More Recent Comments

Friday, December 06, 2013

Do you understand this Nature paper on transcription factor binding in different mouse strains?

I've published a few papers on the regulation of transcription of a mouse gene and students in my lab have done the standard promoter-bashing experiment to define transcription factor binding sites. I did ny Ph.D. in a lab that specialized in DNA binding proteins. I've kept up with the basic ideas in eukaryotic gene expression in order to teach undergraduate courses on that topic and in order to write appropriate information in my textbook.

I've been interested in genome organization for several decades and I've been following the literature on pervasive transcription and transcription factor binding in whole genome studies. I'm reasonably familiar with the techniques although I've never done them myself.

I'm not bragging; I'm just saying that I know a little bit about this stuff so when I saw this paper in one of the latest issues of Nature I decided to look more carefully.
Heinz, S., Romanoski, C., Benner, C., Allison, K., Kaikkonen, M., Orozco, L. and Glass, C. (2013) Effect of natural genetic variation on enhancer selection and function. Nature 503:487-492. [doi: 10.1038/nature12615]
Almost everyone knows about the major problem in this field after the ENCODE publicity fiasco of last year. The "problem" that it's relatively easy to identify transcription factor binding sites but it's quite another matter to determine if they are functional.1 The title of the paper suggested to me that these authors had found a way to address this problem by looking at natural variation between inbred mouse strains. If they detect differences in transcription factor binding sites, will these be correlated with differences in the regulation of identifiable (annotated) genes? Or, will the evidence suggest that those binding sites are just "noise"? Will the strain differences affect the production of low levels of transcription (i.e. most of "pervasive" transcription)?

As soon as I read the abstract it became apparent that the authors were not addressing the most important issue.2 They were looking at something else. But what, exactly? Here's the abstract ... see if you can guess.
The mechanisms by which genetic variation affects transcription regulation and phenotypes at the nucleotide level are incompletely understood. Here we use natural genetic variation as an in vivo mutagenesis screen to assess the genome-wide effects of sequence variation on lineage-determining and signal-specific transcription factor binding, epigenomics and transcriptional outcomes in primary macrophages from different mouse strains. We find substantial genetic evidence to support the concept that lineage-determining transcription factors define epigenetic and transcriptomic states by selecting enhancer-like regions in the genome in a collaborative fashion and facilitating binding of signal-dependent factors. This hierarchical model of transcription factor function suggests that limited sets of genomic data for lineage-determining transcription factors and informative histone modifications can be used for the prioritization of disease-associated regulatory variants.
After reading the paper very carefully (three times!) I think I know what they mean but it sounded a lot like gibberish the first time I read it. I'm curious to know what the rest of you think. Do you understand what this paper is all about after reading the abstract?

Maybe you have to read the introduction to get a better idea. Here it is.
Inter-individual genetic variation is a major cause of diversity in phenotypes and disease susceptibility. Although sequence variants in gene promoters and protein-coding regions provide obvious prioritization of disease-causing variants, most (88%) genome-wide association study (GWAS) loci are in non-coding DNA, suggesting regulatory functions1. Prioritization of functional intergenic variants remains challenging, owing in part to an incomplete understanding of how regulation is achieved at the nucleotide level in different cell types and environmental contexts2,3,4,5,6,7,8,9,10,11. Recent studies have described important roles for lineage-determining transcription factors (LDTFs), also referred to as pioneer factors or master regulators, in selecting cell-type-specific enhancers12,13,14,15, but the sequence determinants that guide their binding are poorly understood. Previous findings in macrophages and B cells suggest a hierarchical model of regulatory function6, in which a relatively small set of LDTFs collaboratively compete with nucleosomes to bind DNA in a cell-type-specific manner (Fig. 1A, a and b). The binding of these factors is proposed to ‘prime’ DNA by initiating deposition of histone modifications that are associated with cis-active regulatory regions (Fig. 1A, b and c) and enable concurrent or subsequent binding of signal-dependent transcription factors that direct regulated gene expression6,13,15,16 (Fig. 1A, c–e). In principle, this model provides a straightforward framework that allows non-coding variants to be classified with respect to their ability to directly perturb LDTF binding and their potential to exert indirect effects on binding of other LDTFs and signal-dependent transcription factors. To test the validity of this model and its ability to explain effects of genetic variation on transcription factor binding and function, we exploited the naturally occurring genetic variation between the inbred C57BL/6J and BALB/cJ mouse strains (~4 million single nucleotide polymorphisms (SNPs) and ~750 k indels17) as an ‘in vivo mutagenesis screen’.
Does that help?

I may post an English translation in the comments after a few days but for now I'd like to hear from you. Have we reached the stage where Nature articles are all but incomprehensible to people who aren't actively working the specific field? (This is a main article, not a "Letter.")

