Managing editor, Condensates.com
|Type||Kitchen Table Talk|
Dewpoint and Condensates.com welcomed Carlos Castañeda from Syracuse University for an incredibly thought-provoking talk on April 28 as part of our Kitchen Table Talk series. Carlos has extensive experience in protein quality control. He obtained his PhD in Molecular and Computational Biophysics from Johns Hopkins University, studied with David Fushman at the University of Maryland during his postdoc, and became a professor at Syracuse in 2014. In that time, he biophysically characterized ubiquitin linkages, developed enzymatic methods to study ubiquitination, and determined how ubiquitin modulates phase behavior.
Carlos has mastered the use of a wide range of biochemical, biophysical, and cell biology methods to carefully tease apart the molecular mechanisms that drive protein quality control. The Castañeda lab showed how ubiquitin can modulate phase transitions in a way that may release proteins stuck in stress granules so they can be sent for degradation by the proteasome. And they have gone on to show how this phase behavior is important for neurodegeneration by determining the effects of ALS-mutations in vitro and in cells.
In his talk, Carlos shares all the molecular details that allows some polyubiquitin chains to modulate phase separation more than others. We were all fully immersed in his talk and many attendees had great questions and hung around for a lively discussion afterward. Carlos also generously provided written answers to all questions he didn’t have time to answer during the show—you can find those here. Enjoy this thought-provoking talk and discussion in our recording of the event below. And if you want to start or continue a conversation with Carlos, he welcomes emails at email@example.com.
Create an Account or Sign In to view the video.
Jill Bouchard (00:00:00):
So without further ado, it’s my great pleasure to introduce today’s speaker and my former collaborator, Carlos Castaneda, from Syracuse University. Carlos got his PhD in Molecular and Computational Physics, from Johns Hopkins University. For his postdoc, he studied with David Fushman, at the University of Maryland. And he became a Professor at Syracuse, in 2014. He’s done quite a bit of work in the field of protein quality control, including biophysical characterization of ubiquitin linkages, developing enzymatic methods to study ubiquitin and best of all, determining how ubiquitin modulates phase behavior.
Jill Bouchard (00:00:35):
In fact, Carlos’s lab, along with some of my former colleagues from the Taylor Lab, at St. Jude, showed how ubiquitin can modulate phase transitions of a proteasome shuttle protein. And this may be a mechanism that releases protein stuck in stress granules, so they can be sent for degradation. And when I was in the Mittag Lab, at St. Jude, I was fortunate to join some of these efforts, in understanding ubiquitin modulated phase separation, as well.
Jill Bouchard (00:01:00):
And the Castaneda Lab has also gone on to show how neurodegenerative disease mutations affect phase behavior. And somehow, Carlos has managed to do all this, during a pandemic, with three kids under the age of five! We commend you. And Carlos’s lab techniques are equally impressive. His lab uses a wide range of biochemical, biophysical, and cell biology methods, to carefully tease apart the molecular mechanisms underlying all the complex interactions involved in protein quality control. And I’ll let Carlos fill you in on all those fascinating details. So Carlos, thanks again for joining us. We’re so excited to hear your story about ubiquitin-mediated phase transitions in protein quality control. The floor is yours.
Carlos Castañeda (00:01:44):
Thank you so much, Jill. That was quite the introduction. I don’t even know how to begin after hearing all of that, but I think a lot of you just heard the first part of my talk, so great. No, I really want to thank Jill. I do want to think Dewpoint Therapeutics, as well, for giving me the opportunity and giving us, really, the opportunity to tell you about some of our recent work that is unpublished but particularly, looking at the role of ubiquitin and also polyubiquitin chains, on how they modulate phase behavior of the shuttle protein that’s involved in protein quality control, called UBQLN2…
Carlos Castañeda (00:02:22):
So I wanted to kind of start off with talking about what is protein quality control and why we should care about it. And it’s really essential for maintaining cell health. And I’m trying to depict in this image here, on the top left, a normal operating cell where our proteins are interacting with each other. They’re behaving correctly. And a lot of times, these proteins are maintained through functional protein quality control mechanisms, that ensure that the proteins are not mis-folded, or if they’re no longer needed, that they can be degraded.
Carlos Castañeda (00:02:56):
And you can think of this as a balance, as this protein homeostatic balance. And one example, one really classic example of protein quality control mechanisms that many of us hear about, especially in biochemistry, is the ubiquitin/proteasome system. And so that’s what I’m depicting down at the bottom, whereby a protein can get covalently modified with a ubiquitin chain, that these are a series of ubiquitins that are attached to each other and also attached to a substrate protein here.
Carlos Castañeda (00:03:28):
And it’s this polyubiquitin tag that is recognized by different proteasomal receptor proteins. And that is the signal for sending the substrate to the proteasome, where it will be degraded into short peptides, and then the ubiquitins themselves would also get chopped into individual monomers and eventually, recycled into the cell.
Carlos Castañeda (00:03:52):
So this is a classical example of protein quality controlled mechanism. However, we know that under certain conditions, we can disrupt this protein homeostatic balance. Some ways that we can disrupt that is for instance, by exposing your cells to certain types of stresses. Aging is another things that can occur, too. Some of our cells live a very long time, and over time, those protein quality control mechanisms can get dysregulated, as well as disease causing mutations.
Carlos Castañeda (00:04:24):
And this can disrupt this balance. And what we often see are the accumulation of protein aggregates and dysfunctional proteins. So one really interesting example of this is in neurological, neurodegenerative disorders, whereby we’ll see an accumulation of protein containing inclusions. You can think of these as protein clumps, and these are very characteristic of disorders such as amyotrophic lateral sclerosis, so ALS, Huntington’s, Alzheimer’s, some of these other disorders.
Carlos Castañeda (00:04:55):
And what I’m showing here are images, actually, from a hippocampal neurons, staining for UBQLN2. And you can see, hopefully, with my laser pointer, these red splotches of aggregates that basically stain for this protein. And this is our protein of interest, UBQLN2. And it accumulates into these so-called aggregates and these types of inclusions. What’s interesting about this protein, though, is that both the wild type and the mutant form of it can accumulate into these inclusions. And oftentimes, when we see these inclusions, we’re not really sure whether they’re symptomatic of the disease or whether they’re causal of the disease. And so we’re really interested in trying to tease this part out.
Carlos Castañeda (00:05:40):
So our training was, or my training was in molecular biophysics. When I started my lab here, at Syracuse University, a few years ago, we really wanted to look at the overall structure, this functioning of this protein, called UBQLN2. And so I’m going to abbreviate this protein here, so you’ll see this over here. And I want to emphasize that UBQLN is not the same thing as ubiquitin, and, I unfortunately, realized that the names look almost exactly the same, UBQLN versus ubiquitin, so there’s an L in UBQLN, and there’s an ubiquitin. There’s a T in ubiquitin.
Carlos Castañeda (00:06:17):
The reason why the names are so similar are because UBQLN has ubiquitin-like motifs and ubiquitin-binding motifs, as well. So this protein is a 600 amino acid protein, and it has an ubiquitin binding domain, so this is called a UBA domain, whereby it recognizes this ubiquitin chain on say, an ubiquitinated protein, like I was describing earlier.
Carlos Castañeda (00:06:43):
And it has an ubiquitin-like domain, that can interface with the proteasome, shown here on the left. And so it’s considered to be a shuttle protein, whereby you’ve got this UBL-UBA architecture, and it can take a substrate and target it to a proteasome, so you can think of UBQLN as acting as this middle person in this pathway. But what was really interesting to us was that the middle part of the protein wasn’t well characterized, at the time when we started working with this. There were little bits of data that suggested the middle parts are interacting with heat shock proteins, chaperone proteins.
Carlos Castañeda (00:07:16):
And when we started looking at some of the sequence characteristics of this, it contains largely intrinsically disordered segments that I’ll talk about it a little bit. The other reason why I got interested in this protein is that it seems to be involved in multiple protein quality control pathways: the proteasome degradation pathway, as I mentioned, autophagy, which is another ubiquitin-mediated degradation pathway, as well as ER, so endoplasm reticulum degradation pathways, too.
Carlos Castañeda (00:07:47):
And as I mentioned, this protein can be mutated into ALS. Turns out that the most of the ALS-linked mutations are harbored into this proline-rich region shown here, which rests pretty close to the ubiquitin-binding domain. And so we wanted to characterize what this protein is doing. We wanted to see if we could try to look at the structure of this protein. And when we started working with this protein, we made this early discovery, that the protein undergoes liquid-liquid phase separation, in vitro.
