VIDEO: Ankur Jain on Nucleic Acid Phase Separation

Ankur Jain (00:03:47):
So what we are interested in is how RNA contributes to the formation of these bodies. B, what are the principles by which RNA are specifically recruited or excluded from these bodies? Finally, once RNA does get into one of these bodies, how does it change it’s folding, its ability to base pair, bind proteins, and so on? Today, I’ll share with you some insights that we have gotten on how nucleic acid themselves phase separate. Most of this work is done by two incredible students, an incredible postdoc in my lab, Sumit Majumder, and an undergraduate student, Daniel Stein. I’m just doing the talking head for their work. Feel free to interrupt in between if you have any questions. It also goes for the people on the chat, joining us on Zoom.

Ankur Jain (00:04:41):
So, I’ll start with a quick refresher on complex coacervation. Nucleic acids bear charge and they readily undergo charge-mediated phase separation called complex coacervation. The idea here is that if you have a mixture of two oppositely charged polymers, if you mix them together, polymers will bind each other purely by electrostatic interactions, low valent signs will be released in the solution and this entropy gain, when low valent signs are released in the solution essentially drives complexation of these two ions. These complexes crash out from the solution. You end up with the polymer-rich phase, which contains a vast majority of the polymer and the polymer-depleted phase.

Ankur Jain (00:05:32):
Here, I’m showing one example from a paper from Sarah Perry, where these complexes are solid-like. They precipitate. They are this elastic rubbery state. Now, if these complexes or this mixture, if it can reach equilibrium, then the ion pairs will constantly remodel. As a result, these complexes will have liquid-like properties. This state is usually referred to as the coacervate state, which is the terminology that I’ll be using for the rest of the talk.

Ankur Jain (00:06:06):
Now, this term coacervate was coined about a hundred years ago, and I’m showing some excepts from this paper from 1929. de Jong and Kruyt, they observed these spherical clusters when they mixed gum arabic, gelatin and some other denatured proteins. From these excerpts, it’s clear that they were struggling to what exactly call them. They are unmixed, but it’s not really unmixing. It appears that there was also some conversation about whether these are phase separated or not. There was some discussion where somebody perhaps called that these visible separation may or may not imply that they’re phase separated. Eventually, they chose the term coacervation, which comes from coacervare in Latin, just means to pile up, and also goes along to show how this discussion over terminology has probably stood the test of time and has been ongoing for the last hundred years or so.

Ankur Jain (00:07:10):
Okay. These coacervates are all around us. They find many applications in the day-to-day products that we use. There are a part of conditioners, creams, for drug delivery, for making artificial or synthetic fats and so on. They’re also all around us in biology. One of my favorite example comes from the sandcastle worm. So this tiny marine creature lives underwater, and it builds this nest by gluing together bits of sand. So here are individual nests or individual dens of these worms. They are built by putting together individual grains of sand one at a time. In low tide, you can probably see it along the coast of California. They decorate the shore. Here’s an example of a larger colony.

Ankur Jain (00:08:06):
The way this animal puts glues together these grains of sand is quite interesting. It’s a very Herculean task to build a glue that’ll work under water. The solution these animals have come up with relies on polyelectrolyte complex coacervation. They have this glue, which is like epoxy. It’s a two-component glue, which do not see each other when it is inside the worm, secrets it out, mixes the two things together and applies them on individual sand grains. Now, the task is to keep this glue together underwater, and that is achieved by complex coacervation of peptides. These peptides are fairly low complexity. These serines get charged with phosphorus, get phosphorylated, and there is a complimentary set of proteins, which bear a positive charge.

Ankur Jain (00:09:03):
Now, moving back to nucleic acids, nucleic acids also are charged. If one mixes DNA or RNA with an appropriate polymeric cation, one observes complex coacervation. Here, I’m showing one example, where I’ve taken a single-stranded DNA, polyT. The DNA is labeled, but it’s a homogenous solution of DNA so you don’t see any inhomogeneities over here. If you mix it with a poly-cation like spermine, which is a metabolite present in virtually all cells, basic peptides, proteins, poly-lysine, supercharge GFP, get your poly-cation of choice, one ends up with a DNA-rich phase and a DNA-poor phase. The signal here is arising from a fluorophore that was conjugated to the DNA. We observe about a thousand-fold enrichment of DNA in these condensed phases and the rest of the solution is essentially depleted.

Ankur Jain (00:10:01):
These type of experiments have been done for several decades now. In the more recent times, Matt Tirrell, Sarah Perry, Christine Keating, and Evan Sprujit have used these systems to encapsulate things, to make peptide-RNA coacervates and so on. Now, one outstanding question in complex coacervation field, which may be pertinent to our biopolymer-based coacervates is how do short-range interactions, such as hydrogen bonding, cation-pi interactions, pi-pi interactions affect coacervate properties. This simple polyT single-stranded DNA provides us a relatively robust chassis to build on directional base-paring interaction, interactions where we know the binding energy, and we can then start proving, ask questions about the coacervate properties.

