VIDEO: Jonathon Ditlev on Disease-linked mutations in shank2 result in aberrant condensate physical properties, composition, and function

Condensates.com welcomed Jonathon Ditlev, PhD, Scientist, Molecular Medicine and Cell Biology Programs, The Hospital for Sick Children, Assistant Professor, Biochemistry, University of Toronto to deliver our second Kitchen Table Talk of 2025. During his talk, Dr. Ditlev discussed Disease-linked mutations in shank2 result in aberrant condensate physical properties, composition, and function.

Dr. Ditlev has done extensive work studying phase separation and neuronal and immunological signal transduction systems. He is a biochemist and biophysicist with a PhD from the University of Connecticut and did his postdoctoral studies at the UT Southwestern Medical Center, where his work centered on phase separation and the stoichiometry within as a way to explain signaling activity in actin filament clusters and signaling during T cell activation. In his current role, he continues to do groundbreaking work to study the association between phase separation that is at the level of biochemistry or biochemical reconstitution, but also cell biology in membrane receptors and signal transduction both in neurons and immune cells. In addition, his team now looks at the effect of phase separation on local RNA translation and interactions between different condensates.

Click here to watch the engaging discussion and read the transcript below.

TRANSCRIPT

Diana Mitrea: So welcome, everybody. Good morning, good afternoon, good evening, depending on where you’re joining us from. I’m Diana Mitrea, Head of Scientific and Corporate Communications here at Dewpoint Therapeutics, and welcome to Kitchen Table Talk number 39. We welcome you here, together with my co-host, Kamran Rizzolo, Associate Principal Scientist in the Chemistry and Drug Discovery department at Dewpoint, and we’re here in the Kitchen Table in Boston.  

Before we begin, I have a few housekeeping announcements. This is a live event, so please be courteous. Keep your microphone muted, keep your camera on if you’d like, and nod along with the speaker. Enjoy the talk. During Q&A, we’re going to call you and ask you to unmute. At that point, please unmute your microphone and ask your question. Ideally, we would keep the questions to the end, or at least to the breaking points that Jon will have during his talk. After every section there’s going to be a break for Q&A. 

Without further ado, I’m going to pass the baton to Kamran to introduce today’s speaker. 

Kamran Rizzolo: Thanks very much, Diana. It is with great pleasure that I get to introduce today’s KTT speaker, Dr. Jon Ditlev. I first met Jon during my postdoctoral studies back in 2019 at the University of Toronto. I met him actually at the official Condensate Journal Club, which back in the day, was a room with 10 to 15 people, and we would meet once a month. So now, six years later, my old department, I realized, has a whole class called Biological Condensates, which Jon teaches, and also other folks that are in the Journal Club. So, it’s a reflection of not just how much the field has progressed, but also how it’s evolved in my old department, too, which is very cool. 

As we’ll hear today, Jon has done extensive work studying phase separation and neuronal and immunological signal transduction systems. He’s a biochemist and biophysicist by training with a PhD from the University of Connecticut. Afterwards he did his postdoctoral studies at the UT Southwestern Medical Center with Professor Michael Rosen. His work there centered on phase separation and the stoichiometry within as a way to explain signaling activity in actin filament clusters and signaling during T cell activation. 

Jon’s current role is as an assistant professor at the Department of Biochemistry at the University of Toronto, and he also holds a position as a scientist at the Hospital for Sick Children. There he continues to do groundbreaking work to study the association between phase separation that is, at the level of biochemistry or biochemical reconstitution, but also cell biology in membrane receptors and signal transduction, both in neurons and immune cells. In addition, his team now looks at the effect of phase separation on local RNA translation and interactions between different condensates. So, I’m excited to hear more about the evolution of this story. Without further ado, please join me in welcoming Dr. Jon Ditlev. The floor is yours, Jon. 

Jonathon Ditlev: Kamran, thank you for the very kind introduction, and Diana for organizing this talk, and also for everyone that’s here for taking time out of your day to hear some of the work that’s going on in my group. And so today, the goal that I have is to talk about some of this work that we’ve done looking at this postsynaptic density protein shank2 and how missense mutations in its intrinsically disordered region can affect some of the biophysical properties and the function of the condensates that form. 

But before I begin, I’d just like to acknowledge the people who actually do the work in my group. Specifically, Dr. Liyanage, who spearheaded this project shortly after joining my group as a postdoctoral fellow, Steve, Gaddy, Xusheng, Victoria, Ana, and Sam have all done work as well, assisting Leshani and her project. And then, of course, thank the funding agencies that fund our work. 

For people that are joining us that are not as well versed in condensates as much of the audience, I thought I would just talk a little bit about cellular organization. And so historically, cellular organization has been thought of as being the cells organized by membrane bound compartments, such as mitochondria, the endoplasmic reticulum, the Golgi. But more recently, in about the past 15 years, there’s been a lot of work that’s shown that there are many non-membrane-bound compartments that also form, and these are termed biomolecular condensates. 

And as I want you to see, they can form in the nucleus, in the cytoplasm, but also on membranes. So even though they’re non-membrane-bound, they can be membrane associated. And just as some examples of well-described condensates include the nucleoli PML bodies, P body, as well as T cell receptor clusters, which I’m going to use as an introduction to membrane-bound condensates. 

