 All right, thank you. Good morning everyone, a big thank you to the organisers for giving me the opportunity to present some of the work that we've been doing on the self-assembly kinetics of linear virus like particles. I apologise for my voice, I'm a Dutch person so I don't do well at elevated temperatures and under the influence of sunlight. Anyway, so these virus like particles consist of a strand of a double-stranded DNA which have been encapsulated by a tri-block recombinant biosynthetic protein and I'll show you the chemical structure later on in my talk. This biosynthetic protein which actually was designed to mimic the functionality of the code proteins of a tobacco mosaic virus. Many moons ago we investigated theoretically both the stability and the assembly kinetics of a tobacco mosaic virus and that led us to design this protein. But before going on, here are my buddies in crime. There's an experimental group who did all the experiments. There's the group of Renko de Vries at Wageningen University which is in the east of the Netherlands. The actual experiments were done by Amando Hernandez-Cathia who is now at Northwestern. We've actually, I mean, left two years ago. We've since done more experiments. I'll be reporting on the end of my talk hopefully. The theory gang is welcomes us of Willem Kegel who is part of the audience and actually Willem is the chap who drew me into physical virology. So maybe, I don't know, 12 or 13 years ago he sent me a manuscript about the self-assembly of tobacco mosaic virus in fact and I thought it was rubbish. But we wrote a fantastic paper about it. No, no, no, no, no, no, as you mentioned. Anyway, at the time, Daniela Kraft who now is at Leiden University was a PhD student with Willem doing experiments. Her background is theoretical physics. So she turned to the dark side but she wanted to have like a little toy theoretical problem. So the first theory for the self-assembly kinetics was done with Daniela. Very recently, another chap, Mella Punter, who was a master student in my group, he extended the model that Daniela and Willem and I coughed up so many years ago. Okay, so what do we want? Oh, funding. It's rather important. Funding virus-related topic is not that easy in the Netherlands, I think. I mean, so we got some funding from NWO, which is the Dutch equivalent of the NSF and the now defunct Dutch Polymer Institute, which is always extremely generous and I used to have a project with the previous speaker from the HFSP and there's a chap sitting over there who got a salary out of it. Anyway, so our aim was to encapsulate DNA. The reason, I mean, way in the background is gene therapy, double starling DNA, using a synthetic protein. Of course, viruses are the ultimate encapsulation machines. They're made to infect things. So we thought, why not imitate what nature does so well? Here on the left is an image of a CCMV, which is a plant virus. It's very, very small. It's just shy of 30 nanometers across. About half the viruses have this shape. Or, for instance, look at tobacco mosaic virus, which is actually the first virus to be discovered, I think. That one is not longer. This is about 300 nanometers. Now, if you look at the internal structure of CCMV and TMV, it's actually quite different. They're very different. I mean, the shape difference is obvious, but also the way that the genome was encapsulated is very, very different. If you look at CCMV or any other small plant virus, it consists of this shell, you know, of protein. And inside this shell, you've got a cavity. And this cavity holds the genome. This is the genome, where, for CCMV, you've got three different particles that they all have, roughly, 3,000 nucleotides of single-star irony inside. Now, the driving forwards, so actually, I should say, both these viruses have been reconstituted in vitro, which means that you can disassemble the viruses into the proteins and the genome. You can purify them and then mix them into the same test tube. You swirl, I presume. That's what I think that experimentalists do. And, lo and behold, it will self-disable. And the driving forwards, I mean, both, I mean, certainly for CCMV, in the first instance is electrostatics. So the inner surface of the viruses is very highly positively charged. And this is also the reason why you can just yank out the RNA and replace it with anything else that is negatively charged, like particles or droplets or whatever. In fact, I don't know if Adam Slotnik is in the audience yet, no? But actually, Adam Slotnik encapsulated DNA. This is interesting because, you know, this particle is 30 nanometers across. DNA has a persistence length of 50 nanometers. It is no way that you can get DNA in this very, very, very small lumen. So the protein is easy. It adapts to the template that you offer it. But it still has this preferred curvature that is, you know, like 30 nanometers. So what this protein does, it will grab hold of multiple strands of DNA and form a tube with a diameter of 30 nanometers. Our goal was to encapsulate single RNAs, not bundles. And one reason would be that, well, you know, if you have a long thing, long thing object that's more easily absorbed by cells than a fat one. So why not try the trick of tobacco mosaic virus? Well, tobacco mosaic virus doesn't have a lumen. It has a hole which does run through the entire center of the object. But the RNA is not encapsulated in this lumen. It's actually embedded in the capsaid itself. Right? This TMP will not encapsulate DNA, double-stranded DNA. It will do single-stranded if you force it to. So our goal was now to try and combine, you know, the properties of TMV and the