I wasn’t really expecting this tweet to get such a response. It clearly hit a communal nerve. Probably in part because I was deliberately provocative 1/
2/ Did @ryanlisterlab really show that DNA methylation has no role in transcriptional repression? Of course not. But they do show that it does not seem to be enough by itself to change the expression of genes
3/ Why is that observation important? Because it makes us think a bit more critically about something that plagues #epigenome-wide association studies, the idea that environment changes DNA methylation, gene expression and cellular memory.
4/ That’s a sloppy and pervasive model in #epigenetics research. Many colleagues are shocked to find that DNA methylation is actually the default for most of the genome. Where DNA methylation *isn’t* found tends to be the interesting sites
5/ It’s not 100% true, but a good way of thinking of these unmethylated sites is that they represent where transcription factors bind in modules at cis-regulatory loci, and become unmethylated. See Dirk Schübeler’s work for examples.
6/ This idea that DNA methylation represents a ‘footprint’ of TF binding overcomes a major headache in the DNA methylation-centric vision of environmental responses: DNMTs do not have the inherent ability to choose where they methylate
7/ A perfectly reasonable model for environmental influences mediating changes in DNA methylation is through the intermediate steps of signal transduction->TF activation/ectopic binding->new DNA methylation patterns. Small RNAs could also be candidate mediators.
8/ The other problem, of course, is that a change in DNA methylation is easy to find when studying pools of cells. Not because cells necessarily change their DNA methylation, but because of confounding effects like cell subtype proportions, functional variants etc
9/ No EWAS to date has unequivocally shown that 5mC changes are free from these confounding influences. It’s frustrating when these studies get published and the authors over-interpret the findings. Everyone at this stage should know about and acknowledge confounding effects
10/10 We can do better #epigenetics research if we stop assuming direct mediation and causality by 5mC in environment/cell memory relationships. That was the goal of the original tweet, a plea for better starting hypotheses and interpretation in these kinds of studies
