VIDEO: Priya Banerjee on Phase Transition and Hollow Condensates

Priya Banerjee (00:04:34):
If you peek inside these organelles, you will see at the molecular level, there are protein and nucleic acid molecules. They’re in a condensed state of matter in a way that resembles a soft matter, but they’re in equilibrium with a dispersed phase, which is depleted of these polymers. So this reversibility could be a key, because then cells could manipulate such reversibility to create or dissolve such condensates based on their need.

Priya Banerjee (00:05:02):
I must admit that when I was a graduate student, I read a lot of literature. I still read a lot of literature, although we are appreciating phase separation in biology in the last 10 years or so. There are papers that talked about phase separation as a mechanism to compartmentalize sub cellular space as early as ’90s where Harry Walter and Donald Brooks put together this nice hypothesis where they predicted phase separation is extremely common due to high crowding and confinement in cells.

Priya Banerjee (00:05:37):
Some aspect of phase separation could be very nicely explained by a theory put together by Flory and Huggins in 1940s where the free energy of this mixing which tells you how stable this solution is. So suppose we have a polymer and a solvent, the free energy can be expressed in two terms; entropy and interaction enthalpy.

Priya Banerjee (00:06:02):
Now the entropy tells you how disordered the whole system is and you can guess that when you are forming a condensate, it’s more ordered and randomly distributing. So this mixing entropy term is always unfavorable. So what drives the separation is this interaction term which is commonly called the chi parameter, and for a given simple polymer solvent system, the interaction parameter or chi parameter could be simplified by this mean theory model where you take into account of all possible interactions, polymer-polymer interaction, polymer-solvent interaction and solvent-solvent interaction. And you might see that this condense state is stabilized if there’s too much polymer-polymer interactions or polymer-polymer interaction dominates this, or the polymer hates the solvent, so polymer doesn’t want to spend more time with the solvent so it goes to a condensate.

Priya Banerjee (00:07:00):
In terms of protein-protein, protein-RNA interaction, the description of these interactions are well known from protein chemistry, chrystography and other literature, that protein and RNA side chains or DNA side chains could encode interactions such as electrostratics, you have a negative charge, interactive positive charge, you can have pi-system that can interact via pi-pi-interaction. The cation can interact with an aromatic system which would be cation-pi and then hydrogen bonding is what drives DNA double helix formation, could also play a role.

Priya Banerjee (00:07:37):
So all these interactions are what characterize in terms of folded protein, and how they stabilize their three dimensional structure, but for disordered systems such as the protein and RNA molecules which have these repeat sequences, they do not fold into any specific type of structure, we might have to revisit how these interactions come to play through phase separation.

Priya Banerjee (00:08:00):
There could be other forces, such as molecular crowding, as Harry Walter predicted in 1995. All these together come in to play a big role in phase separation, as you might expect, and these interactions could be encoded by polymer sequences. That’s why it’s important to understand the driving forces so that we can characterize phase separation under physiological condition versus phase separation in a disease state.

Priya Banerjee (00:08:32):
Typically, when you talk about phase separation, some aspect could be very well described by a diagram called phase diagram. So in this simple one component system, and by one component I mean you have a polymer and a solvent, so a protein and a solvent. This line is called a coexistence line, it separates the single phase region from two phase region. You can determine this line called Tie line which will tell you the composition of the condensates in the droplet versus in the dispersed phase. What you have on the y-axis here is interaction strength that could be tuned by temperature, salt or pH of the system.

Priya Banerjee (00:09:19):
It could also be a function of protein-protein interaction, protein-RNA interaction that could be tuned by say a post-translational modifications as well as partner or small molecule binding. And again you can get in and out of this one phase regime by just concentration change as well so volume fraction and protein concentration. Most of the systems that we study or you think about in terms of a cellular condensate, we have a multi-component. So when you also need to think about how phase diagrams look like in more than one component system.

Priya Banerjee (00:10:03):
So we have the volume fraction of the second component here, you have volume fraction of one component here. And in certain cases we see complexities beyond what I described in the previous slide. For example, here you see closed loop. So the closed loop suggests that within this regime II, you have phase separated condensates. It’s suggesting in addition to the interactions and changing all these solution conditions you can again modulate phase separation by simple changing and concentration of the two polymers as well as their related compositions.

Priya Banerjee (00:10:43):
So supposing you have protein and RNA system you can tune their composition without doing anything else, without doing post-translational modification, or temperature, salt, or environmental changes. You will still able to get in and out of this condensation, de-condensation phase. And that is something I’ll focus on later part of my talk.

Priya Banerjee (00:11:02):
And you can study experimentally by looking at this light scattering intensities which tells you there are condensate formation or not. And you have the ratio of the protein versus RNA here. So you fix protein concentration, you vary RNA concentration. You could see there’s condensates forming and dissolving.

Priya Banerjee (00:11:21):
So this particular phenomena is known as re-entrant phase transition which could be extremely common for all cellular systems where phase separation is driven by protein-protein or protein-RNA interaction where more than one components are present. So to start in re-entrant transition this work was done when a student, as a postdoc in a Ashok’s lab, we looked at this protein, which probably needs no introduction. It’s the hydrogen atom of phase separation. It’s a protein called fused in sarcoma, FUS.

Priya Banerjee (00:11:55):
So, this protein is associated with multiple different pathways of RNA metabolism. It is also mutated in ALS disease. The protein is typically a nuclear protein, but it also goes to the cytoplasm where it localizes within stress granules.

Priya Banerjee (00:12:14):
The protein is an RNA binding protein, I’ll spend a lot of time talking about FUS and RNAs its natural partners. So, when we looked at RNA driven phase separation of FUS, we could see this re-entrant transitions. You see this rise and fall of light scattering intensity, very simple technique, but you still get what you want to see how phase separation is regulated by RNA. We looked at two different nonspecific RNAs. These are homopolymeric, poly(U) RNA with a very massive length versus a 40-mer.

