Associate Principal Scientist, Dewpoint Therapeutics
|Type||Kitchen Table Talk|
On April 12, Dewpoint welcomed Ben Sabari, expert in transcriptional condensates and my former bay mate from Rick Young’s lab. Ben is currently an assistant professor at UT Southwestern where he studies transcriptional regulation and condensates. Before moving to Texas, Ben did his postdoc in Rick Young’s lab at the Whitehead Institute, and before that he got his PhD in chromatin biology and transcription from The Rockefeller University with David Allis.
In the Young lab, Ben spearheaded the experimental work on condensates, which led to one of the first papers showing a link between phase separation and transcription. Since then, Ben has authored many papers diving into the details of this subject, and in the video below he shares some of his latest work uncovering the molecular mechanisms of transcriptional regulator partitioning. We all enjoyed this stimulating talk as part of our Kitchen Table Talk series, as you can tell by the extensive discussion at the end, and I hope you do too.
Create an Account or Sign In to view the video.
John Manteiga (00:00:00):
Yes, and thanks so much everyone for joining today. It’s my pleasure to introduce our speaker, Ben Sabari. So Ben is currently an assistant professor at UT Southwestern where he is doing some fantastic work covering the mechanisms of condensate regulation in transcription. Before moving to Texas, Ben did his postdoc in Rick Young’s lab at the Whitehead. That’s where I actually got to spend a ton of time together with him, because we were bay mates for over three years there. When Ben joined the lab, I was a second-year grad student. As soon as I met him, I knew that I’d learn a lot from him just from being around him.
John Manteiga (00:00:35):
He came in with his experience in chromatin biology and transcription from his PhD work at Rockefeller, with David Allis. Using that, he kind of immediately became a scientific force in the lab and he was always known for asking the toughest questions at lab meeting, and getting pretty animated sometimes. So in Rick’s lab, Ben really spearheaded the experimental work on condensates, when he was first getting started in Rick’s lab. All that work led to a publication of one of the first papers showing a link between phase separation and transcription. Since then, Ben has authored many papers diving into the details of this subject, and today he’s going to share some of his latest work uncovering the molecular mechanisms with functional partitioning of transcriptional regulators. So with that, I’ll hand it over to Ben.
Ben Sabari (00:01:22):
Great. Thanks, John, for the really kind, kind intro. And yeah, I was really … Looking back on that time, it was really fabulous to be your bay mate, but also be in the lab in those really animated lab meetings. I sometimes really miss that, so it’s really fun to be at least virtually among colleagues again from that time. And so yeah, as John mentioned, I started my lab at UT Southwestern. I started in January 2020, so a little over three years ago now. What we’re really interested in studying is nuclear organization by biomolecular condensates, and really how the components of the transcriptional machinery are brought together at specific genomic loci to regulate gene activation. And if I would put one word on what we’re really excited about, is specificity and how specificity comes from these weak multivalent interactions that condensates are known to use…
Ben Sabari (00:02:21):
And so with that, kind of to jump into my talk, which will be in three parts. We’ll have an intro in why one should care about gene regulation, and then why one should care about condensates in that context. Then I’m going to tell a short vignette on a paper, the last paper published from Rick’s lab, that was really inspired by work from Ann and Isaac, which you all know, on the role of transcription factors in DNA regulatory elements and multivalency, and how that leads to specific genomic loci being the site of condensate formation. And then finally, the bulk of the talk will be on a paper recently published earlier this year on how charge blocks within disordered regions enable functional partitioning of transcription regulators in the context of condensates.
Ben Sabari (00:03:11):
And so I guess to this audience I don’t have to say why we should care about condensates, but why should we care about gene regulation? This is really the question that drives me and my thinking, is how do you turn on a gene at the right time in the right place? And I think this example of development tells us why we should even care about that, right? We all start from a single cell. The vast majority of the cells in our body come from that cell, and have a single same genome, right? But obviously carry out very different functions. The way that works for the most part is that each cell is activating a different program of genes, and so this process of turning the right gene on at the right time is critical for development. It is often dysregulated in disease, and is just something that we need to know kind of the mechanistic details of.
Ben Sabari (00:03:59):
That’s exactly what we’re interested in. We’re interested in kind of, how do the molecules at play regulate and inform the cell of what their identity should be? And if you boil all of this down, what does it take or what does it mean to have gene regulation or gene control? It’s about getting this RNA Pol II enzyme, for the sake of mRNA, to the right place in the genome, to the right base pair out of six billion base pairs, and for it to do its job well, and for it to process to be elongated and actually synthesize this mRNA that then starts the process of gene regulation. And I’m just writing out for you the kind of key steps. There are many regulated steps. There are many players involved in getting Pol II to do this really, really complicated process.
Ben Sabari (00:04:42):
And then furthermore, if you kind of look at models where all the players or even some of the players are trying to be drawn in, the point I want to get across to you is that this is a really multi-step process requiring many proteins, that has many positive and negative regulatory steps for the enzyme to actually get to the point where it’s transcribing. And we know quite a bit about this process. Biochemistry has taught us, from purification efforts, a lot about what are the protein players involved.
Ben Sabari (00:05:10):
We can draw these beautiful animations of how polymerase and their accessory machinery are recruited by DNA-binding transcription factors to particular genomic loci, and all the pieces kind of fly in at the exact right time that they’re needed, and that the thing is paused and needs some secondary positive regulatory information from the enhancer, and then polymerase is off to the races.
Ben Sabari (00:05:29):
And of course, there’s a lot right about this animation but there’s a lot of things missing. And for the sake of this talk, what I think is really the key thing that’s missing here is that the nucleus is not this very empty space with things just actually finding one another in this sort of magnetized way. But it’s this immensely crowded environment, not just with protein but also DNA and other macromolecules.
Ben Sabari (00:05:47):
And so what we should be imagining is that whole process occurring in a space that looks more like this, right? And you might be familiar with this video simulation of bacterial cytoplasm, but the protein concentrations are very similar to what one expects to find in a eukaryotic mammalian nucleus. And so now this gives this idea that what might be very important is the spatial organization of material within the nucleus. Let me see. Okay, and this kind of comes at …
Ben Sabari (00:06:28):
When we think about the biochemical process we typically draw it out, like I’ve drawn out the process of transcription, in this linear way. It’s very nice to kind of see it and understand the steps, and for textbook it’s really important to learn all the A, B through N processes that are required to be recruited for the reactions to occur. But of course inside the cell, all these are present in various concentrations, and they’re all really scattered about and diffusing and trying to find one another. Worse yet is there are all these other proteins in the way, right? All these unrelated proteins that are just kind of potentially bumping away the right things.
Ben Sabari (00:07:03):
So the hypothesis that we pursue is that this selective compartmentalization offered by condensates allows for a dramatic enhancement of the rates of reactions. And because these are dynamic assemblies, dissolving them could return the reaction rate back to its basal conditions. And just to very importantly state that of course there would be some basal rate without the condensate, however the condensate dramatically enhances that, and the ability to tune that is really I think a powerful feature of these dynamic compartments.