Carlos Castañeda (00:08:17):
And so what we did was we purified this protein in E Coli, and then we basically put this into buffer. And you can see with the microscopy images, here, the dynamic ability of the protein to form these beautiful looking liquid-like droplets. And you can see how rapidly they fuse with each other. And again, you can see that UBQLN is enriched inside of these condensates versus the outside. And the important thing to remember here is that this protein phase separates as we increase salt, and under physiological-like conditions, and at concentrations that are pretty close to physiological concentrations of UBQLN.
Carlos Castañeda (00:09:01):
Okay, the other big discovery we had made, at that time, in collaboration with Paul Taylor’s lab, as Jill mentioned, during my introduction, is that the endogenous form of the protein is recruited into stress granules. And stress granules are an example of one of these biomolecular condensates that we see inside cells, and so what I’m depicting, on the left-hand side here is UBQLN2 under no stress conditions. And you can see how it’s largely diffuse in these three cells here.
Carlos Castañeda (00:09:34):
And it reorganizes into this puncta-like state, upon stress, upon different types of stress, so arsenites, which is a common stressor that’s used in the field to mimic oxidative stress, heat shock, and some of these other stressors as well. But you can see that it co-localizes nicely with this stress granule marker, EIF4G1.
Carlos Castañeda (00:09:59):
My undergraduate, Jules Riley, who’s currently at Penn right now, in collaboration with Heidi, we were able to image these cells. These are actually the G3BP line, the G3BP is fluorescently tagged with GFP. And I’m only showing this to kind of illustrate how these stress granules have this very dynamic liquid-like behavior, very similar to what we see with the UBQLN2 condensates, that we saw in vitro, so that you can kind of connect to these two observations.
Carlos Castañeda (00:10:34):
So with these things in mind, as being a molecular biophysics lab, we really wanted to figure out, what are the molecular determinants for how this protein undergoes phase separation? And I think we’ve learned over the last few years, that there are many components. There are discrete components that can drive this phase behavior. One is a responding to different kind of conditions, so for instance, increase in the salt, change in temperature, and that’s what we see here.
Carlos Castañeda (00:11:08):
But if we zoom into these droplets and we start to look at the protein components here, you can see how a very common characteristic are that they contain a lot of multivalent interactions. And these multivalent interactions are what really drive phase behavior, and these multivalent interactions can be present in intrinsically disordered segments, as well as folded domains, as multiple labs have recently shown over the last few years.
Carlos Castañeda (00:11:39):
So with this knowledge in place, we wanted to tease out what goes on inside UBQLN, and so the quick take-home messages from our prior work, the first is that our phase behavior of UBQLN requires oligomerization via this domain called this STII-II domain, shown here. So I’m depicting the domain architecture again, in UBQLN, also highlighting here, two disorder predictors, as well, a prion-like prediction algorithm and highlighting for you, that STII region, over here, so you can see how it has even some prion-like characteristics and how this middle part of the protein is largely disordered, with the exception, of course, the folded UBL and ubiquitin associating domains.
Carlos Castañeda (00:12:29):
We’ve gone to characterize this using turbidity assay experiments, which we use as a proxy for phase behavior. And so what this experiment does, we’ve essentially put our purified protein into a cuvette, on the UV-Vis spectrophotometer. We ramp up the temperature, and as you can see, that this protein, the full-length protein phase separates as you go up in temperature. And this behavior is reversible, consistent with what you would expect for a phase behaving system. But any time we removed this STII region, what occurred was that the protein was no longer able to self-associate, and it also lost the ability to phase separate under any of the conditions that we tried.
Carlos Castañeda (00:13:12):
So this is a summary of a lot of data, but I hope you can see that that STII region is really critical to this. We also ended up using a smaller construct where we only contained this C terminal half of the protein. And you can see how it even has more robust phase behavior under these conditions, as opposed to a smaller construct here, 450 to 624, which has a little bit of the STII region, so it still can phase separate, but the moment that we remove that, this protein is no longer able to phase separate under any conditions that we’ve tried. So this is takeaway message number one.
Carlos Castañeda (00:13:51):
The second takeaway message has been accumulating over the years. We’ve been using this sticker and spacer framework hypothesis that’s been popularized a lot recently by Rohit Pappu’s lab, but it’s been known a lot in the associative polymer literature, for those of you that are in more polymer physics. And what this hypothesis says is that you can think of these biopolymers as containing stickers and spacers. And the stickers are really the components of the protein that are dictating the driving forces for phase behavior. And so what we ended up doing was, we took the C terminal part of this protein, so residues 450 to 624, and we were able to collect the NMR spectrum of this. So you can see a ton of peaks on the left-hand side. Each of these peaks represents an NH, a backbone amide of the protein.
Carlos Castañeda (00:14:48):
You’ll notice there’s a ton of peaks in the middle part of the spectrum. That’s characteristic with it being intrinsically disordered. And you see that there are peaks on the periphery. These primarily correspond to the folded ubiquitin-binding domain of this C terminal construct. And what I hope you can see, when we change the protein concentration of this protein, we go from 50 to 600, you’ll notice that some peaks move. You’ll notice that some peaks attenuate completely. And that provides a lot of rich information on the parts of the protein that are involved in self-association, that we think also promotes its ability to phase separate.
Carlos Castañeda (00:15:31):
And so what we’ve done here, on the right-hand side, is we essentially quantified these shifts onto this chemical shift perturbation diagram, shown here. And you can see these gray bars indicate those residues that disappear, as we go from 50 to 600 micromolar. You can also see that there are hot spots of residues that are impacted. Some of them include the proline-rich region, which ends up being interesting for the disease link perspective of this protein. But they also involve the ubiquitin-binding domain, so there’s a hot spot here, and there’s a second one over here.
Carlos Castañeda (00:16:09):
So with that in place, we wanted to ask the question, what happens when we add ubiquitin? And we know that ubiquitin binds to this ubiquitin-binding domain. And we did the an NMR experiment, where we’re able to quantify that binding, so what we do here is we’ll take this construct of 450 to 624. We will add in unlabeled ubiquitin to it. And then we can monitor those peaks that move. Those peaks that move correspond to that binding that’s taking place between ubiquitin and the UBA domain. And what was interesting was how two of the hot spots, two of the largest CSPs, correspond to the very same hot spots that we’ve determined to be important for self-association and for phase separation for UBQLN2.
Carlos Castañeda (00:17:01):
So we hypothesized that the binding of ubiquitin to this part here would affect phase separation behavior, because it’s taking away some of the stickers that are involved in mediating the phase behavior of UBQLN.
Carlos Castañeda (00:17:17):
Okay, and so we did this really cool experiment that I never get tired of watching this video, where we had UBQLN2 droplets, right? And all we did here was, you can imagine fluorescently UBQLN2. And we added ubiquitin to the far side of this well, and you can see how, over time, very quickly, those droplets disassembled. You can see this diffuse signal of UBQLN, suggesting how ubiquitin was driving the disassembly of these UBQLN2 condensates, in vitro, right? So this is all in a purified protein test tube or microscope well here.
Carlos Castañeda (00:17:57):
Okay, so this is actually known as polyphasic linkage. This is a concept that has been popularized a lot by Wyman and Gill, starting from a paper back in the 1980s. And what this theory, what this linkage idea is, is that a ligand, right, is going to affect the phase transitions of a protein, right?
Carlos Castañeda (00:18:20):
And our case here, with ubiquitin is essentially acting as a modulator of the UBQLN2 phase behavior, right? And as we titrate in ubiquitin, so these are different stoichiometric ratios of ubiquitin and UBQLN, you’ll see that we start to see these UBQLN2 droplets start to disappear. There are less of them, and by the time you reach these concentrations, you no longer see any droplet behavior.
Carlos Castañeda (00:18:48):
We can quantify this, and we end up doing a lot of turbidity assay experiments, where we put our protein into the cuvette, and then we do these temperature based experiments. And you can see, that as we add in more ubiquitin to a constant concentration of UBQLN … so you see many of slides will have these numbers on here … you’ll see the onset of phase behavior goes to higher and higher temperatures. And you can quantify this boundary, and we use those temperature-concentration phase diagrams as a way to quantify this behavior.
Carlos Castañeda (00:19:23):
And you can see here, that as we increase the amount of ubiquitin, that phase boundary increases to higher and higher temperatures. And this is a wonderful example of how a ligand can modulate, right, the phase behavior of a protein. And we think, in the case of part of this protein, that we’re taking some of the stickers. I highly recommend those of you that are listening to also be reading a recent publication by Rohit Pappu’s lab, where Kiersten Ruff and Furqan Dar recently published a nice framework of thinking about how ligands, in general, different types of ligands can affect phase behaviors, of different types of proteins.