Priya Banerjee (00:12:49):
We also looked at the same phenomenon. With cellular RNA, we see re-entrant transitions with poly(A) RNA as well. So this is basically telling you it’s very promiscuous in terms of RNA binding which existed, the evidences is the RNA binding being promiscuous for FUS was already well known. We now show that they can regulate, RNA molecules can regulate in a similar way, how phase separation is regulated for FUS, and if you look it under a microscope-

Diana (00:13:19):
May I ask a question?

Priya Banerjee (00:13:20):
Yes.

Diana (00:13:22):
Have you looked at the influence of RNA structure with… Does the RNA need to be unstructured to see this phenomenon?

Priya Banerjee (00:13:32):
So we looked at to some extent, most of the studies are unpublished. So, we looked at how RNA structure influences this process. So, for example, you can look at how G4C2 G-quadruplex can regulate this. With some RNA, if you get phase separation, then you can still see re-entrant but if you form aggregated structured then it’s not very easy to dissolve them with excess RNA. And that could be a kinetically driven process as well because for liquids, the dynamicity is very high for solids, we could not test more than a few hours, but RNA structure not only changes how wide or how much transition window you have for these cases, if you can also change the dynamic nature of this condensate.

Diana (00:14:24):
Thanks.

Priya Banerjee (00:14:24):
Yeah. So, then, what you have here is basically within the regime II, you have this condensate strength. You understand the mechanism of this process, I looked at the RNA binding domain and I’m always fascinated to see all these arginine-rich domain in proteins, which bind RNA. In fact, the arginine-rich domains that we see in FUS are quite common in eukaryotic RNA binding proteins.

Priya Banerjee (00:14:57):
These papers argue that RGG/RG type of protein are present in more than 50% of eukaryotic RNA-binding proteins and from the business perspective, what I see here is this poly arginine-glycine repeats and poly-proline-arginine repeats which are formed due to hexanucleotide expansion that’s seen in the c9orf72 G4C2 mutant but are about the same.

Priya Banerjee (00:15:22):
They form this toxic polypeptides and they are toxic because they cause the disease they can also kill the cells if you put them in the extracellular medium and from the perspective of all these arginine-glycine repeats or arginine-proline repeats, we see that they are also present in several proteins that are highly functional in RNA metabolism. So, to understand the model of phase separation in this specific case, or to be specific the range and transition for this protein, we thought it is not a bad idea to look at what this arginine-glycine repeats are doing. So, we started with a very simple polymer, which is this RGG box of FUS which is 34 residue long protein.

Priya Banerjee (00:16:10):
And you could recapitulate the re-entrant behavior with poly(U) RNA or other similar RNAs. And one thing we did here is we know that the concentration of both the polymers and got a lot of turbidity diagrams like this and microscopy based evaluations of their phase separation, so, that we could create something we call a thermodynamic state diagram where phase separation could be investigated in the two dimensional concentration planes. So, you have the protein on your one dimension and RNA in the other dimension.

Priya Banerjee (00:16:45):
And the color intensity tells you the scattering intensity, more intensity means the propensity of phase separation is higher and this dotted line which gives you a look inside this closed loop although you don’t see the closed loop, you see just part of the close to be a half ellipse. You see how the two phase regime which is shown here by this Roman-plated or Greek-plated two, you have phase one or regime I here, or regime III, they’re completely homogeneous mixture, but then you have this phase of separated regime, they’re separated.

Priya Banerjee (00:17:25):
And again this arrow’s suggesting that you can do the entire transition from a homogeneous mixture to condensate to dissolution of condensates, just by varying RNA. We are not changing any protein structure or protein post-translational modification, or salt, or anything, we’re just going to regulate phase separation by composition. With that in hand, we wanted to ask a very specific question. For example, we have this finite regime where you see phase separation. I’d like to know about what regulates this phase separated regime. How wide can we be in this specific case? And what regulates the width of the two phase regime? So that’s one question.

Priya Banerjee (00:18:10):
And the second question, I’d like to know that if I move around within the space of the regime by changing concentration and composition of the mixture, does it matter– does it change the condensates’ physical properties? For the first part, we did a very simple studies, Ibraheem, a graduate student in the lab he thought it is all about electrostatics. And that is because arginines are positively charged, RNA molecules are negatively charged, so many people thought electrostatics play a big role. We thought that too, but one simple test of that model could be replacing all these arginine with lysine, which is also positively charged, and the charges don’t change if you do simple mutations.

Priya Banerjee (00:18:54):
So in one study, Ibraheem changed all this arginine. So we took… a designed polypeptide, where you have 10 arginine and we replaced all with lysine we did not change the charge and we wanted to see if it changes how wide the phase separator regime is. And we see it matters a lot. So, the green line suggests the arginine have wider regime of separation than the lysine. And this is done with again poly(U) RNA homopolymeric RNA. And you see that when you try to dissolve it with RNA, by adding RNA to this condensates, you’ll see that it was shifting the overall boundary of dissolution almost tenfold.

Priya Banerjee (00:19:41):
So, there is a sequence code embedded within arginine that is missing in lysine that regulates the width of the two phase regime. We now ask the following question, what happens if you change RNA? And the first experiment that we did is changing poly(U) RNA. So poly(U) RNA is is a polymer of uracil, which has one membered ring and the other class of RNA bases have two membered rings. So, polyadenine will be a nice test RNA where you have these two membered ring which provides more pi-electron.