So how do these biomolecular condensates form? Well, they usually form through multivalent weak interactions, although some interactions can be very strong as long as they’re highly dynamic. As an example, I’ve got the blue molecule, multivalent blue molecule and multivalent orange molecule here, and these can undergo two types of interactions. The first is a high affinity interaction, which the two biomolecules will zipper, and the result of that is for the formation of a heterodimer. However, they can also undergo these multivalent interactions and form a protein or protein nucleic acid interaction network. 

When interactions between the proteins or nucleic acids become more favorable than the solvent, these networks can undergo phase separation and the formation of a liquid within a liquid. Importantly, not all phase separation forms liquid condensates, as you’ll see later in my talk. There are condensates that form liquid rapidly, but then rapidly undergo a transition into a gel-like state. And this has been shown now for roughly about 15 years. 

Importantly, as I mentioned earlier phase separation can occur both in solution and on membranes. If we take this multivalent protein, that’s phosphorylated in three spots, bound by an adapter protein, which is bound by another adapter protein we can form condensates on membranes, if one of the proteins is attached to the membrane. If we take those same proteins and mix them in solution at high enough concentrations, we can form condensates in solution. One of the important aspects of this is that condensates on membranes form at about an order of magnitude lower concentration than in solution or in the cytoplasm. And this allows cells to organize condensates on membranes at much lower concentrations than are usually required, or usually we think of in solution. 

As an introduction to membrane associated condensates, I want to tell a story that Dr. Xiaolei Su and I spearheaded when we were postdocs. And this is how phase separation regulates biological processes at the T cell immunological synapse. At the synapse, we can see here a junction between two cells, the antigen presenting cell in blue, the activated T cell here shown in red. And then, where these two cells contact one another, this protein ZAP-70 localizes from the cytoplasm of the T cell to the membrane, and at the membrane ZAP-70 phosphorylates this adapter protein LAT, as you’ll see in a second, and that is thought to drive signaling at the immunological synapse. 

If we take an on-phos view using TIRF microscopy of a T cell and look at what happens when a T cell receptor contacts an MHC peptide, or in this case an antibody, we see that the T cell receptor that that binding activates kinases at the membrane those kinases, phosphorylate proteins, including LAT, LAT microclusters form. And if you watch the video, you’ll see that LAT microclusters form at the edge of the synapse and then are moved towards the center of the synapse and that this process is actually very important for T cell function, and the proper response of the T cell to antigen presenting cells. These LAT microclusters have been linked with calcium influx, actin polymerization, and also downstream Erk phosphorylation through genetic studies. 

And Xiaolei and I asked some questions, how do these clusters form? Is phase separation possibly the mechanism that drives their formation, and also is phase separation responsible for some of these functions that have been previously described in the literature? And so, genetic work has shown that there are many, many proteins involved in T cell activation. I’m just highlighting a few of the proteins here in the signaling pathway, particularly kinases that phosphorylate the T cell receptor in LAT, adapter proteins that bind LAT, as well as more adapter proteins that can activate actin polymerization. 

The first question that Xiaolei and I asked is, what are the minimum requirements for phase separation of LAT, if it is in fact phase separation? And so, what we did is we attached a fluorescently labeled LAT to a model lipid bilayer, and then we dropped Grb 2 and Sos into the solution. And when we did that, we saw the rapid formation of clusters, and further experimentation showed that these clusters are fully consistent with phase separated condensates. 

And so, we then asked, okay, so we can do this in vitro, but is this important in cell biology?  Because there’s a lot of biochemistry that we can do that has very little relevance towards what happens in a cell. And so, what we did is we took LAT knockout T cells. We re-expressed LAT with different numbers of tyrosines on it and then looked at whether these different LAT species could form clusters on the cell membrane as we saw in vitro, and what we found is a very high correlation between our cellular results and our in vitro results. And furthermore, we linked that condensate formation with Erk phosphorylation, showing that condensate formation is correlated with successful cellular signaling in T cells, and this will be very important for what I discuss later in the talk.  

We also found that LAT clusters could sort molecules. And so, in these experiments that were done on model bilayers, we formed LAT clusters and had the phosphatase CD45, which targets LAT in T cells also attached to the membrane, which is where it’s localized in cells. And what we found is that LAT clusters could exclude CD45 from their internal environment, thereby protecting LAT from dephosphorylation, allowing the clusters to be maintained on the membrane. And so, we did some further work, and found that this was an electrostatic repulsion, that CD45 is a highly negatively charged protein and LAT clusters are highly negatively charged protein clusters, and the electrostatic repulsion actually allowed for the sorting of these molecules. 

We also linked LAT cluster formation directly with actin polymerization. In this movie, you’ll see upon addition of ATP, ZAP-70 comes to the membrane, LAT clusters form, and then actin is polymerized from the clusters. Lindsey Case did some very nice work in a paper back in 2019, showing that it was actually the dwell time of these actin nucleation promoting factors, as well as Arp2/3, that allowed for the formation of these actin filaments at the LAT clusters, and that this was regulated through phase separation. 

Another thing that came out of this extended study that we continued looking at was, we asked how these LAT clusters are moved by the cell. More generally, how these LAT clusters interact with their local dynamic environments. And one of the first things that we noticed is that LAT cluster composition changes as it moves from the edge of the synapse towards the center of the synapse. And you can see this here LAT is present in the clusters, as it should be, from its inception to where it enters the center of the immune synapse. However, the adapter protein Nck seems to dissipate as the LAT cluster gets closer to the center of the synapse. 