Carlos Castañeda (00:20:06):
And this will come in handy, I think, for the rest of this talk. So the next big question and really the point of why I’m here is to tell you about some of our recent work about polyubiquitin chains and how interactions affect phase behavior of UBQLN2.
Carlos Castañeda (00:20:22):
So here’s a picture of ubiquitin. You can see it’s a small compact protein. It has a hydrophilic patch, that is really important for how it interacts with other proteins. It also contains a bunch of lysines. These lysines end up being really important, because the reaction of these different ubiquitinating enzymes, these ubiquitins can get attached the substrate protein, via an amine. But the other thing that can happen is that these ubiquitins can also for an ubiquitin chain on these substrates.
Carlos Castañeda (00:20:56):
So for example, you could have a protein chain called K63, and it’s called that, because the K63, the amine on that side chain of one ubiquitin can be covalently attached to the C-terminus of another ubiquitin, making this longer and longer chain, all right? And so these enzymes, there are linkage specific enzymes that can make different types of chains. And these different chains can code for different kinds of signaling outcomes in the cell.
Carlos Castañeda (00:21:25):
Monoubiquitination, typically, can modulate the behavior of a protein. K11 and K48 linkages are really prominent for the ubiquitin proteasome systems, so targeting a protein for degradation that way. 63 is involved a lot in autophagy but not exclusively, so it’s also involved in other mechanisms. You also have linear chains, so called M1, where the end-terminal amine of the protein can actually make a linear chain, and that’s also involved in autophagy and sometimes signaling.
Carlos Castañeda (00:21:57):
So we wanted to ask whether we could actually build these chains and actually biochemically and biophysically characterize the effects, on a quantitative basis. So Thuy, my postdoc/lab manager, as well as Yiran, graduate student in the lab right now, worked on this project. And this has been a huge undertaking, where they together were able to make large milligram quantities of these chains. So I’m just illustrating here how you can do this. You can take monoubiquitin. You can add these E1, E2 specific enzymes. And you can control for making dimers, trimers, and tetramers, using chain terminating and removable mutations.
Carlos Castañeda (00:22:43):
And so that’s what we can do here. You can see the formation of tetra-ubiquitin, for K63. These are just a gel to kind of illustrate the different types of chains we’ve been able to make, dimers, trimers, tetramers, that we then want to throw into solutions with UBQLN2 and see how they affect these behavior, okay?
Carlos Castañeda (00:23:06):
So what we did was we made dimers first, so these are K48 dimers and K63 dimers. And I hope you can see that if we quantify the UBQLN2 phase behavior under these conditions, you’ll see that we increased the number of ubiquitin units, and we can increase the range over which we see these behavior of UBQLN. I want to keep in mind that the ratios that I’m going to be showing here are ubiquitin to UBQLN ratios, over here. So just keep that in mind when you’re thinking about this. We did this on purpose, because we’re obviously changing the number of ubiquitins with these different types of chains. The longer the chain, right, it has more ubiquitin units.
Carlos Castañeda (00:23:54):
So that was one takeaway, right, increasing the number of units, increasing the valency, in other words, really can drive more phase behavior of UBQLN2 for a greater concentration range. But the other thing that we noticed, especially with K48 and K63, was how with K63, we actually promoted that phase behavior over a much wider range. These numbers, you can see here, at 1.2 to 1, we’re pretty much done with K48. But we can see K63 tetra-ubiquitin base behavior with UBQLN, almost at four to one, which believe it or not is one to one, if you’re thinking about UBQLN to K63 tetra-ubiquitin concentrations, right?
Carlos Castañeda (00:24:41):
So that’s a really drastic change on the phase behavior, and the thing that’s really important to recognize here, these molecular weights of these two proteins are identical. The only thing that we’ve changed here is the site at which the linkage has occurred, right? Instead of it being K48, here we’ve got K63, okay? So that’s the only difference between these two types of chains.
Carlos Castañeda (00:25:06):
We quantify this further using these turbidity assay experiments. I’m illustrating for you here, a lot of work that Yiran did, whereby she was showing how when we add … We’ll focus here on this tetra-ubiquitin K48 UB4. You can see how this phase behavior has eventually gone from just decreasing the phase behavior. And you can start to see a little slight enhancement of K48. But it’s really prominent with K63 tetra-ubiquitin.
Carlos Castañeda (00:25:38):
In fact, you can see how it’s almost a bit like reentrant phase behavior, in a sense that we’re switching, right? We’re promoting phase behavior and eventually inhibiting phase behavior, as you start going up to higher ubiquitin to UBQLN ratios. And we quantity this further using another ton of work, where we made these phase diagrams at different concentrations of UBQLN and illustrating for 63 and for 48 UB4, you can really see, at the very low concentrations of UBQLN, 12 1/2 micromolar … So we’re keeping that fixed, and we’re adding K63 or K48 … K63 really promotes this phase behavior towards physiological temperatures. You don’t see that for K48. You can start to see K48 show up in these plots. But by and large, 63 is really driving this phase behavior over a much wider range than K48 tetra-ubiquitin is.
Carlos Castañeda (00:26:41):
Why is this interesting? Why is this really exciting for us? So over the last couple of months, we’ve been focused on looking at the literature. There’s actually a recent pre-print, by Paul Taylor’s group, that came out a few days ago, whereby, they’re showing that K63-linked polyubiquitin actually accumulate in heat-induced stress granules. And so these images UTOS cells, this is endogenous polyubiquitin, you can see where you have these heat-induced stress granules. We see this enrichment of K63. It turns out to be a lot of K63 ubiquitinated G3BP and not so much for K48. And this is replicated again when he does a knockout and then adding in this G3BP. And you can see this enhancement with this K63 in the stress granule but not so much for 48.
Carlos Castañeda (00:27:40):
Similarly, Buchberger’s group, in a paper that was published a couple of months ago, has performed SIM microscopy. And then here’s looking at G3BP heat-induced stress granules in HeLa, and you can see how K48s are on the periphery. K63s tend to be on the periphery but also on the inside of these granules. And so I thought this was a really interesting observation, given that we know that UBQLN can go into stress granules, given that I just showed you that K63 can really change the behavior of UBQLN’s phase behavior versus K48, right?
Carlos Castañeda (00:28:19):
Okay, so we wanted to figure out what’s going on, on a molecular level, right? So what is happening? Is there a change in binding affinity? Is there a change in the conformation of the chain? Does that impact how this is working, the accessibility of the binding patch for ubiquitin, right? So we turned to doing NMR again, and we switched this time, instead of focused on full-length, we went to our smaller construct, 450 to 624. And we were able to recapitulate the same behavior, whereby when we add 48 chains to 450, we see a similar effect, with 63 really extending this regime of phase behavior, whereas 48 does not do this. And so we were able to use this shorter construct to really map out what was going on when we add these different ubiquitin chains.
Carlos Castañeda (00:29:19):
Okay. So we did NMR titrations, again. The first one I’m going to show you is looking at UBQLN, so this is looking at this 450 to 624 construct. And you can see, that as we add in, right, these ubiquitin chains, you can see that the chemical shift perturbations end up being very close to each other, right? They’re very similar, so we think this means that the binding interactions are similar, across these two types of chains. We also monitored the binding affinity by tracking these residues, as we titrated in these ubiquitin chains. By and large, they end at the same place. But if you pay really careful attention, you’ll notice that there is a change. These curves on the K48 side look like they’re a little bit more sigmoidal, right, than they do over here.
Carlos Castañeda (00:30:14):
So we were really curious about why that is and what’s the origin of that. Well, what I haven’t told you yet are what these conformations of these chains look like. So from these two chains, I want to impart that these images that you’re looking at are representative. They are not the only images, these types of tetra-ubiquitin chains. I don’t want you to leave this talk, thinking tetra-ubiquitin always looks like one of these two forms. That’s not the case. These ubiquitins chains, if there’s anything I learned from my postdoc, it’s that they’re very dynamic, right? K48 can exchange between open and closed conformations. K63 can occasionally become compact. A lot of things can happen that can change these conformations of these chains.
Carlos Castañeda (00:31:05):
So what I’m depicting for you here are structures from Cynthia Wolberger’s lab, where I’ve labeled one of the ubiquitins, so you can kind of get an idea for where they are. I’ve also highlighted the hydrophobic patch, so you can see where it is. And I think you can see that the hydrophobic patch, the binding part, is not able to access, say, another protein, because it’s interacting with the hydrophobic patch of another ubiquitin. Not so for K63, whereby these hydrophobic patches are already exposed, right? So what we think needs to happen is in order for K48 tetra-ubiquitin to interact with something, it needs to open. It needs to open a little bit, so that this patch can bind to a receptor protein, in this case, for instance, the ubiquitin binding domain of UBQLN.