Ankur Jain (00:10:59):
So, what do I mean by that? We take a string of DNA and incorporate a few GC-rich patches. So here I have a 90-mer DNA, where I have incorporated three strings of GGATCC. It’s palindromic sequence. The GGATCC will base pair and it’ll provide a short-range crosslinking sites. The charge on the DNA is the same. These bases do not essentially change the charge per unit length of the DNA polymer. Now, if you mix these DNA with poly-cations, once again, we see complex coacervation and they can start relating the properties of coacervates with the binding energy. So the idea would be these DNA in the presence of poly-cations will form coacervates. Hybridization will crosslink DNA and change the properties of coacervates, which we can measure.

Ankur Jain (00:11:57):
So, does it actually occur? One property of coacervation or complex coacervation that it is sensitive to low valence cations. For instance, if you make coacervates out of pure polyT DNA, if we increase the amount of sodium chloride, a monovalent cation, these coacervates are going to dissolve. All the DNA will end up in the solution phase beyond 20, 30 millimolar sodium chloride. Now, if you take another DNA, in this case, which can base pair by having GGATCC site interspersed in the polyT backbone, these coacervates are more resilient to salt-mediated dissolution.

Ankur Jain (00:12:45):
We can play this simple game on and on. Here is another sequence where we have increased the length of this hybridization patch and hence hybridization energy. These are resilient to about 50 millimolar salt. We can play this game day in, day out. Base pairing is fairly well characterized and build these kind of phase diagrams as a function of binding energy and find the critical salt concentration, after which, these coacervates will survive.

Ankur Jain (00:13:17):
One point that I would like to emphasize here is that we don’t have a good way to assess what will be the binding free energy when the DNA has entered into the coacervate phase. What I’m plotting here on the X-axis is just an input parameter hybridization energy based on nearest neighborhood type calculation and some arbitrarily chosen solution condition. So put it in other words, this delta G is an input parameter, which we get from theoretical calculations. What we are measuring is the coacervate behavior. On the Y-axis, we are getting the experimental readout. Okay. So the take-home message from this slide is that base pairing perhaps can stabilize these complex coacervates. There can be an interplay of electrostatic interactions and stacking or hydrogen-bonding interactions.

Ankur Jain (00:14:12):
Now, what about the behavior of the polymer itself, DNA itself within this coacervate? One assay which has been used to assess polymer behavior is fluorescence recovery after photobleaching. Most of you are perhaps already familiar with this game. You photo-ablate a small region, watch how long it takes for that bleached fraction to recover and that reports on molecular mobility of the polymer. Now, when we do this experiment with polyT, we photo-ablate a small region and watch recovery. It recovers within a fraction of a second, much faster than our instruments response time. Play the movie again, photo-ablate a small region, it almost instantaneously recovers.

Ankur Jain (00:14:56):
Now, we can start building base-pairing interactions similar to the data I showed you in the previous slide. We, once again, end up with these spherical coacervates. However, in this case, the recovery timescales are about a thousand times slower, it recovers over the course of about 200 to 300 seconds. Because this is DNA, base paring is fairly well-characterized. We can play this game in, out, day in, day out. If you take an extreme case, where we are looking at… You can see dinucleotides repeated several times. We end up with these solid rubbery elastic precipitates, which do not exhibit fluorescence recovery upon photobleaching, indicating that the DNA molecules are just stuck in space and do not move around. So these are two extreme cases, and one data point in the middle. We can design any kind of desirable… If you have a recovery time in mind, you can go back and engineer base-pairing interactions to provide you that in the coacervate.

Ankur Jain (00:16:06):
Here, I’m showing collected data on about eight sequences. On the Y-axis is the fluorescence recovery. On the X-axis is the time on a log scale. As we move from left to right, we are looking at DNA with increasing hybridization energies. Now, how does this recovery timescale change with increasing binding energy? When we do that, we see a more or less exponential dependence. On this graph, on the Y-axis is the recovery timescale, the characteristic time extracted from the FRAP recovery curves. On the X-axis, again, is our input parameter, our theoretical estimation of hybridization energy of two DNA strands. We start seeing this more or less straight line on a semi-log plot indicating the timescale varies exponentially as the hybridization energy.

Ankur Jain (00:17:07):
Now, these FRAP timescales can be converted into polymer self-diffusion coefficient by using this kind of a relationship. If we know the width of the bleached spot, we can use a simplifying relationship to extract polymer self-diffusion coefficient. Once again, when we do that, we see this exponential scaling of the diffusion coefficient with the hybridization energy. So as we are increasing the hybridization energy, polymer diffuses slower and slower, we can start making predictions on what will be the diffusion coefficients if we know delta G. Now, again, just to emphasize this D-self is scaling as exponentially with delta G. I’m not showing the temperature data, but you get the point. As we increase the temperature, the polymer diffuses faster.

Ankur Jain (00:18:00):
So these base-pairing interactions change phase behavior. They change polymer mobility. You can also start asking questions about the macroscopic properties of these coacervates. There is another assay, where we can watch the fusion of two droplets. These fusion timescales are dependent on, A, the size of the droplet itself. These droplets have higher viscosity than the surrounding medium. Then surface tension is going to try to bring the droplets closer. So viscosity is going to prevent that. This relaxation timescale will be directly proportional to the droplet size, viscosity of the droplet and inversely proportional to the surface tension.

