 All right, should we get started? All right, well, it is my great pleasure to introduce this year's Steambach lecturer, Professor Michael Rosen. The Steambach Lectureship honors the legacy of Harry Steambach, whose pioneering work in metabolism identified vitamin D deficiency as the cause of rickets and represents the biochemistry department's most prestigious honor and recognition we bestow upon a scientist. Professor Rosen did his undergraduate studies at the University of Michigan, where he was selected as a Winston Churchill scholar to pursue post-bacc studies in Cambridge. He received his PhD at Harvard University in Stuart Shriver's Laboratory in 1993, where he did pioneering structural and biochemical work on immunofillings that clarified the molecular basis of these important receptor, ligand, receptor interaction pairs. As a postdoctoral scholar in Tony Pawson and Lewis Kay's laboratories at the University of Toronto, he developed novel techniques for studying large proteins by NMR and used these methods to investigate the structural basis of intramolecular interactions such as auto inhibition, which play a key mechanism in cell signaling logic. These studies telegraphed the broad direction of Professor Rosen's own independent research group, which seeks an understanding of the cell that bridges different length and time scales. His group's early body of work from Memorial Sloan Kettering provided an amazing view of how the nanoscale activities of signaling proteins can direct the microscale organization of act and polymerization within the cell. Professor Rosen moved to UT Southwestern in 2001 and was named an HHMI investigator in 2005. Since then, his group's continued investigation into the multivalent organization of signaling complexes led to the groundbreaking discovery of biomolecular condensates as major players in the organization of molecules within the cell. Since then, his group has produced a focused, elegant, and biochemically grounded understanding of the biophysical properties of these fundamental structures within the cell and has identified key ways in which they can be regulated to control important aspects of cell biology, ranging from chromatin structure to the cytoskeleton. In addition to their importance for basic science, Rosen is exploring translational applications targeting phase separation as a founding scientist of phase medicines, which has recently raised more than $80 million in capital. Professor Rosen has received countless honors and recognitions, including most recently the Wiley Prize and Kaiser Award, and he was recently elected to the National Academy of Sciences in 2020. And so without further ado, let's welcome Professor Rosen for the first of two exciting lectures on this exciting and groundbreaking work. Well, boy, thank you very much for that, that very kind, overly kind introduction. It really is a great pleasure here to give the Steenbach lectures. From what reading I've done about Harry Steenbach, he was one of the founders of the field of nutrition, which of course matured into or developed into the field of biochemistry, which is really where I'm grounded. Steenbach also was one of the early people who argued for the importance of commercialization as a way of taking basic science research and enabling it for the public good. And Warf here basically developed from his desire to take his understanding of vitamin D and production of vitamin D and enable at the time the cattle industry but also treatment of children with rickets. And as you just heard, one of my hopes is that we will be able to develop our understanding of, as you'll hear about, biological phase separation into new ways of developing therapeutics through phase medicines to cure things or at least help diseases including ALS and certain kinds of cancer. So again, it's wonderful to have the opportunity to spend a couple of days here with you in Dr. Steenbach's honor. So research in my lab for about the past dozen years or so has focused on, I can move beyond this, has focused on a family of cellular structures, cellular compartments that we term biomolecular condensates. These are compartments primarily in eukaryotic cells, probably also in prokaryotic cells that by definition concentrate molecules, certainly proteins, certainly RNAs, probably small molecules as well, in the absence of a surrounding membrane. So these are distinct but related in some functional ways to the canonical membrane-bound compartments, nuclei, mitochondria, Golgi, et cetera. They're sort of better understood and in textbooks. As I've tried to indicate here, what my lab has taken to calling the egg picture, there are many, many different kinds of biomolecular condensates throughout the cell and they are involved in a huge range of biological processes. Everything from signal transduction at membranes to the organization of RNA and metabolism of RNA and various compartments to certain aspects of gene expression. Over the years, condensates have also become associated with many different human diseases, most notably neurodegeneration, which results from mutations of certain proteins that are known to be resident in certain kinds of condensates, but also more recently it's become clear that defects in molecules resident in condensates lead to things like cancer and skin diseases and moreover that one can treat other kinds of diseases through targeting condensates involved in the relevant biology. So understanding condensates we think is going to be relevant both to understanding basic biological processes but also hopefully will provide, as I said, insights into creation of new kinds of medicines. A key feature of condensates is that their components are quite dynamic. They both move within the molecular molecules that are in them, move rapidly within them, but also exchange between the concentrated condensate structure and the surrounding cytoplasm or nucleoplasm. But yet somehow, although they are dynamic, they are distinct. So different components concentrate in different kinds of condensates which then endows them with different kinds of biochemical activities. And in general I think it's fair to say that these appear to function as kind of intermediates between macromolecules that are organized and function on nanometer length scales and canonical membrane bounded organelles that again are organized and function on micron length scales. Now when we first got interested in this area and it's been about a dozen years now, we set out a series of questions that we were hoping to answer over time. So first, what is the physical nature of these condensate structures? How is it that they're able to form and still exchange molecules with the surroundings but yet be persistent over time? Second, what is their chemical nature? What kinds of molecules make them and can we understand how they might be regulated and function according to that chemical composition? And again, closely related to that, how can they be regulated? Because no organization in biology means anything if cells can't control it. And then finally, perhaps most importantly, what are the biochemical and cellular functions that arise from a formation of these condensed structures and how do they arise from the physical properties, chemical composition and regulation? So what I'm gonna tell you about this afternoon is a series of stories. At the beginning, I'm gonna give kind of a historical introduction how we got involved in this work that will take us through essentially, I think key aspects of physical nature, chemical nature and regulation. And then I'm gonna move