Priya Banerjee (00:20:14):
And when we did that experiment we saw even dramatic changes. So, you have arginine having this two phase regime. At some concentration, which is pretty low — 0.1 mg/mL, it becomes impossible to dissolve this condensates by adding poly(A) RNA, suggesting that it’s not just the RNA as much as the protein structure sequence is the RNA sequence as well, that plays a big role here and again, lysine are wider than what you have in the previous poly(U) RNA. We’re still way smaller compared to what arginine two phase regime looked like. And in a simple turbidity plot, you could see that it’s almost hundred fold or more than, I would not say not hundredfold it’s like 50 fold change in the amount of RNA that you need to dissolve this condensate.

Priya Banerjee (00:21:08):
So, we looked at several other works that talked about how arginines could distinctly interact with RNA or DNA compared to lysine. And we also did some calculations based on quantum mechanical models. And we found that all these results went to a trend, where arginines are far more sophisticated in interacting with RNA bases than lysine and these interactions come from the fact that arginines, although it is positively charged, similar to lysine, arginines have pi electrons in the side chain that can mediate cation-pi and pi-pi interaction with a double membered ring way better than when you have a single membered ring because the availability of pi system is much larger here.

Priya Banerjee (00:21:58):
While you go to lysine, lysine does not have such pi electrons which can still do some cation-pi but it’s much weaker because the network of interaction here is electrostatic cation pi pi-pi, while lysine can only do some cation-pi and no pi-pi at all. So, these trends led to a hypothesis that there will be a rank order of interactions when we have this full interacting members, Arg-Adenine, Arg-Uracil versus Lys-Adenine and Lys-Uracil, and what you will see that maybe this widening of the two phase regime is due to the short range interactions beyond electrostatics.

Priya Banerjee (00:22:39):
So, we started off looking at how electrostatic interactions are important and we found that electrostatic interactions may not be the only thing here, there could be short range interactions that adds complexities to these interacting polymers, how they phase separate. So this leads us to hypothesize that arginines are much more interacting than lysine. And we must do experiments to sort of validate that as an experimental, experimentally if that’s what we want to do. So, we thought it would not be a bad idea to look at condensate dynamics if we can actually see [outside change inaudible 00:23:16].

Priya Banerjee (00:23:19):
To look at condensate dynamics, we implemented a technique which is called Correlative FRAP- Fusion techniques. So, in Fusion, we are looking at micron-scale dynamics. So, we have optical traps, where we can trap two droplets into traps and then we can bring one trap to the other in a very controlled fashion and then you will see that at some intermediate length, they are going to merge because of coalescence-driven fusion. And when that happens, Trap-1 loses the droplet and it undergoes a relaxation.

Priya Banerjee (00:23:54):
So this particular process could be detected very efficiently by looking at the force relaxation from Trap-1, which offers you more than two orders of magnitude, compared to normal video microscopy-based analysis. So that’s why it’s a little bit easier to do these measurements using optical trap. But at the same time, what you can do is you can measure on these droplets you can implement a technique called fluorescence recovery after photobleaching very commonly used in the field.

Priya Banerjee (00:24:24):
So these are all say single protein molecules, you bleach them with very high pulse of laser. So the molecules are now irreversibly sent to a dark state. Now other molecules can come and diffuse into the volume of bleach and then you see some sort of recovery based on diffusion. Although FRAP has its own limitations, but nevertheless it provides you some information on the nano-scale diffusion. So the bottom line is with Trap induced fusion, you get some micron-scale dynamics with FRAP you get some nano-scale motions. And what we’re trying to look here is basically whether we have arginine versus lysine differences. So, in a typical experiment, this is how it works.

Priya Banerjee (00:25:11):
It’s two optical traps, they are flowing condensates, you can trap them under flow. And once you see you have a definite size that you want to work on, you can turn on the fluorescence channel and you will be able to now control the two traps and you see how they’re undergoing fusion. And at the same time since all under fluorescent microscope you could do FRAP measurements on the same droplets as well.

Priya Banerjee (00:25:37):
So, this was done to study if arginine versus lysine differences are manifested equally at a different length scale of condensates and we see that indeed, with arginines, if you have poly(A) RNA, you have relaxation time which is 350 millisecond compared to lysine with poly(U) RNA which is 2.8 minutes and again, they are undergoing fusion way faster than arginine-rich polypeptides, suggesting again that the interaction strengths are somehow controlling dynamics in the micron scale.

Priya Banerjee (00:26:12):
We can quantify such effects and see that indeed poly(A) RNA is usually higher between poly(A) and poly(U) and lysine versus arginine also are different and expected based on the phase diagram analysis I showed you. And utilizing FRAP measurements, we see that indeed, for arginine with poly(A), there is almost no recovery and we are looking at the polypeptide recovery in this case, and then with lysine with poly(U), you see a very fast recovery and these two are intermediate.

Priya Banerjee (00:26:45):
So, this particular FRAP curves follow the same rank order as we expected based on interactions. So, overall what we learn from this data that the interactions that govern phase separation also shows some trends in the condensate dynamics not only in micron-scale but also in nano-scale. Going from synthetic condensate to FUS condensate we see similar trend. So again, we took the RNA binding domain of FUS and with poly(A) and poly(U) diffusion. The diffusion shows again one order of magnitude difference and this is just the RGG3 domain of FUS and using FRAP we see two-fold difference in diffusion rates.

Priya Banerjee (00:27:32):
Now one thing I wanted to mention in the previous slide going back here that we have to be careful when we analyze FRAP as a stand alone because you see, there is no diffusion at all for arginine with poly(A) suggesting that this could be a gel-like condensate, which may not be true because you see they’re still undergoing fusion in 350 millisecond timescale, suggesting that there are some decoupling between the micro-scale dynamics and the nano-scale dynamics.