Carlos Castañeda (00:31:57):
So what we did was we switched the experiment. So now we’re going to look at the ubiquitin side, and we can selectively label this ubiquitin, that I’m going to call the proximal ubiquitin. It’s the last ubiquitin. Its C-terminus is free. That’s why it’s called proximal, because you can imagine it would be attached to a substrate, so it’s proximal to the substrate. And if you look at the comparisons of the chemical shifts of this ubiquitin versus monoubiquitin, you’ll see that there are a lot of shifts that correspond to residues 8, 44, and 70. These are the hydrophobic patches. These are these yellow spheres that you see, right?
Carlos Castañeda (00:32:43):
On the other hand, K63, you can look at the chemical shift perturbations, and you really don’t see a whole lot happening here. That’s because these patches are well-exposed, right, when you compare these to monoubiquitin. Okay, so what happens when we add UBQLN? So I just told you, right, looking at this proximal ubiquitin of tetra-ubiquitin, you can see that these residues are not in the same place as they would be in 63. That’s shown here, on the NMR spectrum, of looking at residues 40 and 41, which are pretty much at this hydrophobic interface.
Carlos Castañeda (00:33:24):
But as we titrate in UBQLN, what we think is happening here is that K48 tetra-ubiquitin is opening up, to accommodate the UBA binding that’s happening here, right? And what’s interesting, you’ll notice that the shifts are right on top of each other, really indicative that they’re binding in a similar way. But what had to happen, in the beginning, is that we needed to change. We needed to open up this K48 tetra-ubiquitin, so that the UBA domain can bind. When that happens, now you start to see that the micro-environments of these two proximal ubiquitins are really similar to each other. And that makes sense, given what we’re saying about how we think we’re having to open up this protein.
Carlos Castañeda (00:34:12):
All right. So this is what we think is happening, and this we think is the origin behind that sigmoidal behavior that we see with the K48 titration, when we’re looking at it from the UBQLN side. We then wanted to ask whether K11 and other chains, like M1, how do they behave when we add these to UBQLN droplets? Why do we focus on these? Well, because we hypothesize that these accessibilities of these hydrophobic patches really matters for how these different types of chains would modulate phase behavior. You can see with K11 dimers, shown here, that these hydrophobic patches are kind of close to each other. They’re kind of compact, the overall conformation is, whereas with M1 chains, they look really similar to K63 in that they’re extended, and these patches are more exposed, right?
Carlos Castañeda (00:35:09):
So we then made these chains, and we repeated the same experiment that I showed earlier, where we added K11 tetra-ubiquitin or M1 tetra-ubiquitin, to UBQLN, and what I think you can see is that K11 and K48 had really similar behavior. You can see that they’re both largely driving the elimination of phase behavior, by the time we reach 2.4 to 1, ubiquitin to UBQLN ratios. But M1 actually extends the regime, where we see phase behavior of UBQLN, out to even higher ratios than what we saw for K63, so leading us to think that maybe these more extended chains can promote this phase behavior of UBQLN, whereas these compact chains of K11, K48 are inhibiting the ability of UBQLN to phase separate.
Carlos Castañeda (00:36:09):
We quantify this even further by making these phase diagrams. I think, once again, you can see that what’s happening here is that we are promoting phase behavior, so you can see this downward trend, as we initially titrate in ubiquitin, right? Eventually, as this starts to come back up, because they’re going to eventually drive the disassembly. And you can see how that is quantified on this plot over here.
Carlos Castañeda (00:36:37):
So we think, these 2, M1 and 63 chains, are providing this multivalent platform, right? It’s a multivalent ligand to promote phase behavior of UBQLN. Not so much for K48 however, so we hypothesized that if we increased the accessibility to these ubiquitin units, that would promote phase behavior. Phase behavior that is now partially driven by homotypic interactions of UBQLN, right, and the higher concentrations, where we’re now involving heterotypic interactions, as well, all right, heterotypic meaning UBQLN, as well as polyubiquitin.
Carlos Castañeda (00:37:26):
And so we decided to turn to a construct that was developed by the Shu group at UCSF. This is a HOTag. This is a 30 amino acid peptide that can make a well-defined tetramer. And what we do here is we essentially added this little glycine residue linker, in this ubiquitin shown here. And you can imagine that here, these ubiquitins are really accessible. And our hypothesis is that this would drive more phase behavior over an extended range, than what we see for the other types of chains.
Carlos Castañeda (00:38:05):
And so when we did this experiment, this is exactly what we saw. So we see here, this HOTag ubiquitin construct promoted this phase behavior over much, much larger concentration range than M1 or K63, and far more than K48, right? I want you to think of this HOTag construct as the multi-monoubiquitinated protein, because in many ways, you can imagine, here’s a protein that can have ubiquitin, right, at different sites. You can have different lysines that can be involved in the ubiquitin attachment. And so you can make this multi-monoubiquitinated protein.
Carlos Castañeda (00:38:46):
And in a way, you can see how that is really modulating the behavior of UBQLN to phase separate. And you can see this, again, reflected here, whereby as we increase the HOTag protein here, we titrate in more and more. You can see that the regime over which we’re seeing these droplets extend to a far greater range. You also see that these droplets are getting larger. I think a lot of this has to do with where we are in the phase diagram of these different plots. And I’d be happy to talk about that afterwards.
Carlos Castañeda (00:39:23):
We then wanted to ask, what’s going on with polyubiquitin side of the story, right? Where is polyubiquitin? Is polyubiquitin found inside these different types of UBQLN2 condensates? And so we fluorescently tagged the ubiquitin chains, and tagged the UBQLN. You can see, at these low ratios of .1 to 1, that the UBQLN is clearly in the droplets, definitely need UBQLN to be in the droplet. But I think you can see that K48 tetra-ubiquitin has a very minor enhancement inside the droplet verus outside. This is definitely not the case as you go to the HOTag, where you can see there’s clearly more of this protein, this ubiquitin chain, in the droplet, than you find outside.
Carlos Castañeda (00:40:15):
And as we titrate and we go to higher stoichiometric ratios, one to one in this case, ubiquitin to UBQLN, you see it’s even clearer now, that K48 really no longer has a preference for being inside or outside the droplet. And same thing with these different types of chains here too. You can still see that the HOTag and M1, these tend to be in the droplet more so. You’ll notice that there’s an increase in the dilute phase. And that’s because we’re titrating in more and more of the ubiquitin chain. So there’s clearly going to be more that’s present in the dilute phase.
Carlos Castañeda (00:40:57):
We can quantify this. And so we ended up doing this by a centrifugation experiment, where we would induce phase behavior of UBQLN or UBQLN and polyubiquitin chains. And we would centrifuge. We’ll get the dilute, the dense. We can measure the concentration of the protein in the dilute phase. We can measure the concentration of a protein in a dense phase, by essentially, isolating the pellet and resolubilizing it with urea and then measuring the protein concentration in the dense phase.
Carlos Castañeda (00:41:33):
These are the kinds of things that we now look at. We were inspired by do this, partially by Kiersten’s work, recently published in PNAS, really is a good idea to monitor, directly, the concentrations of the proteins that you see in these droplets, right? I think you can clearly see that, as you add these different types of chains, you’ll see this promotion of the phase behavior. The concentration, the saturation constitution, which your driving behavior, decreases from here all the way down to really where we are for physiological concentrations of UBQLN2.
Carlos Castañeda (00:42:15):
This is really promotive for the HOTag, and in fact, the order of this is exactly the same order as we saw with the enhancement of phase separation. 48 really doesn’t do much to that saturation concentration of UBQLN, before eventually driving the disassembly of those droplets. So K48 generally destabilizes phase behavior, but you can see, with these other chains, these saturation concentrations initially decrease before they come back up.
Carlos Castañeda (00:42:46):
We can also do this by looking at the polyubiquitin signal. And when we measure the dilute phase… So the dilute phase are these solid lines. The dotted lines are the dense phase measurements of these polyubiquitin chains. You can see how the dense phase of these polyubiquitin are significantly lower than the dense phase concentrations of UBQLN. And again, the order of which you see this corresponds at a same order that we saw for the phase, whereby the HOTag is really the one that’s driving more phase behavior over a greater range. That’s exactly what we’re seeing here. But even still, even when we look at these, this orange and the purple lines, shown here, those dense phase concentrations of ubiquitin are nowhere close to where UBQLN is, and that’s an interesting observation in that, most of the UBQLN is not bound to ubiquitin.
Carlos Castañeda (00:43:48):
In fact, you have UBQLN maybe at least twice as much or three times as much as the ubiquitin concentration, as you see here. So that, again, I think, speaks to this balance of interactions that are taking place inside the droplets, where we have homotypic and heterotypic interactions that are modulating the phase behavior of these polyubiquitin UBQLN droplets.