Ankur Jain (00:18:44):
We can watch these droplets fuse. When they initially start just kissing each other, the aspect ratio would be close to two if the droplets are of similar size. When they have relaxed to a spherical geometry, the aspect ratio will be one. When we plot this aspect ratio as a function of time, we see an exponential relationship similar to what many, many other people have described before. This allows us to extract the characteristic timescales of fusion. So each data point is a characteristic timescale extracted from an exponential fusion curve and we have done many experiments for droplet pairs of different sizes. These plots allow us to extract this ratio, viscosity to surface tension, also known as the inverse capillary velocity.

Ankur Jain (00:19:36):
Again, we can play the same game. We can do this for many, many DNA sequences and get this inverse capillary velocity as a function of delta G. To our surprise, we, again, see this exponential scaling. The ratio of viscosity and surface tension scales exponentially with the hybridization energy. This is a bit surprising and I’ll cut out some additional data here, but suffice it to say that what we find is that the surface tension of these droplets is not substantially changing when we change the hybridization patch and just the viscosity alone is scaling with the hybridization energy. So we see that phase behavior, polymer self-diffusion, the macroscopic properties of these coacervates are scaling with the interaction energy of these short-range patches.

Ankur Jain (00:20:35):
I’ll show one more piece of data. What happens if we mix two distinct polymers, in this case, a polyT DNA, which does not form any base-pairing interactions or appreciable base-pairing interactions, and another DNA, which can self hybridize? So here is one such movie. At time, T, is equal to zero, we inject a poly-cation in the solution which triggers complex coacervation. The two DNA strands are labeled with different fluorophores. What you’ll see over time is that in these complex coacervates, the two DNA strands are separating from one another. They’re spontaneously partitioning into two distinct compartments. We end up with a compartment, which is enriched in polyT, and another compartment, which is enriched in a sequence that can base pair.

Ankur Jain (00:21:33):
We can zoom into these droplets. So here in the red is the non-base-pairing sequence. In green is a sequence that can base pair. And what you can appreciate is that they form this crescent geometry with certain interfacial tension and the two DNA are self-segregated, the two compartments. Again, we can play this game as a function of binding energy. In the data that I’m going to show you for the next minute or so, the top slide with top image will be the non-base-pairing DNA, T-90. The middle row will be the base-pairing-competent DNA. The bottom will be the overlay. So in this case, the interaction patch is just for HGC, really low binding energy and the two DNA are virtually overlapped with one another.

Ankur Jain (00:22:27):
So we increase the binding energy. Again, we do not see substantial compartmentalization. Beyond a critical value, critical difference in hybridization energy, we start observing this compartmentalization–two DNA spontaneously going to two different compartments. As we increase the difference, we see a greater and greater partitioning. PolyT prefers to be on the polyT-rich phase. The base-paring competent DNA prefers to be in the other phase. If you have a keen eye, you’ll also notice that it’s not just the partitioning, but also the contact angles are changing. There’s some interesting scaling with binding energy in this scenario.

Ankur Jain (00:23:14):
Now, one can measure how much DNA is in the polyT-poor phase and polyT-rich phase, right? So if you have a compartment like this, we can calculate what is the concentration of our favorite DNA in one compartment versus the other compartment, right? That can give you information on partitioning. What we find is that up to a critical difference in binding energy, this phi-A to phi-B ratio is one, or in other words, we do not see any compartmentalization. As we start increasing this difference, we can reconstruct this entire coexistence curve, where the solution splits into a polyT-poor and a polyT-rich phase. It’s a really cute experiment and it took Sumit a really long time to get this data and he’s really, really proud of it. You may have seen these coexistence curves as a function of temperature or as a function of salt, if you’re dealing with complex coacervates. What I want to emphasize here, we are looking at this coexistence curve as a function of difference in binding energy experimentally.

Ankur Jain (00:24:35):
Okay. So that was a lot of physical chemistry. This scaling, exponential scaling with interaction energies, this is to the best of my knowledge is not really observed for complex coacervates, but there are precedents for neutral polymers. This famous theory from Rubenstein and Seminov states that when you have associative polymers, which can crosslink with one another at multiple sites, the lifetime of this association bond essentially dictates the macroscopic properties of the coacervates. So let me rephrase that. For this polymer to diffuse within a semi-dilute solution, in this case, a neutral polymer, which can get crosslinked at multiple sites, the diffusion will require breaking of polymer-polymer crosslinks.

Ankur Jain (00:25:28):
The lifetime of this bond will scale with the binding energy or the energy of that bond. Essentially, tau-bond would be some microscopic time, tau-naught multiplied by E to the bar E by KT, where E is the interaction energy. What we are finding is that this tau-bond in our case or this E in our case with DNA is just the hybridization energy. What this means is that perhaps the DNA, the charged polymer within our coacervates is essentially behaving like a neutral polymer in semi-dilute solution. Depending on how you want to think about it, either it’s really trivial or really profound. Again, I’ll repeat it. A charged polymer, charged DNA in the coacervate phase is following the same rules that you would expect a neutral polymer to have in semi-diluted regime.