So, we asked what function might that have? And so, we took a look at dynamic, active networks. And John Hammer did some really nice work with this in his group. And what you’ll see here is that at the T cell synapse and on phosphate of the T cell synapse there’s two dynamic, active networks. The first network is an Arp2/3 generated network in which dendritically branched actin filaments form at the membrane and then push backwards. 

There’s also a formin-generated actin cable network at the center of the synapse that telescopes inward, leading to directed motion across synapse. So, we asked whether this change in LAT cluster composition might allow the LAT clusters to be dragged from the edge of the synapse and then pushed towards the center. If you recall right at about this edge where the two networks transition, Nck falls off. So Nck and N-WASP are not included in the clusters at this point. We reconstituted this, using dynamic, active networks on model membranes with members from Satyajit Mayor’s group and Ron Vale’s group, and Khuloud Jaqaman did a really nice job analyzing this data for us. 

And what we found is that complete clusters or clusters that have Nck and N-WASP bind to wet actin filaments, while minimal clusters as would be found near the center of the synapse, do not tend to bind in wet filaments to the same degree. When we induce motion through myosin activity, we find that complete clusters with Nck and N-WASP are dragged with the filaments towards the center of these actin asters, while the minimal clusters, containing only Grb2 and Sos are pushed around by the actin filaments. 

So, this to us generally indicates that condensate composition can not only determine the function as we all think about it, but it also can determine the interaction with local environments, which likely is important for the neuronal synapses as I’ll talk about in a minute. So, I’d like to take a break if anybody has any questions about some of this introductory work that Xiaolei and I did previously. I see Jeff has a question. 

Jeffrey Toretsky: First of all, it’s really good to see you, and it’s nice to see this kind of replay of this early work. And I have a couple of students with me, so they’re enjoying it, I’m sure. The question I have you ever looked at extracellular signals? Like you had Lck in your first cartoon, and Lck is, you know, associated with CD4. And obviously, we know CD4 is important for T cell recognition signals. Have those played a role in what’s happening inside the membrane of the cell? 

Jonathon Ditlev: Yeah, so this this is actually I think, an underappreciated aspect of T cell signaling. John Hammer and Jan Burkhardt have done some nice work looking at how the tension of the opposed membrane and opposed cytoskeleton, how that affects T cell signaling. But by and large there has not been much work looking at how the organization of the opposed cell that might activate CD4 or bind CD8 or PD1 or CD28. How that organization plays into T cell signaling, and I think this will be important, not only for understanding T cell signaling, but also for CAR-T cell work that my colleague Xiaolei has really spearheaded in his group. At least in my mind, an active question is, how does the opposed membrane regulate what’s happening in in the T cell and carrying that over to neurons as well in the future? 

Jeffrey Toretsky: Thanks. 

Jonathon Ditlev: You came to a neuronal talk. And all I’ve talked about is T cells to this point. Why have I started this introduction? Well, back in 1968, Gerald Edelman suggested that there may be similarities between immunological synapses and neuronal synapses, and that we probably can learn something about neuronal synapses from immunological studies, as well as learn something about immunological synapses from neuronal studies. This was long before biological phase separation came to the forefront of our minds. Me and a few other people have asked the question, well might phase separation be a general mechanism that regulates the functional organization of both immunological and neurological synapses? So, I just showed you that phase separation plays a very key role in immunological synapse formation and function. Can we apply what we’ve learned in the immunological synapse to neurological synapses? 

And so, what is the neurological synapse that I’m going to be talking about? I’m specifically focused at the connection between an incoming axon and dendritic spines on a receiving neuron. Neurons in the brain have roughly 10,000 dendritic spines per neuron, which means that they are able to integrate a lot of information coming in and generate an action potential, which they shoot out an axon. 

Within the neurological synapse there’s two terminals, the presynaptic terminal, which contains several condensates, including the active zone, synaptic vesicle clusters, as well as some other RNA granule type structures. And then there’s the postsynaptic terminal, which is also filled with condensates, including the postsynaptic density, RNA transfer granules, there’s GIT/PIX condensates that Mingjie Zhang has shown. And if you want more information on this, Khuloud Jaqaman and I wrote just a brief description of how condensates play a role in signaling in the dendritic spine, but also Mingjie Zhang’s group has done some really nice work and written a couple of really nice reviews about how condensates can control the passage of information from one neuron to the receiving neuron. 

Now, I just want to highlight the postsynaptic density. So, there’s been quite a lot of genetic work done on this structure. And roughly, you know, proteomic work has shown that there’s at least a thousand proteins that can localize to this membraneless structure, and perhaps most strikingly, over 60% of the proteins that localize to the postsynaptic density have been linked with neurodevelopmental and neuropsychiatric disorders. 