Ben Sabari (00:07:37):
Going back to gene activation, it’s been observed for really decades that the components of transcription and transcription itself are in fact compartmentalized within some type of structure within the cell. Going back to the late ’90s with Peter Cook, observing that foci of RNA are formed in cells, and then to really beautiful live cell imaging from Ibrahim Cisse and Bob Tjian, and then to the work that my colleagues did in Rick’s lab and really a whole host of folks in sort of the last five years. But the enigma was always, how are these compartments assembled, right? These are really giant assemblies relative to the individual components that are concentrated there, and we’ll get back to that kind of length scale idea.
Ben Sabari (00:08:24):
And so the reason why it’s kind of hard to imagine how these are assembled is because I would say the field of transcription at least has been very, very focused. I would say at large, the field of molecular biology has been really focused explicitly on a single type of molecular interaction, these stable and stoichiometric interactions that enable the formation of complexes. And of course, we’ve learned an immense amount and we have, really thankful to this paradigm, the last 50 years of advances in biomedicine.
Ben Sabari (00:08:50):
But about 10 years ago, a little bit over 10 years ago, we were taught that there are other types of interactions that do very different types of assemblies. These are simplifying these dynamic and multivalent interactions that enable the formation of something that we’re calling condensates, right? And condensates, in the context that they’re incredible different complexes … Again, I don’t have to belabor it for this audience … But they can form network-expanding interactions. They’re not limited by length scale, and they can form these large compartments that we’re observing in cells.
Ben Sabari (00:09:24):
And we’re going to focus quite a bit on these intrinsically disordered regions, but I like this figure from this Brangwynne review because I think we shouldn’t forget that the underlying principle is these dynamic multivalent interactions, not necessarily disorder. Of course, Mike Rosen and many colleagues have demonstrated that actually structured multivalent interactions with multivalent ligands can do the same kind of thing. But with that said, we’re really focusing today on these IDRs, and in large part because the transcriptional machinery, which we know and love for its beautiful structured and folded domains, contains a lot of intrinsic disorder. It contains a lot of large regions of disorder, many of which are sort of known for a very long time to be critical for the function of those proteins, and many others have just been sort of ignored.
Ben Sabari (00:10:12):
And so the other key thing that I want to get out of this slide is this idea of length scale. It’s so hard to draw that on a slide, so I’m going to attempt to do something here where this paradigm I would argue, on the left, is focused very much on this kind of 10 nanometer length scale, which would be a very large complex. Whereas even a small condensate that we observe of the transcription machinery itself is a 300 nanometer diameter. So in volumetric terms, these are dramatically different sized assemblies. The interactions at play to lead to the assembly of this are going to be very different. Right now, the current hypothesis is that it’s these weak dynamic interactions.
Ben Sabari (00:10:53):
So now to sort of bring this back into the context of transcription, if we consider just the stable and stoichiometric interactions, we have a model that looks very much like this, as I showed you earlier. This is something you might find in textbooks. All of this is absolutely true and happening, but at this higher length scale and using all these different disordered regions, or other multivalent interaction domains, these assemblies are able to then have a higher length scale organization into a condensate, enabling the very high local concentration of all the components at specific genomic loci dictated by regulatory elements such as enhancers and promoters.
Ben Sabari (00:11:29):
And again, this is not just what we’re finding but it’s really over the last 10 years a lot of work, experimental data, demonstrating that these types of things are existing in cells. And because you have something like this, the idea that this is happening in this kind of simple, linear way has to be revisited where now, instead of a single polymerase and a single of any of these factors, you have a large amount of them. And instead of one polymerase firing at a time, you can get these large bursts of transcription which is actually exactly how that’s observed, how transcription occurs in cells.
Ben Sabari (00:12:03):
And so like I said at the very outset, my lab is very interested in how specificity works in the context of condensates. And we really understand to great detail how specificity works in the context of complexes, but really that is lacking for us I think in the context of condensates, and that’s what we’re excited to explore. We’re excited to explore this in two contexts. One is, how do you get specificity of genomic locus? Because for transcription, it’s really important for something to happen at the right gene and not at the wrong gene, so how is that established? And then once something is formed, how is it bringing in the right pieces, right? Just bringing in a single concentrated factor will not allow you to push through this biochemical process. It requires to bring in all the factors necessary to actually increase the rate of this reaction.
Ben Sabari (00:12:57):
Okay, so now I’ll tell you just a short vignette on that first part, specificity of genomic loci. This was really a wonderful collaboration with Krishna Shrinivas, when he was a graduate student with Arup Chakraborty. He’s now a fellow at Harvard, and again was inspired by the work by your own Anna Boija and Isaac Klein on how transcription factors and their activation domains are critical for this formation of condensates. Krishna started making these simulations with this question of, how do you get certain DNAs to nucleate condensates, and others there’s not nucleated condensates? And he built this model, a three-part model of a DNA with various transcription factor binding sites. The transcription factor that bound with relatively high affinity and then a coactivator that bound with relatively weak affinity.
Ben Sabari (00:13:47):
What he found was that when you added DNA to the system, this sort of model of DNA to the system, at different concentration regimes, you got the nucleation of something that we call a condensate in that model, that super-stoichiometric recruitment of all these factors to the number of binding sites. And in the absence of DNA, this thing was completely diffuse. And so this is kind of one of the movies from the animation. You can see that you get this large formation, and what he could do in the modeling is just turn off the transcription factor-DNA interaction. What you’ll see is that this thing kind of falls off the DNA and then eventually dissolves. So under certain concentration regimes he found the formation of these assemblies required the DNA as a surface, or a seed, or a scaffold, whatever language you prefer.
Ben Sabari (00:14:32):
And so I thought that was very exciting, and using tools that Anne and Isaac had developed for their previous paper that was published at the time we were starting this, we set out to develop this kind of in vitro system where now we could do exactly what Krishna was doing in the modeling, but in a kind of experimental way. So a DNA of specified sequence, the coactivator which we’re going to be using this mediator subunit that’ll come up really again in the next part of this, and a transcription factor OCT4. And what I hope you can appreciate in these two rows is that there’s either DNA or no DNA added, and these are concentrations of MED1. And when you have no DNA, you have no droplet formation, but in the presence of DNA you’re getting droplet formation at these fairly low concentrations.
Ben Sabari (00:15:16):
I was able to do that same experiment over a very wide range of concentrations, and what you can see in the green line is what’s without DNA. Eventually at some point, yes, you do get these things to form condensates, but this becomes in the micromolar regime. The cellular concentration of MED1 is actually in this kind of 100 nanomole regime, and in the absence of DNA you are not getting any formation of condensates without the DNA. But with the DNA you are getting this really robust droplet formation, suggesting to us that there’s something really important about the kind of tuning of all these components to be fairly low concentrations in this cell, so that they only nucleate and phase separate at DNA elements.
Ben Sabari (00:15:57):
And there’s a lot more data in this paper. I encourage you to check it out if you haven’t read this. And so the rest of the paper is really like, “What about the DNA is interesting?” Because we can make different types of DNA that worked and didn’t work. And Krishna could iterate really quickly in his model, and he found that there’s something really important about the density and number of transcription factor binding sites on the piece of DNA, before which you would not see any kind of super-stoichiometric assembly, and after which you would see this kind of assembly. And so that needed to be tested in vitro, and so here I’m just showing you a snippet of the data, but at the 78 nanomolar concentration, well within the physiological range of the protein, if we add the same length of DNA, the same concentration of DNA, but one has very few binding sites and one has many binding sites, you can see that only in the context with the many binding sites do you get the droplet formation.