into some more recent work for my lab that gets into what I think is really this key issue in particular, what are the biochemical functions of condensates and what kind of biochemistry is arise uniquely from assembling molecules into these higher order structures? So I will begin, as you heard from Scott, with previous work in my lab before we stumbled into this area, focused on understanding structural and biochemical aspects of signaling pathways that control the actin cytoskeleton. And we were particularly interested in signaling pathways that controlled this actin nucleating machine called the ARP-23 complex, which had been discovered around 2000 by Tom Pollard's lab that binds to an existing actin filament and can grow a new actin filament from the side. And in that process, that nucleation of actin filaments, ARP-23 complex controls a vast array of cellular processes. Now, ARP-23 complexes are activated dominantly by proteins in the Wiscott-Aldrich syndrome protein family or the WASP family. And many of them, like this molecule, N-WASP, have long proline-rich regions. And so N-WASP is a molecule that has about 150 amino acids stretch in it that's sort of a long disordered loop. Fully half of the amino acids are proliens. And so there are many, many, roughly nine, as I've shown here, different binding sites for sarcomology three domains, SH3 domains, which I suspect virtually everybody here knows, is a widely expressed signaling module that binds to proline-rich peptides. Now, suffice to say, we published a paper in 2008 that suggested that oligomerization of N-WASP proteins should be a means of potentiating their activation of the ARP-23 complex. And so we started looking for molecules that we thought might engage this proline-rich element to cause oligomerization. And so in particular, we started studying a protein called NCK, which is, again, a widely expressed adapter protein that contains an SH2 domain, a sarcomology two domain that binds to proline-rich motifs, and contains three SH3 domains that are known to bind to, as I said, these proline-rich motifs. And in fact, a lot of studies before we got into this had shown that NCK binds to the proline-rich region of N-WASP and is involved in controlling N-WASP activity. And we kind of just looked at that and thought with our hands, well, if you've got something that has three, a SH3 domains on it, and you've got a ligand that has nine binding sites, that should be a system that can create oligomers. You have one NCK that might bind one N-WASP molecule, there'll be other sites that are open and another N-CK could bind and so forth, so you could get polymerization. And as I said, we published this paper in 2008 that showed that oligomerization would be a means of potentiating this activation pathway. And so we started fiddling around with these two proteins in the lab, and then one afternoon, my student, Wei Chen Cheng, basically had this weird observation. And that is, she showed that when you mixed NCK and N-WASP at low concentrations, not much interesting happened, but as you increase the concentrations of the two at some very sharp point, the solution went cloudy. And eventually, after a couple of days of trying to figure out what this cloudiness was, we put it under a microscope and we found that when you mix these two things together, you get these little liquid droplets. And we could watch them float around in solution, we could watch two of them come together and fuse into a larger droplet. So for example, this, if you can see there in the back, this sort of figure eight is almost certainly two droplets that are in the process of fusing when we snapped this image. And moreover, the proteins become hugely concentrated in these droplets. So if we label either of them with a floor or floor, we could get an image like this, which is matched to the bright field image there. And you can see that the green fluorescence, in this case, I believe, of n-wasp, is much, much higher than in the surrounding liquid. And approximately in this case, about a hundredfold increase in concentration in the droplets than in the surroundings. And we thought to ourselves, huh, you know, we thought we were just gonna make oligomers, tetramers, octomers maybe, but we've made this weird structure. And what could that mean? Well, of course, while we were doing all of this biophysics and structural biology in my lab, we were doing our best to follow the cell biological literature on these molecules. And what we knew is that a lot of people had shown that when proteins in this whisked altered syndrome protein family, this wasp family, when those molecules are used in cells, very often if you label them with floor floors, you can see they make these very bright foci. They get assembled into some kind of high concentration structure. And those structures, it's easier to see here with these invadipodia are sites at which actin is assembled to cause changes in cell morphology and cell adhesion. And so we thought, well, you know, we have this unusual process of forming these liquids of this phase separation that concentrates molecules. Maybe this process is what's going on here in cells. Could this be a mechanism to explain that process? And so we started to think, you know, how could all of this occur? And eventually we made a connection to polymer physics and polymer chemistry that I think has really given us a very strong grounding into how some of these cellular structures may form. And this literally dates back to polymer chemistry, actually from steam box time. This is polymer chemistry that was Paul Flory, Walter Stockmeyer back in the early 1940s, you know, after some of steam box major discoveries, but still coincident with his time here. Because those pioneers in polymer chemistry asked the following question. They said, if I had some binary interaction to make a complex. And at the time they were starting the reactions of alcohols with carboxylic acids to make ester bonds. But all of the theory they developed, all of their experiments could absolutely be generalized. So here, you know, red circle interacts with a green circle to make two concentric circles would fit their theory. They asked what happens if those interacting entities are presented in multivalent arrays. So for example, if we have a trivalent version of the red guy and a divalent version of the green guy. And what they found, both with elegant quantitative experiments, but also really beautiful analytical theory, is that molecules of this ilk undergo what are called sol-gel phase transitions. Now, the name they chose for these was terrible because one is really not looking at a transition from a soluble thing to a gelatinous thing. This is a connectivity transition. That's important for people to understand the difference. What they found was, at low fractional binding or in a non-covalent system, as I'm trying to indicate here with concentrations on the two axes, at low concentrations, these molecules will assemble into small complexes, dimers, trimmers, tetramers, but basically small oligomers. And back in the systems they were looking at, those were soluble in solution, so they called this the sol phase. But as one increases fractional binding or increases concentration, there's a very sharp line beyond which these molecules will assemble into huge macroscopic polymers. And in the day, the polymers they were looking at were gelatinous in solution, so they called this a sol-gel transition, but as I said, it is in fact a connectivity transition from small oligomers to huge polymers. So anybody who's ever poured in