Carlos Castañeda (00:44:13):
So where does this leave us? Where does this leave us, in terms of thinking about the physiological significance of this? What we think is that these different types of chains are driving different types of behavior for phase separation. So I showed you with K11 or K48, these chains will largely drive the disassembly of UBQLN droplets. M1 or K63 chains can largely drive the promotion of phase separation, before eventually leading to disassembly, right? So what’s interesting are that K11 and K48 chains are really involved in signaling things for the proteasome, whereas K63, especially, and sometimes, actually, M1 chains are involved in signaling things for autophagy.
Carlos Castañeda (00:45:02):
So we’re postulating here that maybe you could have, say, a stress-induced condensate, where you’ve got ubiquitinated proteins. UBQLN could, for instance, enter via its phase separation capabilities, but depending on how it interacts with these chains, you can have different outcomes. You can either have disassembly. Maybe UBQLN can extract this protein out, or maybe it can drive another type of process, whereby for instance, it could be the onset of an autophagosome. Why do I bring that up?
Carlos Castañeda (00:45:36):
Because there have been some really recent interesting papers that have shown how p62, which is another shuttle-like protein, has a UBL UBA-like architecture, it works together with K63-like chains to make a pre-autophagosome structure, which will then recruit other components to make the autophagosome, which is more of a membrane-bound organelle, right? Interestingly enough, there’s another shuttle protein that prefers to interact with K48 and make a stress-induced nuclear body over here. What’s interesting about this protein is that it actually has multiple ubiquitin binding domains that might be necessary for it to promote phase behavior with K48.
Carlos Castañeda (00:46:23):
So we think that these different types of chains are modulating phase behavior outcomes, that you could see inside cells, and we’ve been using the UBQLN protein, as an example of that. And so we wanted to ask, then, well, let’s try to mimic what the inside of a cell could look like, right? Let’s imagine, we’ve got UBQLN around. Let’s imagine, we’ve got a lot of ubiquitin around, right? So we’re going to have an in vitro experiment here, where we’ve got 50 micromolar UBQLN, 50 micromolar ubiquitin. But what we’re going to do is we’re going to add E1 and E2 enzymes that are going to make K63-linked chains. And what we’re going to do is we’re essentially going to add them and start recording and watch, over time, as to what happens.
Carlos Castañeda (00:47:10):
So we have tagged here UBQLN, right, so just UBQLN is fluorescentally labeled here. And what I think you can appreciate are the formation of these droplets that are occurring, over time. These are UBQLN droplets. It makes sense that the experiment started with no droplets, because we just have monoubiquitin around them. By the time we reach this point now, we have a lot of UBQLN droplets that are formed.
Carlos Castañeda (00:47:38):
We look at this really carefully, and you can see how as we see those droplets form, you can see the appearance of these ubiquitin chains over here, right? And that’s coincident with what we think is occurring, whereby we are changing the polyubiquitin and monoubiquitin ratio. We’ve made a polyubiquitin chain. We still have a lot of monoubiquitin around, but we’re making K63 polyubiquitin around.
Carlos Castañeda (00:48:05):
This is exactly what we think is happening inside cells, where you have this modulation of activity, based on the types of chains that are present. So with that, I’m going to end with my take-home messages of how monoubiquitin can drive disassembly of these droplets, right? But these different types of chains can render different outcomes, where K11 and K48 can drive the assembly, but K63 and M1 can actually promote phase behavior with heterotypic interactions.
Carlos Castañeda (00:48:35):
And we think that these different roles of flexibility of the chain, the accessibility of the stickers, and the balance, they all contribute to phase behavior of the protein. And with that, I am going to thank my group, because without this group, I would not be able to tell you these awesome stories. So here we are in our pandemic photo. I want to thank our collaborators, especially Paul and Tanja Mittag, for also being really great scientists and helping us to think about these problems, as well as our funding sources, and thank you for listening. I’d be happy to take questions.
Jill Bouchard (00:49:17):
Worldwide applause for you, Carlos. Thank you so much. It was very beautiful talk. There’s a couple of questions I’m just going to summarize briefly, about the droplet sizes and if your turbidity assay picks up on that and if the K63 and the 48 are similar in size, if you can comment on that.
Carlos Castañeda (00:49:40):
So we’re asking about these types of droplets that you see, I think, in these microscopy images.
Jill Bouchard (00:49:45):
Yeah, I think so.
Carlos Castañeda (00:49:47):
These droplets are getting larger and stuff. Yeah, so all of these experiments are done in a single temperature, right? So they’re 30C, and you can see that they’re at 50 micromolar concentration, right? And I think what’s happening, right, the reason why we’re seeing these larger droplets, with some of these, is because of where we are on the phase diagram of this system, right, because we’ve got those tie lines. And I think what is happening here is that those concentrations, the dilute and dense, right, they are going to change a little bit, based on whether polyubiquitin is around.
Carlos Castañeda (00:50:29):
And what can happen is that the droplet size can change as well, right? And I think that’s what we’re seeing here, because if you look at some of these phase diagrams, that you can imagine, here we are at 30. With 48, we’re not really that deep, okay? This is not the exact same concentration as what I showed on those microscopy images, so bear with that for a second.
Carlos Castañeda (00:50:52):
But let’s just imagine this was 48 and this was 63, for example. You can see that we are deeper in the 63 diagram, where here, in this case, the M1, than we are with the other one. And that would affect the droplet size. At least, that’s the idea that I think I have, or that we have, about why we’re seeing that. The turbidity assay experiments, yeah, so we typically just monitor at 600 nanometers here. It certainly could be affected by droplet size. I could imagine that. However, we do see that a lot of the trends that we observe, with the turbidity diagram, we still can recapitulate when we do our microscopy, when we’re spinning down the samples, right, and getting our dilute and dense, and we still see similar trends. I hope that helps.
Jill Bouchard (00:51:53):
Great. For Michael and Fettah out there, if that doesn’t resolve, go ahead and shoot us another one, and we’ll get you in towards the end, then. But we’re going to just start by Charlotte Fare has a couple of really thought-provoking questions, and we’re going to unmute her now, hopefully.
Carlos Castañeda (00:52:09):
Okay. Hi, Charlotte.
Jill Bouchard (00:52:14):
We’ll get there. Hold on just a second. Or did she… yeah, she’s here. Charlotte, can you get in there?
Charlotte Fare (00:52:27):
Jill Bouchard (00:52:29):
Charlotte Fare (00:52:29):
Jill Bouchard (00:52:29):
Charlotte Fare (00:52:29):
Can you hear me all right?
Jill Bouchard (00:52:30):
Carlos Castañeda (00:52:30):
Yeah, I can hear you just fine.
Charlotte Fare (00:52:31):
Okay, great. That was a great talk. So the first question I had was sort of related to this idea of droplet size, and I was wondering if sort of the mesh size or the pore size of the droplets is larger in the extended ubiquitin chain droplets, because that might be one reason why the droplets look bigger, just because there’s more space between the molecules.
Charlotte Fare (00:53:00):
And then the second question that I had in the chat was sort of related to that, if the droplets formed by 48 and 63 ubiquitin chains are differentially dynamic, because that would sort of fit with having more space?
Charlotte Fare (00:53:18):
And then the third question I had was if you expect the behavior of UBQLN to change when the ubiquitin is actually attached to a protein substrate, sort of like a tethering.
Carlos Castañeda (00:53:36):
Okay, so these are great questions. So for the meshwork kind of question, this is something we’ve been thinking a lot about recently. And I’m happy to hear people’s suggestions on this because we are trying to do some tracking experiments, some single molecule tracking experiments, where we’re putting in some beads, right, that we can then kind of maneuver within these droplets and try to get a sense of whether their dynamics are a little bit different, in the presence of say, 63 or say, one of these other types of chains, to kind of get at that question, because I think you bring up a really good point of how these different types of chains could be modulating or could be affecting the inside of the droplets.
Carlos Castañeda (00:54:23):
I think, if you guys don’t mind me sharing. We’ve been trying to do an experiment (this is actually quite hot off the presses) where we were using FRAP, so fluorescence recovery after photobleaching experiments, to kind of look at UBQLN inside of droplets and polyubiquitin in the droplets. The tricky thing about these experiments is I think many of you are pointing out what droplet size is. And we want to ensure that we’re looking at things on a similar scale. And so we actually tried really hard to make sure that these droplets are about the same size, so we are not biased by that.
Carlos Castañeda (00:55:04):
But what I think you can see is as we add in these different types of ubiquitin chains, UBQLN’s fluorescence recovery takes longer, as you would kind of expect, if the motion of UBQLN is sort of slowing things down. You can see, with the ubiquitin chains, right, they actually kind of speed up a little bit, as we’re going up to higher ubiquitin to UBQLN ratios. But then again, you’re also moving towards the regime, where you’re going to drive the disassembly of these droplets.