Ben Sabari (00:16:44):
And then we wanted to know whether that had anything to do with transcription or function, and we were able to use a classic reporter assay. This is a luciferase-based readout where you can have a promoter, and then on the other side of the plasmid is synthetic … An enhancer sequence. Typically people are inserting a sequence from the genome to test whether it has enhancer activity, but we took the approach of inserting synthetic DNA that was used in the in vitro experiments. Here we can do very kind of a nice set of single increasing binding sites to see whether there was any point at which, before which and after which, there was a big change. And really dramatically with luciferase output, we found that between four and five you had this big change, or step function change, in the amount of luciferase signal, an indirect measure of how much that gene is being transcribed.
Ben Sabari (00:17:33):
And so this was sort of where that ended, and we’re still following up on this in our lab, but maybe next time I come I’ll give you more on this topic. But really, I want to leave you with this idea that there are many ways to modulate binding valency. There are many ways to tune this DNA surface, even this chromatin surface. We talked about in this subject of DNA binding sites. Also you can have different types of RNA, nascent RNA being transcribed from that locus, which can tune fairly dramatically whether that region has a lot of interaction potential. And then not to forget chromatin, my first sort of subject of interest, but PTMs on chromatin are known to be forming multivalently, and so large acetylation regions, large methyl regions can also be before and after such a threshold and I think are really exciting to consider when we think about how genomic loci do or do not form condensates.
Ben Sabari (00:18:27):
And so kind of coming back to cells and sort of bigger picture here, empirically we just know from data that’s been around, and many different condensates form in many different types of genomic loci. For them to be functional, they must be forming at these specific genomic loci, and so we are focusing very explicitly on sort of clusters of enhancers, super-enhancers. But other regions such as gene bodies, localized to speckles, insulated-loop anchors, polycomb bodies, heterochromatin, the constituent of heterochromatin, and of course the nucleolus, all of these must form at a specific genomic locus or else they won’t be doing what their job is meant to do.
Ben Sabari (00:19:10):
And so we’re very excited to kind of keep pushing on this, and kind of keep discovering what is critical about each of these elements. We believe it’s something to do with multivalency, to switch before and after some kind of threshold, but more is to come. And so I’ll use this opportunity to then sort of jump to say, once these are formed, the next kind of big question of specificity is, how do they recruit all of the components that are necessary for their job, right? You might imagine that the ones that are involved in transcription must be bringing in other things than the ones involved in these other functions. That was a question that we started our lab asking, and sort of thinking about, how do we actually address that kind of question?
Ben Sabari (00:19:50):
And so as I was saying, we kind of start with these observations from cells, right? That there are these condensates. This is a particular staining for this MED1 protein, that is the largest subunit of the mediator complex, the central element, the central coactivator in transcription. You can see that it’s forming these foci throughout the cell, and what we want to know is, what’s going into these foci, right? The first time I saw this image it was like, “Wow, that’s that genomic locus and what’s going in there?” And those are very difficult questions to ask, actually, in an unbiased way. And so we turned to things that are a little bit easier to handle, and that’s sort of biochemical reconstitution.
Ben Sabari (00:20:28):
But in my lab I really try to emphasize always, any time we’re doing any kind of biochemical reconstitution, we have to remember why we’re asking the question and why we’re using that particular protein, and that’s because we found something interesting in cells, and we found some interesting phenotype in cells, and always bring back the observations. So in my lab we try to … Really got a crisscross between observations and cells, observations in vitro, and kind of bring them back and forth to make sure that they’re both telling us something that’s true about actual biology.
Ben Sabari (00:20:55):
But the point is that we could take this region of the MED1 protein, it’s IDR has about 600 amino acids–I’ll tell you more about it on the next slide– fused to GFP as a recombinant purified protein, add it to a soluble nuclear extract, right? So imagine we can just squeeze out all the proteins from this nucleus in the correct relative concentrations, and this protein forms these beautiful droplets. And then we can ask, what from that extract is going into that droplet, and what is being left behind?
Ben Sabari (00:21:27):
So again, just to reference ourselves, so this is the MED1 protein. It’s a fairly large protein, and then this is the C-terminal IDR that we’re using. It is the most disordered of the region, but there is an extended disordered region which we’re kind of not including. So just for the aficionados, we’re using this final 600 or so amino acids of the protein. Again, we could put it into a nuclear extract. It forms these droplets, and then proteins from the extract either partition into that droplet or get left behind.
Ben Sabari (00:21:55):
And because these are denser phases of matter, we can very simply pellet these by centrifugation, by G-force these will collect at the bottom, and then collecting with all the proteins that partition into them and in the supernatant will be left behind all the things that don’t want to be in the pellet. We can take that supernatant and we can take that pellet, and we can submit it to mass spectrometry. And then for every protein identified we have a partition ratio, or the signal that is found in the pellet divided by the signal that is found in the supernatant.
Ben Sabari (00:22:24):
And the first thing we did when we got this data set is look at, what are the most partitioned proteins? What are the proteins that have the very highest partition ratio, the very highest pellet over sup ratio? And we were really struck by how that list contained many really important transcription regulators. If you take that list and do gene ontology analysis on it, what you can find is that you get all these categories in being enriched for RNA Pol II binding, POL II transcription machinery binding, and so on and so forth.
Ben Sabari (00:22:55):
And what those factors were were RNA polymerase itself, and all of these positive allosteric regulators of RNA polymerase that without it, there could be no processive elongation. And so these are proteins like CTR9, part of the PAF complex, IWS1, SPT6. RPB1 is the largest subunit of polymerase itself.
Ben Sabari (00:23:16):
And what we found to actually have very low partition ratios were a lot of very well-studied, phase-separating proteins like FUS, nuclophosmin, HP1 alpha and PTBP1. So we were really excited to see that kind of specificity. It wasn’t just like any sticky protein was getting partitioned. And we can validate that by a Western blot adding this really critical control where we left out the MED1-IDR and just tested, were we getting these proteins because they’re just inherently insoluble in the extract, and that’s certainly not the case.
Ben Sabari (00:23:45):
And what you can see here is without or with MED1, in the supernatant fraction all of these marked in green … Again, these factors from here … They are left inside the supernatant without MED1, and then when we add MED1, they’re effectively removed quantitatively from the supernatant, and they appear in the pellet. Whereas factors like FUS or PTBP1, under any conditions tested, they just remained in the supernatant and are never found with any signal in the pellet. So it seems as though what we’re finding from this in vitro context is that there’s some selectivity.
Ben Sabari (00:24:19):
Why do we get so excited about PAF and SPT6 in polymerase is because again, like I said, these are very important positive allosteric regulators of RNA Pol II. They bind, cause conformational changes to Pol II and will enable it to actually be elongated. They work in antagonism to a whole complex called negative elongation factor complex, NELF. What you can see from these beautiful structures from Seychelle Vos, when she was with Patrick Cramer, is that … And this sort of supports a lot of beautiful biochemistry by many, including Bede’s former advisor, that NELF is exclusively binding RNA Pol II, and mutually exclusive to the PAF and SPT6 complex. So you can see how NELF kind of wraps itself around the Pol II and it sort of would … These PAF complexes and SPTs could not bind at the same time.