So, this might be the most disease rich structure in the human body, which means that there’s a lot of space to operate. And so, a closer look shows that the postsynaptic density is scaffolded by multivalent proteins. There’s roughly 17 or so scaffolds that have been recognized to be required for postsynaptic density formation and specifically 4 family members. So, we’ve got the Homer family, the MAGUK family, which includes PSD-95, DLGAP family, which has several family members, as well as the shank family, shanks, 1, 2, and 3. And each of these proteins can be found in any given postsynaptic density at different levels. One of the complex issues with looking at neurons is that each postsynaptic density is completely unique. And Seth Grant did some really nice calculations and found that there are at least 14 trillion combinations of molecules that can occur within any given postsynaptic density. Understanding each postsynaptic density will be very difficult; however, we can look at general principles using some of these proteins. 

Mingjie Zhang actually did some very nice work in the late 2010s showing that protein fragments of each of these scaffold members can lead to the condensation on the membrane linked with an NMDA receptor subunit tail. He showed that through the reconstitutions that model PSDs can interact with and promote polymerization of actin filaments just as Xiaolei and I showed in T cell at the T cell synapse. And this is dependent on cortactin and the Arp2/3 complex. He also showed that model PSDs can sort molecules much like we saw with immunological synapse and LAT clusters in CD45. In the model PSDs, CaM kinase II, which is highly enriched, maybe the most abundant protein in the postsynaptic density is enriched in these condensates, while gephyrin, which forms inhibitory condensates, is excluded. In a follow-up paper from his group, Mingjie and his colleagues showed that this is driven through electrostatic repulsion, just as we saw with LAT in CD45, really reinforcing that we can learn a lot about neurological condensates from general principles established in immunological condensates or immunological synapses. 

So, my group, when we started in 2019, right before the pandemic, which was excellent timing, one of the goals was to reconstitute the postsynaptic density, but not using protein fragments as I’ll get into a little bit, but using the full-length proteins. And so we set out to purify some of these proteins. Homers and MAGUKs have been, purification has been established; however, DLGAP or GCAP and shank proteins had, to the best of our knowledge, never been done before. So, you know, starting in a pandemic, let’s purify proteins that have never been purified before. That’s great. So long story short, we’ve been able to purify DLGAP, shank2 and shank3. 

And what we found is that DLGAP on its own can undergo phase separation. It quickly forms a gel, and as you’ll see, shank2, and then you won’t see shank3. We can purify the full-length version. We can work with it; we can phosphorylate it; we can do different modifications on it. However, they also form condensates readily, and I’ll get into that in a bit. And so that led us to really rethink how we’re going to study this, because if each of these proteins can form condensates on their own, we get into some complex coacervation rapidly. And what we really wanted to do was link some shank functions with downstream, some shank mutations and functions with downstream effects. 

What we did is we minimized the system, using only Homer and shank. Even though the PSD is a membrane associated condensate, we’re looking solely at in solution information here. The first thing that we did is we asked whether shank2, much like shank3, as Mingjie showed earlier, could undergo phase separation. We see it at 50 nanomolar and so, using the full-length protein, these shank2 and homer1undergo phase separation at very, very low concentrations, but they are physiologically relevant. 

We then asked, well, these are spheres, they must be dynamic right? And so we performed FRAP and as you can see in the recovery chart, there’s no recovery. So we photobleached a section of these condensates and then over the course of 10 minutes and beyond, there’s little rearrangement, if any, and there’s no exchange of molecules from what we can see with the outside environment. What this at least suggests is that these condensates likely form as a liquid, forming, minimizing the surface to volume ratio in spheres, but then they rapidly, within about a minute transition to a gel-like substance. 

Phosphorylation is a common modification. We know that, based on online databases that shank2 is able to be phosphorylated at many sites. And so we chose MAP Kinase to phosphorylate shank2. And we hypothesized that maybe phosphorylation will regulate the dynamics of these condensates, as well as possibly their formation. And so we performed experiments with phosphorylated shank2 and Homer1. And what we found is that phosphorylation does very little to alter the dynamics and formation of these condensates compared to the non-phosphorylated version. 

To me, this was kind of disappointing, because you know, in my training, in my signaling world, phosphorylation does everything. And here it appeared to do absolutely nothing. So that was kind of a bummer. But what we have is a usable platform that we can now perform functional experiments with either phosphorylated protein or non-phosphorylated protein. 

The shank family proteins have historically, canonically been linked with actin polymerization and controlling spine morphology through increase or decreasing actin polymerization, depending on the rate of signals coming in from the presynaptic bouton. As I mentioned, we had a bit of downtime because of Covid, and so we did some sequence analysis with Dr. Forman-Kay here at Sick Kids, and what we found is that shanks 1 and 3 gene ontology terms linked with their intrinsically disordered regions, were clustered very well with actin machinery, suggesting a direct link between the gene ontology of their IDR and downstream function. However, shank2 surprisingly contained go terms that were linked with RNA translation. And so search the literature a little bit, there’s been some work on shank2 not nearly as much as with shank3. But some of those studies found that there were translation differences in neurons that express shank2 mutants. And so we thought that maybe RNA translation in the spine is linked with shank2. And so shank2 knockout neurons are hyperconnected, suggesting that there’s additional RNA translation to generate these proteins that occur when shank2 is not there. So maybe shank2 has a role with dampening or inhibiting RNA translation. Shank2 mutations also alter the global translation, you know, continuing with our idea that maybe there’s a link between shank2 and translation. And importantly, the computational analysis showed that shank2 actually clustered with FMRP and CAPRIN1, which the Foreman-Kay lab, just a few years ago, showed to regulate RNA metabolism. And so we thought, hey, this is great. Here’s a link, maybe shank2 can regulate local translation through FMRP and CAPRIN.  