properties of CCMV. So the easy, so having physical driving for the assembly and still have used the tricks that TMV use in order to encapsulate something. Because as we shall see, that's not so trivial. Because it's a quasi-linear object. And in fact, we already knew this about, I mean, quite a while. We tried to encapsulate single-stranded DNA, homopolymeric. So it cannot self-hybridize. So it's a polymer, like a homopolymer. And we tried to do that with a synthetic molecule, which is a naftali derivative, which has three functionalities. It's got a functionality that will allow it to hydrogen bond with the single-stranded DNA. It got this planar group. This planar group will allow it to interact with itself and self-assemble along the single-stranded DNA. And then there is a group attached in order to increase the solubility in water. And now this homopolymeric DNA will not form a helix, which means that you can monitor helicity as a proxy for how much stuff is absorbed onto the DNA. Because, I mean, if you scratch your eyes a little bit, actually quite a bit, then this looks a little bit like adenine, so ATCG. So we mix them together. And what happens is we see a CD signal coming up, and the CD signal tells you that there is a co-assembly into a helical structure of the DNA and that let's call that a code protein. And so this is what happens. So on the vertical axis, we've got the normalized a CD signal as a function of the concentration of this object relative to the concentration of the template for three different template lengths. So this is the longest, so 40 nucleotides, 20, 10, and 5. So we're adding here more of the code protein. And what we're seeing is that the CD signal increases, so it gets packaged. The DNA gets packaged. Actually, it doesn't really get packaged at all. It's really difficult to get full 100% coverage of the DNA. And actually with hindsight, that makes sense, of course, because it's like a one-dimensional absorption process. Entropy rears its ugly head. I will create vacancies. So DMV is smarter than this. At the time, we're not as smart as DMV. By fitting the experimental data to like an eising-like model, we can extract stacking energies and binding energies. By doing the same experiments as a function of temperature, we can even extract and apply the model. We can extract enthalpies as well. And so here again, the CD signal versus now temperature for one concentration of 0.25 millimolars at this stoichiometric ratio. By fitting the theory to the experimental data, we get enthalpies of, let's say, about 20 KBT. Now, you can see that there is a reasonably well-defined elongation temperature. So what you can now do is plot this elongation temperature as a function of the concentration of the code protein, the quotation marks code protein. And that gives us these straight lines, which you expect, actually, Vanthoff's law. And now the interesting thing is what you can do is you can also not add the template. And these objects will also self-assemble without the template. And but it does that at much, much, much lower temperatures, except for this one. So this is G1. That's this one. At large enough concentrations, actually what happens is that the self-assembly takes over from the templated assembly. So what you have is a competition between templated assembly and self-assembly. This is something that if we want to design a virus, a linear virus, we have to make sure that does not happen. Also, okay, so how does TMV do it? So TMV doesn't follow this route. It doesn't rely on linear type absorption. What TMV has is what is known as an origin of assembly sequence, which is the same thing as rating the packaging signal. And assembly actually happens bi-directional. It starts at this packaging signal and then runs that way. And then later on, in the last part of the assembly, the assembly goes in this direction. Let us forget that this is about 900 nucleotides upstream. The RNA cell is slightly over 6000 nucleotides. So we can forget about the last part. The last part is boring. Now looking at images like this one, what we're seeing is indeed linear object growing as a function of time. And if you focus onto one of those images, you see a rod and what you see is two strands of RNA coming out. This shorter strand is this bit over here. The longest strand that's all of this that has not been encapsulated yet. So this is one of the views that people think is how assembly happens. Under conditions of neutral pH and physiological salt, the protein itself is not in monomeric form. It's in this 34 CP bilayer disc. Actually, if you reduce the pH a little bit, this bilayer disc morphs into what they call a lock washer. At low pH, this lock washer will then self-assemble into helices of various lengths. But we are at neutral pH, so this is a stable form. This means that this form is a high energy structure. Now what the original assembly does, it pokes a little finger. It's a stem loop type thing, interestingly enough. It pokes a little finger into that hole and it uses a conformational change in the protein. And it's this conformational change in the protein that actually causes this bilayer disc to morph into this lock washer. What next happens is that either monomers attached to this lock washer or complete bilayer discs, which then also morph induced by the first one that has already morphed. So this is a luster. So actually turning this into this costs a lot of energy. Turning this one into that costs less energy because of the presence of the first one. And what happens is that the remaining part of the RNA strand is drawn in through the hole in the middle and finds itself lodged in