NEW YORK – Researchers at Western University in Canada and their collaborators have discovered genome-wide DNA methylation signatures for several neurodevelopmental Mendelian disorders, which they plan to roll out clinically to complement existing molecular tests.
The so-called epi-signatures might be particularly useful in testing for several syndromes at once and to help classify variants of unknown significance. In addition to constitutional disorders, they might find applications in cancer diagnostics.
For their study, published earlier this month in the American Journal of Human Genetics, the team, led by Bekim Sadikovic, an associate professor of pathology and laboratory medicine at Western University and director of the clinical molecular genetics laboratory at the London Health Sciences Centre, analyzed DNA methylation across the genome in patients with 14 Mendelian disorders and found specific, though overlapping, epi-signatures for seven of them. They also developed a machine-learning algorithm, allowing them to screen patients for these syndromes with high sensitivity and specificity.
Sadikovic said he has been involved in epigenetics research for a long time, starting as a graduate student, and had always been interested in its potential for clinical diagnostics. Up until recently, he said, clinical DNA methylation testing had been restricted to a limited number of disorders — including imprinting disorders, such as Prader-Willi syndrome and Angelman syndrome — and usually involved single-gene tests.
However, the advent of DNA methylation array technology several years ago opened up new possibilities. "For the first time, it allowed us to systematically, using very precise technology at relatively high throughput and low cost, generate genome-wide DNA methylation data," he said. "What I wanted to know is, is there [clinical] utility beyond these targeted assays?"
Just like array-based technology replaced fluorescent in situ hybridization (FISH) in the past to test for copy number alterations in the clinic, or like next-gen sequencing panel and exome tests replaced single-gene tests, genome-wide epigenetic testing might someday supplant single-gene DNA methylation tests, he said.
In addition, genome-wide DNA methylation testing might lead to the discovery of new imprinting loci that could serve as molecular disease markers. Again, this could be similar to CNV arrays, which, after they became the clinical standard for copy number variant detection, helped discover dozens of new disease-causing deletions and duplications, he explained. "The utility of these arrays came with putting them into diagnostic labs."
His group's first goal, Sadikovic said, was to validate the DNA methylation array technology for well-characterized epigenetic disorders, and over the past few years, they published several studies showing that specific epi-signatures are associated with diseases such as autosomal dominant cerebellar ataxia, fragile X syndrome, Floating-Harbor syndrome, alpha thalassemia/mental retardation X-linked (ATRX) syndrome, and Kabuki syndrome.
In some sense, an epi-signature could be viewed as a sort of phenotype that results from the underlying genetic defect, Sadikovic said, and he has started referring to it as an "epi-phenotype."
Last summer, his group also published a clinical validation study of a genome-wide DNA methylation assay to test for several imprinting disorders, which might replace separate locus-specific molecular tests.
For their latest study, the researchers asked whether they could find epigenetic signatures for a wider range of syndromes, in particular those caused by defects in genes that are involved in the so-called epigenetic machinery, for example in DNA methylation, histone modification, or chromatin remodeling. Their hypothesis was that those gene defects would result in disease-specific DNA methylation signatures.
In order to gather large enough patient cohorts for the project, Sadikovic's team at Western University partnered with researchers at McMaster University, the University of Ottawa, the University of Montreal, the Greenwood Genetics Center in South Carolina, and the Care4Rare Canada consortium.
In total, they analyzed peripheral blood DNA samples from nearly 300 patients with one of 14 Mendelian conditions, all neurodevelopmental syndromes, that have been associated with defects in epigenetic regulation, as well as from approximately 650 healthy controls. To generate the DNA methylation profiles, they used Illumina's HumanMethylation450 bead chip or Infinium methylation EPIC arrays.
Developing the database of cases and controls took several years, Sadikovic said, and having good normal reference cases is particularly important because DNA methylation differs between men and women and changes with age. "All this needs to be accounted for when you're trying to identify an epi-signature in any individual patient," he said.
For nine of the 14 conditions, the researchers were able to find epi-signatures, which were specific to the disorder but overlapped to some extent. For seven of the nine disorders — the remaining two had too few patients — they then developed a classification algorithm, based on a multi-class support vector machine (SVM), and found it to be highly sensitive and specific for the seven syndromes.
Sadikovic said that the overlap between the epi-signatures of some conditions, which did not extend to control samples, warrants further study. "What is really interesting is that in some cases, in reference to normal controls, one condition may be showing hypermethylation, and another condition may be showing the opposite effect, hypomethylation. We don't really know what that means," he said, noting that functional studies might provide new insights into genotype-phenotype correlations in these patients.