Why might that be? Just a brief simplified look at FMRP regulation translation. So Julie’s lab did some really nice work looking at how different forms or different modifications affect FMRP-dependent translation. What they found is that when the IDR of FMRP is phosphorylated, translations dampened, compared to non-phosphorylated version, and then when it’s methylated, translation is increased and so simplified version is phospho-FMRP suppresses translation. FMRP promotes translation. 

Importantly, the kinases and phosphatases that act on FMRP, specifically PPA2 and CK2, are localized specifically within the postsynaptic density. And I just want you to think, for the purposes of this talk, that FMRP is a switch. So upon phosphorylation, it suppresses translation, and then upon phosphatase activity, it can return to its non-phosphorylated state and then promote translation. 

So we asked, does shank2 with molecular features similar to FMRP modulate RNA metabolism, in this case, translation through interactions with FMRP? And so once again, we work with our shank2 and Homer model simplistic system, and we ask three questions. So number one does shank 2 regulate translation? Number 2, if yes, how? And then 3, do disease-linked mutations alter shank2 modulated translation, as has been suggested by some of the observations in neurons that express shank 2 mutants? 

So our hypothesis is that it’s the phase separation specifically of shank 2 that regulates RNA binding, protein, localization, FMRP, and the phosphorylation state of the RNA binding proteins to control translation. Furthermore, disease-linked mutations in shank2 can alter these dynamics. 

Our first question was do shank2 condensates affect translation? So what we did is we used a luciferase translation assay. We premixed shank2 condensates, dropped them into the rabbit reticulocyte lysate, allowed translation to occur, and then measured the amount of luminescence from luciferase following translation. And what we see is that in our control sample, which contains only luciferase, RNA, and rabbit lysate, we have a lot of luciferase translation, as indicated by its luminescence. However, as soon as we add shank2 and Homer1 condensates to that mixture, the translation is dampened by about 5 orders of magnitude, showing that shank2 does something to inhibit translation, as we have hypothesized. 

And we did some microscopy. And we did show that these condensates remain, even though they’re moved from just the solution based to a lysate. So they’re very robust condensates. Okay, so they can modulate translation in vitro. And then we asked, do disease mutations alter the enrichment of FMRP or phospho-FMRP? 

And so the first question is does FMRP go into condensates? So we used a physiological buffer and looked at whether FMRP could localize to shank2 condensates, and at 150 millimolar salt we saw very little enrichment. There might be a very little bit if you look closely, but by and large FMRP remains, it doesn’t really detect the condensates. 

However, at 50 millimolar sodium chloride we see good enrichment of FMRP in the condensates, suggesting that electrostatic interactions drive its enrichment in condensates when there’s low salt shielding of whatever interacting motifs there are between FMRP and shank2. 

So then we asked, because it’s electrostatic, maybe increasing the negative charge of the intrinsically disordered region of shank2 will affect FMRP enrichment. And so we did several experiments in which we looked at phospho-shank condensates, phospho-FMRP, and then different combinations of the two, and, as you can see from our measurements on the right, there’s no significant change of FMRP enrichment in shank2 condensates, regardless of phosphorylation status of either protein. 

And so once again, my whole thought process that phosphorylation matters, for everything was completely blown up, which is great. I don’t mind being proven wrong, but it also makes us wonder well how, if phosphorylation is not really doing anything in these wild type condensates, does it matter for function? That’s just a question that I want you to remember, as we go through the rest of this talk. 

So we showed that FMRP can be enriched in condensates. What about RNA? RNA has never really been detected in the postsynaptic density. There’s been some omics work that’s found that microRNAs can localize there, but by and large, the postsynaptic density is thought to be a protein-rich environment. However, if you look at electron micrographs, you’re not going to be able to see RNA anyway. So whether it’s there or not is still is not clear from cellular data. However, in vitro we can see that RNA readily enters shank2 condensates when FMRP is present. All of this data allows us to build a model in which we think that shank2 condensates are regulating RNA translation through the absorption of FMRP and mRNA preventing translation from occurring outside of the condensate. 

Okay, this is great. We have an experimental platform. We can look at different aspects of it microscopically. We can look at functional outputs through translation assays. What about disease-linked mutations, because there are a lot of them? And so each of the lines here show different mutations that have been sequenced from patients in shank2, and shank2 is now recognized by the Simons Foundation as highly correlated with autism. 

Today, I’m going to focus on three specific mutations. First, I’m going to talk about the R958S and A1731S. And then we have a patient line of induced pluripotent stem cells from a patient that has a heterozygous truncation mutation. And so we’re asking whether we can apply what we’ve learned biochemically to an actual patient’s cells. 

And also, importantly, a lot of these missense mutations are found in the intrinsically disordered region and so shank2 the isoform that we’re using is about 1,450 residues. And so the obvious question is, what will a single point mutation do amongst an IDR of about  1,000 residues, and Rick Young just released a paper looking at missense mutations in different condensate, forming proteins, and found that just a single missense mutation can alter its localization. So just keep that in mind. 