Ben Sabari (00:25:06):
And so we then probed where NELF was in this context, and we’re really struck by this data where again we’re looking at a pellet without or with MED1. And so with MED1 pellet, and you can see all of the factors that we’ve been talking about. But if you look at NELF relative to input, we find very little if any signal coming into the pellet. As you can see even in a real overexposure, it’s a real tiny fraction of input whereas these are effectively 100% of input.
Ben Sabari (00:25:36):
And this suggested to us that we were creating, at least in this in vitro context, with just an IDR of a protein, no discernible structure formation of this thing. A compartment that’s selectively recruited positive functional proteins and excluded negative regulators of RNA Pol II, and also excluded these other unrelated factors that are very abundant in the nucleus. And so it’s sort of reminiscent of our initial hypothesis that, “Wow, this is something pretty interesting to follow up on.”
Ben Sabari (00:26:04):
And of course, anything we find in vitro we want to come back to the cell, and we started by looking at ChIP-Seq data. If you’re not familiar with ChIP-Seq data, we’re just looking at where along the genome does a protein actually bind, by using antibody techniques and sequencing? And what I want to sort of turn your attention to is that we draw these kind of things as having multiple enhancers really because in the cell we find that many of these most important cell identity genes do in fact have these kind of multiple enhancers clustering with one another, that are interacting with the promoter of genes. In this case, we’re looking at the cluster of micro RNA 290 in mouse embryonic stem cells.
Ben Sabari (00:26:43):
And what I want you to appreciate is that this locus that has actually the very highest amount of MED1 in this cell, you’re also getting the very highest amount of RNA Pol II CTR9 SPT6, but actually relatively low levels of NELF. We wanted to know whether this was something that was sort of genome-wide. And so what we did is we looked at these factors, and we looked at, what are the regions that have the very highest amount of MED1? It’s in orange, and the lowest amount of MED1, in gray. And those areas that had the very highest amount of MED1 also had the very highest amount of polymerase, CTR9 again the PAF complex, and SPT6.
Ben Sabari (00:27:18):
If we looked, actually ranked all of the genes that were associated with MED1 enhancers, from low to high as is deemed down here, what we found was really strikingly in the lowest fit of the RPKM values of the CHiP data, that you are getting some increase, and there was some apparent threshold above which you got this really dramatic increase in these amounts of RNA Pol II CTR9 and SPT6. Again, that followed the actual transcription of those loci suggesting that the concentration of these factors actually meant something functionally. And so this led us to then set up experiments that test this hypothesis, and we wanted to now know, is the IDR itself enough, sufficient for the partitioning and for the gene activation as we’re observing with the MED1 in the context of mediator?
Ben Sabari (00:28:08):
And to do this, we turned to a very powerful experimental model that I’m sure many of you are familiar with, but it’s a cell line developed by David Spector quite a while ago now, where it’s an integrated Lac operator array. This is a really beautiful system, because it allows us to concentrate any factor that we want at fairly high concentrations at a specific defined genomic locus. And so by doing that, we can fuse a Lac repressor to a fluorescent protein and use that as a control, and then also have the MED1 IDR there.
Ben Sabari (00:28:41):
And strikingly, what we find when we probe those cells with antibodies against polymerase, CTR9 and SPT6 is that if you look at that CFP LacI focus, okay, and that’s blown up in the corner here, when you do the IF you see no overlap with polymerase, CTR9 and SPT6. When MED1 is there, you start to see polymerase, CTR9 and SPT6. But if you look at the negative regulators NELF and HP1 alpha, you see the opposite. NELF seems to be there. HP1 alpha seems to be there. But when you have MED1, it’s kind of creating an environment where those are now being excluded.
Ben Sabari (00:29:15):
Strikingly, and in interesting agreement with the CHiP-Seq data, if we now plot many cells for the intensity of CFP at that focus, meaning the amount of CFP LacI that’s there, equivalent to the amount of mediator, or MED1, they were observing in the CHiP-Seq data, there is an apparent threshold before which these factors are not being partitioned, and above which these factors are really highly partitioned. They control any concentration we looked at, any CFP intensity we looked at, we don’t see anything like that. Suggesting that there’s some emergent property, there’s some switch at the concentration, where now after a certain concentration, after the recruitment of a certain amount, you’re now getting partitioning.
Ben Sabari (00:29:59):
And so finally what we did was use this beautiful system again, and after the Lac array there is a reporter that expresses a transcript that has a 24x MS2 repeat on it, which allows us to visualize the RNA. We can ask whether creating this environment with all those factors including Pol 2 meant anything for transcription. What I’m hoping you appreciate here is when we have just the control, we see no accumulation of RNA by MCP or indirect recruitment, but when we have the MED1 IDR there we’re seeing substantial amounts of mCherry-MCP being recruited, suggesting that that reporter is being activated by this high localization of MED1.
Ben Sabari (00:30:42):
And so we have now a growing model where now, when you can put this large, high concentration of MED1 at the particular genomic locus, you can now be switching gene activation states from being repressed and stopped, to now being very active. And we next wanted to ask again … This is sort of still in an artificial system … Is this important for something that could be called a bona fide gene activation event? Not something that we kind of built up in an artificial way.
Ben Sabari (00:31:18):
And to do this, we turned to a very specific cell model, and this is the 3T3-L1 cells that can be treated with a cocktail of compounds to undergo adipocyte differentiation. And the reason we do this, to make a long story short, is because it has been previously published by Bob Roeder and others that MED1 is required for the differentiation of these 3T3-L1 cells. And by other groups, it’s been shown that during the differentiation process, MED1 forms these very large clusters of enhancers, genomic super-enhancers, at key adipocyte genes. And so we thought, “Wow, this is really a beautiful model for us to ask. We know that MED1 is required. Is the IDR required?”
Ben Sabari (00:31:59):
And to do this, we developed cell lines using CRISPR. We developed three independent cell lines that targeted early exons to regenerate a knockout, and then three independent cell lines that cut off the IDR. It’s this beautiful thing. The IDR is encoded in the final exon of this gene, so there’s no nonsense-mediated decay. What we see is the non-targeting, you have MED1. The knockout, you see very little signal or no signal for MED1, and the delta IDR you’re getting this sort of truncated factor, truncated protein.
Ben Sabari (00:32:29):
And then when we force this cell line to undergo differentiation, the non-targeting activates these key cell identity genes, Adipoq and Fabp4, but the knockout as expected does not, and the delta IDR also is incapable of activating these gene programs. We do a time course, this is in kind of stalled differentiation, they’re really just unable to differentiate. And then we can look at the product of differentiation, which is fat cells that have fat deposits, actual adipocytes, by using this BODIPY stain, and the non-targeting control forms these beautiful adipocytes, where the knockout and delta IDR are unable to.
Ben Sabari (00:33:04):
So it seems like the MED1 IDR and likely all of its ability to do the things that we were saying–and we have quite a bit more data in the paper. I encourage you to look at that I’m not showing you here–that it’s all critical for the actual activation of these new genes during the cell state transition. And so here I’m just listing a few summary points: These condensates have preferential partitioning. There’s something about these genomic loci with highest MED1 that seemed to be recruiting the highest amounts of these partitioned factors. The high-level concentration is leading to transcription in a reported-based system, and the IDR seemed to be required for this activation during a kind of bona fide gene activation event, or really a host of gene activation events during a cell state transition.