acrylamide gel, basically you start monocrylamide and bisacrylamide and you start them polymerizing, you're making polymer chains that are gradually cross-linking, the solution gets more and more viscous and then instantaneously, if you don't pour it into the caster fast enough, the whole thing solidifies into one gel. You just did a sol-gel transition. Okay, so I think people can sort of relate to that. And again, they were looking at acids and alcohols making polyester, but all of this conceptually should hold every bit for proteins or macromolecules as well. Now, what they also developed, a little bit later in the 1940s, was an idea about solubility of these systems. And the idea as follows, the key concept rather is that as polymers get longer, they become less soluble in solution. And so if you have an eppendorf here with a solution of monomers in it, a little meniscus at the top, as those monomers start to polymerize, assemble into larger oligomers, they become less soluble. And eventually there's some critical length that's achieved that decreases the solubility to the point where they crash out into a separate phase. And since polymers are dynamic molecules, that phase tends to be a separate liquid phase. Now, for a non-covalent system, these two processes are thermodynamically coupled, such that if one starts with, let's say, some concentration of the red molecule here in solution and one titrates in the green molecule, as you add more and more green and move off to the right, the complexes that one makes become larger and larger. And at some point out here, they can reach the limit of their solubility too long to still be fully soluble in solution. And they crash into a second phase and they become highly concentrated in that phase. And so within that phase, they can jump into this regime so that now they've undergone a sol-gel transition in addition to a solubility transition. And I hope that this is clear to people. I've always found it to be sort of beautiful physical chemistry. And so the basic idea is that multivalent interactions can drive polymerization, move you across this plot to the left. And that, as oligomerization increases, solubility inherently drops. And at some point, one can then undergo phase separation to produce a separate concentrated phase and within that phase, you can move into the gel regime. And again, all of this was beautifully worked out all the way back in the 1940s. What they didn't know back then is that proteins consisted of multiple modular domains and that proteins might be able to do this. But when we made this connection, we started to think, maybe this is what's going on. Certainly, this is likely what's going on in our test tube and perhaps this could explain what's going on in cells as well. Now, science benefits tremendously from serendipity. So just a few months after we had these initial discoveries, there was what's now viewed as absolutely seminal paper in this field by Cliff Brangwin, who was in a postdoc in Tony Hyman's lab. So Brangwin and Hyman really get tremendous credit for this. They were looking inside C. elegans' oocytes and found that P granules, which was a compartment that was known, a member in this compartment at the time was known, they found that those behaved as liquid droplets. So for example, they could see cycles of fusion and fission when two P granules hit each other. They fused into a larger one. If they applied force to the cell, they could see some of them get pulled apart and separated. If they applied shear force in a particular direction, they could see a P granule kind of smear out in the direction of the force. And when the force was removed, they would round back up again very much the way liquids should behave inside the cell. And they could see rapid exchange of these components with the surroundings by photobleaching, all consistent with liquid-like structure of these P granules. In addition, the assembly and disassembly processes were suggestive of a critical concentration, sort of sharp line between one phase and two phases that I indicated previously. And so what they proposed is that P granules in the C. elegans embryo basically behave as base-separated liquid droplets. And they showed similar behavior just a couple of years later for the nucleolus in Xenopus suicides. And so when we saw actually the first of these two papers, we asked ourselves, well, what are the molecules within these structures? And I literally didn't even know these things existed before then to show my level of naivete. But sure enough, when we looked in them, what we found is that they are highly enriched in multivalent macromolecules. So for example, molecules that contain multiple RNA binding domains that can interact with repetitive elements of RNA. Many of them also contain so-called low-complexity sequences or amino acid sequences that are enriched in small numbers of different types of amino acids. And those can be, as I'll say more in a second, weakly adhesive. Moreover, I knew from my time in Tony Pawson's lab as a postdoc that there are lots of signaling molecules that are composed of repetitive modular domains. And so when we saw the Branglorn-Heimann paper and integrated it with our work, we really started to think, well, perhaps this process of phase separation doesn't explain just these actin-based structures, but perhaps many of these other kinds of structures, and perhaps many of these higher order things and cells writ large. And so that then led us to this model that we published in our first paper in this area in 2012, where essentially it is concomitant assembly and phase separation of multivalent molecules that could perhaps underlie formation of these, again, structures that we now call biomolecular condensates. Again, just two generic molecules here that are multivalent can interact and form essentially an oligomeric, polymeric network with decreased solubility and therefore a separate phase. Now, since that time, there have been lots and lots and lots of instantiations of multivalency-driven phase separation. So we've done a fair bit of work with multi-domain proteins, both engineered proteins, where we take a single domain and make repetitive versions of it, but also natural proteins that I'll say a lot more about that control the actin cytoskeleton. Other people, for example, Min-Ji Jung in Hong Kong has looked at the protein components of the postsynaptic density, which, again, are it's a whole collection of multivalent macromolecules that interact and shown that they can assemble and phase separate. There's been a tremendous amount of work done on these so-called intrinsically disordered regions of molecules, these polypeptides that don't adopt a discrete three-dimensional structure. And many of these have been shown to phase separate through a variety of different kinds of weakly adhesive interactions, cation pi interactions, charge-charge interactions. My colleague, Steve McKnight at UT Southwestern, has done beautiful work showing that some of these disordered regions can form short so-called cross-beta structures that, again, can occur in multivalent fashion down a disordered chain. And we think all of these are, in many ways, analogous to what's going on with these multi-domain proteins. It's a bunch of adhesive elements strung together on flexible chains. Just in this case, the interactions are of higher affinity involved, discrete binding elements. Here, these are much smaller, weaker interactions, but nevertheless, multivalency is sort of the essence. RNA and DNA through repetitive base pairing elements will also undergo this