the core of the protein, which has undergone this conformational change. The question now, of course, is why so difficult? And actually, I should say, is this not just TMV that does this? This whole bunch of viruses? Yeah. Papaya mosaic virus, clover yellow mosaic virus, papaya ringspot virus and tobacco. I forgot what the R means, a virus. So apparently this is the way to do it. Okay, so a teeny weeny bit of theory, not too much because this helps us to design the protein. Let us just look at a simple language adsorption. So you have template, this is our DNA now because our goal is to template DNA. And so if you have an attractive interaction between the code proteins and the template, then you'll get language adsorption, which is not very cooperative. It is really difficult to get the all of the binding sites covered by code protein. So you could say, well, these proteins might interact. That reduces the interaction energy further, because more negative. So that drives further assembly. So far so good, but still it's a one dimensional process and an entropy will make holes in this. Making holes in a capsid is not so good for the virus because nucleases might just think something to nibble at. That's the end of the virus. Right. So worse, if you allow the proteins also to interact, then they will also self assemble a solution. Oh, not good either. Right. So, so binding will give you on this is good interactions are good and interaction between proteins are bad as well because you get self assembling. Now let's look at this at this red. This is also well in physics land, everything is fair. This is a protein, a code protein, and it's not assembly active. So it's in this in this low energy state. Let us presume that a template now has an origin of assembly where these proteins would preferentially absorb onto and then change the conformational structure of their protein. Right. So more from red to green. It has to pay something for that. But now the protein can recruit other proteins which are of the wrong color, turn them into the right color and gain, whoops. Okay. And gain a free energy of binding, a free energy of interaction, lateral interactions, and then there is a little bit you have to invest to get the process started. But this is a nucleating process. Right. The statistical mechanics of this is really easy. It's called the zipper model. You have in essence three knobs that you can turn. The first knob is what we call the mass action variable. This mass action variable is a product of the concentration of free proteins, 5p in a solution times a binding constant. Or you could say, well, it's equal to the volume fraction, 5p divided by some critical concentration of proteins, like a critical aggregation concentration. The second thing which is important is the stoichiometry. So the stoichiometry tells you what the ratio is of the number of binding sites in your solution and the number of proteins. And stoichiometry turns out to be extremely important to the success of this process. And finally, because this is a nucleated process, it's a cooperativity plays an important role and it's captured by a parameter that we call sigma. And that's again a Boltzmann factor involving this free energy that you have to pay, H, any free energy epsilon of the binding because the very first protein that binds doesn't have a neighbor. So it actually in essence loses that. Okay, so this allows you to calculate the size distribution. So how many proteins are on average assembled onto the templates? So G is the binding energy of the protein to the template and epsilon is the binding energy of two code proteins that are in contact. And then H again is this free energy that you have to pay in order to morph the protein into its wrong shape. It's the stacking energy, yes. Yeah, it's a stacking energy. And then if you look at the stacks in the free solution, that might have a slightly different stacking free energy because the template will position the planar groups in such a way that you're not losing as much entropy. As you would in free solution, but never mind. So you can also write down your rate equations. I'm not going to show them. You know, you have an empty one, and then you have one, you have two, you have three, one can fall off. And in essence, you've got so forward rates, which depends on the number of proteins on the template. This is the backward rate, this is the forward rate. And these two are connected through microscopic reversibility, meaning that they're connected to these energy scales that we have in the problem. And then you stick this into Mathematica or you write a program and you can calculate the non-equilibrium distribution of the number of proteins absorbed onto a template, right? Let's first have a look at the equilibrium distribution because that tells us why it is advantageous to zipper. And the reason is as follows. So on the vertical axis, we've got the base 10 log of the probability of having n proteins on the template relative to having no proteins on the template as a function of the number of binding sites. And so these lines are what we find depending on the ratio of the concentration relative to the critical concentration, right? So this blue line gives you this ratio, we're just below the critical concentration, 0.98. This red line gives you the distribution exactly at the critical concentration 1.02, just above, 1 to almost, just above. So whether you're below or above, you always have an exponentially decaying function, but it depends on from what side of the assembly. So if you're below the critical concentration of your exponential decay makes assemblies even less stable