Carlos Castañeda (00:55:43):
But one thing that is clear is that with K48, for example, we don’t really see that much of an effect here, and I think, with UBQLN, that, I think, resonates with very little K48 enters these UBQLN condensates. So I think K48 really seems to favor the dilute phase, if you will, to kind of use the language of polyphasic linkage. I don’t know. These are early days. I really would like to hear more on what people think about these results and what that means, because we’re head scratching a little bit, just thinking about what all these different types of chains are doing inside the droplets.
Carlos Castañeda (00:56:24):
Your second question, Charlotte, I think, was about if you had the ubiquitinated substrate versus-
Charlotte Fare (00:56:33):
Carlos Castañeda (00:56:34):
… just ubiquitin chains. Yeah. That’s such a great question. That’s exactly where we want to go, for two reasons: one, because I think there are some substrates that will interact with UBQLN. There is a literature out there that talks about certain client proteins that prefer to interact with UBQLN over others. An example could be mitochondrial transmembrane proteins, because UBQLNs are known to chaperone those. And I think there’s some data that suggests that they can interact with the substrate, in addition to interacting with the ubiquitin tag on the substrate verus, say, just the ubiquininated substrate that doesn’t interact with UBQLN, only via its ubiquitin chain.
Carlos Castañeda (00:57:24):
But yeah, that’s a great question, that I wish I had the answer for, as to what those things are doing. Does that help a little bit, Charlotte?
Charlotte Fare (00:57:33):
Yeah, thank you so much.
Carlos Castañeda (00:57:35):
Jill Bouchard (00:57:37):
Awesome. You guys are… I’m loving this conversation. And we’re going to move to Guoming Gao, if you would like to talk.
Carlos Castañeda (00:57:50):
I’m trying to peruse the …
Jill Bouchard (00:57:52):
Yeah, there you are.
Carlos Castañeda (00:57:53):
… the chat.
Guoming Gao (00:57:53):
Hello, can you hear me?
Carlos Castañeda (00:57:54):
Jill Bouchard (00:57:55):
Yes, we can.
Guoming Gao (00:57:56):
Yeah, great talk. Yeah, awesome talk. Thanks for that. Yeah, so my question is actually, so I guess that the size change, so assuming that we have the same concentration all through the whole phase diagram, and we also have the same time, the same incubation time, to do the imaging. And I am proposing that maybe what we see of the size difference is a difference in the coarsening kinetics, right? So is it possible that the ratio between ubiquitin and UBQLN could affect the coarsening dynamics of this phase separating process? Is that a possibility?
Carlos Castañeda (00:58:39):
Yeah, it’s another fantastic question, right? We’ve always been interested in thinking about the kinetics of this process. And yeah, I wish I could say more about the kinetics, because I don’t really know, to be quite honest. That’s a great point, right? There could be an effect that you’re seeing here with the assembly, right, of these different types of droplets, based on these different types of chains, absolutely. And yeah, maybe, M1, K63 are quite different, for instance, than K48. Yeah, I mean, we have, I don’t know, Thuy or someone else in lab can probably speak to this, if they’re there, but when we have done all the imaging, we’ve been trying to be extremely careful with how we’re doing this, with making sure it’s the same time, after we incubate these samples, so that we’re not observing different kinds of effects.
Carlos Castañeda (00:59:44):
I think that these droplets, they do change in size, over time, but that’s mainly because in these microscopy experiments, our samples are just settling, right? You have droplets floating around in solution, and they’re just going to settle. Yeah, so that’s the other complicating factor in trying to interpret some of these data. But that sounds like a really good experiment, maybe something one could do, even with optical tweezers or something, to kind of look at the kinetics of that.
Guoming Gao (01:00:19):
Yeah, thanks for that.
Carlos Castañeda (01:00:20):
Sure, thank you.
Jill Bouchard (01:00:27):
We are going to move to Rahul now.
Carlos Castañeda (01:00:31):
Okay, I’m trying to scan the …
Rahul Samant (01:00:34):
Hello, can you hear me?
Jill Bouchard (01:00:36):
Carlos Castañeda (01:00:36):
Yeah, I can hear you.
Rahul Samant (01:00:37):
Okay, cool. Hi, Carlos, that was just very cool. I was wondering if you’ve had a chance to play at all with branched ubiquitin chains. I’m guessing in vitro, I don’t know how hard that is to do. But if not, I guess I was just wondering if you have thoughts about how that might influence LLPS during different parts of stress, so do you think adding branch points, for example, adding a K48 to a K63 chain might be a way to change the fate of those condensates, when you have different phases, stress recovery, for example?
Carlos Castañeda (01:01:12):
I think that’s a great point. And I know I’ve been thinking about the branch chains, because as I think your work, too, has shown in the past, right? You can have K11, K48 branched chains. I think there’s some literature about that and how that’s actually a really effective signal for proteasome. You can have K48 K63, which would be really interesting to look at here, because they’re kind of opposite of each other. Yeah, I feel like this kind of opens up a world of opportunities, right, to be thinking about how polyubiquitin chains are not just chains that have different linkages and may interact with the binding partners differently, but how these different chains may also affect these droplets form or they don’t form. And maybe you can tune, right, the ratio of ligands. Maybe instead of making it really broad, like you do with some of these longer chains … Let me see if I can get to this point. Yeah, so if you’re looking at this, for instance, right, here we’ve made it really broad, with M1 or this HOTag system, multi-monoubiquitin system.
Carlos Castañeda (01:02:28):
Maybe you had K48 K63 could change the brown and the blue lines, so that it’s sort of in between, or maybe it’s a sharper transition, potentially, right? I think that opens up a world of opportunities to think about these different types of chains and what they’re doing inside cells.
Rahul Samant (01:02:47):
That was kind of going to be my followup. And do you think the effects would kind of be almost like an additive or subtractive effect, or do you think one might win over the other? I guess I have no real thoughts for the biophysical things…
Carlos Castañeda (01:03:02):
I feel like it’s going to be very context dependent. I think that it’s going to depend on how the protein that the ubiquitin chain’s interacting with, how its phase behavior is modulated. Because in our case here, right, UBQLN is phase separating, but another protein could have very different behavior. Maybe it prefers one chain over the other.
Carlos Castañeda (01:03:29):
Before I got into all these different chains, there’s a lot of literature to suggest UBQLN doesn’t care what linkage chain it interacts with. But what’s interesting with that literature, they’re mainly just focused on the isolated UBA domain. And when they did studies with the isolated UBA domain, you didn’t see a preference for 48 or 63. But there are UBA domains, for instance, with Rad23, whereby the UBA domain is linkage-specific. And that could, right? I think the binding affinity differences and how the rest of the protein can play a role. It’s just like this complex map, I feel, how all these different interactions that are occurring, that can really modulate with what you see. It’s like a world of opportunity.
Rahul Samant (01:04:24):
Thank you so much. Cheers.
Carlos Castañeda (01:04:25):
Yeah, great question, Rahul.
Jill Bouchard (01:04:27):
And speaking of other interactions, we have a question from Matt about more titrations. Matt, can you unmute? There you are.
Carlos Castañeda (01:04:36):
Matt Wohlever (01:04:42):
Yeah, can you hear me?
Jill Bouchard (01:04:42):
Carlos Castañeda (01:04:45):
Yes, I can hear you.
Matt Wohlever (01:04:45):
Hi, beautiful talk. I was just curious. You see this wonderful effect with the isolated ubiquitin. Do you see anything with the UBL domain, especially with the interaction of the UBA and UBL domain within UBQLNs?
Carlos Castañeda (01:05:01):
Right. Yeah. That’s a great question, so for this work, we’ve been mainly dealing with the full-length proteins. So I will certainly say that the UBL domain is playing a role, because as you point out, it can interact to the ubiquitin binding domain. It’s definitely a complicated question. We have currently a preprint on bioRxiv, where we are teasing apart the UBL and the UBA and how they contribute to UBQLN’s phase behavior.
Carlos Castañeda (01:05:35):
But that paper doesn’t have any ubiquitin chains added to it. It’s just the UBL or the UBA, and then trying to look at the effects. And it’s an interesting set of results, because it’s such a complicated story. That UBL we find, it actually kind of interacts with the STII domains, weakly. And so when polyubiquitin is binding, let’s say, to the UBA domain, we think that it changes the rest of what’s going on with UBQLN. Because now the ubiquitin-UBA interaction is taking place. The UBA is no longer able to bind to the UBL because the UBL interaction of UBA is much weaker than UBA to ubiquitin.