Priya Banerjee (00:28:01):
So, when we analyze globally FRAP-Fusion data, we see such trends. And we see similar behavior for FUS RNA binding domain as well as the RGG3 because you see, diffusion is only two-fold different while you have the fusion relaxation time almost an order of magnitude different. But overall it tells you that the cation pi pi-pi types of interactions further brings out regulatory elements over electrostatic interaction and they work together to modulate phase behavior as well as the dynamics of these condensates.

Priya Banerjee (00:28:39):
So with this, which was published last year, we concluded that cation-pi and pi-pi interaction oppose the re-entrant dissolution of protein-RNA condensates and this could be an important feature for repeat polypeptides because you add more arginine into it and you might lose the way you regulate their dissolution.

Priya Banerjee (00:29:01):
So the next part of my talk, I’ll be focusing on something we have done in the recent past. And it’s very much a work in progress. And I’ll be happy to get your feedback on this. So we talked about the phase behavior, we talked about the condensate dynamics, how about [inaudible] of their structure because coming from a field of protein folding. Protein folding happens because there is a global minimum and everything goes there. So there is one state where the structure is folded and it might go there, disordered proteins do not fold because it’s pretty flat, because there are no overall stable minimum.

Priya Banerjee (00:29:45):
For phase separation in these two component systems, you see that there is a finite regime where you have condensates. So it’s not like you’re forming condensates at a very specific composition or concentration. You are changing their composition and you might change the properties which we do not know yet. So, the idea is to understand if we form condensates under all these different composition, which is shown here by this dotted line and see similar behavior, and that would be pretty boring to me, but I still like to see that.

Priya Banerjee (00:30:20):
So, what we have done — We looked at the relevance of this problem from biology. And going beyond the simple droplet models is necessary, because, you look at cellular condensates and you see hierarchy of structures. You see that, for example, in nucleolus, you have these sub-structures that are well described by co-existing droplet coexisting multiphasic condensate model.

Priya Banerjee (00:30:50):
Similarly, in stress granules, you see a core-shell structures. Lately, we started seeing a different type of structure. For example, if you over-express this RNA-binding protein TDP43, and you find this engineered condensate in cell nucleus which have nucleoplasm-filled multiphasic structures inside. So they are called regulated condensate. The mechanism is not well understood, but what comes from all these elements is the fact that condensates could have non-trivial structures and there could be subcompartments that can be utilized further to communicate in the cellular environment in a way that might be important for signaling functions and control over gene expression over stress regulation because that’s how multiphasic compartments should work. So to understand the building blocks of this process in a simplified system, we thought maybe RNA-protein mixtures might play a role here.

Priya Banerjee (00:32:00):
So the first experiment that we did almost three years ago, we changed the RNA concentration of the existing condensates to see what happens to the condensates. And what I found that by simple mixing experiment if you have, so I have to play this video. Let me just escape. Yep. So let me just play this video. So what we are doing here is we within the existing condensate that we are changing the RNA concentration, and without dissolving them, and I hope you can see the video.

Priya Banerjee (00:32:36):
So first, you let them role a little bit and then we’re going to change the organic concentration. And what do you see the formation of this multiphasic structures, which are unstable. Because they’re fusing within five minutes they’ll go away. But we see that it’s possible to create even transiently substructures, which you typically see in cellular condensates. And we’re again utilizing a very simplistic model system and it gets to where we have one protein and one RNA.

Priya Banerjee (00:33:11):
So we spent the last two years to chase down these structures, whether we can form them and what type of systems would form them, and we found that pretty much everything that we see undergoing phase transition by heterotypic interaction has this kind of structure. So, first what we have done here is we took another model protein system, protamine which has this repeat R [inaudible] residues. And we created a phase diagram to see if there are specific composition of the mixture or concentration of the mixture where you see these hollow condensates.

Priya Banerjee (00:33:47):
And we did that over a very broad range of concentrations. So we saw that if you’re too low in concentration, you don’t form anything. And then if you have certain concentration and composition regime, you see these droplets and they prevailed for a really wide range. And after that if you keep pushing the concentration you see condensate formation, which are not your droplet-like structures, they are more like your doughnut-like structures. So you see that they have this ring-like appearance under a fluorescent microscope. They are much larger than what you see in a typical condensate as well, which was extraordinary to us because we are now forming these structures just mixing of protein and RNA and when the system undergoes such a structural transition, it becomes a problem for the system because we are creating a larger surface area because we are now creating two surfaces compared to one surface for a condensate.

Priya Banerjee (00:34:52):
So, we wanted to understand what’s going on. But before we do that, I wanted to sort of show you a little bit more how these structures look like under a fluorescent microscope or DIC. So these are all RNA concentration higher than the protein concentration to the protein concentration is 4 mg per mL, 4.4 mg per mL and RNA is about five-fold higher and the internal lumen looks literally like a hollow condensate here. So, this z-stack analysis of images showing that you have some sort of vesicle-like look. Here, if you start with the protein concentration fixed and keep adding RNA your condensates size increase, and then at some concentration of RNA, they become… The poly dispersity goes away, they become sort of monodispersed, and they’re highly covering the overall solution but they resist fusion. [If you keep pushing inaudible 00:35:54] RNA concentration, you see that they become very fuzzy. So they look not very focused because they’re moving along, and it’s really hard to image them.

Priya Banerjee (00:36:07):
And finally, even after some RNA concentration, you see these hollow structures forming. We look at under a microscope to study where the RNA and protein are, SYTO13 looking at the RNA, and protamine is just the protein building block and you see they both go to the rim and the inside part, which is the lumen sort of devoid of these structures.

Priya Banerjee (00:36:34):
We can also study diffusion in a specified regions, for example, in the lumen outside or on the rim. So here by doing diffusion studies, we’re trying to see if it is really a dilute phase within this interior space. And we see that indeed, the diffusion of macromolecules outside and inside they’re in microsecond timescale in T half, versus when you are on the rim, you see it’s in millisecond timescale. So three orders of magnitude difference in their overall diffusion behavior in specified regions of the structures, suggesting that the internal space indeed is similar to what you have outside, which is a dilute phase.

