Thread by @philipcball

Ready for another bit of obscure WTF molecular biology that has an important broader message you’ll see discussed nowhere in the science press? Of course you are! Then here goes. (Warning – long thread! But with a punchline.) It’s about this paper: /1
nature.com/articles/s4158…

The general issue here is how genes are translated into proteins – yes, about the most central aspect of molecular biology there is. Protein-coding DNA sequences are first transcribed into mRNA in the cell nucleus (that is, for eukaryotes like us, which have a cell nucleus). /2

Then before the mRNA is exported from the nucleus to the ribosome that translates it into a protein, it is chopped up and parts of it are spliced back together. /3

As we’ve known since the 1970s, genes contain segments called introns that aren’t part of the code for proteins, and which are removed from the initial "pre-mRNA" transcript. /4

The parts of the pre-mRNA that are retained in the mature translation-ready mRNA are called exons. These are spliced back together by a structure called the spliceosome. Crucially, it can splice exons in different combinations, e.g sometimes omitting an exon here and there. /5

This means that one protein-coding gene can give rise to several (even many) mRNAs, and thus to different proteins. This is a key part of protein synthesis in complex eukaryotes especially, where the variant of protein produced from a gene may differ in different cell types. /6

The choice of how to splice is made by the spliceosome, which does so on the basis of various signals received from the rest of the cell: protein synthesis is dependent on the broader context of the cell and tissue, and indeed the environment. /7

(This is one reason why it is wrong to think of information flow as unidirectional from DNA to protein...) /8

What determines splicing is a host of molecular factors, including proteins and noncoding RNAs – ubiquitous RNAs that don’t themselves encode proteins but are made from (cough) noncoding DNA and perform various regulatory functions, influencing the output of coding genes. /9

If splicing isn’t done properly – if the production of the mature mRNAs is inefficient for some reason – it’s a big problem. As the authors of this paper say, “disruption of mRNA splicing is associated with many human diseases.” So this stuff matters. /10

Well now: it turns out that the nucleus contains various “bodies” – basically lumps of stuff visible in the microscope, typically made from proteins and ncRNAs – that do something to influence the production and splicing of mRNAs. Here they are. /11

If this sounds all a bit like early 20th century cell biology, when researchers saw vague bobs inside cells (like chromosomes) and had no idea and no clue about what they were and what they did and if they even mattered – then yes, that’s exactly what it’s like. /12

The fact that one class of these nuclear bodies is called Cajal bodies tells you as much. (There’s a whole class of small ncRNAs called small Cajal-body-specific RNAS that seem important but no one really knows what they’re for, except that they are regulatory.) /13

The paper’s authors put it brilliantly: nuclear bodies could be structures self-assembled to do a particular function (“structure enables function”), or they could be ... /14

...emergent in the sense that a bunch of molecular components doing a shared regulatory task happen to self-assemble (“function results in structure”). /15

That is a wonderfully honest admission that the understanding of this stuff is still pretty rudimentary. But the key thing is that there’s a level of organization here that is important, unaccounted for, and... /16

... looks nothing like the smooth molecule-to-molecule information transfer the conventional textbook picture of transcription and translation implies. /17

This paper is about a class of these nuclear bodies called nuclear speckles. The name says it all: they are bright spots of stuff revealed in fluorescence microscopy. Yes: biology does not have a language for this kind of thing, where cell processes seem orchestrated... /18

... not by precisely encoded molecular information, precise molecular structures, or specific and selective molecular complexes, but by lumps of stuff that look like dirt on the microscope. Yet we see it a lot. /19

So then, nuclear speckles. Could they actually be where splicing itself happens? That was once speculated, but no – nascent pre-mRNA is not made there. However, it seems that they contain molecular components that diffuse out of the speckles to the spliceosome. /20

As the authors say, the prevailing notion is that “speckles act as storage assemblies of inactive spliceosomes”: store cupboards of spliceosome parts. However, they do more. /21

As these authors have shown previously, speckles are “major structural hubs” that somehow organize the three-dimensional structure of the chromatin (the DNA plus histone proteins it’s wound around) being transcribed. /21

By doing so, the authors argue in this paper, nuclear speckles play a key role in determining the efficiency of mRNA splicing – and thus in the health of the cell and organism. /22

For one thing, genes located close to speckles are more efficiently spliced than those further away – and this effect is contingent on the pre-RNA still being attached to the DNA from which it is being transcribed. /23

This seems to be a result of the increased spliceosome concentration near speckles. /24

What’s more, those genes close to speckles show the greatest changes in splicing efficiency between different cell types, i.e. the “speckle effect” is somehow responding to cell type, and delivering a signal of it to the transcriptional apparatus. /25

Now here’s the thing: as I said earlier, there is barely even the conceptual apparatus to think about phenomena like this in biology. I mean: “near to”? Like speckles exude some kind of “splicing force”? /26

This is summed up, I think, in the paper’s Figure 6, which offers a brave stab at formulating some kind of picture of what’s going on. /27

The general idea is that there’s some kind of affinity between “splicing factors” concentrated in the speckles – proteins, ncRNAs – and the nascent pre-RNAs being transcribed on parts of DNA. /28

