Insight into DNA methylation analysis
The association of epigenetic changes with disease pathogenesis is the subject of much study. CLI caught up with Dr Kathleen Barnes (Oxford Nanopore Technologies) to find out more about how DNA methylation is linked to disease as well as developments in methods of DNA methylation analysis for improving disease diagnosis, prognostic stratification and therapy.
What is epigenetics and how is methylation involved?
If the DNA sequence is comparable to lines of computer programming code, then epigenetics is like the series of punctuation marks that can be used to alter whether or how those lines are read when the program is run. Epigenetics is a fascinating regulatory system. Without changing a single nucleotide of DNA sequence, epigenetic marks govern which genes are expressed or silenced and how chromatin, the complex of genomic DNA and histones, is structured in ways that alter gene expression. Epigenetic changes can be picked up over time in a person’s life, even beginning in utero, such as from exposure to certain pathogens, carcinogens and environmental toxins.
There are several types of epigenetics, and methylation is probably the best known. Adding or removing a methyl group, typically on a string of cytosines in the DNA sequence, can change whether a gene is expressed or silenced. Imprinting disorders happen when an allele from one of the parents is repressed through methylation, resulting in the silencing of the gene associated with that allele. Examples include Prader–Willi syndrome and Angelman syndrome, which are caused when there is hypermethylation or hypomethylation, respectively, of specific alleles.
If you’re familiar with the X-chromosome inactivation process, then you already know about an example of methylation. In most mammals, females have two X chromosomes, but having all of those genes expressed at once could cause problems. Instead, one X chromosome in each cell is inactivated, initially through structural changes that repress transcription and, longer-term, through methylation changes that keep the chromosome silent for the entire life of the cell.
Why is it useful to analyse methylation status?
At the most basic level, analysing methylation status is essential for broadening our understanding of biology. Methylation is such a common mechanism regulating gene expression that it requires thorough characterization to understand how, where and why it is used in organisms.
On a more pragmatic level, the most important reason to evaluate methylation patterns is because they are deeply related to disease, especially current disease. Methylation can be an early warning signal for diseases such as cancer. It can be diagnostic for conditions such as the imprinting disorders mentioned above. Fundamentally, while DNA sequence stays the same over time for most of our cells, methylation profiles change; that gives them great power as a signal for what’s happening in our bodies at any given time.
For example, methylation changes can signal the onset of cancer and are detectable long before other clinical factors become noticeable. Already, distinct methylation patterns have been associated with breast cancer, glioblastoma, lung cancer, thyroid cancer, hepatocellular carcinoma, leukaemia and more. Scientists have even found differences between methylation profiles of a primary tumour and metastatic sites, suggesting that epigenetic data could even prove useful in tracking the progression of cancer. Because changes in methylation appear to be common across cancers, routine methylation analysis could one day supplement or even replace the organ-specific cancer screening protocols we use today, providing a non-invasive, holistic view of whether cancer is emerging anywhere in the body.
In addition to cancer, there are many autoimmune diseases that have perplexed scientists because they are not shared by identical twins; one twin will have the disease and the other will be healthy. In these cases, suspicion has fallen on epigenetics as the driving factor because it would explain disease onset differences between identical twins. Studies have shown that methylation of certain human leukocyte antigen (HLA) genes is correlated with the risk of developing rheumatoid arthritis, and even that specific methylation profiles may be linked to response or non-response to treatment for this disease.
Differences in methylation have also been associated with neurological conditions such as autism spectrum disorder, metabolic disorders such as type 2 diabetes, and aging-related diseases such as Alzheimer’s.
Although it is possible that very few of these correlations will turn out to be causative factors in disease onset or progression, at a minimum these studies suggest that methylation status may be a useful biomarker for early detection of these and many other chronic and complex conditions. That is a compelling reason to improve our technical capabilities for profiling methylation in a broad range of clinical applications.
How is methylation status normally analysed, and are there any limitations/drawbacks or pitfalls of this method?