Priya Banerjee (00:37:20):
We can also do typical liquid-like property studies such as FRAP. So we are going to FRAP our small region here. And you’ll see it recovers very well. We can also trap optically these two condensates and try to merge them. They form multi-compartment vesicle-like structure, and then they are merging. And then here we are trapping a droplet and trying to see if they can fuse with the rim without disturbing it. And it does actually work out pretty well, suggesting that these phases that you see are compatible.

Priya Banerjee (00:37:57):
One of the main feature of the phase diagram is to show that specific composition this happens. So can they actually show that in real microscopy. So, what we have done here is we trap a droplet and just change its RNA concentration. Here in this movie you can see that two droplets are trapped and RNA is being flown and you see that they undergo a structural transition to this hollow condensate. We keep them for a long time, we flow buffer. So, under controlled environment where we just flow buffer nothing happens; with RNA, they undergo this transition. We can also destroy RNA by RNAase treatment. So here over a period of 15 minutes, you can see that overall, this internal lumen is collapsing as RNA is being digested. And at some point, there will be a transition from hollow condensate to spherical drop.

Priya Banerjee (00:38:57):
Finally, again, is it a system specific or is it something we can see over a broad range of systems? So FUS RGG3 does that pretty as well as in protamine, so this is the hollow condensate formed by FUS RGG3. We also talked about a lot in the first part where we replaced arginine with lysine. So, here we replace arginine with lysine. And we see that condensates are still similar in structure as you have in protamine. So, protamine has 21 arginine residues. We replaced all of them to create this variant called protamine-K and it too forms these condensates. So that’s why we suggest that this is a property of systems that undergo phase transition via heterotypic interactions.

Priya Banerjee (00:39:46):
We also looked at synthetic condensate forming system that chemists typically studies such as this PAH polymer, which is a positively charged polymer. With poly(U) it forms condensates. Polyphosphate is another negatively charged polymer. Polyglutamic acid, a typical polyglutamic acid tract, it’s found in many proteins such as NPM1 which Diana studied a lot. It also formed these condensates.

Priya Banerjee (00:40:14):
One of the favorite image that I have is this with cellular RNA. So I initially thought it’s not possible because cellular RNA has everything. So, maybe there will be something else going on, but we can still form these structures with cellular RNA mixtures and in other type of polymers as well. So overall, what we see here is a general behavior for systems under these disproportionate conditions. That would be the important part for the next discussion in here, that disproportionate mixtures drive the formation of these hollow structures.

Priya Banerjee (00:40:50):
So, that brings out this important question, what is the mechanism? Why some systems under certain conditions forming these structures? We had no clue until we found that Shklovskii published this theoretical paper in 2005, where he talked about how disproportionate mixtures can undergo condensation and form anisotropic building blocks. Now, the data is trying to show you on systems they’re pretty much homopolymeric without any defined anisotropic structures, they are different than lipids which are very well known to form hollow structures, such as vesicles or lipid-mimic polymers.

Priya Banerjee (00:41:32):
What you typically do is you create a head and tail. What that means is that even with diblock nature where interactions are different among different building blocks. What Shklovskii suggested you can still do it if you have homo polymers, but you have to drive their condensation to a partial point. So, suppose you have a protein and RNA and you have disproportionate amount you can form tadpole-like structures where part of the RNA chain for example, is only covered by the peptide or the protein because there is not enough.

Priya Banerjee (00:42:07):
So, once you form that anisotropic structures, they can undergo formation of these condensates, which we call micelles. We call them micelles because now if you think about how these condensates will collapse, they can undergo condensation where these heads are joined, but they are free RNA, a tail, which could be on the surface. And that might create some sort of fuzziness in structures because these condensates should not undergo fusion because overall negative charges on the surface or repulsion. So, that will lower the surface energy tremendously. So, that will resist fusion. And again, it’s a surface driven property not a property of viscosity at all.

Priya Banerjee (00:42:48):
And finally, by some mechanism, these micelles can come together to form vesicle-like structures. Where now, these chains which are not condensed, which are not satisfied at all, can maximize their interaction with solvent by distributing them. So, it’s about two surfaces because you now have internal surface and you have an external surface. So what Shklovskii did here — he presented a very nice statistical mechanics model to calculate free energies and gave us some clue how to think about the process.

Priya Banerjee (00:43:23):
To understand it further, if this is really what’s happening, we teamed up with Davit and Murali at Iowa State to do some molecular dynamics simulations. So, they implemented this system of protamine and poly(U) RNA with a hundred uracil residues. And they integrated interactions which are electrostatic and excluded volume, which you typically can expect in polymers, like protein and RNA chains. And indeed, they could form tadpoles at very low chain concentrations. So what you see here is partially condensed head of an RNA chain and the free RNA tail, which looks like what Shklovskii presented in 2005.

Priya Banerjee (00:44:11):
If you increase the total chain concentration now you form condensates, where the surface of these condensates are covered with this RNA chain which is not satisfied. And that would explain why this condensates may not undergo fusion at all, because the overall RNA concentration on the surface will lower their overall ability to fuse because the surface energy is lowered.

Priya Banerjee (00:44:35):
And finally, if you keep pushing these condensates, the RNA tails which are not part of these condensates will seek for more surface area and that will create an additional surface giving rise to the vesicle like-hollow condensate structures. So overall the model of tadpoles work nicely, and it also points out that the intrinsic anisotropy that you have in a lipid, it’s not a prerequisite to form hollow structures like lipid vesicles. You can still form such structures by driving the condensation to a partial regime where you create an extrinsic building block such as these tadpoles in our system.