process of phase separation. A point that's going to turn out to be relevant to my seminar tomorrow is that polymer chemistry tells us, actually, I would say soft matter physics in general, tells us that all of these ideas about multivalency-driven phase separation are completely independent of scale. And so a real neat piece of work came out of Pietro Di Camilli's lab at Yale, where he showed that synaptic vesicles, which are, of course, way, way, way bigger than any of these molecules here, that if you coat synaptic vesicles with multiple, weakly adhesive molecules, they will assemble and produce a liquid phase where you see dynamics and you see exchange. But the basic component of it is a vesicle, something much, much bigger than these. But the basic underlying concept is really the same. And that becomes relevant again to my seminar tomorrow, where I'll talk about chromatin through this lens. Now, one of the very nice things is because there's 80 years of polymer chemistry behind all of these ideas, one can turn to polymer chemistry to think about things like regulation. And so what polymer chemistry told us is that higher valence, the larger the number of adhesive elements in a given molecule, higher valence or higher affinity between the individual elements will promote ligamentization. I think that should be hopefully intuitive. And thus, they will promote phase separation. And so it immediately suggests that if biology could control the valence of the interactions, biology could control this process. And so we've shown that that is indeed the case with this system that, again, I'm going to increasingly say more about. This is this NCK molecule with the 3SH3 domains and the NWAS molecule with the 9 proline rich motifs. In this case, these two molecules, when you mix them together, will phase separate. I showed you droplets from that. But you've got to go to pretty high concentrations. You've got to go to sort of double-digit micromolar concentrations, which is above biological relevance here. But so you can imagine at sort of meaning biologically meaningful concentrations, one could be below that phase boundary. However, NCK is a molecule that has an SH2 domain that binds to phosphatiricines. And what was kind of cool is that when we looked at all of the known NCK binding proteins, over half of them had three or more phosphatiricines that bound NCK. And so when you array NCK, then, on this scaffold, you have bumped up the valence of the SH3 domains, in this case, from three SH3s to nine SH3s. That shifts the phase boundary down such that concentrations that didn't phase separate here now do phase separate. And so the actions of kinases and phosphatases, which put on phosphates and take them off, could then be used by the cell to control where the phase boundary is and thus control whether the system's in a one phase regime or a two phase regime. We have a lot of data there showing those ideas. Moreover, there are many, many different covalent modifications of intrinsically disordered regions that can be used to increase or decrease the affinity of interactions between molecules. And those have been shown, I won't run you through the specifics, but those have been shown to either favor phase separation or disfavor phase separation. Moreover, again, all of this beautiful work from decades old polymer chemistry also can lead us to models that explain the composition of these structures. And so we put forward an initial very, very simplistic model in 2016 where we grossly divided the components of a bona fide cellular condensate into two groups of molecules called scaffold molecules were the ones that were highly valent whose interactions would give assembly and concomitant phase separation. And then, in other words, these are the molecules that create the structure. And so what one can see in terms of genetic experiments, these are the molecules that if you delete them, this higher order assembly goes away in the cell. In addition though, there are a lot of molecules that we know are concentrated in different condensates that are of low valency. And those get enriched in the condensate either because they can directly bind to the condensate of the scaffold. So for example, this one here, the green and the red molecule can bind, for example, or say the green and yellow molecule can bind to one of the blue sites in the scaffold. You can also have client molecules that bind to sites that are not involved in assembling the structure itself. I'd imagine those have, the condensate responds differently to high concentrations of the two of them. Also what we're finding is that there are molecules that can get into a condensate simply because they are physically compatible with the physical properties of the phase. And I'm deliberately vague on exactly what that means. All we know at this point is when we've looked at groups of small molecules, there are many small molecules that will enrich and vitro in a phase-separated condensate that don't have obvious binding sites on any of the components in there. So we think it's gotta be something about just the physical chemical properties of that condensed structure that's more favorable to some compounds than others. And again, we're trying to figure all this out now. What we have come to understand though over time is that this notion of scaffolding client and the distinction between them is absolutely not binary. We even intuited that earlier and that is more precisely one should talk about molecules that contribute to different degrees to forming the structure. So we now think about these as more scaffold-like. But you can imagine you could start with let's say a pentavalent blue protein and say what happens to the tetravalent, to the divalent, one can convert things from contributing strongly to contributing weakly. And so it's probably more correct to think about things as being more scaffold-like or more client-like, but this notion of a binary distinction is certainly not right. Not right exactly. And moreover, one shouldn't think about these in universal terms. In other words, a molecule that can be a scaffold under some conditions could turn out to be a client in others. For example, in the presence of high concentrations of this small client, the blue sites can be sucked up and not able to interact with the multivalent yellow molecule as well. And so under those kinds of conditions, the multivalent blue protein would behave more like a client and less like a scaffold because of the competitors. So again, one has to think pretty carefully when one analyzes, for example, biological data where condensates contain many, many dozens of molecules and the effects of perturbations. Moreover, it's important to note that these are very complex fluids. So the analogies that truthfully people like me and others in the field have made are like, oh, this is like oil and water. That's pretty good to explain to the person sitting next to be on an airplane who happens to be a lawyer or something. But to scientists, the oil and water analogy is a terrible analogy in that oil is a very simple fluid and the oil droplets in your salad dressing are homogeneous. These are absolutely, these are, in fact, we're finding more and more not homogeneous, they're heterogeneous at small length scales. And moreover, the different molecules that are within them behave very differently. Scaffolds are connected to other molecules through multiple binding sites. Clients, it's only a single binding site or a small number of binding sites. So their dynamic behaviors and their responses to change in viscosity, for