Ankur Jain (00:26:38):
So, I’ll summarize some key lessons that we have learned from the simple DNA system. For the rest of my talk, I’ll delve into what insights we can get from these physical chemistry experiments with pure DNA and how they can relate to bimolecular condensates themselves. So, what I’ve shown you so far is that non-ionic interactions, in our case base-paring mediated, but you can draw analogies with protein coacervates, where you have hydrophobic interactions or pi-pi interactions for proteins, they may tune the material properties of complex coacervates. You may have encountered these kind of data in other publications, where TDP-43 or tau may form complex coacervates and certain mutations may change tau-tau interaction energies. As a result, there is a change in viscoelastic properties or material properties of the condensate.

Ankur Jain (00:27:40):
Number two, a kind of a corollary of the first one, complex coacervation in the material science world can be a route to generate micron-sized, self-assembling, DNA-based materials. You’re probably familiar with origamis, where we rely on the specificity of the DNA bond to create crystalline structures of the origami, where the charge on the DNA backbone is neutralized by low valence cations. Scaling up of origamis to larger dimensions, it’s challenging. Here, perhaps by relying on the negative charge on the backbone, we can create micron-sized or even larger materials. Third, again, I want to reiterate this point, charged polymers in the coacervates are behaving like neutral polymers in semi-dilute solution. That is perhaps the reason why some of the existing theories apply so well for protein-based condensates, which may have some component of electrostatic interactions. Yeah.

Diana Mitrea (00:28:52):
That was really beautiful. I just have a question. So I was surprised about the crescent architecture of those droplets. Can you comment a little bit what that means in terms of surface tension and-

Ankur Jain (00:29:06):
Absolutely. So I think Omar Saleh and others have done beautiful work on building these kind of systems with DNA nanostars. This interfacial tension essentially, the geometry that we obtain essentially reflects on the interfacial tension. What the system is going to try to do is minimize its total surface tension, which includes surface area, one interface of the coacervate with the aquasolution, plus some energy which will be at the interface. What we find essentially, maybe I have a backup slide that I can show in a bit, if the interaction energy difference is relatively small, we see a larger interface. The non-base pairing DNA always occupies the concave shape or the convex shape, the smaller side of the interface. As we start making the difference larger, the curvature changes. Is that where you were going or… Okay. There are some interesting theories, or there is only one interesting theory, relevant theory that we can find here. I’d be happy to chat about it after the talk. Other questions? Okay. Yeah.

Sangram Parelkar (00:30:33):
[To follow up on] Diana’s question. So you described the two-phase coacervates. I was just wondering from biological point of view, do you see something like a heterogeneous biomolecular condensate inside the cells?

Ankur Jain (00:30:46):
Maybe in five minutes.

Sangram Parelkar (00:30:47):
All right. I’ll look forward to it.

Ankur Jain (00:30:52):
Okay. The last bit is that the observations that we’re making may have implications for RNA and DNA delivery. If you have done plasmid transfections or delivered DNA or RNA or oligonucleotides to the cell, you’re often relying on cationic lipids or polyethylenimine, some positively charged polymer, and you’re making a complex. What our results show is that this complexation is going to be dependent on the sequence of the DNA or RNA molecule. If you have a GC-rich, let’s say a small RNA that you’re trying to deliver, the same cationic peptide, it may end up in these precipitate-like state, not sure how much of it’ll be available or how availability translates to the material properties of the coacervate that one is making. It may also have some relevance with intracellular condensates based on proteins. We are just extracting insights using DNA as a model polymer and the same rules may apply for protein-based condensates. Go ahead.

Sangram Parelkar (00:32:10):
[inaudible]

Ankur Jain (00:32:25):
Yeah. Yeah. So that is what many of these systems are designed to do when you make plasmid and mix it with polyethylenimine. The cation is selected and the concentration of cation is selected such that it can withstand the ionic strength of the medium. I guess analogous rules will apply when people are trying to deliver nucleic acids in lipid ions. Can you repeat that? Yeah. Okay. So, I’ll move on to what insights can this synthetic system provide on cellular RNP granule scenario, which is where our real interests lie.

Ankur Jain (00:33:08):
Now, I’ll briefly introduce one system that we have worked on previously, which comes from these nucleotide repeat expansion disorders. Just as a quick recap, these are genetic neurological diseases. Some better known examples are Huntington disease, fragile X syndrome, myotonic dystrophy and certain forms of amyotrophic lateral sclerosis. The cause of the DNA here is an unstable repeat, often GC-rich that gets copied too many times. For instance, in the case of Huntington disease, there is a CAG patch, which gets copied too many times. In unaffected individuals, there are fewer CAG repeats and the affected individuals have more.

Ankur Jain (00:33:52):
Now, one feature of these disorders, if one stains for the repeat-containing RNA, that accumulates in the nucleus and these domains refer to as foci. Here, I’m showing data collected for many, many publications. The schema is in blue is the nucleus. This staining is done using fluorescence in situ hybridization, staining for the repeat or the transcript that may contain it. Now, across this 10 or so diseases, what you’ll see is that in virtually all cases, the RNA is accumulating in these clustered domains inside the nucleus. You can recapitulate these domains in vitro or… By in vitro, I mean, in cell culture, by just expressing a repeat-containing transcript, and it correlates this assembly of RNA in these domains requires a threshold number of repeats. Furthermore, it correlates with disease and animal model. Several studies have used these foci as disease biomarkers for testing therapies that may potentially work to rescue symptoms.