1. The problem is compounded my misuse of the word "enhancer." An enhancer is a region of DNA that's known to be required for gene expression. Those who do whole genome studies tend to use the word to mean ANY site of transcription factor binding. These are "potential" enhancers. They could also be junk binding sites with no biological function. I don't think it's a good idea to misuse the word "enhancer" in this manner. (See the title of the paper.)

2. Turns out there's nothing in the paper to suggest that the authors are the least bit interested in whether their results are biologically relevant.


John Harshman said...

I read the abstract, and I have no real idea what the message is. Nature is supposed to be a general-interest journal, which makes it even more important than usual for scientists to write clearly. Why don't the editors of Nature think so?

Steve Watson said...

OK, so it's not just the fact that I'm a layman that makes that incomprehensible. Good to know.

Samphire said...

Let me explain it to you in simple language. Hang on, is that some-one at the door? I'll be back in a minute.

Schenck said...

Clearly we're getting Sokaled...

Is there some variant of Poe's Law, where IF there's no way to distinguish between Sokaling and poorly written jargon-laced garbage, then it may as well be a farce anyway?

Anonymous said...

OK, well bypassing what you included of the Intro and judging by just the abstract (and giving them the benefit of the doubt) I'm assuming they looked at natural variation in TF binding sites and rather than looking at transcription directly, tried to correlate it with disease phenotypes.

Anonymous said...

As an aside, from the Intro: " (For) disease-causing variants, most (88%) genome-wide association study (GWAS) loci are in non-coding DNA, suggesting regulatory functions"
If the hyperbole of the ENCODE results was valid wouldnt we expect ~99% of disease-causing variants to be non-coding?

Claudio said...

Agree with Robert. Either this paper is a Sokal-eske play perpetrated by disgruntled postmodernists or modern molecular biology is completely lost.

Georgi Marinov said...

The paper is actually quite cool.

But it is true that there is a disease afflicting a lot of papers I see these days - the combined pressures of packing a very large amount of stuff (so that the paper is "worthy" of the journal) into the very limited space available, and of making some both "fundamental" and novel claims (again, to match the journal's high standards) makes for some truly unreadable texts. You read them, you understand every sentence, you understand all the experiments, and yet after you've finished you realize that you don't actually know what the paper is trying to tell you and have to read it again and again. Very frustrating. And very common with high-throughput data papers, the data contained in which can be used to look at a lot of things in the same time, making them a lot more susceptible to the syndrome.

Bjørn Østman said...

I could not understand a single sentence of the abstract. :(

Jonathan Badger said...

Because Nature, like the other "Glam Mags" like Science and Cell, has strict page limits -- everything has to fit in 6 pages ignoring the fact that this isn't enough for most modern papers. So clear introductions are not the priority given that the space used there could be used elsewhere.

Larry Moran said...

I'm pretty sure I could re-write that introduction making it much easier to understand and using fewer words. Same with the abstract.

However, the point about space restrictions is quite valid in general. The body of that paper is essentially gibberish to everyone but the experts in the field and that's harder to fix.

zumb said...

The same can be said of most (all?) systems biology papers.

John Harshman said...


That's just an excuse for not doing the work. You want a paper in Science or Nature, you have to do the work to condense it and still make your point clearly. The editors should be making sure that happens, and they aren't. They're relying on reviewers who, most often, are all specialists in that field and are not the best choice for spotting clarity, because they already know the message.

I know it's hard, but you have to do it. And the abstract is especially important, because it's where you have to make the case that the rest of your paper is worth reading.

Anyway, science is communication. If you want to communicate only with specialists in your field, please don't submit to Science or Nature (or PNAS, or any other journal of wide scope).

Jonathan Badger said...

Oh I absolutely agree that science is (or at least should be) about communication and this paper fails at that except maybe for people working on related projects. And yes, there is a real feeling these days that specialist journals are inferior (because of their generally low impact factor) so that people tend to only submit stuff that doesn't make it into prestigious journals even though they are often better places to publish with more space available and a readership who cares about the topics.

caynazzo said...

I have a quibble about the function of an enhancer. As I understand it genes do not require enhancers for expression; enhancers enhance gene expression (transcription levels). In a reporter assay, the way you validate a putative enhancer is by checking whether it effects expression (of the target promoter) in both orientations (multidirectional), whereas a promoter's sequence is direction dependent. Another way to validate a stretch of DNA for enhancers activity is to look for DNA binding proteins via an electro-mobility shift assay.

Unknown said...

They correlated transcription,methylation and DNA binding outcomes with naturally occurring variations of known TF binding sites in two mouse strains. I think. ChIPSeq, RNASeq, and some bisulphite sequencing as well?

I always wonder about these papers, but i have not enough experience to review most of the figures.