than the monomers, I mean than the free templates which sit over here. Exactly at the critical concentration, we have got a flat liner. We've got a flat liner which is below the critical one, the empty one. So you still don't see templated assemblies. You have to go slightly above it and only slightly which means that this process is highly cooperative. You have to go a little bit above it. What we see is that we have complete ones, so an exponential distribution now starting at the complete ones and going down. And this means that intermediates are suppressed. To suppress intermediates is crucial. Why is it crucial? It's crucial because you do not want to waste protein on incomplete viruses. Right? So this is a way of having either no templates covered by protein of only ones that are complete, more or less. And this means also that if your stoichiometry is not right, which means that if you do not have a lot of protein, you're not wasting that protein or putting one or two proteins on a whole bunch of templates. Those few that you have, you put on one, that one survives. So it's a highly nucleated process. There's a lag time. I'll show if I'm not going over time, I might be showing you some graphs as well. And so first of all, the transition is really sharp. So there is a really critical concentration. A second of all, intermediates are suppressed. And finally, even if your protein concentration is just a little bit above the critical one, if your template is long enough, it will actually encapsulate it. So the cooperativity increases with the length of the template, which is why you can increase the length of your RNA of TMV and still have complete viruses. And in fact, there are zillions of patents out there where people put genes on the genome of TMV to produce in tobacco plants. Okay, so first of all, the slope of this line depends on the binding affinity. So the larger the binding affinity, the more steeper the slope. But at the same time, stoichiometry does the opposite way. So if you starve the solution of protein, then the slope goes down and down and down, which means at some point you're going to lose complete coverage. Actually, you will not get any coverage at all. Okay, dynamic. So let us do an experiment. And let's take a bit of DNA, which is considered 51 binding sites. The cooperativity has a value of 0.007, which is a typical level of cooperativity, that you also find in supramilocular polymers or in organic chemistry. It's ethrop... I'll show. So it's this Boltzmann factor involving this free energy barrier that you have. And then the absence of a bond, which also acts like a barrier. Yeah, not so. All right, la-de-da-de-da, whoops. Q is 51, segment 0.007. Sorry, possible, finding an one-dimensional object like that. Do you remember the bigger... Yes, yes. ...in what they mentioned, or the same distribution? Yes. And the level of size of the slope... Yes, I agree, but we're actually making use of the fact that you have to turn the protein into something that is able to absorb. So you're killing entropy through energy. I forgot to mention that, yes. Entropy bad, energy good. So we're killing... In essence, why you don't have a phase transition in 1DE if your interaction range is finite. Right? So by having this... Yes, yeah. It's a different type. I can talk about it after my talk, if you want, because it's a little bit technical. Actually, it's not that technical. Okay, so we start off with only empty templates. And that's here, this axis here. As a function of the log of time, I've scaled the log to this level of cooperativity, because the level of cooperativity dictates your lag time. So that's taken out now. Now, for two different stoichiometry. Stoichiometry is zero, means an infinite amount of protein out there. So you're never depleted of protein. And stoichiometry is 1.5, which means that you haven't got enough protein to cover all of the assemblies. Now, for short times, the probability of finding empty proteins, empty templates starts at one, and then it goes down, and down, and down, and down, it's a song. And it's only at the later stages that you start to see a difference between excess protein and not excess protein, so depletion of protein. It takes a while for the proteins to register that they're not enough of them around. And then if you have excess protein, in the end, you end up with almost no empty templates. If, however, the number of proteins is not sufficient to cover all of the templates, then at some point you start to level off, because some of them will remain empty. This is exactly what we want. Notice there is a little bit of an undershoot. If we look at, on the right-hand side, so in blue, that's the probability you have a complete capsid. So for the stoichiometry of zero, so an infinite amount of protein, it goes like that, and it's really sharp, and whoop, goes like that. However, if I don't have enough protein, I will still be able to make complete viruses. This is the cool thing about this. And it goes up. I overshoot a little bit, and then I go down again. Right? So we have undershoots and overshoots, and here these undershoots and overshoots, they only in the theory happen only if you do not have enough protein in the solution, never when you have an excess protein, never. And this will, I'll show you later if I have enough time, anyway. Well, you can also look at the average coverage, theta, as a function of time, for two different concentrations now. So the red curves are for an excess protein, and the blue ones are for not enough protein for two different concentrations. So the concentration is here, e to the power of three, that's about 20, I