example, will be different between different molecules. And so one has to pay pretty close attention when one makes perturbations to what kind of a molecule are you perturbing and what role does it play in creating the structure if you really want to interpret it. And I think that's something that's becoming more and more obvious as people study more of the biological systems and get away from sort of more of these toy systems that so many of us, including me, have been studying for a while. So with all of that as kind of introduction and background to the field, hopefully framing the problem both for today and for tomorrow, I want to spend the rest of my time talking about sort of that fourth issue on my initial slide and that is what a condensate to do, right? And that is what biochemical functions specifically arise from phase separation, right? Not just this condensate, must be important in signaling because it concentrates the kinase, but like what happens within that structure that makes the kinase somehow different? Is there anything that is different? And a core layer of that is are there any unique modes of regulation that are enabled specifically by the properties of these phase separated structures? And so I'll tell you sort of a longer vignette and then a shorter vignette to get at these next set of questions. So this takes me back to that pathway that I've been talking about on and off here. It's a pathway that controls actin assembly and the putacyte cells of the kidney starts from a single-pass transmembrane receptor called nephrin. That's a huge extracellular domain I'm not showing. Single-pass transmembrane helix and then disordered cytoplasmic tail that's phosphorylated at three sites that bind to the SH2 domain of NCK. And NCK is this adapter protein with three SH3 domains that then bind to the nine proline rich motifs in this n-wasp protein. And what my postdoctoral advisor, Tony Pawson, had shown is that when nephrin is assembled, concentrated at the surface of cells, these punctiform and they basically assemble very potently long actin tails from them. And since n-wasp activates the ARP23 complex to make actin film, it's basically, this is a system to look at how signaling pathways can control the ARP23 complex. And of course just looking at these at the outset, this seemed to us like a system that would phase separate. And so sure enough, if you mix these three proteins in solution, you can robustly induce phase separation. Moreover, if we attach nephrin to a supported lipid bilayer, so a lipid bilayer attached to a glass slide, if we uniformly distributes, but if we add NCK and n-wasp to it, it assembles into these micron scale clusters that have properties very, very much similar to these three-dimensional liquid droplets that we see when everything's in solution. And moreover, if we target nephrin to the plasma membrane of the HeLa cell and trigger its phosphorylation, we very rapidly see analogous punctate structures form that contain both NCK and n-wasp and assemble actin filaments. Now we could study actin assembly in this system by again putting nephrin, NCK and n-wasp on a supported, sorry, by putting nephrin on a supported lipid bilayer, adding NCK and n-wasp to it, and we can see formation of these micron scale puncta. If we add actin and the ARP23 complex to it, what we can see is that the actin rapidly assembles on the membrane and moreover, it assembles essentially always, at least as far as we can see here, on the puncta. And so the question is, why is it that actin preferentially assembles on the clusters? And there's kind of two potential answers. And one of them is sort of, well, duh, that's where the signaling molecules go. Of course that's where the actin assembles. But the other one is more interesting, and that is perhaps within the condensate structure, there's some change in the specific activity of these molecules that enhances their activity toward the ARP23 complex. And so a way to distinguish these would be sort of the following thought experiment. Let's say we have a membrane here that has these dots are supposed to represent individual, let's say, NWAS molecules. And we have a condensed region here that has, we'll say four molecules, but again, the condensates we're forming have hundreds and hundreds, but let's say we have one that's small that has four molecules here. And we have another region in the membrane that's sort of the surroundings, and let's say that has only one molecule. And we can then measure the actin assembly activity here and here, and maybe what we would see is that in the condensate it's higher and in the surrounding membrane it's lower, be consistent with previous data I showed you. Then if we normalize by the number of molecules there, the number of NWAS molecules there, we can distinguish these two mechanistic possibilities because if the reason there's higher activity in the condensate is just because there's four times as many molecules there, when we divide by four, the condensate activity should decrease and become identical to just the surrounding regions of the membrane. If however, something has changed, the specific activity of the molecules here are higher than here, even when we divide by the number of molecules we'll still have higher normalized activity than the surrounding membrane to a degree by which the specific activity has changed, right? Now, the way we're doing these experiments we can't measure number of molecules but we can use as a proxy fluorescence intensity. So if we normalize activity by fluorescence intensity we should be able to distinguish these two mechanistic possibilities. And my postdoc Lindsey Case was able to set up a very nice quantitative assay where again we can make clusters of this signaling system on a membrane, add actin in the arc-2-3 complex to it, see actin assemble, and then for any given spot in the membrane we can look at how rapidly actin assembled there. And so what we see is that inside a punctum we can see a rapid assembly and outside we can see much slower assembly. We can quantify this by looking at the maximum slope of these curves. And then the critical thing is we can take that slope and we can divide it by the local intensity of n-losp at that site, okay? So look at specific activity and even when we do that, so this is now normalized polymerization rate, we see that inside the condensate is appreciably faster than outside. So that means the activity per signaling molecule is higher inside than outside. So it's not just mass action, it's not just a concentration device. Something is different within the condensate. So the question is what? So I think we answered one question but then we set up, I think, an even more interesting mechanistic question, why does specific activity go up? So the first thing that we thought of was perhaps just there's some non-linearity in activity as a function of density. And I could explain afterwards why we thought that might be the case, completely wrong. Because we were able to attach n-losp to a supported bilayer just by itself and look as a function of density at this normalized polymerization rate, it's basically completely flat. And so it's not as though something about just the density of n-losp molecules gives non-linearity such that specific activity goes up at higher density. But we were lucky then to interact with Jay Groves, who's a single molecule microscopist at UC Berkeley. And Jay was studying very much an analogous system where he had a membrane protein called LAT that was phosphorylated on multiple sites. That interacts with a protein