Ben Sabari (00:33:48):
And so now we’re getting back to the most important thing, what is a mediating partitioning? Is there something common among these pelleted proteins? And so we were able to use the proteomics to ask these questions, and what we found after digging through this data quite a bit is that all of the most partitioned proteins had their own IDRs, and so the pellet-enriched proteins were enriched for having IDRs of their own. We could find that CTR9 and SPT6 had IDRs at their termini, which we could clone and purify, as controls. We took factors that were not being partitioned, NELFA, HP1 alpha and FUS. And just to remind you, those factors, CTR9 and SPT6, were being recruited while the others were being excluded.
Ben Sabari (00:34:31):
We could find remarkably in in vitro systems that just these IDRs were capable of partitioning into reconstituted MED1 IDR droplets. And so what we’re looking at is recombinant protein, IDR fused to mCherry of CTR9 and SPT6 and the others listed, and then MED1 IDR and GFP, and I hope you can appreciate that CTR9 and SPT6 partition whereas these others do not. We can do the same type of experiment in this Lac array context, where we’re focusing again on the Lac focus, and CTR9-IDR, SPT6 partition and the others do not, suggesting that these IDRs encode the specificity.
Ben Sabari (00:35:09):
We can show too that the full-length protein requires the IDR to partition into these foci, where if we take the full length of CTR9 or SPT6 it partitions, cut off the IDR, it does not. The IDR alone, it does. And so we start to feel confident to say that there’s something just exclusively about two IDRs that can be highly specific.
Ben Sabari (00:35:31):
And what are the features of these IDRs that enable them to be specific, right? We haven’t really had categories like this to look at. And so again, after really beating our heads against this sort of sequencing data or amino acid sequence data, or what is important about these proteins that partition, we came to this realization that for many of the proteins that partition, they had this very interesting feature of having charged blocks. And again, this is a feature of IDRs that have been reported by others. But I think for CTR9 and SPT6 and a number of other factors we focused on, they have something particularly unique in that they have alternating positive acidic … Or basic acidic, positive and negative blocks along their IDR. Whereas a factor like NELFE has a lot of charge, but its charge is not blocked in this way so it’s just kind of had this alternating RD-motifs.
Ben Sabari (00:36:25):
We tested 12 different IDRs, and we found that actually number of blocks was the best parameter that actually tracked with how well things partitioned in the Lac array context. And so now to test this idea, we did really a lot of mutagenesis in different types of experimentation and I’ll just show you a few examples here. We could scramble the sequence, remove all the basic amino acids, remove all the acidic amino acids, and that really just prevented the partitioning wholesale. But those are pretty large perturbations to make.
Ben Sabari (00:36:55):
So we started getting a little bit more clever here and doing these sort of modified scrambling experiments. I’m showing you some data for SPT6. And so we can set about scrambling the entire sequence. We can scramble the charge sequences, or we can scramble all the non-charge sequences, leaving charge as it is, in three random iterations. And if charge patterning was solely important, and it wasn’t some other hidden motif, these three should still partition where this one should not, and that is exactly what we observed in this Lac array context, and when we purified the wild type and charge scramble and one of the non-charge scrambles, we found something very similar where the non-charge scramble still partitioned relative to wild type.
Ben Sabari (00:37:35):
We can make synthetic versions of this, where we could take … Without getting into too much detail … Sequence features that existed in CTR and SPT6, make libraries of various acidic block, linker, or basic, and then make five different synthetic IDRs all of the same length, varying sequence but same charge patterning. All five of them are capable of partitioning into MED1 IDR in the Lac array context and the in vitro droplet.
Ben Sabari (00:38:02):
And then this was our favorite experiment, where now we could take that NELF, right? That charged protein, again part of this negative factor that was not partitioning, but it has a lot of charge. Can we rearrange the charge to make it blocky, okay, and does that actually partition? In fact, we found it does. When we take the wild type as we had shown multiple times in the previous experiments, and now we create these blocky sequences, and it partitions very well. And again, we can show that exact same thing in in vitro.
Ben Sabari (00:38:31):
And this is a fairly complicated experiment that I don’t usually present, but I think it’s pretty striking in the context of like, how does this partitioning relate to function? And what we could do is now instead of taking just the IDR, we could take the NELFE protein. And why that’s interesting is the NELFE protein recruits the entire NELF complex. And we could do it with the wild type sequence that does not partition, or we can make chimeras that either have the NELFE blocky sequence, CTR9’s IDR, or SPT6’s IDR and ask, “Does that now lead to the partitioning of the entire complex? And if it does, into the Lac array, how does it affect the transcription off that reporter gene,” okay?
Ben Sabari (00:39:13):
So what we’re looking at here is the MED1 IDR, CFP LacI, MED1 IDR cell lines. We’re creating these compartments with the MED1 IDR. The red is whether the NELFE chimera is partitioned, and exactly as we expect, right? When it has the wild type CTR9 it partitions, when it doesn’t, it doesn’t, so on and so forth, and then when we make it… But what I really want you to focus on is this MCP, which is the RNA coming off of that reporter. When we inadvertently partition NELF, right, by putting on the wrong IDR, it goes in and it turns off the gene, okay?
Ben Sabari (00:39:53):
When it now can’t go in, the gene is active, okay? When we put it in by having the wrong one, again it turns it off, opposite. So then the wild type, right, it’s still active because it can’t get access. But when we give it that blocky sequence, it turns off the gene. So this ability to get proteins to be partitioned or not, again encoded in the IDR, seems to be really important for whether or not different enzymatic activities, different allosteric-regulatory activities, are or are not gaining access to their substrates within a condensate.
Ben Sabari (00:40:22):
And so this is a summary here, and one more data slide to share. But it seems like we can really manipulate the different charge patterns here and different sequence features without changing anything about the overall composition, and that can dramatically affect whether or not an IDR goes into a condensate or not, and has a functional outcome.
Ben Sabari (00:40:42):
And so now we really want to kind of come back to MED1 and ask, does MED1’s activity, right? It seems to be really important for activating these gene reporters and for activating adipocyte identity genes. And so do IDRs with similar charge patterning, have similar pattern charge blocks, have this similar function? Can they recapitulate? Can they complement genetically?
Ben Sabari (00:41:07):
And so we first tested a lot of different IDRs now on the Lac array, so all the experiments that we showed you before, MED1 IDR is on the Lac array and so that’s what I’m showing you here, no IDR versus MED1 IDR. And it recruits CTR9, it recruits SPT6, and it activates the reporter. But now we put onto the Lac array three different blocky IDRs, so CDK12 is a new protein that we found just sort of in analysis of protein data … Sorry, just protein sequence. It has a blocky sequence. CTR9 and SPT6 we’ve been talking about.
Ben Sabari (00:41:39):
What I hope you can appreciate is that those proteins do sort of very similar things, and then when we use not blocky version, either charge scrambled versions of CTR9 or SPT6 or the FUS IDR, they just are incapable of doing these three things. And what I want you to look at here is, this is really cool to us, is that when you put the FUS IDR there, there’s a black hole for CTR9 IDR, so there really is some exclusion going on when you have the FUS IDR, so really saying there’s different IDRs create different chemical environments.
