she said. "Again, in many cases, it's not going to be interpretable. When it's interpretable, it will help you."
In the meantime, clinical labs are betting on targeted tests for mitochondrial disease. Baylor's mitochondrial diagnostic laboratory, for example, which has provided molecular diagnoses for the disease for almost 20 years, recently launched a next-gen sequencing-based test to analyze the mitochondrial genome, which detects point mutations, large deletions, mutation heteroplasmy, and deletion breakpoints. To analyze nuclear genes involved in the disease, next-gen sequencing-based testing of gene panels "appears to be practical at this time for clinically or biochemically defined syndromes," according to Wong.
"While panels are a good approach for many patients, NGS analysis of the [approximately] 1,300 genes associated with mitochondrial biogenesis, structure, and function may be needed for patients with unclear multisystem mitochondrial disorders," Wong said. "The current trend suggests that the most efficient approach to molecular diagnosis of difficult-to-diagnose disorders will ultimately be the whole exome, or entire genome sequencing. As the technology improves, the cost of NGS diagnostic approaches will be greatly reduced, with faster turnaround time and high diagnostic yield."
Have topics you'd like to see covered in Clinical Sequencing News? Contact the editor at jkarow [at] genomeweb [.] com.