Priya Banerjee (00:45:18):
So, I talked about lipid vesicles. Now, I think I would like to spend some time to convince you that these are lipid vesicle-like structures, although there’s no lipid. One of the most important part in a lipid vesicle is that the ordering in the lipid bilayer comes from how heads are joined and tails or outside, or vice versa, depending on what type of liquid you have. And that creates the organization within this bilayer creates an ordering that can be termed as liquid crystalline ordering.

Priya Banerjee (00:45:51):
So, we looked for such ordering in the first place, because the simulations suggest that indeed protein-RNA can create some structures, which could be driven by ordering in a specific way. Polarization microscopy detects such ordering. So you use polarized light, and your detectors also can detect polarized light. And if there is specific polarization that comes from refractive index mismatch, you would be able to see some patterns which are analyzed based on some models. So, for liquid crystalline materials which are hollow vesicles you typically see this [inaudible 00:46:31] quartet structure, which tells you that there are ordering on the rim.

Priya Banerjee (00:46:35):
Under a fluorescent microscope it looks like hollow, but if you turn on the polarization, you see this beautiful quarter structure suggesting and there are ordering of lipid, ordering of protein or RNA molecules within the rim, similar to lipid-like molecules. And you can see that also for excess protamine — So we can all not only form vesicles with one building block which is RNA being in excess, we can form the same type of vesicles or similar vesicles where we have excess protein lower amount of RNA and we see similar type of behavior in polarization microscopy.

Priya Banerjee (00:47:15):
And again we do not really understand this behavior at the molecular level yet. So we are doing more experiments and simulation to understand this, but this is remarkable given that none of the building block has any sort of ordering, but overall they can come together and create an ordered assembly.

Priya Banerjee (00:47:32):
So, another very important property of lipid vesicle is a size dependent filter-like property. So, what you have here is basically this vesicle-like rim when you have for a liquid or any other building block, it can only let some molecules to go through while others are not permitted based on their sizes. So, we thought maybe we should test that and see if there is a typical mesh size of these condensate ring. And what we have found here is using dextran probes. So dextran is a polymer that can be found in different sizes. So, we went all the way from 4,400-mer to 155,000-mer. So, these are different chain lengths and molecular weights and they have typical sizes. So for example 10K polymer would be 2.5 nanometer in terms of their overall size.

Priya Banerjee (00:48:31):
So, when we look at their partitioning when we add these fluorescent molecules on the outside and see if they can go through this barrier to inside, we see that for 4.4 K the intensity outside versus inside are pretty similar and these are again, summarized right here. So, the red is the lumen intensity compared to what you have outside and the blue is on the rim. And the rim is glowing a little bit more compared to what you have but nevertheless it goes through. For 10K we see that the inside is a little bit less bright compared to the outside, but the rim still gets something and 10K falls right here. But above 10K, we don’t see anything going on the rim. And you could see that 40k, 70k, and 155k pretty much similar in range. So they don’t even go to the rim, suggesting that there is indeed a size dependent filter behavior here. And the typical size based on this experiment we find is 2.5 to 2.8 nanometers.

Priya Banerjee (00:49:34):
And finally, for any lipid-like vesicles, we want to see if we can encapsulate biomolecules to carry out a specific function. If you’re thinking about biology or for people who are in bioengineering, they would like to deliver things. So you have to encapsulate something, cargo maybe, for that purpose. So we tested the ability of our vesicle-like structures to do the same. And you’ll see that indeed, it can do something similar. So if you have a single-stranded DNA, and this is basically the pro-molecule or the cargo molecule, it goes inside the protamine like vesicles and you see these beautiful [inaudible 00:50:13]-like structures. Double-stranded DNA also goes in and these are the profiles. You can clearly see how they look like and green versus red. Molecules such as BSA goes inside as well, and you could see that it’s excluded from the rim, but included in the lumen.

Priya Banerjee (00:50:31):
Also, you have prion domain of FUS, it goes barely inside, as you could see from here, that partition is not very strong. And control substrates such as green fluorescent protein doesn’t go in at all; it’s always just outside. And here, the inclusion coefficients are sort of ranked. So double-stranded DNA, as most in our case, and then followed by the BSA and single-stranded DNA. Again, we do not know the molecular rules that govern such partitioning yet, but we’re just trying to see if our vesicle-like enclosures indeed behave as similar to what you have in lipid vesicles. And we see a lot of similarities here in terms of encapsulation of biomolecules. So, these data are available in our bioRxiv article, which was also submitted in a journal for publication.

Priya Banerjee (00:51:29):
With that, I’d like to summarize the part two of my talk: that we can create vesicle-like enclosures by phase separation and these are completely lipid free. And these enclosures behave similar to what you see in lipid vesicles, such as local ordering, size-dependent permeability and select encapsulation. Now, one of the major questions we need to understand to go forward is how cells or how biology in general can utilize structures to carry out some function and that’s something we’re doing right now, to understand how multiphasic condensates, in general, can facilitate similar signaling reactions in general or any biochemical reaction in general.

Priya Banerjee (00:52:14):
With that, I would like to acknowledge people who are very instrumental to do all this work that I mentioned today. So, two postdocs in the lab, Mahdi and Richoo did the bulk of work. Ibraheem and Taranpreet did a lot of work, which I talked to you about today. And our collaborators UB and outside UB, our funding agencies at NIH and UB. And thank you again, all for giving me the opportunity to come here through Zoom and give the presentation. I love condensates website, as you mentioned. It’s like one of the most educational thing I have ever seen in this field. So thanks for doing it. And again, my undergraduate students such as Andrew and Hannah, they also had a very good impression of this website. So, again, thanks for your attention. I’d be happy to answer any question that you may have.

Mark Murcko (00:53:17):
Thanks. Wonderful.

Priya Banerjee (00:53:18):
Thank you.