Ankur Jain (00:35:05):
Now, mostly published data, what we showed about five years now is that these GC-rich RNA can form hairpin-like structures. If one makes hairpins longer and longer, these RNA may start having besides intramolecular hairpins, they can start forming intermolecular structures. I’m depicting three different strands of RNA in three different colors. In cells, we developed this minimal system where we put in some test sequence, test repeats, tagged with MS2 hairpin system, which just derived from phage. By expressing a cognate protein tagged with GFP, we can visualize the RNA. We put this system under a drug-inducible promoter so that we can watch these bodies in real time. This works fairly well if you put in 47 or above a threshold number of repeats. We observe RNA accumulating in these domains. In low number of repeats, it does not.

Ankur Jain (00:36:06):
This allowed us to demonstrate that these foci, which were absorbed primarily as static pictures and fixed cells, they are behaving like liquid, or at least RNA is getting exchanged both within these foci and with the nucleoplasm. So, what can our experiments with this physical chemistry experiments with single-stranded DNA tell us about these one class of RNP granules that we are making in cells?

Ankur Jain (00:36:40):
The first is that–first prediction: our model is that these foci or at least RNA recruitment to these foci is based on a combination of electrostatic and base-pairing interactions. Charge is perhaps getting neutralized by either cations in the cell or some polymers which bear positive charge. Now, in test tubes, we find that these coacervates are sensitive to salt concentration. What about in cells?

Ankur Jain (00:37:06):
So here, if we treat Schwann cell, which has these RNA domains with an increasing amount of salt, we expect these foci to dissolve. In this case, we have chosen ammonium acetate as our ion of choice. I can discuss later why we think that is the appropriate choice in this case. As soon as we treat the cells at time, T, is equal to zero, what you’ll see is that immediately these foci disrupt. Over time, we’ll start seeing effects of osmotic stress and so on, but at the short timescales, within seconds after addition of ammonium acetate, the behavior that we are observing, we think it’s just coming because of dissolution of electrostatic interactions. Go ahead.

Bede Portz (00:37:53):
That’s a really cool result. Have you tried this with chelators, like EGTA, EDTA?

Ankur Jain (00:37:59):
We have not done that primarily because we are worried whether how effectively they go into the cells and what other effects that may have. That’s a good point. It’s more or less a toy experiment showing that these bodies are actually susceptible to salt-mediated dissolution. It’s not just these RNA foci, several other bodies. Speckles, in particular, appear to be held together by electrostatic interactions, or that’s what we believe.

Ankur Jain (00:38:32):
Now, the second prediction will be that the biophysical properties of these domains will depend on base-pairing interactions. In test tubes, we see that as we increase the extent of base pairing, we see reduced mobility of the DNA in the coacervate going all the way to rubbery plastic or rubbery solids.

Ankur Jain (00:38:57):
So, how does that happen? Virtually, all of these diseases are caused by GC-rich repeat expansion. I showed you data with CAG repeats. There are diseases caused by CTG repeat expansions. If we express an RNA with CTG or CUG repeats, again, it starts accumulating in these domains. If we conduct analogous experiments now with an AT-rich RNA, which can form similar architecture, when expressed at equivalent levels, we do not observe it accumulating as foci or in the domains in the nucleus. So if we diminish base pairing, we do not observe these bodies forming.

Ankur Jain (00:39:40):
What about increasing the extent of base paring? Here, I’ll first introduce some data on CAG repeat-containing RNA. The RNA of these foci is mobile, and it rapidly recovers if you photo-ablate a small region. Here, I’ve done partial photobleaching. What you’ll notice is that the RNA flows from both directions and fills in this hole. Now, nature provides a cruel example, where the interaction strength has been increased naturally in the case of this ALS-associated repeat.

Ankur Jain (00:40:17):
So this disease, one of the mutations associated with this disease is this GGGGCC, hexanucleotide repeat expansion. The sequence besides forming Watson-Crick base pairing may also form G quadplexes, which are more thermal stable than your standard Watson-Crick base pairing. Here, I’m drawing a cartoon of what a multi-molecular cluster of GGGGCC repeat-containing RNA may look like, can have intermolecular G quadplexes combined with some extent of Watson-Crick base sparing. Now, if you repeat a similar experiment, these RNA also accumulate in these domains inside the nucleus. However, if we photo-ablate a small region, now they do not recover, suggesting that base pairing does modulate the mobility of RNA in these domains or these clusters, or maybe I should say intermolecular RNA-RNA interactions. Again, nothing else has changed between the two experiments, except the sequence of this RNA.

Ankur Jain (00:41:25):
Likewise, this ability to perturb interactions may potentially provide ways to break these aggregates in cells. Here is one example where we have used a drug doxorubicin. It’s an intercalator that destabilizes DNA-DNA base pairs. Also, it’s nonspecific and also destabilizes RNA-RNA base pairs. When we treat cells from myotonic dystrophy type one patient fibroblast… So in the untreated case, we observe a certain number of foci. If you treat these cells with a low dose of doxorubicin, we observe about a 90% reduction in foci number and their size, providing at least a proof of principle that playing with base paring may be a route to disassemble these bodies.