Ben Sabari (00:42:08):
And then finally, the last piece of experimental data, we came back to our 3T3-L1 cell line where we had this knockout clone. We asked, “Can we complement adipogenesis with different chimeras of MED1?” And so you have a non-complementing control. You have the MED1 full-length protein that contains its natural MED1 IDR, and then we could put in the SPT6 IDR. Again, it’s a 200 amino acid sequence versus a 600 amino acid MED1 sequence. It has no sequence complementarity between them, but they both have this kind of blocky charge, and then we have the charge scramble control.
Ben Sabari (00:42:44):
What I hope you can appreciate, and I’m only just showing you this data here or outside of this, but if we look at Fabp4 expression, again this is one of the key adipocyte genes, the MED1 IDR complements but the SPT6 chimera also complements, and we also see that in the actual formation of adipocytes.
Ben Sabari (00:43:02):
And so just to sort of summarize everything here, these condensates can selectively partition the transcription machinery. They are sufficient and necessary for activation of specific gene programs. The multivalent charge blocks on clients and IDRs are required for the selective partitioning, and please check out the paper. There’s a lot more detail in there.
Ben Sabari (00:43:25):
But I want to close with this idea that there really are just many, many open questions. I hope that anything we do opens more questions than it answers, but these are some of the ones that I’m thinking about. I’d love to get your thoughts on what are things that we should be thinking about. But just to go through them briefly, we talked about MED1. There are a whole host of other coactivators that have their own IDRs that actually have very different sequence features to them. Do other coactivator IDRs partition different sets of proteins? How much overlap is there? How much specificity is there between IDRs?
Ben Sabari (00:44:02):
Are these IDRs mistargeted in disease? I think we can say very clearly yes, given there’s all these fusion oncoproteins that have fusions between DNA-binding transcription factors or chromatin regulators and IDRs. What are all the flavors? This is kind of a bigger-picture question, but what are all the different flavors of selective heterotypic IDRs? We talked about some kind of charge complementarity between two IDRs, but we tested other very charged proteins and they don’t work, even ones that have blocks but not exactly the same type of patterning of blocks. So you could potentially have a large degree of specificity even with playing around with how charge is patterned along an IDR, which I think is really exciting.
Ben Sabari (00:44:39):
Given that we’re talking about charge here, PTMs, there’s a clear role for how phosphorylation, which dramatically changes charge, acetylation, which would dramatically change charge, a whole host of these solutions, various things that will modify charge, how did those affect partitioning? And then what I think you all are interested in, is partitioning druggable? I think as we start to get deeper and deeper into how the specificity actually works, can we block points, nodes, along the network, and could that actually cause it to fall apart and lose function?
Ben Sabari (00:45:08):
And just because really, like a lot of things in our field that we’re just at the tip of the iceberg here. I do think that there’s a lot more to discover in the context of how IDRs encode specificity, and I’m excited to keep working on this and chat with you to hear what you’re interested in. And with that, I have to thank my group, six members that first started in my group. Heankel as a graduate student really led the experimental effort here. Prashant did all the in vitro biochemistry, really remarkable purification efforts, and Reshma doing most of the computational work, and the rest of the team helping out in many big and small ways, collaborators and funding. And I’ll thank you for your time, and happy to take questions.
John Manteiga (00:46:02):
Awesome. Thanks so much for the talk, Ben. So we do have a couple questions queued up. So actually Diana had a good … No? Okay. Okay, so Omer, you had an interesting question in the chat. Could you unmute and ask it?
Omer Gullulu (00:46:29):
Hi. I hope everyone is hearing me clearly.
Ben Sabari (00:46:32):
I can hear you at least, yeah.
Omer Gullulu (00:46:35):
Amazing, thank you. That was an amazing presentation, Ben. So I was wondering whether using the nuclear extract is the optimal way to capture the condensate components of the MED1 IDR? Whether some nucleocytoplasmic transcription factors are already in the cytoplasm, that is causing kind of loss of capturing in the final evaluation?
Ben Sabari (00:47:16):
Yeah, that’s a really important thing to consider always, how do our experimental conditions bias us to either find or miss something? And obviously, given how we make the extract, we would miss anything that was in the cytoplasm. But to be honest, the reason to do this is because this nuclear extract preparation is kind of core to the field of transcription, and is how a majority of the proteins we care about were purified from this extract. And so we know going in that this is a extract that contains the pieces required for transcription, but you’re absolutely right. I think we’re very excited to now use … You know, we were doing a particular cell line. We can make extracts from different cell lines. We can make whole cell lysates. We can make cytoplasmic fractions, nuclear fractions, various fractionations of nuclear fraction, get chromatin fractions. Yeah, it’s very malleable in that way, and the particulars of how we did this were that we made a nuclear isolation and then extracted protein from there at 420 millimolar monovalent salt. So we’re also missing a lot of things that are really tightly bound to chromatin, so-
Omer Gullulu (00:48:32):
Mm-hmm, that would be amazing to see the temporal dynamics, which components are getting sequestered out, which are partitioning inside? That would be amazing to see the dynamics of MED1 condensates, mm-hmm.
Ben Sabari (00:48:48):
I agree, yes. That we’re excited to pursue, and we’d be happy to collaborate.
Omer Gullulu (00:48:53):
John Manteiga (00:48:59):
We have another question from Tanzeem in the chat? You can unmute.
Tanzeem Rafique (00:49:10):
Oh. Hi, Ben.
Ben Sabari (00:49:10):
Tanzeem Rafique (00:49:10):
That was a really nice talk. So I was wondering, when MED1 IDR is made blocky, does it have an effect on gene activation? And I have a second question, so I was wondering if the MED1 IDR sequence is conserved in nature?
Ben Sabari (00:49:29):
Yeah, those are great questions. So for question number one, we didn’t do any experiments to make a MED1 IDR sequence itself more blocky. We kind of did experiments to break the blockiness or make these chimeras. So the short answer is that when we made the chimera with SPT6, it maintained gene activation to the point of creating adipocytes, and for the few genes that we investigated.
Ben Sabari (00:49:59):
But for question number two, which is a really important question, I think this is … Anyone interested in IDRs will appreciate this, is that IDRs are kind of positional sequence, very poorly conserved in the sense of you put it in Clustal, you’ll have lots of gaps and changes. But if you look at feature conservation, which we did in the paper, and we asked like, “Are the number of blocks or the position of blocks relative to one another conserved?” They’re conserved a lot, right?
Ben Sabari (00:50:27):
Whereas the positional sequence identity only goes down to 50%, and the particulars for MED1 IDR, it’s like a metazoan-specific thing so it doesn’t exist in … The IDR itself doesn’t exist in yeast, let’s say. But in the species that we were able to observe, positional sequence was going way down but the sort of blocks were there. And if we then look into the blocks, it was very cool. You could see like arginines to lysines switching, or small gaps appearing. But still, the kind of local density of charge maintained, right?
Ben Sabari (00:51:02):
So obviously, it’s sort of amenable to much more mutations than would be a structured domain. And I think as many of us can appreciate, there’s going to have to be new rules and new tools on how to think about IDR conservation, that are going to be very different from our kind of tools that we kind of take for granted, that are based in the structure function. I feel like I’m preaching to the choir here, but that’s how we’re thinking about it.