Emily Wu (00:53:23):
I have a question. Hi, this is very nice. Thank you. I had a question about the RNA concentration. When you compared different type of RNA and you also have cellular RNA where you see the reconstruction of the hollow condensate.

Priya Banerjee (00:53:44):
Right here?

Emily (00:53:45):
Yeah, do you see differences when you put cellular RNA, is that concentration lower or higher?

Priya Banerjee (00:53:51):
Yes. So cellular RNA forms the structure at a lower concentration. We also see a lot of differences in the size and the size of the rim, compared to what RNA we use. So the cellular RNA that we used here is only I think six to eight mgs per mL, and that was good enough to form these structures versus when we used poly(U) RNA, it was more than… at least 11 to 11.5 mgs per mL. So, depending on what RNA you use, yes, absolutely, you see a lot of changes.

Emily (00:54:25):
And is structure, the size of the condensate also kind of smaller with cellular RNA?

Priya Banerjee (00:54:31):
Yes, that is true. So, with cellular RNA. It is smaller than what we… Every size’s smaller than what we see for the poly(U) or poly(A) type of RNA.

Emily (00:54:45):
Great, thank you.

John Manteiga (00:54:49):
Hi Priya.

Priya Banerjee (00:54:49):
Hello.

John (00:54:50):
Well done, I love the videos.

Priya Banerjee (00:54:52):
Thank you.

John (00:54:53):
I have a question about the partitioning of the different dextrans and also the specific like BSA versus DNA et cetera.

Priya Banerjee (00:55:01):
Yes.

John (00:55:01):
I’m wondering how the… So these like size cut offs, how do those correlate to the specific partitioning of the different, BSA and single-stranded DNA et cetera. Do they, do other ones break the rules here? Because they have features, or?

Priya Banerjee (00:55:18):
Yeah. So for the dextran probes we sort of think about these probes not interacting with the system at all. So, because they’re supposed to be not. So the experiments that we had done, we have done here is a little bit different here versus the encapsulation. So, here, we form the condensate and we add the dextran outside and ask the dextran, can you go inside? And that would be the true size-dependent partitioning behavior. And you see that for 4.4K to 10K, it can go a little bit but for higher order, it does not. For encapsulation, what we did, we had our carbon molecules in solution, and we formed the vesicles. And then we sort of flush out the system to see if we can still retain them. And typically, that’s what you do. We haven’t done complete separation of the outside medium. But we see that when we do that, we can still keep some of these molecules trapped within these. So typically, to answer your question these two experiments were done in two different ways.

John (00:56:32):
So if you had to guess if you would add these different things after the droplet was already formed, do you think they would look similar, or different?

Priya Banerjee (00:56:41):
That’s a great question, one thing we see is very different. So you see the correlation between what goes on the rim versus what goes inside. So for example, in dextran 4.4, you see a spike, that because it goes inside, as well as it goes on the rim. You can actually see that the blue and the red have some sort of correlation. It’s higher, lower, higher, lower and so on. So if it doesn’t go to the rim, it doesn’t go inside. That’s the bottom line. Here, this is something very interesting.

Priya Banerjee (00:57:08):
For example, if you take BSA — and we see that the rim is completely dark compared to outside or inside. So the rim there is a dip. So I don’t think if you add BSA outside, it’s going to go inside because it doesn’t go through at all. It doesn’t go on the rim. And that might be why it’s being trapped here because it just cannot go through it. And that could be due to interaction or due to just exclusion that we do not know. And we think the way this is behaving is different than the dextran.

John (00:57:39):
Awesome, thanks.

Alicia Zamudio (00:57:42):
I’m really curious if you think that these vesicle-like structures may have some physiological role of trafficking within the cell.

Priya Banerjee (00:57:53):
Well, I can only speculate without data. And I would love to but so what I would suggest is, there are not much literature about the structures being forming. But in germ granules I think there are a couple of papers on Oskar condensates where they see that the Oskar protein, which binds to RNA can form the hollow condensates, which they call germ granule. And they play a big role in trafficking of mRNA. And one of the defining feature of that particular work to us was when they mutate the protein and reduce the RNA binding affinity, those structures don’t form.

Priya Banerjee (00:58:40):
So, we think that there might be a lot of roles in how storage and metabolism could be regulated by forming structures, because it offers a different micro environment compared to what a condensate offers because of the high density of macromolecules within a condensate versus the lumens are basically in a dilute phase inside. And yeah, so I think they might play a big role and without much data here I may not be able to go anywhere beyond speculation.

Diana (00:59:19):
Thanks Priya, that was excellent.

Priya Banerjee (00:59:22):
Thank you.

Diana (00:59:23):
I was curious about… I think it’s really fascinating, when you were talking about the localization of the tails that are from the tadpoles. I’m just curious if you can comment a little bit about the preference?

Priya Banerjee (00:59:43):
Yes.

Diana (00:59:43):
Why would the tails prefer to go inside versus outside and where switch might happen. For instance, you were showing that if you tune the ratio between the RNA and the protein, you go through hollow condensates, full condensates and back to hollow. Do you know if during that transition, there’s also a change in the directionality of the tails now that’s the tail [through RNA or protein? crosstalk 1:00:00:10].

Priya Banerjee (01:00:12):
That’s a great question. We probably do not have enough data again in our simulation to answer your question. But I’ll start off by saying when we looked at these structures and talk to some of the colleagues who are working on lipid vesicles, lipid traps for a long time, they didn’t believe that these are the data. It’s an April Fool’s joke, these are forming by lipids. And that made us to think about and go into the literature of lipid mimicking polymers that form vesicles and they are protein vesicles that are formed mimicing lipid-like structures and what they have done typically is creating these two half of a polymer. And the two halfs they call it diblock polymers, two halves have different interaction propensity or different interaction with the solvent.

