Researchers found that dense allele-specific DNA methylation (ASM) mapping in normal samples plus cancer samples reveals possible candidate regulatory sequence polymorphisms (rSNPS) that are difficult to find by other approaches.
In a study published in Genome Biology, researchers found that dense allele-specific DNA methylation (ASM) mapping in normal samples plus cancer samples reveals possible candidate regulatory sequence polymorphisms (rSNPS) that are difficult to find by other approaches.
In an interview with Cancer Network®, Benjamin Tycko, MD, PhD, of the Center for Discovery and Innovation at Hackensack Meridian Health, discussed the study findings and how they can be applied in a practical way.
CancerNetwork®: First off, could you explain your recent study which found signposts and DNA pinpointing risks for cancers and other diseases?
Tycko: So, first, this was a large study. We're a small lab, we're small group, but it was a large study. And so, we did Agilent SureSelect methylation profiling in 24 human samples, mostly normal tissues, plus 81 whole genome bisulfide sequences from normal cancer samples. The normal samples were enriched for immune cell types and brain cell types. We then been purified neurons from human autopsy brains and so on. And the cancers that we focused on were multiple myeloma, B-cell lymphoma, and glioblastoma multiforme, which are all busy clinical services at HUMC. It’s a collaboration between HUMC and the new research institute, which we call HMHCDI.
Really, the project started with analysis of the normal cells and tissue types. And at the beginning we said, yeah, let's throw in some cancer samples, maybe something interesting is going to come out of that. And I’ll get back to that point.
So, [Catherine] Do, who's the first author on the paper, she noticed one day – there’s more ASM in these cancers than in any of the normal samples. And so, we said is that just random noise fluctuations, doesn't make any sense, or can we explain it some logical way, and thank goodness we could explain it in a logical way. And it turns out to be due to well-known epigenetic phenomena in cancer that is global hypomethylation, plus focal gains of methylation and in polycomb occupied poised chromatin, basically. So, [Catherine] could extract that conclusion from bioinformatic methods that are now available to everybody to use.
I like to make the analogy to the movie “Waterworld,” which is affectionately known as the best bad movie ever made. And in that movie, the waters have risen to cover everything initially, except for Mount Everest. And then the waters start to slowly recede. So, in the normal cells and tissues, as those waters recede, you start to see mountains. And those are ASM signals in the normal cells and tissues, but then in cancer the waters recede further, that’s global hypomethylation, and you start to see not only the mountains, but all these skyscrapers start to be visible. And then you even have really tall cancer specific skyscrapers – these are these gains of methylation in the polycomb poised chromatin regions. It’s unequivocal, it's absolutely the explanation. Seventy percent of the gains of ASM and cancer are due to the receding waters and 30% are due to these very tall skyscrapers, which are the gains of methylation and poised chromatin. So that was a big relief, we said, so we can write a paper.
Then, the mechanism was important because [Catherine] was able to show that the mechanism, the fundamental underlying mechanism, is the same in cancer and normal. And that was also a big relief, because it meant that we could use the cancer ASM… to be informative about non-cancer diseases. And the mechanism turns out to be something that we suggested years ago in a human genetics paper. And other people agree that ASM reflects binding of transcription factors to one allele and not the other. Basically, the functional snip is a snip that destroys a transcription factor binding site… So, we really can use all the data. And we might find an interesting, functional snip for rheumatoid arthritis that we only discovered in the multiple myeloma, but it's still valid because it's a transcription factor binding effect. And so, at one stage, in B-cell development, this was relevant and might be thereby connected to this autoimmune disease. Cancer immuno-surveillance becomes accessible to us because we have a lot of T-cell data as well.
And then there was… two other points. So is the ASM mapping, this is very technical, is the ASM mapping non-redundant with other post GWAS modalities. And we show that that's the case… the majority of ASM signals are not well annotated by ENCODE active chromatin, or by allele-specific ChIP-seq or even eQTLs, even though these phenomena are enriched among those. And so, and we even found ASM signals in what we call chromatin deserts, gene deserts a lot of course, but also chromatin deserts where there is no ENCODE signal at all, no active chromatin. And that's very interesting, I think, to be pursued because these signals might be footprints of prior activity in these regions; methylation being sort of meta-stable, not as stable as we used to think. But it's still a meta-stable mark.
The very last thing is maybe even the most exciting, which is that… and [Catherine] got after me to pursue this actually that she found de novo ASM in the cancers due to somatic mutations in noncoding regions producing ASM, that's never seen in a normal sample because that mutations never found in a normal samples. Does that put us ahead in trying to understand noncoding mutations in human cancers? Yes, because she was able to show that some of these de novo ASM regions were due to mutations destroying transcription factor binding sites. So, there you can really study that, you can use CRISPR methods and so on to really ascertain that this is functional. So, the burden of these noncoding mutations in the myeloma cases is 400 to 1000 such mutations per case, how are you going to sort through those hundreds of mutations to tell the clinician, this one might be important.
So, how would you sum up the overall implications of this study?
So, I think that it has implications for public utility. The way that we’ve provided these maps as Genome Browser tracks, as well as NCBI, has public utility for every other lab doing post GWAS studies, not only for cancer but for other diseases ranging to Parkinson's disease, schizophrenia, and type two diabetes and so on. Public utility, it never goes away, that's why we like to do these kinds of studies. People have DNA in their freezers, they can expand this map more readily than by doing ChIP-seq where you need fresh material and so on. This can be done on archival material, this is very useful, even potentially paraffin blocks, right? Very useful in a practical way.
And then the second is for this diagnostic sort of genomic medicine application, both in other diseases and in cancer where one would like to understand the… which of these hundreds of noncoding mutations might actually have a function. And that I think… overall the utility is to advance genomic medicine.
Are there any sort of next steps planned for the study?
I think that we're already applying it to COVID-19. I mean, I'm not kidding. So, this has, we have interesting signals in the COVID-19 risk regions. We're going to go back to cancers, particularly the same ones really that we started with: myeloma, B-cell lymphoma, and glioblastoma… these are important cancers. So, we have to… they have a large burden, I mentioned of these noncoding somatic mutations, I think it's going to be fascinating to identify the functional ones.
Do C, Dumont ELP, Salas M, et al. Allele-specific DNA methylation is increased in cancers and its dense mapping in normal plus neoplastic cells increases the yield of disease-associated regulatory SNPs. Genome Biology. doi:10.1186/s13059-020-02059-3.