Ankur Jain (00:42:19):
Last, I guess, would be that self-associating RNAs may segregate within foci, question that was asked a few minutes earlier. To demonstrate that, we used these two different sequence of RNA, CAG repeat, which forms foci G4C2 repeat that forms foci, but our hypothesis is that the foci of CAG repeat RNA are held together by CAG-CAG interactions. The other ones are held together by G4C2-G4C2 interactions. So when we co-express the two RNAs, we should observe foci, and that’s what we do. I’ll show you data in the next slide. Here, what we are doing is fluorescence in situ hybridization. This is the nucleus of one cell. We are staining for CAG repeat-containing RNA using a CTG probe. We’re staining for G4C2 repeat-containing RNA here with an appropriate probe. When we overlay the two images, at least with the diffraction-limited microscopy, we observe that the two RNA are colocalizing.

Ankur Jain (00:43:32):
They are forming foci, and foci are likely going to the same site. However, now, the resolution of diffraction-limited microscopy is limited to about half a micron. If you use better tricks like stimulated emission depletion, or STED in this case, here, I’m zooming in on one such punctum. And in magenta are the CAG repeat or CAG repeat-containing RNA. In green are the G4C2 repeat-containing RNA. Now, what you’ll start appreciating is that they are segregating from one another in distinct domains. At the bottom is another such example. Again, the scale bar is 500 nanometer. We are zooming into a single punctum. We can look at the line profile. If we just draw a line against this region, it becomes more or less obvious that they are segregated from one another. Although we don’t have sufficient Z resolution, they may be stack on top of each other. So this is an underrepresentation of this aggregation.

Ankur Jain (00:44:40):
Okay. I think I’m almost out of time and I’ll just summarize here. So three key take-home messages that I want to convey today. Number one, short-range interactions can tune properties of electrostatic interaction driven complex coacervates. In the case of simple DNA system, we see the scaling with the lifetime of DNA-DNA bond. I think these kind of experiments may allow us to get away from more or less droplet zoology to building a framework of how we relate interaction energies to the behavior of these coacervates or droplets.

Ankur Jain (00:45:21):
Number two, these complex coacervates or the DNA within these complex coacervates is behaving like neutral polymer in semi-dilute solution, which allows us to bring in existing theories that were devised for semi-dilute solutions and start assessing complex coacervates. And last, this may have some implications for cellular RNA granules, and most notably, we think that pathogenic foci of repeat-containing RNA are forming via a combination of electrostatic interactions and base paring interactions. That provides us some insights potentially on how we can target these foci and disease may have some lessons for how RNA or even proteins may partition into cellular RNP granules.

Ankur Jain (00:46:12):
I’ll thank the people who did all the work. Majority of the work that I showed you today was done by Sumit Majumder and Daniel Stein. Parts of work about repeat expansion disorders were conducted when I was in Ron’s lab at UCSF. Some experiments, and measurement of viscosity were conducted by Sebastian Coupe in Nikta Fakhri’s lab at MIT. Of course, I’d like to thank the funding sources that allow us to conduct this work. Thank you all for the attention, both here and on Zoom, and I’ll be happy to take questions.

Jill Bouchard (00:46:53):
Awesome. Thank you so much. Who has questions?

Bede Portz (00:47:04):
Use a mic so the people at home can hear you, Jesse.

Jesse Lai (00:47:08):
Hello. Thanks for your talk. Can you comment on how the formation of those RNA foci affects RAN translation or the translation of the flanking gene and whether or not when you dissolve it with… I think you used ammonium acetate. How does that affect the repeat proteins?

Ankur Jain (00:47:32):
Yeah. So maybe just to bring everyone on the same page in these repeat expansion disorders, there are two types of RNA toxicity that have been observed, if that’s the right term I’m using. One, the RNA can accumulate in the nucleus of these domains as foci. Another peculiar observation has been that these repeat-containing RNA get translated or produce protein products in all three frames without necessarily requiring a canonical open reading frame or a start codon called repeat associated non-AUG translation, so what we find is that because the RNA is stuck in the nucleus, it does not undergo RAN translation.

Ankur Jain (00:48:12):
The RAN translating RNA are actually in the cytoplasm of the cell. I’m happy to chat later on how we think that the species that are resulting in RAN translation, the RNA molecules that are resulting in RAN translation, they are distinct from the ones that are trapped in the nucleus. It’s a great question, though. It was a big corundum to us. If the RNA is going inside the nucleus, why is it producing protein products? Not just regular protein products, its aberrant translation products. So the two distinct pools of RNA that we find, one in the nucleus and the other in the cytoplasm.

Bede Portz (00:48:52):
Erik has a question.

Erik Martin (00:48:57):
I have a question that comes up observing these structures here, which is to say that there’s a lot of empty spots in them. If I’m understanding your model correctly, essentially it should just suck up all RNA, regardless of whether it’s what you’ve put into the cell or not.

Ankur Jain (00:49:16):
Yeah.