Carlos Castañeda (01:06:25):
So you can have that UBL now interact with the STII regions. And as I showed in the very beginning, one of those STII regions, you need for phase behavior. So then, you can imagine you’re kind of occluding one of the STIIs that you need to modulate phase behavior. And so that could modulate the phase behavior of UBQLN2, even further.
Carlos Castañeda (01:06:51):
So it’s a complicated thing, and so one of the things that we want to do is to kind of look at what’s the conformation of the protein–in the droplet, outside the droplet, with ubiquitin, without ubiquitin–to kind of get at that question, because that is an excellent point, right? I think the next stage and one of the things my graduate students are trying to do now is we are focused on the ubiquitin chain. What happens now, if you throw in interactions involving the UBL. And what does that do? Because we can imagine that can even further modulate all of this behavior that we see here. And that’s a really exciting project, I think, for the future. It’s a great question, Matt. Hopefully, that helps answer it.
Jill Bouchard (01:07:37):
We’re going to move to Kamran now.
Carlos Castañeda (01:07:43):
Kamran Rizzolo (01:07:45):
Hi, Carlos. Excellent talk. I’ve heard you talk in the past. It’s a pleasure to hear your stories again. So this is a very intriguing topic to me. I used to work with the proteasome storage granules in yeast, and they appear under quiescence. And what we did is we actually tried to map its composition, right? And what we find is that there was monoubiquitin in there. It was just proteasomes, mainly, and monoubiquitin. And that was a requirement to make the PSGs. And we did show that.
Kamran Rizzolo (01:08:26):
In addition, there’s that Saeki paper, which you mentioned, where they also require ubiquitin in the proteins to form these also proteasome-like sort of foci, right, in the nucleus. I think it’s colorectal cancer cells, if I remember right.
Carlos Castañeda (01:08:44):
Kamran Rizzolo (01:08:45):
So it’s really interesting to see, right, that there’s this sort of different function for ubiquitin. In some cases, it’s required to make the condensate, and then in other cases, it dissolves it, right?
Carlos Castañeda (01:08:59):
Kamran Rizzolo (01:09:01):
And you did touch a little bit upon that, at the end of your talk, so I just was wondering if you can just speak a little bit to that. Thank you.
Carlos Castañeda (01:09:07):
Yeah. Absolutely, and I think that, again, is why it’s so interesting to see how context-dependent these interactions are with polyubiquitin. And I think that that’s actually one of the reasons why I got really interested in Kiersten’s paper from Rohit Pappu’s lab, because they’re trying to establish this framework idea that a multivalent ligand, depending on how it interacts with another protein that has, say, stickers and/or spacers, that can lead to very different outcomes, whereby that same ligand could act as a promoter of phase behavior, but depending on the system, depending how it interacts, it could act as an inhibitor of another system, for example.
Carlos Castañeda (01:09:53):
And I think that’s exactly what we might be able to see with these different types of condensates, that sometimes require polyubiquitin chains. Or it sometimes requires specific types of chains, Or maybe they have different preferences for mono versus poly. Yeah, I think that’s the idea, and that’s where it gets exciting for me, just to think about how these different proteins will behave with these chains.
Carlos Castañeda (01:10:23):
I don’t think it’s easy to predict until you kind of can characterize what’s going on with each of the components.
Kamran Rizzolo (01:10:30):
Carlos Castañeda (01:10:32):
Yeah, but that’s a great point, Kamran. I like that.
Kamran Rizzolo (01:10:35):
Carlos Castañeda (01:10:36):
Yeah, hopefully, that helps.
Jill Bouchard (01:10:39):
All right. We’re going to try for Yifan. I’m asking to unmute, see if you’re up for it. Maybe if not, we’ll go to Bede. I know you guys have been dying to talk to each other. Yifan, are you there?
Carlos Castañeda (01:11:01):
I can read his question, though. It’s how the volume fraction of the dense phase is determined, is that right?
Jill Bouchard (01:11:10):
Carlos Castañeda (01:11:10):
Yeah, so this is a tricky thing. So we’ve been working on ways to assess the dense phase accurately. I think my lab mates are probably online. And I don’t know. They could probably answer this question better than I can.
Carlos Castañeda (01:11:29):
But one of the things we do with the dense phase is we will resolublized it, with urea. So we resolublize it with a certain amount, right? And then we’re essentially using our pipetter, to figure out how much volume of that dense phase that we have, once everything is resolublized. We’ve been really careful to try to do this multiple ways, using different fluors, for example, because it’s really tricky to do this if you’re trying to measure both ubiquitin and UBQLN. So we have fluorescently tagged UBQLN one way and then a different fluorescently tagged ubiquitin. And then we will switch those and repeat it again and see if we get similar results, similar numbers, from that. And it turns out that we do.
Carlos Castañeda (01:12:20):
So that was encouraging. We’re trying to find other ways to kind of get at those measurements, because they are kind of tricky, right, to get at the dense phase. But we’ve been being really careful, I think, just to make sure that we’re not overdoing this or changing the numbers too much. I hope that helps.
Jill Bouchard (01:12:45):
I’m sure it does.
Carlos Castañeda (01:12:46):
I’m happy to talk more about that offline, too, if you want to do it by email.
Jill Bouchard (01:12:50):
That’s a great point, so we will make sure to include your email on the post, just so anybody can follow up with questions. Bede, if you are still around, go for it.
Bede Portz (01:13:02):
Hi, Carlos. Wonderful, as usual, with the talk.
Carlos Castañeda (01:13:05):
Bede Portz (01:13:05):
So I know you’re interested, also, in proline isomerases. And at the intro, you mentioned that this PXX motif in UBQLN is enriched for ALS-associated mutations. Do you know if any of them are predicted to disrupt recognition and action by proline isomerases, and have you started to explore what they do to your system?
Carlos Castañeda (01:13:31):
Right. It’s like the next big project is to figure out. So I think the answer to the second part of your question, yeah, that’s a really good question. How do the mutations, do the mutations affect how ubiquitin can modulate phase behavior of UBQLN? I think that’s the next thing we’re going to do. And we’re starting to think about that now, because it’s a complicated thing. In our Structure paper in 2019, we were able to show that when we had our 450 constructs and we introduced the point mutations there, when we added monoubiquitin, those droplets of UBQLN did disassemble, when we added ubiquitin.
Carlos Castañeda (01:14:22):
But that being said, I don’t know if that would be the same, right, for these different polyubiquitin chains and UBQLN2, and full-length UBQLN2 and introducing in those same mutations. So that remains to be seen.
Carlos Castañeda (01:14:41):
Right. So your other question has to do with the isomerase and whether the sequence within the PXX region is considered to be a substrate, potentially, for an isomerase. Yeah, that is a great question. So we think that there are at least two spots where you could potentially see a phosphorylated residue next to a proline, which is really reminiscent of some of the motifs that a proline isomerase would recognize.
Carlos Castañeda (01:15:18):
But the catch is that many of the mutations are to a serine, a threonine, or sometimes, a histidine. There is currently no evidence that those particular serines or threonines are phosphorylated. But the reason why there isn’t any evidence of that is because the traditional way of doing mass spectrometry, where you use, say, trypsin, right, to cut things up, there’s no lysine or arginine anywhere near that PXX segment. So you can’t cut up that sequence enough, so you could even make it fly on a mass spec.
Carlos Castañeda (01:15:58):
So we’re thinking about that a lot, because it’s a hypothesis that we’ve had for a while, that maybe those mutations are getting phosphorylated. And maybe that, right, is being recognized by a proline isomerase. And that is another …
Bede Portz (01:16:14):
Carlos Castañeda (01:16:15):
… another layer, exactly. Right.
Bede Portz (01:16:18):
Cool. Thank you.
Carlos Castañeda (01:16:19):
Yeah, thanks Bede.
Jill Bouchard (01:16:20):
Awesome. This conversation and discussion is so wonderful. And unfortunately, there’s some people that even gotten their question in yet.
Carlos Castañeda (01:16:27):
Jill Bouchard (01:16:28):
So for the interest of time and we’re starting to see people drop off, we’re going to just have one more question with you, Carlos, and then we’re going to send you the remaining, since you’ve graciously agreed to answer more questions, via email. We’ll make sure to post those with the video for everybody, along with your email address, so people can also followup with all of the stimulating conversation.
Carlos Castañeda (01:16:54):
Yeah, this has been great.
Jill Bouchard (01:16:55):
Erik Martin, you are going to be last but not least. I see you’re unmuted, so I’ll just be quiet.