called GRAB2 with an SH2 domain, in this case two SH3 domains. And that interacts with a protein called SOS that has a proline matrix. It's an absolutely analogous to nephrin, NCK, n-losp, but just different set of molecules. And what Jay showed is that if you look at the dwell, the membrane dwell time, in this case of the SOS protein, if you looked outside the clusters, that membrane dwell time had a distribution. If you look at single molecules and their lifetime distribution, those molecules dissociate from the membrane very fast. Inside the clusters though, SOS has this long, long tail of increased dwell time out to tens of seconds. And that's because within the condensates there's higher connectivity than outside. And I'll show you data on this in a moment. Moreover, they developed a theory that suggested that increased dwell time could enhance or should enhance downstream signaling if the signaling is slow, multi-step and somehow driven out of equilibrium. This is classical kinetic proofreading. So what they argued is that this system is analogous to kinetic proofreading where dwell time here is analogous to the off-rate in systems like the ribosome. And so when we started to think about our system we realized it had all of the ingredients. Activation of the ARP2-3 complex is slow. It takes tens of seconds from binding of an ARP2-3 complex to creation of a daughter filament. There's a whole bunch of different conformational changes that have to occur and binding interactions that have to occur. And moreover, there's an irreversible step somewhere in this where ARP2-3 complex permanently latches on to the mother filament. In biology there are molecules that have to take that apart. But just in equilibrium it's irreversible. And so we thought maybe membrane dwell time could be the thing that would regulate this system. And sure enough, what we found across a variety of different perturbations is that, again, these are the data that I showed before. Inside has higher specific activity than outside in a condensate. And moreover, we always saw that the dwell time was higher in the condensate than in the surrounding regions and across a lot of different perturbations. And so basically paralleling the grove's idea of what we came to understand is that NWASP has to remain at the membrane long enough for ARP2-3 compass to go through all these steps and get turned on. And that if it falls off in the middle, the system has to reset back to the beginning. And that essentially is a recipe for dwell time dependence of flux through the pathway. Now, one of the neat experiments that my postdoc Lindsey Case did that supported this idea and also comes up with interesting regulatory concepts is the following. So she asks, what happens if I have a fixed density of nephrin on the membrane, a fixed amount of NWASP's solution, and I titrate in NCK? So if you walk through this, at low concentrations of NCK, any given NWASP molecule is gonna be attached to the membrane through nephrin through only a small number of bonds. And so dwell time should be short. At the other end, with a huge excess of NCK, any given NWASP molecule, yes, is bound to many different NCKs, but when there's an excess, only a small number of those are actually bound to nephrin and therefore bound to the membrane. So at this end, there'll be low membrane connectivity, again, dwell time should be short. In the middle, when the component concentrations are more balanced, any given NWASP is attached to multiple NCKs and they themselves are attached to nephrin in the membrane. So dwell time should be high here. And so we should see dwell time and maybe bell shaped activity. And sure enough, that's what Lindsay found. So again, fixed NWASP, fixed nephrin density on the bilayer and titrated in NCK. And you see that the dwell time basically goes up and then down as you go to excess and really strikingly, if you look at specific activity, specific activity goes up and then down. And Lindsay was able to do this and look at a correlation between dwell time and specific activity and boy, certainly there's a very strong correlation. I'm not gonna claim linearity because we don't have any kind of specific molecular model to explain this, but certainly there's a real strong correlation here. And she was able to do a series of different titrations and all of them basically can be aligned with each other correlating specific activity in membrane dwell time. And so we think this is a very strong argument for the fact that A, dwell time is really the key parameter. B, it also has, I think, a really interesting conceptual point to make about how activity can be controlled. Because normally when we think of controlling specific activity, we think about allosteric changes that change the inherent chemistry of a system. And this kind of a system where a condensate can form across a wide range of stoichiometries, simply changing the relative stoichiometries can change specific activities and thus change the signaling output from a system. And so we think it really, this is a new means of controlling protein activity that arises from these kinds of stoichiometrically undefined assemblies, whether they form through phase separation or not actually it doesn't matter at all, just they have to be able to form through a lot of different stoichiometries. This is something that we can observe actually in cells too. Lindsay rigged up a beautiful cellular assay where she can simultaneously image nephrin, NCK, n-wasp, as well as actin filaments. Just an example, here's a cell that contains these foci and there's one this punta that shows all three molecules in it. And again, she was able to photobleach these as a function with different NCK intensities and again see this bell-shaped fluorescence recovery. Also looked at actin intensity as a function of NCK and again it's bell-shaped and finally we can plot dynamics against actin intensity and we see again a reasonable correlation there. So both in vitro and in this cellular system, we think this idea that relative stoichiometry can control specific activity is going to be relevant. So we think that cells could use this as a way of fine-tuning a fundamentally switch-like system, in other words. One could imagine the cell in a single-phase regime and actin assembly activity is low. Signal comes in, causes assembly of the signaling molecules into a two-phase regime into clusters and then activity jumps up from here to here, but then within this two-phase regime, the cell could adjust the relative stoichiometries of the molecules or the degree of phosphorylation turns out to be functionally equivalent and then change activity within this sort of higher activity envelope. Of course, one could imagine cells needing to be able to make a switch-like decision but not wanting to give up complete control and this could be a way of controlling at the high end of the switch. So just to conclude this, we've shown that phase separation increases membrane dwell time and when downstream processes are slow and driven out of equilibrium, activity then will correlate with dwell time and because of stoichiometry's ability to control membrane connectivity, stoichiometry controls dwell time and thus stoichiometry controls activity, which is, we think, a new way of looking at control of signaling envelope. The last thing I'll note is that these behaviors are very likely quite general. There are lots of signaling systems that are composed of collections of multi-valent macromolecules. Many of them require activity through slow multi-step processes and so this should be something