Erik Martin (00:49:16):
So it kind of gets me thinking that the only reason why these are discrete puncta and not just huge conglomerates of RNA is that a lot of RNA is chaperone by hnRNPs, et cetera. So can you make some connection between sort of perhaps if you eliminated some hnRNPs or protein chaperones, can you modulate the size or structure of these?

Ankur Jain (00:49:35):
Yeah. So I assume you are referring to these images and why do we see these essentially blank regions? So here, we are using STED, but that still has a resolution limit of around 40, 50 nanometers. We are observing these exaggerated structures while just probing our own RNA for interest. There are plenty of other RNAs inside these structures. Notably, we know MALAT1 is also there and likely several other cellular RNAs that we have not stained for. Of course, there are going to be a large number of proteins, as well hnRNPs. We would know some SR family proteins go there. So these empty spaces that we are observing are likely just occupied by many, many at such proteins. If you have a good target in mind of which are hnRNPs we should hit, we’d love to do that. We are trying to hit few of them. We don’t see essentially any change in structure.

Erik Martin (00:50:38):
Speaks to the fact if RNAs aren’t chaperoned by proteins, then they’re more likely to end up in condensates like this. Is that-

Ankur Jain (00:50:48):
That’s a great point. That’s a great point. So what you are saying essentially, maybe if I can rephrase, if the RNA doesn’t find a protein-binding partner, they are more likely to base pair. I would love to test that directly, but we don’t have a good way to do that. How do we distinguish between an RNA that is just not base paired versus not bound by another protein? Easier to do at bottom up than in cells.

Bede Portz (00:51:15):
Balaji.

Balaji Olety (00:51:18):
Thanks. My question was if you observed any biases in these foci formation between nuclei and cytoplasm. The reason I ask is because many of viral genomes have CpG suppression in their genomes and I was wondering if that was an evolutionary way to basically prevent these foci formation. My second part of the question was are there any cellular proteins that suppress these foci formations specifically in the cytoplasm, which have been shown that there are certain such as ZAP, for example, binds to these CpG islands and prevent viral transcription? So on those notes, if you have any biases between cytoplasm and nucleus.

Ankur Jain (00:52:01):
Right. So maybe I can split into… I’ll answer the second part first. So, are there any proteins which are modulating maybe the export of the RNA? Several labs have done screens that are interested in RAN translation modifiers, and they’re often nuclear pore components or the RNA transport machinery shows up. We are also running some screens, but we are not able to find anything beyond those standard RNA export factors. There’s nothing specific to the CAG repeats that we could identify so far or specific to G4C2 repeats either. Can you remind me the first question again?

Balaji Olety (00:52:41):
[inaudible].

Ankur Jain (00:52:43):
Right. So, is there any bias between… Or what determines the partitioning between nucleus and cytoplasm or-

Balaji Olety (00:52:53):
[inaudible].

Ankur Jain (00:53:06):
Right. Yeah. So we do not see foci in the cytoplasm, at least in these cases. The RNA in the cytoplasm, if it is there, it appears to be well dispersed. Now, a key caveat there, which we recently encountered, is that once those repeats start getting translated in the cytoplasm through RAN or non-RAN translation, they start coaggregating with the peptide products. So let me just repeat that. Once the CAG repeat-containing RNA goes into the cytoplasm and starts getting aberrantly translated, the RNA coaggregates with the peptide and we do not know yet whether it’s entirely base-pairing-mediated, protein-RNA-mediated, or if there’s a third game at play. Maybe relevant to this talk, what I can share is that we see that G4C2 RNA is more in the nucleus than CAG repeat-containing RNA. So that identical repeat length and RNA copy numbers, more base pairing will lead to more retention in the nucleus. We think primarily these foci are forming via transcribe the RNA, the RNA get clustered and just are retained and do not make it out to the cytoplasm.

Bede Portz (00:54:28):
There’s a question online from Gable. Do you want to unmute yourself and ask?

Ankur Jain (00:54:34):
Omar, thank you. [See Omar Saleh’s paper about DNA droplets here.] I’m so glad that you are at the talk. Anyway, go ahead, Gable.

Jill Bouchard (00:54:40):
He doesn’t have a mic. So the question is about temperature.

Bede Portz (00:54:47):
You have this melting and annealing protocol through which you generate these in vitro. Can you comment on the effects of temperature on the behavior of the RNA coacervates?

Ankur Jain (00:54:57):
So in this case, in all the data that I’ve shown you today with DNA, we do not do any denaturation or annealing. We had to do it in our previous paper because RNA purification protocols intrinsically involved precipitating RNA, by precipitating I mean crosslinking the RNA and making base pairs by salt and polyamines and so on. Here, we can buy commercial RNA from a vendor, which comes with decent amount of salt and there is not substantial clustering. So there is no annealing protocol involved in the data that I showed you today.

Bede Portz (00:55:34):
Sidharth

Sidharth Sirdeshmukh (00:55:34):
Well, that was a pretty awesome talk. So have you considered assessing the changes in material properties when you reconstitute different repeat-containing DNA and RNA together? Because you show the experiments where you’re reconstituting DNA with different repeated sequences and then experiments with just RNA, but how about both together?