And so we first looked at whether FMRP localization was altered by these mutants. On the left here, if we look at dephosphorylated proteins, we see that FMRP is well enriched in the mutant condensates phospho-FMRP enrichment doesn’t really seem to change very much when normalized to shank2 intensity. However, when the 958S mutant is phosphorylated, we see a drastic reduction in FMRP enrichment in the condensate. This will come up later, but note that it seems to only enrich the edges of the condensate. 

We also looked at the 1731S mutation, and we saw the exact same phenotype in which, when shank2 mutant was phosphorylated, FMRP is not very well enriched, and any enrichment seems to occur at the edges of the condensate. If we look at just to compare visually, we see good enrichment of FMRP and wild type condensates, however, in these mutant condensates. When they’re phosphorylated, there’s little enrichment. And this plays out very well when we quantify the relative fluorescence of FMRP compared to shank2 in the condensates. There’s a significant decrease in FMRP. 

We also wanted to look at RNA and whether the pattern of RNA enrichment follows FMRP enrichment. And what we see is that yes, when condensates, when the mutant condensates are phosphorylated, the amount of RNA within the condensate also is diminished when normalized to shank2, just as FMRP. 

Okay, so here comes a so what, who cares? This is great biochemically, we can use mutants and look at localization of different proteins. But what about in neurons? And so what we did is we teamed up with James Ellis group here at Sick Kids and looked at, induced pluripotent stem cell derived neurons at four weeks. And so the key with these experiments is that we’re looking at a neuronal environment, not so much that we’re looking at dendritic spines. We’re doing that work now. 

What we found is that when we express mCherry-shank2 and mNeonGreen-FMRP, in either shank2 wild type or shank2 knockout neurons, we get good co-localization of the two proteins together, which is very similar to what we see in vitro. However, when we express the 958S mutant, we can see a complete localization change of FMRP. And importantly, it doesn’t matter if there’s wild type copies of shank2 in the background or not, the two proteins don’t co-localize with each other in these neurons. We see the exact same molecular phenotype with the 1731 mutant in which FMRP does not co-localize well with shank2. And when we just ran Pearson correlation, we can see a significant difference in the co-localization of FMRP with shank 2 condensates in these neurons. Importantly, this also suggests that in a heterozygous situation, as you’d have with a wild-type neuron that the mutation dominates. 

Even though, which is important because in the disease these missense mutations are heterozygous in the patients. And so these data suggest that even though you have only a heterozygous missense mutation, it can dominate the behavior of condensates in cells. And as I mentioned, we’re currently performing functional assays, looking at the functional results of this mislocalization of FMRP in these neurons. 

Okay, so going back to the biochemistry. As I mentioned, it seems that in the with mutant condensates, FMRP localizes only to the edges of the condensate. This at least suggests that there might be a physical change in the variant containing condensate. And so the question comes up, well, how do we study this? One of the obvious experiments to do is looking at fluorescence recovery. But as I showed you earlier, within about a minute, these condensates turn into a gel. And so we have a technical limitation in that. We can’t really look at molecular, at the dynamics of molecules in these condensates simply through FRAP, but here at Sick Kids we have an oscillation rheometer, which allows us to look at the viscosity and elasticity of these condensates. And so, looking at the viscosity of the condensates, using the rheometer, what we find is that the disease-like mutants are more viscous than the wild type condensate, even though they’re in a gel-like state. And when we look at elasticity, we see the storage modulus, which is an indication of the ability of the material to store elastic energy. We see that both mutants have a higher storage modulus than the wild-type condensate, and if we look at the loss modulus, which is the amount of energy dissipated, we see that there’s a distinct difference between mutant condensates and wild type condensates. 

So this allows us to conclude that even though the dynamics within the condensate are very, very low, that mutant condensates are more viscous and elastic than wild-type condensates. And so we’re able to build out this model now that includes the disease mutations in which we see that phospho-FMRP is well enriched in mutant condensates; however, RNA and FMRP are diminished. And that’s linked, at least correlated, with an increase in viscosity and elasticity compared to wild type. And so the functional output of these condensates, if RNA translation occurs outside of them is that RNA translation should go up because FMRP, non-phosphorylated FMRP, promotes translation and its phosphorylated counterpart that inhibits translation is mostly in the condensate. And that’s exactly what we find. 

When we reconstitute mutant condensates in our rabbit reticulocyte lysate assay, we find that translation is increased 15- to 30-fold, depending on the mutant compared to wild-type shank2. And we think that this is the first molecular insights into why shank2 mutant neurons see an increase in translation. It seems that some of these missense mutations, at least this class of mutation in which erines are generated in the intrinsically disordered region can lose their ability to inhibit translation, which is its wild-type function. 

So I mentioned that we have a patient here at Sick Kids that is heterozygous for an R841X truncation mutation, and we thought, we have the we have the iPSC, let’s take a look and look at FMRP localization. And so what we see is a pattern very, very similar to the missense mutations that I showed earlier, in which we here we express wild type shank2 in these neurons that at least genomically, have a truncation mutant, and what we see is a complete mislocalization of FMRP away from the condensates, just as we saw previously. 