that's widely seen. In fact, Jay Groves showed exactly the same behavior in the paper that we ended up publishing back to back with the RASMAP kinase system, sort of the lat grab two sauce RASMAP kinase system I just mentioned before. So at least for an N of two, that's general. So I see that I have run out of time. I will not run you through the next vignette and you're gonna laugh a little bit that I thought I could get through all of this, but I'll make one quick point. We've looked at another system where we have recruited enzymes into a phase-separated compartment. Doesn't matter exactly what those are or what the compartment was. But in that case also, we have shown that condensates don't merely concentrate molecules, that there are changes in activity beyond that and in this other system, it turns out that within the condensate, the KM of the key reaction step decreases. And so we see higher activity, not just because the molecules are concentrated, not just because of mass action, but because of this change in KM. Now, we've only looked at these two systems in quantitative fashion. And in both cases, we found that the condensate doesn't merely assemble molecules, doesn't merely concentrate them, but actually changes elements of their reaction that further modulates activity. And given that sort of randomly chose these two, not quite randomly, but these are just two out of many, and we're not chosen for that specific effect, my bet is that this is quite general. That many different condensates that concentrate different enzyme activities, enzyme cascades are going to do things beyond merely bringing molecules together, that there will be elements that further enhance or further repress activity. And I think that's one of the real exciting challenges for those of us in the field to sort out. So I'll just finish to go back through those initial questions that I laid out. And lay out our idea that multivalency-driven phase separation, liquid-liquid phase separation, may drive the formation of some biomolecular condensates. An important point to make is that I don't want to claim that every dot you see in a cell gets formed this way. I think there's actually too much literature on that. Like, oh, we see a dot, it's formed by phase separation. Clearly, there are other ways of bringing molecules together and to foci you can see under a light microscope, by some smiles in the audience. I don't want to over claim this, but I do think that there's good evidence for at least a few of these that this process of multivalency-driven phase separation is relevant. And my guess is over time, as people study more and more of these in quantitative detail, some of them will look, oh yeah, that's the mechanism, and others will be, no, it's something different. But for those that do form through phase separation, I think we've gone a long way to answering these sort of initial questions that we laid out almost a dozen years ago. That is, what is their physical nature? Well, they are, again, for the ones that form this way, they are phase-separated polymers of multivalent proteins in their ligands. This would allow them to be both persistent, but also exchange molecules with their surroundings. What's their chemical nature? The composition can be controlled by interactions among scaffolding, client-like molecules, and therefore the interaction network between the various species within them. I showed you an important, we think, mechanism of regulation by changes in valency and affinity through covalent modifications, and there are many, many examples of this, both in vitro and in vivo now. And finally, one of the biochemical functions we're just, I think, beginning to scratch the surface here, but we've shown that reactions can be accelerated by mass action just by concentrating molecules, but also somehow modulated by various physical properties of the polymer matrix. And again, my bet is that turns out to be general as well. So I'll leave you with those ideas and just thank the wonderful collection of people I've had the good fortune of working with over the years. All of our initial work, sort of laying out some of these early concepts, came from a collection of really four graduate students in the lab. Salman Binani, Sadeep Manjade, Weichen Chang, and Pilong Li. Weichen and Pilong are now running their own laboratories and Sadeep and Salman are both postdocs and going on the job market now. The story of dwell time dependence was really exclusively done by a terrific postdoc, Lindsay Case, who just started her own lab at MIT about a year ago. The story I didn't make time to tell you about on the other enzyme system is a graduate student named Ben Keeples. I mentioned Jay Groves, who's a wonderful colleague and collaborator on all of this. There is the rest of my group earlier this spring as the mask mandates started to go down. We felt like we could get together at least outside. So I thank all of them for the really wonderful work in this area. And again, it's a real privilege to be here and to be able to give the Steenbach lectures and have the opportunity to interact with all of you. So thanks very much for your attention. I'll be happy to take any questions. Yeah. Good. How small do you think the collection of client molecules and scaffold molecules can be and still give the bend? Yeah, sorry. I wondered how small do you think the collection of scaffold molecules and client molecules can be to still give the benefit of phase separation in terms of dwell time? I mean, does it have to be tens of thousands of molecules or could it be a couple of dozen? Invite me back in a year. That's a terrific question. My background is structural biology. I think of tetramers and hexamers, right? And when do you go from a molecular system that properly should be understood through what we've all known in love as small discrete numbers of oligomers and at what size do you start seeing we'll say emergent properties? And I mean that in very specific terms. In this case, when does the dwell time go up? How big does it have to be? And this is a question that's relevant in lots of these systems. It comes up with transcription a lot. There's a lot of contention in transcription whether one should think about transcriptional foci as just a collection of a small number of molecules or sort of the phase separation thinking the better way to look at it. The answer is gonna turn out to depend on the specific nature of the system, the interactions between the molecules, the function that you care about. We are just starting to look at this with the actin system. We're making DNA origami of different sizes and putting nephrin on them and we will ask when does dwell time go up? But I think this is again, you've hit on sort of one of the forefront questions in the field which I think is great. We'll get to it eventually. I hope that long answer to a short question. So you presented this as a really fundamental process and I really liked that. It made me wonder about the evolution of this. Has anyone begun thinking how far back in the history of life do you begin seeing these types of concatenated domains and potential properties? Yeah, yeah, yeah. So people think about them in prokaryotes. Obviously the imaging gets harder as you get to something that's only two microns long. But there's certainly evidence of multivalent molecules in prokaryotes that will in vitro will phase separate or form higher order assemblies and form at one pole of the bacteria. So it seems like this is something that's very old. And in fact, if you look back in the physics literature, there