Since the study was submitted, the team has expanded the project to additional conditions and continues to add more.
Sadikovic said he and his colleagues are now looking to launch a clinical whole-genome DNA methylation screen to rule out conditions with specific epi-signatures, including imprinting disorders and fragile X syndrome.
Some might argue, he said, that such a test is not needed because genetic testing already exists for these conditions and is offered by dozens of laboratories, including his own.
However, he said, the test could serve as a parallel screen for those conditions because genetic tests do not always provide an answer. For example, they frequently report variants of unknown significance (VUS), and the DNA methylation test, if it detects a disease-specific epi-signature, could help classify such variants as pathogenic.
In a report published last November, his team already explored this application of DNA methylation profiling for Kabuki syndrome. "It would give us an opportunity to provide a functional interpretation to that VUS, and not in a model organism or in pedigrees that co-segregate but rather using the patient's own DNA methylation profile as a functional assay," he said.
Also, a DNA methylation screen could assign a disease to patients where genetic testing came back negative, for example, because they harbor mutations in a promoter or have an intronic variant that was not covered by the gene test.
Sadikovic said his team is looking to make the test technology more broadly available for clinical diagnostic services over the coming year and is negotiating with a couple of private laboratories at the moment.
His team would retain the data analytics as part of a partnership, but testing might be delivered through a partner. "We're looking at a couple of different models," he said. "The key here is, the technology itself is not anything that is hard to access — this uses the Illumina Infinium arrays and iScan or HiScan instruments as part of the procedure. The key is really the algorithms, which are dependent on the database."
The cost of genome-wide DNA methylation testing is similar to that of current single-gene epigenetic tests for imprinting orders, he said, and equivalent to other array-based tests. Reagent costs alone are on the order of a few hundred dollars, he said, and the test "could be viably done for well under $1,000."
Sadikovic said he sees DNA methylation profiling mostly as a complement to existing tests rather than a standalone technology for diagnosing patients with developmental delay or intellectual disability. The main diagnostic technologies currently used for those patients are microarrays and exome sequencing, he said, but they do not deliver an answer in many cases. "There is still a lot of heritability that's not being uncovered by these classical genetic approaches, and you'd think that clinical epigenomics, particularly DNA methylation profiling, is going to add some percentage to understand heritability in these patients," he said.
Nobody really knows what fraction of Mendelian disorders might have specific epi-signatures, and thus be diagnosed with a DNA methylation test, but it could be many. "We went after the low-hanging fruit … but part of this effort is going to really assess this, which is something that hasn't been systematically looked at," he said. "You can speculate that there are other genes that are maybe not directly involved in chromatin that may have an epi-signature," he explained, such as genes associated with transcriptional activation and cell-to-cell signaling.
"Once we introduce this in a clinical setting and start generating larger and larger databases, with clinical information linked to them, much like what happened with microarray testing and now exome sequencing, we will uncover what other genes or conditions may have these epi-signatures," he said.
Under a research protocol, his group has already applied whole-genome DNA methylation profiling to hundreds of patients with a broad spectrum of developmental delay or intellectual disability who would normally just get a CNV microarray or an exome test, with interesting results. "We are now identifying recurring cases with an epi-signature where a genetic condition is not known," he said. In addition, some patients appear to have epigenetic changes in specific regions, such as hypermethylation of a promoter, which could also be diagnostic.
Over time, DNA methylation testing might thus lead to the discovery of new genetic disorders, based on these epi-signatures. "I think the best way forward, from a discovery standpoint, is to put [methylation testing] into routine clinical use in these patient populations that normally get genomic profiles and allow the data itself to give us additional utility," Sadikovic said.
Besides inherited disorders, his group has started to use DNA methylation profiling to develop diagnostic approaches for somatic cancer. In particular, the researchers have applied it to prostate cancer, ovarian cancer, and myelodysplastic syndrome and expect to have a publication about the prostate cancer results this year. "We can use this type of approach to very sensitively and specifically identify prostate cancer cells relative to the normal tissue, way beyond anything that's so far being published, so I'm kind of excited about it," he said.
The other area where he expects the technology to be useful is for analyzing circulating tumor DNA, and his group has an ongoing project to explore this.
Sadikovic said he is unaware of other diagnostic labs currently offering similar genome-wide DNA methylation testing, though other research groups have also published studies on epi-signatures, for example for Kabuki syndrome.
"What I'm really excited about is the potential of moving epigenomics out of the research realm into clinical use," he said. "It's something I have always believed in."