For biological experiments, the most widely used method is bisulfite sequencing. Most DNA sequencers cannot distinguish between a methylated base and an unmethylated one, so a bisulfite conversion step is needed to change unmethylated cytosine bases into uracil, leaving methylated cytosines as they were. When the sample is sequenced before and after this conversion step, a comparison of the cytosine bases that changed versus those that didn’t allows scientists to track methylation.
Obviously, this is a clunky approach. It requires two rounds of sequencing as well as the bisulfite conversion process, increasing experimental time and costs, and compromising the quality of the DNA. A better approach would be to directly detect the methylation marks as DNA is being sequenced. This is now possible with two commercially available sequencing platforms, but most scientists still use platforms that lack this feature.
In clinical settings, methylation is more often detected with microarrays, but these have their own limitations. Unlike the unbiased discovery that is possible with DNA sequencers, microarrays can only look for things that are already known, and represent only a subset of the entire methylome. Microarrays do
Epigenetic modifications of DNA and histone proteins, such as 5′ methylation of cytosine bases, affect chromatin packaging and therefore alters accessibility of DNA, resulting in changes in gene expression (Vector Milne, Adobe Stock)
not allow users to detect novel or unexpected methylation marks. Additionally, the regions of DNA they analyse are highly targeted, which further narrows the chances of discovering unexpected but relevant information. For ultimate clinical utility, the ability to detect methylation genome-wide would be quite powerful. Finally, methylation microarrays tend to be time-consuming, taking several days to return results. There are many clinical applications for which methylation profiles could prove more useful if they could be accessed much faster.
For example, the surgical resections often needed for tumours of the central nervous system are guided by the tumour type (some require more extensive resection than others). But determining the tumour type is a challenge, and often can only be done during the surgery itself. Researchers in The Netherlands recently demonstrated that a new approach to methylation profiling can be done quickly enough that it could be deployed during surgery to inform the resection protocol. They used nanopore sequencing, which directly detects methylation status as it sequences DNA, to create a methylation profile. Those data were run through Sturgeon, a neural network tool they developed to classify these tumours. In real-time examples performed while surgery was underway, the team was able to accurately identify tumour types for most samples with an overall turnaround time of less than an hour and a half [1].
Not all clinical applications for methylation will require that kind of speed, but there are many cases – from early cancer detection to disease diagnosis – that will benefit from getting results sooner than would be possible with legacy approaches.
How might methylation analysis be improved and what benefits would this provide to patients?
Because methylation offers a biological snapshot of what’s happening within each person at the time, it has great potential for use as a clinical tool. With sufficient clinical research studies to elucidate methylation profiles and their clinical impact, along with the proper analysis tools to generate these profiles, it should become possible to use methylation as the basis for diagnosing disease and stratifying patients by prognosis. This will also give us the opportunity to verify accurate methylation signatures derived from target tissue that is more accessible, such as circulating blood, saliva, buccal swabs, a cervical swab, for better testing procedures. It’s even possible that understanding methylation could support the development of advanced gene therapies to cure imprinting disorders and other conditions caused by aberrant methylation.
None of this will be possible without significant improvements to our current clinical tests for methylation. We will need to achieve unbiased, genome-wide methylation detection to enable the identification of novel or rare methylation profiles. This effectively rules out microarrays and suggests that a sequencing-based approach will be important for methylation analysis going forward. We will also need rapid, streamlined workflows – no more of these cumbersome bisulfite conversion steps and the need to sequence every sample twice to spot evidence of methylation. Direct detection of methylation must be the goal, and it should happen on a platform that can run samples and generate data quickly enough to provide actionable results in a clinically relevant time frame. Lastly, we will need technologies that can detect all forms of methylation; most current approaches focus only on 5-methylcytosine (5-mC) methylation. Although 5-mC is the most widely studied type of methylation in human genomes, there are indications that other types may be important and should be identified whenever possible.