were physicists who were arguing for phase separation as an important concentration mechanism back in the RNA world long before any of this sort of biological stuff really blew up over the last decade. So yeah, people are certainly thinking about it and the idea is that this may be beautiful for sure. Thank you so much. I was actually wondering is it possible to maybe capture like a structure? Because I know you said you had a structure background. Have you thought about trying to capture the structural like formation of how these coordinates look like? Because I'm assuming there's like some conformational change that occurs during this transition period. Yeah, come to my talk tomorrow, I'll show one image. We've begun looking at chromatin, kind of say phase separated chromatin and trying to understand the organization of nucleosomes within them. And it's really too early to answer your question, but short answer is yes. We're very, very interested in that. Polymer chemistry would predict that in order for a polymer to phase separate, it's gotta be in called a solvent regime where it collapses on itself at low concentrations. And what polymer chemistry predicts is when you get a lot of those collapse molecules together instead of being self-satisfied, they're satisfied interactions with others and they extend out. And so there's an anticipation that macromolecules, disordered chains at least, will be more compact outside of the condensate and more extended inside. But again, people need to do the analysis. Hi, great talk. Yeah, so I was curious about your metabolite kind of foray into that area. Do you have any clues about whether proteins that metabolize whatever you're detecting if they're nearby or if it's dependent on how concentrated the metabolite is in your system, whether you're seeing it in a condensate? And then I guess related to that, how do you even get a clean pull down or whatever of these condensates? Yeah, so you're asking about small molecule metabolites? Right. Yeah, so essentially, do we know this, if I understand correctly, do we understand anything about the small molecule composition? I guess the ones that you mentioned are appearing that you don't necessarily have any interactions with, with the... Oh, I see. The molecules that get in there without an obvious binding... Exactly, right, yeah. Yeah, so I'll tell you what we're doing. We're not doing that with natural condensates. We're doing that with synthetic, these engineered systems. And what we've done is we've added either a metabolite library, that we get from a metabolomics core facility, it's what they use to calibrate their instruments, or we've added an FDA approved drug library of about 1,200 compounds to them. And we've developed a mass spec assay that allows us to measure the partition coefficient. What's the concentration in the condensate and what's the concentration in the surrounding bulk? And what we found is that there's a surprisingly large number of compounds that enrich in the condensate and up to like 1,000 volts, so very high degrees. And when we've tried to use ITC to measure binding affinity between those compounds and the proteins that are in there, we don't see strong interactions. And so that's, I was deliberately vague on that. That's why we think it's probably some matching of gross physical properties of now thinking about this as a liquid droplet, rather than, oh, it's binding to something specifically there. But obviously those two effect, for a molecule that does bind, we think it will be both the physical properties and the direct binding that matter. And of course, you could imagine that biotech companies thinking about targeting condensates care a lot about that, so we're hopeful that we'll be able to sort of reveal physical principles of that eventually. Very cool, thank you. So maybe we just step, oh, was there one more? Oh, okay, sorry. Hi, so I had a question from what I got from this talk was that you made it seem that more so than the clustering effect, the biophysical properties created from the condensate is what causes change in like protein activity. So I thought toward bio-catalysis, if we understood what the biophysical properties of the condensate was, would it be possible to place these proteins directly in a solution that mimics those biophysical properties so that you're always actively getting the boost that the condensate provides without actually ever having to form the condensate? Yeah, we're hoping eventually we'll be able to get to that. So the idea is if we can figure out, for example, the compound that concentrates a thousandfold, if we can figure out what physical features of the droplet are the ones that allow it to concentrate in there so strongly, could we just make a solution that mimics those and if that concentration changed activity, would we see that change in activity? My guess is we are at least a PhD thesis or if not more away from having that understanding, but again, that's what I'm hoping to sort of the next period in my program. For sure, it's a great question. Hi, so when I look at these pictures of condensates, I always wonder what's in between them and how does what is in between them affect the activity within the condensate? So for example, if you generate a condensate in one cell type with the same molecules, will it behave the same if you generate it in another cell type that has a different kind of cytoplasm, say more ER versus, I don't know. Has anybody started playing those games? So people have done artificially. So Ned Wingrin and Zemriga Tai had a paper, actually a few years ago now, where they artificially created condensates by coupling, basically co-assembling two enzymes in a cascade and they were able to show and it was, so it was a branch point enzyme and two downstream enzymes that give different products and they showed that if they coupled these two together in a condensate that they could make essentially a functional oxytrofe for the other one. In other words, all the flux went down this pathway and none the other. The way they modeled it involved looking at diffusion of molecules from one condensate to the other. So the size of the condensate and the spatial distribution of the condensate and the flux through that compartment essentially outside of the compartment all played into the degree to which you got this arm going and not that arm going. Which is to say, yes, different cytoplasm or cytoplasm versus nucleoplasm, for example, with different components could make a difference. But to my knowledge, that's the only paper that's really looked at sort of those spatial issues. This is another kind of question. So if you were to take something like a P granule and put a metabolic enzyme into it, would the activity change? My bet is yes. We haven't done that yet. But in fact, we are reconstituting P bodies in vitro. And one of the things that we wanna look at is the activity of the P body components. But then the story that I didn't have time to show basically we rigged up an enzyme assimilation cascade to go inside a condensate. And the question is if you then took some exogenous enzyme or enzyme cascade, would it also change? We think that condensates tend to be more hydrophobic than the surrounding region. Because it is hydrophobic molecules that concentrate more strongly. And so reactions that respond to that, if they're in the condensate, could be accelerated that way. Again, these are on the docket of things to do. Interesting questions. Any more questions? All right, well maybe we should just continue the discussion outside for the reception.