Ankur Jain (00:56:06):
Well, what we were trying to do is use DNA instead of polystyrene and get insights for how proteins and RNA may behave. I guess what you are asking is perhaps if you have double-stranded DNA and it’s getting transcribed, how would that affect this formation or something?

Sidharth Sirdeshmukh (00:56:28):
I guess more along the lines that are basically RNA and protein, they form condensates that enable transcription to take place. So what I’m thinking is that because it seems like if we would use single-stranded DNA [inaudible 00:56:47].

Ankur Jain (00:56:59):
Yeah. So these things are called R-loops and they are observed in several cases. We haven’t tried to do that in our hands, but there are reports that in repeat expansion diseases, there are R-loops, which may be contributing to fragility of DNA or breakage of DNA of these repeat sites.

Shruti Jha (00:57:20):
Is it working? Oh, do you think the reason why peptides are making some sort of aggregates with the repeat sequences is some kind of peptide-based mechanism of translation regulation, like the peptide acts as a feedback loop and that’s why it’s just like over and over?

Ankur Jain (00:57:41):
It’s possible. We just don’t know enough about the system yet. Yeah. It may be a cell’s way of shutting down translation. Still in the works, so I won’t be able to comment beyond that.

Shruti Jha (00:57:50):
Thank you.

Sangram Parelkar (00:58:00):
Can you guys hear me? Yeah. So Ankur, I’m very intrigued by the data, which is shown for the fusion timescale, and it depends on the size of the condensates or the droplets. Could you share some more information on that? I thought you mentioned that as the size becomes bigger, it becomes difficult for fusion to happen. Is that right or is it the other way?

Ankur Jain (00:58:21):
No. So as the droplets get larger, it’s not more difficult, but the time that it takes for the droplets to fuse gets larger and larger. So two droplets, which are micron size, they will fuse faster than two droplets of same material, which are, let’s say, a millimeter in size.

Sangram Parelkar (00:58:38):
Okay. So I was taking more from the cellular context. When you look under the microscope, you see multiple condensates. Given time, do you suppose there might be an upper limit to fusion events also?

Ankur Jain (00:58:50):
Right. Well, in the cell you have true barriers that prevent diffusion, like cytoskeleton, or in the case of nucleus, you have a lot of DNA, which may just physically occlude droplets from coming next to each other. These models and theories are for solutions of polymers, solutions or droplets of polymers in pure solution.

Sangram Parelkar (00:59:12):
Makes sense. Thank you.

Xiaobo Ke (00:59:13):
I have a question about those doxorubicin for inhibiting the DM1 condensate formation, stress granule formation. We know that molecule is a DNA intercalator and also is the inhibitor of DNA polymerase. So, what do you think is the mechanism of action in that particular-

Ankur Jain (00:59:35):
Yeah. So the timescales that we are looking at, we don’t think there is a substantial change in transcription rate. These are done within a couple of hours after addition of drug, and we don’t think there is any change in RNA transcription rates. We don’t have the data in patient cells, but we can do experiments in our synthetic cell lines where we can regulate transcription. What would be the implications for the mechanism? You think it’s just physically putting it in excess, it’s just disrupting many, many base pairing interactions and one of them happens to be in our RNA of interest. It also impacts nucleoli and many other places where you have substantial RNA secondary structure or RNA-RNA interactions.

Bede Portz (01:00:27):
Okay. I think that’s-

Jill Bouchard (01:00:28):
Avi, had a question on Zoom. Avi, do you want to go ahead. Are you still there?

Ankur Jain (01:00:41):
Go ahead.

Jill Bouchard (01:00:41):
There he is. Oh, you’re muted, I think.

Avinash Patel (01:00:52):
Can you hear me now?

Jill Bouchard (01:00:53):
Yep.

Ankur Jain (01:00:54):
Yeah.

Avinash Patel (01:00:55):
Great. Hi. Hi, Ankur. Nice to see you. Well, not see you, but hear you. Anyways-

Ankur Jain (01:01:02):
Likewise.

Avinash Patel (01:01:02):
You showed the STED images, right? So now we know that we have some got extra information, ultrastructure information about these repeat expansion RNA. So, can you tell us a little bit as to how you’re thinking of using that extra information in order to get an extra understanding of the disease or in terms of using that information for developing therapies? Just would love to hear your comments on that.

Ankur Jain (01:01:33):
That’s a great question. To be very honest, I haven’t thought about what implications it would have for disease itself. It does certainly tell us that these RNA are segregated and likely that is the reason that we don’t see toxicity on short timescales, where we have foci. They sequester certain proteins, but they’re just likely clustered outside from… make a small subregion in the speckle. I have to think a bit more about what implications it would have for the disease. That’s a great question. We are just doing it as more or less as a proof-of-concept experiment at this stage.

Avinash Patel (01:02:17):
Wonderful. Thank you.

Bede Portz (01:02:19):
All right. Well, join me in thanking Ankur for visiting us and a great talk.

Ankur Jain (01:02:26):
Thank you, everyone.

Jill Bouchard (01:02:29):
Yes, fabulous talk. Thanks again for coming to our kitchen table and thanks everybody on Zoom and in the room for joining today. See us again in a month or two. Thanks, again.