So then we asked the question, what about FMRP and phospho-FMRP levels? So if you recall in our model depending on the ratio of the two, you’ll either promote translation or inhibit translation. So if you have more phospho-FMRP outside of the condensates, you’ll inhibit translation. In these neurons, when we run Western blots, we see that the ratio of phospho-FMRP to FMRP is actually increased, meaning that there’s more phospho-FMRP in these neurons. And when we rescued the R841X mutation by replacing it with a complete copy of shank2, we can see the ratio return to wild-type levels. This allows us to build this model in which we have high FMRP and phospho-FMRP outside of the condensate thinking about our localization data. And so what we should see is a diminishing of mRNA translation experimentally. That’s exactly what we see CaMKII, which is an FMRP targeted mRNA as the indicator, and what we see is that in the R841X mutant, which has more phospho-FMRP available, CaMKII translation is diminished, matching our model. Importantly, though the mRNA counts are very similar between the two cell types. So your rescued cell type has the same number of RNA molecules as the mutant cell type, indicating that this regulation occurs at the protein level, providing more support for our model, that shank2 can regulate translation and that disease-linked mutations in shank2 can alter that translation downstream. 

In summary, we showed that full length shank2 and Homer 1undergo phase separation at physiological concentrations. And these condensates are not dynamic. They do not recover well following photobleaching. FMRP can be enriched in shank2 condensates, and that that enrichment is sensitive to salt concentration, suggesting that electrostatic interactions are the driving force behind its enrichment. These condensates can repress translation in vitro, the 958S and 1731S mutations dysregulate mRNA enrichment both in vitro and in neurons. These mutations also lead to an increased viscosity and elasticity of the condensate compared to wild-type shank2. Expression of these mutants and wild-type background cells shows similar aberrant FMRP enrichment, suggesting that these mutations are, in fact, dominant which, really provides a mechanism for understanding missense, heterozygous mutations that are often found in patients. 

Phosphorylation of the shank2 mutants promotes RNA translation compared to wild-type which also provides some insight into the mechanisms that explain observations and neurons, but also the truncation of shank2 that deletes the IDR alters the ability of it to interact with FMRP, and it leads to a dysregulation of the ratio of phospho-FMRP to FMRP leading to repression of CaMKII translation. 

With that I’m happy to take any questions you might have. 

Diana Mitrea: Thank you so much, Jon. We have a question from the chat. Mikhail, would you like to unmute and ask your question? 

Mikhail Koksharov: Thank you for the very interesting talk. I’m just curious, from the general point of view. What is the protein concentration inside these droplets when you do them in vitro, or when they are inside living cells? And basically, are they more dense than the cytosol or less dense? 

Jonathon Ditlev: I’ll give you a two part answer. The first is that we haven’t measured the concentration within the condensates of shank2. And that is because there’s a technical limitation of the microscopy we’re doing in that the background fluorescence when we image the condensates is low enough that it’s within the noise. And so we can’t actually calculate the ratio of enrichment for shank2 in the condensates. We could do some fluorescence calculations and actually get to that as we’ve done with other condensate systems. The second part is that in the neurons there’s not a uniform concentration of shank2 across the cell. So because these cells are so highly polarized, the proteins are not well expressed in the soma or body of the neuron, and then transported out or diffuse out to the dendritic spines. The dendritic spines are actually the centers of protein translation. Within any given dendritic spine, you’re going to have a higher concentration of shank2 compared to anywhere else in the neuron in terms of the physiological concentration, I’m not sure what it is for shank2, but I can tell you so if you look at another postsynaptic density protein Mingjie Zheng has said that PSD95, which is highly enriched in the postsynaptic density and dendritic spines, is roughly 100 micromolar. And so you know, even if shank2 is a hundredfold less than that, it’s still at about a micromolar in the dendritic spines, which is well above the critical concentration for phase separation on its own. 

Kamran Rizzolo: I have a question. Excellent talk, Jon. Fantastic work. You had mentioned that shank1, and I think shank3 was the one you can purify. Did you guys have a look at these other 2 shanks, and in addition, are there any known mutations that are linked to these diseases in these proteins? 

Jonathon Ditlev: So first question. So we’ve attempted to express shank1 and shank3 in bacteria. Shank1, we have been unsuccessful. It only expresses like the first half of it. And then we don’t actually know if there’s like some rare codon stop or something, so we’ll probably do that in insect cells. Shank3, there’s actually a student in my group, Gaddy, who has successfully purified shank3 using similar methods that we use for purifying shank2 , and he’s found that truncation mutants have a very different phase behavior than the full length and he’s actually done some BioID in cells with Jiang Yun, showing that missense mutations in shank3 alter protein networks, specifically looking at actin machinery. So different missense mutations in shank3 alter the condensate composition with regard to actin, which actually fits with those mutations because there’s changes in actin polymerization and dendritic spine size in neurons that express shank3 mutants. In terms of diseases, these two mutations, the 958S and 1731S have been found in autistic patients as well as schizophrenic patients. Depending on various genetic factors and also protein expression factors, the presentation of disease can be different, even though the same protein is mutated. 

Kamron Rizzolo: Got it. Thank you. 

Diana Mitrea: Lindsey, go ahead and ask your question.  

Lindsay Case: Hi, Jon, I was just curious like, it’s really interesting that the mutants often have a phenotype in the heterozygous background and have you explored that at all biochemically? If you mix your wild-type shank with the mutant shank do they co-condense? Does the mutant like dominate the viscosity and elasticity? Or is there some like intermediate material property? And how have you explored that? 

Jonathon Ditlev: Yes, so we haven’t explored it. And so that is actually on our list to do. Because, you know, in cells we see this dominant effect, it’d be interesting to see if a similar effect happens in vitro. For sure, these are great ideas. 