Tanzeem Rafique (00:51:28):
John Manteiga (00:51:35):
Awesome. We have another question from Matt, in the chat?
Matt Good (00:51:40):
Hi, Ben. Really neat story. My lab enjoyed reading this paper. I was curious whether you have observed discrete MED mediator, subunit, IDR condensates? In short, do you think MED1 is acting as an insulator against RNA Pol II negative regulators from binding at loci? Or yeah, I guess how much have you explored other subunits and IDRs for partitioning negative regulators?
Ben Sabari (00:52:12):
Yeah, that’s a really insightful question because right, we have to remember that these are parts of a very large complex, and many of the other mediator subunits contain their own IDRs, which we did end up testing a few of them. Using the IDR metrics that we … Effectively, we use MobiDB, which we find is the most stringent IDR predictor. But anyway, when we use that one, MED1 is by far the largest IDR. MED15, 26 and 14 have very small, small IDRs on them that are non-terminal, I believe. And we did purify those, and they did not phase separate in our assay condition, not to say that we couldn’t find conditions that they would again, but in the conditions that MED1 did in the extract and so forth.
Ben Sabari (00:53:01):
But obviously if you take other IDR predictors, and again we could argue what that means, like meta-predict, then there’s way more disorder, and so there are much larger regions of disorder. We haven’t really followed up on those, and again I think that’s a whole other conversation. But your base question of like, do different mediator subunits do different things for either partitioning or capturing different parts of the transcription machine area, I think is very exciting. It’s something that has sort of I think not been studied, and we’re thinking about it. Nothing kind of concrete coming right now, but I think it’s a very interesting question.
Matt Good (00:53:35):
John Manteiga (00:53:39):
We’ve got a bunch of questions queued up in the room here, so we’ll start with Bede, then go to Phi right after?
Bede Portz (00:53:47):
Yeah. Hi, Ben. Great talk, as usual. Sorry if I missed this, but what is the RNA concentration and sort of the nature of the RNA in these lysates? And have you considered or attempted titrating in RNAs that you transcribe in vitro?
Ben Sabari (00:54:04):
Yeah, yeah, great. Yeah, so the short answer is that there is RNA in these extracts, and that we found that batch to batch, extracts have different amounts of RNA and different quality of that RNA being degraded or not, which ends up being very important. And so making a sloppy extract will give you sloppy results, so we’ve kind of in time been very careful with things like making things fast and keeping the on ice, the sort of things you’d be wanting to consider.
Ben Sabari (00:54:37):
So in that kind of sense, more RNA seems to be better for the droplet formation, and that’s just total nuclear RNA that’s isolated. We have found that if you just purify total RNA in a sort of standard way and add it to proteins, certain proteins, that will induce the phase separation. And we also have done synthesizing specific RNAs, and that also does the trick. But what is specific about which RNA does the trick versus which RNA does not is something we have no clue about, but I bet there’s something very cool there.
Bede Portz (00:55:15):
So have you sequenced the RNAs that co-condense with … When you seed with MED1?
Ben Sabari (00:55:22):
Yeah, we’re doing that now. We don’t have really any results yet. It’s sort of confusing to … Again, when we’ll see the data we’ll know how to analyze it. But off the bat, we’re going to have to come up with whole new ways of trying to understand what the sequencing means. So we’re kind of still thinking about that, right? Like with the proteins, we could think about amino acids and their charge parameters. But with the RNA, and I’ve been trying to talk with colleagues, and maybe you guys have some insights. What are the kind of parameters we look for to do clustering and things like that? But it’s something we’re interested in.
Phi Luong (00:55:57):
Ben Sabari (00:55:58):
Phi Luong (00:55:58):
This is Phi. Rick Young once said that all genetic mutation in transcription could be linked to a condensate mechanism. And so when I think about DNA-based condensates such as like cGAS and TREX1, there’s a beautiful partitioning condensate mechanism with that. So based on your charge mutations, could you see that in this type of system, or conceptually in other charge-based condensate mechanisms?
Ben Sabari (00:56:34):
So what was the particular example that you mentioned? I missed the specific.
Phi Luong (00:56:38):
There’s an autoimmune disease where cGAS, the leading binding enzyme, and TREX1, the exonuclease, there’s a mutation at TREX1 that impacts the partitioning into the cGAS condensates, that impacts the resolution of it. So basically, TREX1 is unabled to degrade the cGAS condensates, right? So, and here you have like a charge-based partitioning mechanism, right? So is there any human mutations that is tied to like a charge-based partition mechanism in your system?
Ben Sabari (00:57:12):
Yeah, yeah. You know, I think the answer is probably yes. It’s harder for me to imagine point mutations having a big role in the kind of things that I was describing. I wouldn’t say it’s not going to happen, because biology is crazy. But what I think is more interesting as we talk about human diseases is this recent paper from Denes, another Young lab colleague, where he showed this really remarkable thing where the C-terminal region of I think it’s an HMG high-mobility group protein, through a frame shift mutation, flips from being … Again, I may mess up the details, but flips from being either basic or acidic, to being the opposite through the kind of like nonsense coding that occurs in a frame shift.
Ben Sabari (00:58:00):
To me, that seems very exciting, or some kind of random insertion of sequence, right, in sort of a larger … So to me, the idea of causing a frame shift at the C-terminus to create a new tail, to me that’s like, “Whoa.” That’s going to be really interesting implications for what we’re finding out. Obviously there are IDR fusions, so fusion events, are also quite exciting to think about in this context, if you have an IDR of a certain type and that gets fused to be a very different type. That to me seems like the low-hanging fruit here. But I bet that someone will identify a point mutant that is working in this particular way, yeah.
John Manteiga (00:58:44):
Awesome. We’ve got two more questions. Let’s start with Edgar, and then move on to Avi after that.
Jill Bouchard (00:58:49):
And somebody from the chat, too.
John Manteiga (00:58:50):
As well as one from the chat after that.
Edgar Boczek (00:58:52):
Bede actually asked my question already, so thanks very much, Ben. I’ll just use the opportunity to thank you. It was a great talk. It’s a great paper. I think it’s going to have a big impact on the community. Very, very nice to see.
Ben Sabari (00:59:12):
Thank you. That’s very kind of you.
Avinash Patel (00:59:18):
Maybe just for the … Hi, Ben.
Ben Sabari (00:59:19):
Avinash Patel (00:59:20):
Fantastic talk. In the interest of time I will just ask one, right? So just, you showed that the specificity exists in the transcriptional condensate that is MED1-based. But is there a specificity in general in terms of the gene output from these kind of condensates? I can’t imagine every transcription that is going on has this kind of specificity built in. So do you think there are certain genetic loci that are way more specific compared to others? Any thoughts on that?
Ben Sabari (01:00:01):
Yeah, yeah, that’s a really … Again, another important question that gets at this idea that I think in the transcription field, we talk about what regulates transcription at large, right? As if it was one process and not 12,000 individual biochemical processes. And I think as much as we can start recalibrating our thinking to say, “These are 20,000,” or however many genes are actually on at the same time, different kinds of activation events happening, we’re probably … Or in different classes that are more than a dozen, we’re going to be in a better place to think about this. And I think you’re absolutely right about what you said, right? It’s possible, if you have a gene that just needs one mRNA expressed every five months, it’s incredibly stable RNA, and it just always needs a housekeeping thing, you may not need to form such a thing, right?
