Seemingly contradictory ATAC-seq conclusions from a Science journal paper?

Question

[solo7773]

Reference

The chromatin accessibility landscape of primary human cancers, M. R. Corces et al., Science 362, eaav1898 (2018). DOI: 10.1126/science.aav1898

This is a recently published paper in Science. Because I am not studying this sort of things, it's hard for me to understand some of those authors' opinions.

The Fig. 4A describes the concept the authors used to identify transcription factor occupancy from ATAC-seq data. In the paper, they say "TF binding to DNA protects the protein-DNA binding site from transposition while the displacement or depletion of one or more nucleosomes creates high DNA accessibility in the immediate flanking sequence". Here I think I can understand it correctly: (1) TF binds to this region, and then (2) it blocks Tn5 binding, so (3) consequently there is a deep notch of ATAC-seq signal, and (4) this notch region tells us where a TF binds to.

Later near the end of this article, the authors said:

(1) For example, if a noncoding somatic mutation causes the generation of a TF binding site, this mutation could lead to an increase in chromatin accessibility in cis and a concomitant increase in the observed frequency of the mutant allele in ATACseq as compared with that in WGS (Fig. 7A). (2) Similarly, a mutation that inactivates a TF binding site can lead to a decrease in chromatin accessibility and a concomitant decrease in the observed frequency of the mutant allele.

From Fig. 4A we've known that when TF binds to DNA, the ATAC-seq signal plummets. However, the authors now say the generation of TF binding site increases chromatin accessibility. How could this be possible? Doesn't the generation of TF binding site make chromatin inaccessible?

[afli]

Maybe they mean the generation of TF binding site make chromatin more accessible near the binding sites. If there is a week binding site, then in this site the chromatin is generally not very accessible, when the site is strong for TF binding, then in this site the chromatin may be opened by the loading of the TF, though at the direct binding site it is not opening due to the binding of the TF.

Aifu.

Answer 1

[ATpoint]

The biological concept is that certain TFs like for example the hematopoietic TF PU.1 binds to the chromatin and opens it up (e.g. by recruiting chromatin remodelers) so that other factors can come in and start their actions like initiating or preparing a region for transcriptional processes. But this then involves the opening of an entire region of like 200-500bp. In contrast, and this is what you are mixing up, is that the TF binding itself protects a tiny part of this open region from Tn cutting events, but this is something like 50bp or so. I attached a plot where you see that in condition blue upon treatment, you see an increased footprint of a certain TF indicating increased binding, but this is a very local event (see the scale on the x-axis). In contrast, the increased binding is also associated with higher overall accessability of the vicinity (see that the flanking regions of this 200bp window show increased readcounts on the y-axis). So do not mix up chromatin accessablity induced by TFs with local binding site protection due to the presence of the TF itself. The latter is simply the consequence of the physical presence of a protein denying access to the ATAC-seq Tn.

[solo7773]

Thanks ATpoint. Your answer is great. It's concise and clear. It's also a good introduction for me to know how ATAC-seq works for TF discovery.

[ATpoint]

You're very welcome.

Answer 2

[WouterDeCoster]

Not contradictory to me. A TF binding indeed blocks Tn5 at that position, but increases accessibility nearby (in cis).

[solo7773]

Thanks WouterDeCoster. Tn5 binds to the TF site, as shown in Fig. 4A, downstream and upstream of it. (1) Won't it be sequenced by paired end sequencing? If this region is sequenced, the motif center in Fig. 4A should be bulging, or at least flat, not hollow. (2) Is ATAC-seq a magic method that never sequences the TF binding DNA? If true, I agree that "the generation of TF binding site increases chromatin accessibility and the so called TF footprint will be observed", but it causes another problem. Since the mutation-generated TF binding site is not sequenced, how to calculate variant allele frequency (VAF), as shown in Fig. 7?