Priya Banerjee (01:01:05):
One way to think about, Rohit Pappu published a lot of work on how differential solvation could drive phase separation versus gelation. We think that is a fascinating idea to be explored here. Because if you think about this tadpole, so I’m hoping you can still see the slides, right?

Diana (01:01:25):
Yes.

Priya Banerjee (01:01:25):
So, if you think about that you only have condensed the head partially. So, this is an electrically neutral condensate. So, the interaction with solvent would be different than this charge still and if you have to think about it, you might think that the charge still will interact solvent more favorably than these specific part where you have a protein-RNA complex.

Priya Banerjee (01:01:49):
So when they are joined together, you might see a preferential orientation of these molecules or these complexes, partially condensed complexes to maximize how they interact with solvents. And that might actually tell you how they should orient. For example, when they go to the surface, the RNA chains can maximize their interaction with the solvent. And when you have higher concentration of these chains, they now can even have another surface where dominating coincide.

Priya Banerjee (01:02:25):
And we actually see, both in our experiment and in simulation again, that’s not part of this manuscript that we presented, but we see that RNA will be higher in concentration compared to outside to inside. And that is because of this limitation due to favorable interaction with each other or with the solvent. But how this transition occurring from one state to another, we do not have a clue to be honest here, because this is a fascinating phenomena that we just found, and we’re trying to study everything that we can to understand the mechanism.

Diana (01:03:04):
So, earlier, you were showing that at least the diffusion is the same in the lumen and the outside. So do you expect that the composition, I don’t know salt concentration or whatnot is different in the lumen or is more of a re-entrance [crosstalk 01:03:19]-

Priya Banerjee (01:03:19):
So RNA concentration, we do not know if the salt concentration is different. FCS cannot detect the differences to be more specific, but when they look at the RNA intensity profiles, we see a little bump inside and outside. And the simulation showed the same thing. And that is something that we’re starting right now to understand if it is indeed more enriched in RNA inside than outside. And if that is true, then it might be a way to pack on an inside. So, that’s something diffusion cannot detect. But when they did some simulation on mean square displacement of a [pro inaudible 1:00:03:59] particle they put that’s also part of the manuscript. So what they found that they can detect. In the simulation, they can detect that the outside is more mobile than the lumen.

Priya Banerjee (01:04:12):
And combining with microscopy measurement that we had, we see that the outside although the intensity is pretty low compared to the rim, the outside versus inside, there’s a ratio of five to one. But again, the intensity inside is pretty low because again, they look hollow and it plays basically you’re counting darkness. But you can also see right here, there’s a little bit more red than the signal has to be quantified. So you’re trying to do some inner state analysis where you can quantify the signal from inside versus outside. But the lumen actually is a problem because it is so bright, and you need a larger structure to look at those.

Diana (01:05:02):
That’s really interesting. Thank you.

Mark (01:05:08):
The other thing that’s interesting when you think about arginine versus lysine, you also think about different post translational modification-

Priya Banerjee (01:05:15):
Yes.

Mark (01:05:16):
… opportunities as well. And so it’s always interesting when thinking about results like yours to then add on the layer of what if some of the arginines are methylated versus not, that sort of thing.

Priya Banerjee (01:05:29):
So we were studying with a colleague here who specializes on arginine methylation with several arginines he studied in human. And we see some results that is striking similarity with the arginine versus lysine. So when I was describing this data to him, he thought was really interesting. And we looked at some of the stuff that he’s doing, and we see a nice correlation and again, unpublished data, so I can’t talk a lot, but we think arginine versus lysine could be relevant to understand arginine methylation for arginine. So it may not perturb the phase separation by a great deal but it can change how condensate dynamics is regulated.

Mark (01:06:15):
Yeah, the strength of the interaction if you have an arginine that’s model or di-methylated, the pi-pi stalking interactions can be quite different?

Priya Banerjee (01:06:25):
Yes.

Mark (01:06:26):
In the mono- and di-methylated arginine versus the normal arginine.

Priya Banerjee (01:06:30):
And at the same time, you may not change the electrostatics. So, basically, you are changing arginine to lysine where you are selectively eliminating only one or two in the hierarchy without effectively eliminating phase separation.

Mark (01:06:46):
Exactly. And it can slightly change the dynamic motions, it can change the strength of the interactions

Priya Banerjee (01:06:54):
Yeah.

Mark (01:06:54):
So the dynamics of the complexes that are forming and unforming all of that can be affected just by a few methyl groups.

Priya Banerjee (01:07:03):
Yeah. And that gives you handle to regulate functions.

Mark (01:07:10):
Exactly.

Diana (01:07:18):
Excellent. Does anyone else have any questions for Priya?

Mark (01:07:25):
Sometimes we think of questions later. So maybe we’ll get back to you later.

Priya Banerjee (01:07:28):
Definitely no problem. I’ll try to do my best to email back as soon as possible. But I think there is a chat in the chat window, I saw something.

Mark (01:07:37):
Or someone who had to… someone’s thanking you for your talk.

Priya Banerjee (01:07:43):
Perfect.

Mark (01:07:44):
Matthaus from our Dresden lab.

Diana (01:07:51):
Yes. Thank you very much. This was excellent. And we’ll get back to you with more questions.

Priya Banerjee (01:07:57):
Thank you, Diana. Thanks for the opportunity again, and a very nice meeting with all of you. So stay safe and healthy and hope this will be over soon.

Mark (01:08:05):
Yeah, likewise, thanks so much.

Priya Banerjee (01:08:07):
Yeah. All right. Take care.

Diana (01:08:09):
Bye.

Priya Banerjee (01:08:09):
Bye-bye. Take care.

Diana (01:08:13):
Bye. It was good seeing you.

John (01:08:14):
Bye-bye.

Mark (01:08:14):
Thank you. Bye-bye.