Sayantanee Niyogi: Hi, Jon, great talk! This question could be a little far-fetched, but just that you mentioned the connection between shank2-driven condensates with autism, and you know some, you know schizophrenia. So you’re in the hospital. You have samples. Could this work be, you know, extended to detecting in, you know, biological samples like blood, CSF, and used as a kind of a biomarker to track or diagnose this kind of things where you find this very interesting connection? 

Jonathon Ditlev: Yeah, absolutely. You know from this, I think along these lines, from these experiments, we’re actually seeing that there may be classes of mutations that have very similar physiological patterns, even though the mutations are very distinct. One of the intractable things about studying the postsynaptic density and thinking about it in terms of correcting or rescuing the disease is that there’s tens of thousands of individual mutations in the 600 or so proteins that have been linked with disease. But if there’s individual classes of mutations, and maybe we would get more information by looking at the variety of patient samples that we have access to, we might define classes that could be potentially corrected using a single corrector, whether that’s a peptide or a small molecule. It would be actually interesting to be able to classify postsynaptic density linked diseases in different bins, rather than looking at each mutation individually. 

Diana Mitrea: I have a question. That was a wonderful talk, Jon. Thank you so much. So I was curious for these shank2, Homer, FMRP condensates is the effect on the translation global, or is there some form of selectivity? 

Jonathon Ditlev: So in the in vitro assays, it’s global. But I think in the neurons, there is some structural selectivity built into where these proteins are actually expressed and where they condense. And so FMRP is packaged into RNA transport granules and then shipped as part of the translationally repressed RNA to the polarized regions of the cell, whether it be the axon, terminal or dendritic spines. There the RNAs that are transported are translated locally. If you think about a whale neuron that can be up to 30 meters long, you’re never going to be able to get a protein from the soma out 30 meters, and so to overcome that they ship the RNA and then translate it locally. The experiments previously that have looked at expression of proteins have looked at it globally, but in reality, it’s really only happening in the dendritic spines by and large. And that’s simply through the localization of where these condensates are and then diffusion barriers driven by the structure of the neuron. 

Diana Mitrea: Okay. So going back to the story of filtration and exclusion of specific compounds or proteins. So in the first example, you talked about electrostatic repulsion. I was wondering, is that sort of the general rule? Or how does the lattice size contribute to that? Has anyone looked at it? 

Jonathon Ditlev: That’s a very, very good observation. So electrostatics, at least from what people have looked at, and maybe it’s they’ve looked at it, because it’s the easiest thing to look at is a very large contributor to the sorting of molecules when there’s not specific binding motif interactions or domain motif interactions. We originally thought with CD45, that maybe you know, CD45, which has a fairly large intracellular domain is unable to penetrate, because the lattice size is too small, you know the effective pore size is too small to allow it to be in. We didn’t rule that out as a possibility, because the electrostatic effect was so large. When we used just generic proteins with different charges on them, we were able to move these generic proteins, either outside or inside, depending on their charge, so it could very well be that that the intracellular domain exclusion is linked to both charge and size. We don’t know. You know, some of the some of the really nice work with microrheology on microscopes shows that dextrans with different sizes can either penetrate or not penetrate, depending on the effective pore size. And so that’s absolutely a possibility with sorting molecules as well, I will say, with postsynaptic density, CaMKII is a gigantic, holo enzyme 12 subunits that’s readily enriched in the condensates, while gephyrin, which is much smaller, is readily excluded. So that’s probably the post-synaptic density, even though gelates very rapidly, I’m not sure the effect of the pore size there, but I think it is a very viable potential regulator of sorting of molecules. 

Kamran Rizzolo: One thing I noticed throughout the talk, is that the proteins, as you mentioned don’t change at all when you change the phosphorylated state. But then, with these mutations to Serine or Threonine, you have this whole story right? So question for you, moving forward whenever you see a mutation that drives disease, and it’s Serine, Threonine, very intriguing that it’s, going to you know, have an effect on phase separation, and condensate behavior.  

Jonathon Ditlev: Yeah, and you know, we so there’s actually several Serine / Threonine mutations, we just, we just looked at these two, but there are several within the shank2 IDR. One of the difficulties with working with shank2, is that it has not been studied very well. And so we use MAP kinase to phosphorylate the protein. One of the issues is that we did that based on predicted phosphorylation sites. So MAP kinase, experimentally, as far as I know, has not been shown to phosphorylate shank2 at all. And it only phosphorylates, very specific residues and so some of the Serine 3D mutations don’t lie within, say a MAP kinase predicted phosphorylation regime. It’s a technical limitation, because, you know, we’re an academic lab. We can’t test all 30 kinases that are localized to the PSD to phosphorylate all of these sites but based on what we’ve seen in our experience with some of these mutations is that mutation likely will change the behavior of the condensate, whether through compositional changes or physical changes. And then it just because there’s so many kinases there, there’s likely there’s a high likelihood that it would be phosphorylated. 

Thank you so much guys. 

Diana Mitrea: If there are any other questions, feel free to submit them through the contact form on Condensates.com. Join me in thanking Jon again for such a wonderful talk, and the recording is going to be available on Condensates.com within a week. Have a great day, everybody, and see you next time.