And so those parts of the chromatin are pulled into the speckle – where there are plentiful components of spliceosomes. It is basically like the coalescence of two liquid drops. /29

As the authors say, “Locally concentrating pre-mRNAs, genomic DNA and spliceosomes at speckle-proximal regions leads to increasing splicing efficiency.” /30

Crucially, none of this involves precisely programmed molecular interactions: it is fuzzy and non-specific, as the authors say involving “multivalent contacts”: lots of unspecific (or low-specificity) stickiness. /31

In other words, the ideas here are essentially the same as those invoked in liquid-liquid phase separation in cells, which has been seen or posited in a wide range of processes, from stress response to gene transcription itself. /32

Indeed, in this view speckles really do seem to be like particularly dense liquid droplets. /33

The language appropriate for such things is not about computation and information, but that of condensed matter and physical chemistry: concentration, diffusion, affinity, condensation, phase transitions. /34

And yet these more fuzzy, collective concepts are here potentially governing the thing that has generally been conceptualized in the former, informational terms: which genes get transcribed, and which proteins get made. /35

And this, as I argue in How Life Works, is fully in line with what we should expect from metazoan molecular biology, where such apparent fuzziness provides both robustness – an insensitivity to fine details – and versatility,... /36

... in that outcomes (like what protein isoform is produced in splicing) can integrate and respond adaptively to a wide range of signals. /37

My contention is that it then becomes unhelpful, and indeed probably impossible, to think of all this as somehow being a readout of information encoded in the genome. /38

Sure, there is information in there that influences the process, for example by tuning the promiscuous affinities of the molecular components. But…. /39

…that sort of framing only gets is so far – which is not very far at all. In a probably rather poor analogy, it would be a bit like applying an information-theoretic analysis to the text of Hamlet in the hope of thereby understanding its cultural reception and impact. /40

So the biology revealed by this fascinating study of nuclear speckles looks at this early stage to be bafflingly complicated and fuzzy. But I’d contend that it totally fits with broader principles we can see emerge elsewhere... /41

... and we can find conceptual frameworks for understanding them.

There is a constant stream of this sort of thing in the literature, which is way too complicated to expect that the science press will pick up on it. /42

But as a consequence, I feel that there’s a gradual but significant shift going on in the whole way we should think about molecular (and not just molecular) biology that is just going underneath the radar. Needless to say, it’s why I wrote my book! 43/43

Oh, and of course I have to say - that's this book.
press.uchicago.edu/ucp/books/book…

"blobs", of course.

@Chinahand4 In other words, I think the cognitive dissonance here is precisely because you started out with the thought "Ah, another mangling of the CD".

@Chinahand4 But of course it's not just about Watson. The misconception is very widespread, even within science - I see it all the time.

@dallandrummond I think it is precisely this sort of work that will furnish some (by not means all!) of that language, and I'm very excited by that. I also think these are early days for it. I should have made it clearer that this is exactly the kind of direction I'm glad to see things moving.

@dallandrummond I waas struck that the authors of that paper don't make an explicit connection with the LLPT/condensate literature, which I suspect is an indication that this "language" has not yet diffused far enough!

@dallandrummond ...to get much sense of what is happening. The story tends to be, loosely, that protein synthesis is an essentially digital process. So what's this about vague blobs being near other blobs & somehow tugging bits of chromatin here and there? What's in control? Where's the program?

@dallandrummond What I wanted to suggest is that there *is* a way to make sense of such a fuzzy, collective process - & it is precisely along the lines of the condensate work you mention. But I think we are only just beginning to understand the "language" there, eg with Pappu's "sticky patches".

@dallandrummond It's a different way of thinking that is not yet second nature in mol biol, which is perhaps why the authors of the paper don't make the connection (or perhaps they have other ideas). More broadly, this is not yet a clear and confident narrative. Believe me, it is little known...

@dallandrummond ...outside relatively specialist circles. And especially it is not yet clear how to fit this within a broader picture of metazoan complexity and how that evolved. I'm very glad folks like you are working on that.

@dallandrummond Is that not exactly what you were looking for?
(It crossed my mind to add the link to the nice Brangwynne & Shin Science review, but I thought heck, it's only a Twitter thread.)

@dallandrummond And sure, there are variations in terminology, but it is surely obvious what that reference to LLPS is pointing to.
en.wikipedia.org/wiki/Biomolecu…

@dallandrummond ...while I will indeed be glad to DM to try to clarify, I think it would need to begin with an acknowledgement that your rather cutting initial intervention imputing a lack of awareness of an entire field, while totally understandable when you hadn't seen the entire thread...

@dallandrummond ...was misplaced and mistaken for that same reason. As you now see, a big part of the point of the thread was precisely to point out the connection to that field. Agreed?

@maxabhaase And also this. And low specificity of regulatory microRNAs. There's a lot of work on how low-specificity weak interactions can produce condensates. And some cross-talk seems to be beneficial for evolving new pathways. But yes, many questions too!
quantamagazine.org/a-lobby-where-…