think, 20 times the critical one, and e is about three times the critical concentration. So you see that the larger the concentration, the quicker the assembly goes. And this is also anyway what you expect, because the thermodynamic driving force is larger. You can't see it, but there's a little bit of an overshoot over there. So again, if you don't have enough protein, you get an overshoot. It is very, very small here, but the largest overshoot we can get is about 100%, and they're very, very long lasted. So the relaxation time, it takes, I mean, it takes a very long time for this maximum to go away again. And I think this, you know, crawls back a little bit to what some people have called the pseudo law of mass action, because if you think that maximum that I can see and you cannot see, because I believe it's there, you know, it takes a very long time to even out. So if you take that as your fraction occupied size, and you calculate back your binding free energy, then the binding free energy that you get is wrong. Okay, so there are overshoots, undershoots, depending on the concentration, depending on the stored chemistry, overshoots in theta, and that's the thing that we're going to measure, we only see up here, so for large values of lambda, so in the conditions of excess protein. Okay, I'll come back to that hopefully in about five minutes time, if I have that time. Okay, let's go back to our TMV protein. So if TMV protein has functionalities, one functionality is one part of the protein has to wrap itself around the RNA and do the death now as it were, it undergoes a conformational change. There's a part of the protein that is involved in the interaction with neighboring co-proteins and then there is a part that imparts colloidal stability, we call this shielding colloidal stability to the particles, because you don't want them to drop out of solution. So this is our well, I mean if you make something in the lab it doesn't look as nice as when nature makes something, but it does a trick, so we have a binding sequence, it's basic, it's like 12 lysines, it's positively charged. There is a silk-like block which consists of a certain number and times a silk-like sequence going in adenine, blah, blah, blah, and we can vary the number of those sequences, one, so zero, one, two, all the way up to 18. And then we've got our shielding block which is a collagen-like block of a 400 amino acids which imparts this colloidal stability to the complexes, because you're neutralizing something that is charged, so you need your colloidal stability. Now it turns out that this silk-like block is able to form either a beta sheet or a beta barrel, depending on how big this N is and depending also on the concentration. If we for instance take four of those units, then this object will not undergo this conformational change into a beta sheet and we know this because we do circular dichrism. And what happens is here, so this is a two and a half kilobase pair of double-stranded DNA, this is an AFM image, and essentially what happens is you have, because this is positively charged, it will just adsorb like in the usual way, in the Langmuir way. It's not actually packaging the DNA, and in fact if you expose this to nucleases, the nucleases will kill the DNA. If you take 10 of those sequences or 12 or 14 or whatever, then what happens is that this is the same length of DNA. Now it's packaged into something that is a lot shorter than the original length, so somehow it's been packaged in this object. If you expose this to nucleases, it's protected. And so it shows that having this conformational change is kind of important in order to be able to protect the RNA, in this case the DNA. Okay, dynamics, because we're interested in dynamics, so this is what happens in, if you do an experiment, so you mix the artificial co-protein with the DNA, and every so many minutes you take a little bit of a volume, when you deposit it on the surface, and you do your AFM, you clean it, etc. And so this is a short time, somewhat longer time, somewhat longer times. So we're actually seeing growth from one end. It starts to grow at one end. Hang on, we didn't put in this origin of assembly. I'm good, thank you. We didn't put in the origin of assembly. So why does it start at one of the ends and not somewhere in the middle? Well, look at its tail. This is a big, it's a polymer. It's a coil, it's a random coil. Radiative generation, 50 nanometers. So you're absorbing, actually this goes back to a question that Robin was asking about an hour ago. You know, if you take a polymer, you try to put it onto the surface, this polymer senses the presence of the surface. So the surface in a way repels it because of entropy. So if it absorbs towards the end, you don't have that yet. So even though we did not build in this origin of assembly, by accident we did. But hey, who cares? Okay, now I'm almost done. So we did, as a function of time, measure the length of this growing virus-like particle. This is for the CS10B, so we have 10 of those sequences as a function of time. These bars are the experimental data, and in red is our theory. Now, I will not bore you with how we did the curve fitting, because we had many parameters. The curve fitting is extremely difficult, exactly because you have so many parameters. You have to just believe me that we didn't do anything wrong or fibbed or whatever. I'm not going to tell you that. It's less than what string theory is, I think. Anyway, the cool thing that you already see is this exponential distribution for late times. That's this exponential distribution, this inverted exponential distribution I showed you that came