Expected correlation between ATAC data and Chip-Seq histone marks

Question

[bipin]

I am analyzing a gene knockout dataset for which I have ATAC and Chip-seq data for the histone marks - H3K27Ac, H3K4me1 and H3K4me3.

Since the histone marks represent enhancers is it correct to expect that the parts where we are getting a chip signal should also have an atac signal to indicate open chromatin?

Also, what happens first the modification of histone or the chromatin opening?

[Devon Ryan]

Where ever there's a histone you won't have an ATAC peak, since it's not open by definition.

Answer 1

[i.sudbery]

At a high enough resolution the enhancer histone marks will flank the the open region. This is generally visible in averaged metagene type analyses. However at low resolution you should see these marks at enhancers (H3K4me3 is more a promoter mark than an enhancer one, though you will see some at enhancers). However, I wouldn't go so far as to say that those marks are causative for open chromatin and you'll definately see open chromatin without them, and I guess you'll also see those marks without open chromatin.

As for which comes first.... I'm not sure anyone know? The mechanism by which enhancers become active is still and open research question. It is known to involve pioneer transcription factors, which act to open up the chromatin, but whether those factors recruit histone modifying enzymes, or the modified histones help with the recruitment of the pioneer facts, or both things are happening, I don't know.

[Carlo Yague]

As for which comes first.... I'm not sure anyone know?

The egg or chicken, right ? I don't have an answer either but concerning H3K4me [edit: in yeast], it seems that the mark is transcription-dependent (initiating pol II recruits the methylansferase Set1), so it should come after the transcriptional activation of the gene .

[i.sudbery]

don't have an answer either but concerning H3K4me, it seems that the mark is transcription-dependent (initiating pol II recruits the methylansferase Set1)

That would be K4me3, bot K4me1 I think. Although enhancers do show some transcription (eRNA), I seem to remember it has been shown that this transcript is dispensable for enhancer activity.

[Carlo Yague]

My bad, I generalized the situation in yeast where there is only one H3K4 methyltransferase for me1, me2 and me3 (Set1, whose recruitment is transcription-dependent). But in mammals, its much more complex with like seven H3K4 methyltransferases and what you wrote might be true. I edited my comment above accordingly.