Erik Martin (01:17:03):
Cool, I’m glad to have the distinction of a final question. Carlos, it was really interesting data. I don’t think I’d see a lot of this thus far. But I think my question is kind of echoing back to the questions you were getting at the very beginning. And one thing about this system that I find really interesting is that when you’re adding these ubiquitin polymers into the UBQLN droplets, you’re really kind of adding a polymer that has very different hydration properties than the UBQLN. And I was wondering, that seems to have very intriguing implications on sort of the surface tension of the droplets, as you change these ratios and potentially sort of the coalescence or coarsening kinetics, as you’d mentioned before. And I guess, when I saw the different sized droplets, that’s what was occurring, to me, that might be the source of it. And I was wondering if you’d kind of considered these properties of this system.
Carlos Castañeda (01:18:06):
That’s a really good point, too. So when you’re thinking about the hydration properties, one of the things we’ve been trying to do is trying to get a sense, at least, using scattering techniques, solution scattering techniques, about these chains and how they are moving around. And so we are doing radius of gyration measurements. And it was very clear that K48 is far more compact than, say, K63 or M1 chains, we think, based on the data that we have, thus far.
Carlos Castañeda (01:18:41):
Yeah, so that’s a really good point, right? That’s another factor that could be impacting how these droplets are forming and maybe have an effect on what other people were saying, too, about the mesh size of these droplets, when you’ve got ubiquitin around these polyubiquitin chains and UBQLN. So maybe, yeah, maybe there is. I’d love to find a way to look at that and to answer that question directly.
Erik Martin (01:19:15):
I mean, I think maybe some IR measurements could kind of get to this point, but yeah, it’s a very interesting problem to focus on. But yeah, I always thought, ubiquitin is great, because it could be progressively added into droplets. And you could kind of maybe even maintain some kind of homeostasis, because you’re progressively adding something more soluble into a droplet, that might be progressing towards sort of a gel-like or arrested state. So yeah, those observations are really cool.
Carlos Castañeda (01:19:49):
Yeah, that’s a really good point, too, because one of the things we wanted to think about, down the road, is over time, what do these droplets look like, right? And you might be onto something there, where these different types of chains could certainly affect those kinetics, over time, right?
Carlos Castañeda (01:20:07):
And then that has even further implications, right, for thinking about what these chains are doing, inside a biomolecular condensate in a cell, right? If you had a prolonged situation, where a condensate exists longer and you have one chain enriched versus another, in these kind of condensates. So that’s an intriguing idea. That’s a really good idea. Do you want to work on it with me?
Erik Martin (01:20:33):
Can it be so?
Carlos Castañeda (01:20:37):
Yeah. No, it’d be really fun, because we really want to get a better sense of what’s happening, what the structure of this complex is inside the droplet, and that’s a really hard thing. We initially thought we could maybe do this with NMR. The unfortunate truth is that while other systems, you can use NMR to really probe the dense phase, UBQLN, I think, oligorimizes to the point where we just don’t see the signal anymore. So that takes that technique out. And so we’ve been thinking a lot about okay, can we model this? What’s the best way to model this? And so anybody who’s still out there: I’d be definitely interested in thinking about ways to model the system, because I think you are hitting on all the things that can modulate these behaviors that I think are really, really exciting to think about down the road.
Erik Martin (01:21:33):
Very cool. Thank you.
Carlos Castañeda (01:21:33):
Yeah, thanks, Erik.
Jill Bouchard (01:21:36):
Well, all I have to say is wow. That was one heck of a thought-provoking talk and very many questions in response to that. And so please be expecting some more lively discussion, via email, and posted online soon. And thank you, again, Carlos, for giving our community such a great treat today. It’s really been a pleasure for all of us.
Carlos Castañeda (01:22:01):
Great. It was a lot of fun, really appreciate everybody’s comments and thoughts on this. Again, if anybody has more ideas, please, we’re happy to talk about it, because I think we really want to get at this. I think there’s a lot of cool physics that are happening here.
Jill Bouchard (01:22:18):
Definitely. Very, very true statement. Well, with that, we will round out the day. And thank you, again, and we’ll see you all the next time we do one of these.
Carlos Castañeda (01:22:29):
Yeah, looking forward to Emily’s next. It’ll be great.
Jill Bouchard (01:22:32):
Yes. Great. Sounds good. And thanks again, Carlos.
Carlos Castañeda (01:22:38):
Yep. Thank you, Jill. Thanks, everybody.
Diana Mitrea (01:22:39):
Thanks so much, Carlos.
Carlos Castañeda (01:22:40):
Yep, thank you.
Jill Bouchard (01:22:42):
Carlos Castañeda (01:22:43):
Diana Mitrea (01:22:43):
Question from Elliot Dine: What are the dynamics of UBQLN2 recruitment to SG? Does it match the dynamics of Ub and Ub linkages in SG?
Carlos’ Response: This is a very good and complicated question to answer. Cellular localization of UBQLN2 appears dependent on concentration. Endogenously expressed UBQLN2 are found in stress granules but exogenously-introduced (overexpressed) UBQLN2 form UBQLN2-bodies that do not colocalize with stress granule markers. These bodies can sometimes form without stress (aside from the stress of having to deal with transfected DNA). We are actively working on approaches to elucidate dynamics of endogenous UBQLN2 in stress granules, and polyUb in turn…
Question from Fettah Kosar: How do you quench the Ub-UBL interaction at different times to do the gel runs?
Carlos’ Response: I assume this question refers to the last data figure in which we show the K63 chain reaction in the presence of UBQLN2. For that experiment, we set up a big reaction on ice, and aliquot a small fixed amount into multiple microfuge tubes on ice. We then put the tubes with aliquoted samples into the 37C incubator and put some of the reaction into a glass bottom dish and started imaging immediately at 37C. Every 3 min, we would add the same volume of 2x SDS-PAGE buffer into a tube with aliquoted sample to quench the reaction (the reason for aliquoting first is that when phase separation starts to happens, the droplets can adhere quickly to the tube surface and we “lose” protein in the solution over time).
Question from Harihar Mohan: Do K63 linkages also determine UBQLN2’s involvement in other forms of autophagy, for e.g., chaperone-mediated autophagy?
Carlos’ Response: A very interesting question, but not to my knowledge at this time.
Question from Elliot Dine: Hi Carlos, there’s good evidence from Noda lab and Roland Knorr’s recent paper that autophagy requires gel-like and not liquid-like substrates, and the K63 FRAP results might support that idea, though that’s probably over interpretation.
Carlos’ Response: I really appreciate this comment. Interestingly, UBQLN2 fluorescence recovery is a lot faster than p62 with K63 chains in Sun et al, 2018 (although they photobleached most of the droplets and we only did a portion). Preliminary microrheology experiments also show that UBQLN2/chain droplets are viscous.
Question from Fettah Kosar: With K63 Ub4, droplet sizes are relatively larger in the middle ranges of polyUb:Ubiquilin ratios. Curious.
Carlos’ Response: We imaged droplets of UBQLN2 with chains at the bottom of the coverslip after a 15-minute incubation period. When the solution is first mixed, we look at the solution and see many small-sized droplets (similar for all chains). These droplets quickly fuse, become larger and settle to the bottom of the coverslip. For lower ratios of M1 Ub4 or HOTag6-Ub, we would see a lot more droplets in solution in the beginning than for UBQLN2 alone or with shorter chains. We think the main reason for the difference in sizes has to do with there being more smaller droplets that fuse together to make bigger droplets for solutions with M1, HOTag6, and K63 than for K48 chains. These observations are consistent with our temperature-based turbidity results (a separate experiment) where the onset of UBQLN2+K63 phase separation occurs at lower temperatures than for UBQLN2+K48 under identical experimental conditions.
Question from Fettah Kosar: All images under the 0:1 column are the same image. I am wondering why, since each of these images represents a different experiment.
Carlos’ Response: We performed the imaging experiments several times, with very similar results. For these experiments, SEPARATE samples were prepared of UBQLN2 and designated amounts of polyUb chain, and allowed to incubate for the same amount of time prior to imaging. For this particular figure, we did all of the imaging at the same time and thought that it would be less confusing to use the same starting image for the UBQLN2-only sample. That seemed to have created more unintended confusion. Let us know if this does or doesn’t clarify the issues here.
Question from Michael Assfalg: How does turbidity depend on droplet size?
Carlos’ Response: We do think there is a dependence. We think there are more droplets that also get larger due to fusion over time and increasing temperature. The “turbidity” curves do flatten out at certain temperatures for each mix. When we assemble the low-concentration arm of the temperature-concentration phase diagram, we mainly look at the midpoint of the transition and not the absolute absorbance value of the readings. There are certainly caveats, but together with our dense and dilute phase calculations, we are confident that we are observing changes in saturation concentrations of UBQLN2 as a function of different polyUb chains.