out of the equilibrium model. So for late times we were able to extract the energies that we need to put in. We did the same thing for a protein with 14 of those sequences, and again, so these are the experimental data, and these are our curve fits, which are not brilliant, but they seem to follow what we see in experiments. And I have to say, so these are AFM-based measurements, your statistics are not that good. So one of those bars here could represent maybe 10 viruses, excuse me, virus-like particles. So your statistics are not that good. Anyway, so we can, from our curve fit, we can extract the energies. So the binding energy to the surface, the protein-protein binding energy combined gives us minus 17 kT for both. The free energy barrier to assembly is 1, 6 kT, and 3 for the other one. And we can also back out to the forward rate, which is in minutes of four times the center power of nine. Now, what these numbers mean, well, we have to do more experiments to see changes in these parameters to be able to interpret what we are seeing. What we can do, of course, is compare with our curve fitting, what we did ages ago, to the experimental data on TMV. And interestingly enough, so the binding energies are very similar. So we get 17 kT for TMV, an energy barrier of 7 kT, and a forward rate, which is actually a factor of 10 smaller than for our artificial volume. Okay. I mentioned statistics, and this, I have a little, how much more time do I have? I have two more slides, this one, and then the next one, and then I'll shut up. So I will not show you some more new theory, unfortunately, even though that's the more interesting. No, no, no. Experiments are cool. So what we did is twice, okay, so how can we get better statistics? You know what, I mean, doing AFM is not going to help you. So what we did, we replaced the DNA by a proxy genome. And this proxy genome is optomechanically active, which means that if you stretch it, its luminescence changes, which is really cool. And this is what we find, right? So by looking at the luminescence, by looking at two peaks and taking the ratio, we call this the fraction of a torte structure. Torte is like, do you stretch on it? So this is at a concentration of 0.06 micromolars for stoichiometry. So F plus here means the number of positive charges relative to the total number of charges. So if F plus is a half, I have perfect stoichiometric matching of the total number of positive charges and negative charges. Let's look at 0.1. So I don't have enough positive charge. I don't have enough protein, so my system starts the protein, time equals zero, goes up, the signal goes up, and I goes up a little bit more. At exact stoichiometric ratio of 0.5, that's the green curve, it goes up, it sort of seems to level off a little bit, and whoop, it takes off. And the same is true if we have too much protein there. Now, this now we interpret as initially as a Langmuir-type adsorption. So the first thing that these objects do is they just indeed adsorb onto the DNA without doing anything to the DNA. You have to wait a little bit. This is the nucleation process. After a while, they start to think, hmm, maybe I want to package it. And that's when the zippering happens. So there is a nucleation process, and then zipping and packaging of the DNA. And so we did the same experiments at a concentration 10 times higher, and then where things started to happen. And this is also, I think, where our model protein starts to diverge from the actual TMV code protein. Because for instance, if we look at the blue curve here, that's at a stoichiometric ratio of 0.7, so we have 20% more charges positive charges in your solution that you actually need. You know, you've got your lag time, and then it goes up, starts to zip, and then, oh dear, I have an overshoot. And my first reaction was, I predicted this. Except this happens at excess protein, where my theory tells me, no, Paul, there's no overshoot. Okay, so how do I weasel myself out of this? Because I will. First of all, we know that if we only have protein in the solution at high enough concentrations, it will form assemblies. So there's parasitic self-assembly. And second of all, we know, and I don't have time to explain this, by using a different optomechanical probe that is able to measure to probe whether or not I have one template encapsulated, two or three or four. So we know that what happens here is also a co-assembly of templates. So instead of encapsulating just one proxy genome, and we're encapsulating many more. Which is also what happened with CCMV, remember? The early experiments of Adam Slotnik. Okay, with this, okay, I'm not going to show this. No, no, that's, no, no, no, no. We have a theory, blah, blah, blah. Does the theory explain the data? Yeah. Okay, so finally, so protein polymers can be designed to mimic co-proteins of linear viruses, but you're cool. Our model tri-block protein co-polymer successfully encapsulates DNA. And in fact, it transfects, which is kind of, I didn't mention that. I think that allosteria and directional assembly are crucial if you want to encapsulate a linear object. So you have to beat entropy. So the kinetic zipper model describes the time evolution of encapsulating DNA reasonably well. We predict over and undershooting the conditions of excess DNA, oh, I should have, excess DNA, not excess protein, and overshooting under conditions of excess protein may occur in competition when you have, for instance, micellization. That's what a model actually tells you that it didn't have time to explain. And with that, I would like to thank you for your attention. And is there any questions? Just let me know.