Avinash Patel (01:00:56):
Ben Sabari (01:00:58):
You may be okay with the basal rate of transcription, right? And that’s what I was trying to say in the intro. There is going to be a basal rate that happens through diffusion. You can’t argue with that. But if you are a gene that needs to be on at a specific time, at a specific developmental window, at a specific what have you, and you need to … And that RNA is under extreme negative regulation, meaning its RNA is incredibly unstable, you need to be damn sure if you’re transcribing that thing with very high burst frequency and amplitudes. And in those cases I think are the kind of … Those are the kind of most obvious places you’re going to need such a thing.
Ben Sabari (01:01:34):
And then the question becomes, how many different types of those exist? Because you might imagine, you need to be doing different pathways at the same time, and maybe one needs to be shut down and the other needs to continue. And so there certainly will be differences, right? And there’s beautiful work coming out from folks like Alex Stark and a bunch of people demonstrating that what we used to call the general coactivators or general cofactors, actually seem to have quite a bit of specificity for different enhancer types. So meaning, enhancer type A is going to be more dependent on coactivator A, and enhancer type B on C, D and E, and they won’t be kind of talking to one another. Which is dramatically different from how it’s written in the textbooks, which give the impression that transcription is a singular thing, okay?
Avinash Patel (01:02:21):
Ben Sabari (01:02:22):
Does that kind of answer your question?
Avinash Patel (01:02:24):
Ben Sabari (01:02:25):
The real answer is, it’s very complicated and exciting, and there’s a lot of work to be done.
Avinash Patel (01:02:33):
Yeah, great. Thank you.
John Manteiga (01:02:37):
Great. So now let’s go to Severin, then Constanza, and if there’s still time, we can get to Bede and Alan after that.
Severin Lechner (01:02:44):
Yeah. Hi, Ben. Thanks for the talk. So just a quick question, so you quickly mentioned this drug cocktail used for differentiating these cells to adipocytes. I was wondering what these drugs are and whether they are like … Whether you think they’re involved in changing PTMs on those condensate partners or something like that?
Ben Sabari (01:03:05):
Yeah, so it’s a cocktail. I’m going to blank on the exact details, but there are certain hormones, certain … It’s like insulin, TZD, so it’s like what it basically is doing is activating PPAR-gamma very strongly, and repressing other pathways. Very likely it’s causing pleiotropic signaling cascades that do lots of very interesting things. I don’t know if there’s data explicitly on that cocktail leading to changes of PTMs of MED1, but I know others in other systems have shown that MED1 is heavily phosphorylated in a signaling-dependent manner. And those people have mostly been studying androgen receptor, estrogen receptor signaling, and so we just haven’t looked at phosphorylation of MED1 in the context of adipogenesis. But in other signaling pathways, it has been shown. But none of these are like kinase inhibitors or agonists. It would be through some kind of indirect way, if that was happening.
Severin Lechner (01:04:15):
John Manteiga (01:04:16):
Is Constanza here?
Constanza Enriquez (01:04:27):
Yes. I just was wondering if you have considered the relevance of spacers without charging your IDR sequences, if maybe if you considered something else in the IDR important for the selectivity? I mean, it could be like disorder-to-order transitions maybe, and thank you.
Ben Sabari (01:04:52):
Yeah, you know that’s also a very interesting question. The only thing we’ve done to look at that is by doing these kind of non-charged scrambles. So that effectively means scrambling the linkers, and that’s … In the three random iterations, that seemed to not have a very large effect. And then the other thing is, we made these five independent synthetic IDRs, which again their linkers are totally different from one another. So from those data, I would say there’s nothing kind of like jumping out at us as obvious. What we didn’t do is modify the length of the linker, you know? Which I would think would have a dramatic effect. We tried very hard to always make comparisons with IDRs of the same length. So those kinds of things I think would be pretty interesting, but the matrix of experiments becomes kind of infinite, to kind of really properly explore that. But if you have particular ideas, I would love to chat because I do think that’s an interesting area that we haven’t explored yet.
Constanza Enriquez (01:06:02):
Bede Portz (01:06:09):
Hey again, are your extracts biochemically active such that ATP-driven processes, namely like ribostasis and proteostasis machinery, could be spiked in and you could thus see the potential changes in the condensate interactome?
Ben Sabari (01:06:28):
We haven’t done anything like that. We haven’t done even the … So when you say biochemically active, it’s sort of different from the spiking in. We’ve toyed around with trying to do this I think exciting experiment to see whether the transcriptional activity in the extract is going into the pellet, but we really haven’t spent a lot of time on that. I imagine it would be fairly straightforward if you have an active complex that you’ve purified, to spike it in to see how that changes things. You know, we’ve thought about not exactly what you said, which sounds exciting, but to put in kinases or phosphatases let’s say, or RNAs, and see what happens. But yeah, we haven’t really made progress on those particular experiments. But yeah, the kind of … The potential to modulate is what’s kind of exciting to me about the biochemistry, but it also is infinite, so it’s like you kind of have to pick. But I think those would be really interesting to pursue.
John Manteiga (01:07:38):
Great, and I think our last question here, Alan?
Alan Underhill (01:07:42):
Yeah, that was a great talk. So that dovetails with the concept of kinases in modulation, but with p67 it’s phosphorylation that regulates the charge patterning. So I’m just wondering if similar things at play in terms of lysine, arginine, with alternating blocks of like serine threonine residues or something like that?
Ben Sabari (01:08:02):
Yeah, I think the answer is it’s going to be true.
Alan Underhill (01:08:05):
Ben Sabari (01:08:05):
We don’t have a good example of it, but I know others have data to things like poly polyserine patches, right? That under certain signaling become phospho … Sort of very … So they basically become phosphoserine tracks, and that sort of changes dramatically the charge patterning of the region, again with acetylation. I think chromatin is probably the best example here, where we know, right?
Alan Underhill (01:08:30):
Ben Sabari (01:08:31):
We know for a fact that the way we activate a genomic locus is to negate a ton of lysine charge by acetylation. You know, acetylation isn’t coming in and hitting one lysine-27, right? It’s acetyl spray. It’s like a spray paint, right? So you are just changing the biophysical nature of that specific location within the genome, and I think that is just really interesting. I think people like Mike Rosen, with Bryan Gibson’s paper, have started to address that, but I think that’s going to be an area that’s exciting.
Alan Underhill (01:09:03):
Right. Thanks very much.
Jill Bouchard (01:09:07):
Well, that was a fantastic discussion, and Ben, great talk. Thank you so much for sharing all your work and ideas with us. John, do you have anything else you want to say? I can move my mic over to you.
John Manteiga (01:09:25):
No, just thanks for sharing, and all these awesome questions are evidence of how exciting your work is. I think you had offered to answer any straggler questions if people had them. We will kind of compile those and send those to you at a point in the future. So yeah, thanks so much.
Ben Sabari (01:09:44):
Yeah, absolutely. Thanks for the invitation. I had a really great time, and thanks for all your excellent questions.
Jill Bouchard (01:09:49):
Yeah, so thanks again, and anybody who wants to come back to our next Kitchen Table Talk, we’ll have one in the next month or two. So, thanks again, Ben. Have a great day, everyone.