There is a lot of work to be done to achieve all of these goals, but none of them is out of reach. With enough dedication to empowering methylation analysis in clinical labs, I am confident we can realize the promise of methylation for improving patient care.
For further information visit Oxford Nanopore Technologies (https://nanoporetech.com/)
DNA methyl transferase I (DMNT1) methylates DNA (Juan Gärtner, Adobe Stock)
not allow users to detect novel or unexpected methylation marks. Additionally, the regions of DNA they analyse are highly targeted, which further narrows the chances of discovering unexpected but relevant information. For ultimate clinical utility, the ability to detect methylation genome-wide would be quite powerful. Finally, methylation microarrays tend to be time-consuming, taking several days to return results. There are many clinical applications for which methylation profiles could prove more useful if they could be accessed much faster.
For example, the surgical resections often needed for tumours of the central nervous system are guided by the tumour type (some require more extensive resection than others). But determining the tumour type is a challenge, and often can only be done during the surgery itself. Researchers in The Netherlands recently demonstrated that a new approach to methylation profiling can be done quickly enough that it could be deployed during surgery to inform the resection protocol. They used nanopore sequencing, which directly detects methylation status as it sequences DNA, to create a methylation profile. Those data were run through Sturgeon, a neural network tool they developed to classify these tumours. In real-time examples performed while surgery was underway, the team was able to accurately identify tumour types for most samples with an overall turnaround time of less than an hour and a half [1].
Not all clinical applications for methylation will require that kind of speed, but there are many cases – from early cancer detection to disease diagnosis – that will benefit from getting results sooner than would be possible with legacy approaches.
How might methylation analysis be improved and what benefits would this provide to patients?
Because methylation offers a biological snapshot of what’s happening within each person at the time, it has great potential for use as a clinical tool. With sufficient clinical research studies to elucidate methylation profiles and their clinical impact, along with the proper analysis tools to generate these profiles, it should become possible to use methylation as the basis for diagnosing disease and stratifying patients by prognosis. This will also give us the opportunity to verify accurate methylation signatures derived from target tissue that is more accessible, such as circulating blood, saliva, buccal swabs, a cervical swab, for better testing procedures. It’s even possible that understanding methylation could support the development of advanced gene therapies to cure imprinting disorders and other conditions caused by aberrant methylation.
None of this will be possible without significant improvements to our current clinical tests for methylation. We will need to achieve unbiased, genome-wide methylation detection to enable the identification of novel or rare methylation profiles. This effectively rules out microarrays and suggests that a sequencing-based approach will be important for methylation analysis going forward. We will also need rapid, streamlined workflows – no more of these cumbersome bisulfite conversion steps and the need to sequence every sample twice to spot evidence of methylation. Direct detection of methylation must be the goal, and it should happen on a platform that can run samples and generate data quickly enough to provide actionable results in a clinically relevant time frame. Lastly, we will need technologies that can detect all forms of methylation; most current approaches focus only on 5-methylcytosine (5-mC) methylation. Although 5-mC is the most widely studied type of methylation in human genomes, there are indications that other types may be important and should be identified whenever possible.
There is a lot of work to be done to achieve all of these goals, but none of them is out of reach. With enough dedication to empowering methylation analysis in clinical labs, I am confident we can realize the promise of methylation for improving patient care.
For further information visit Oxford Nanopore Technologies (https://nanoporetech.com/)
The interviewee
Kathleen Barnes PhD Senior Vice President,
Population Health & Precision Medicine at Oxford Nanopore Technologies
Oxford Nanopore Technologies US HQ, New York, NY 10013, USA
Email: Kathleen.Barnes@nanoporetech.com
Reference
1. Vermeulen C, Pagès-Gallego M, Kester L et al. Ultra-fast deep-learned CNS tumour classification during surgery.
Nature 2023;622(7984):8a42–849 (https://www.nature.com/articles/s41586-023-06615-2).