Structure Club
Structure Club invites scientists to present one of their papers to a broad audience. Listeners will hear about cutting edge science from the scientists themselves.
Structure Club
Peter Cherepanov
Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.
Peter Cherepanov discusses his recent paper, “Integrase anchors viral RNA to the HIV-1 capsid interior”
Welcome to Structure Club. I'm Scott Sang. I'm a professor at Florida State University where I use Cryo EM to determine the molecular mechanisms of membrane remodeling.
SPEAKER_01And I am Ashwant Francis, an assistant professor at FSU. My lab focuses on the structural biology of viruses and how they interact with cells. Together, Scott and I created the Structure Club as a general club podcast and YouTube channel, where the papers are given by the authors themselves. Today's speaker is Peter Chernobyl. Welcome, Peter. Thank you. Thank you for having me. Peter got his PhD in medical sciences from the Catholic University in Belgium in 2000 and completed his postdoctoral training with Alan Engelman at Harvard Medical School. After that, he started his independent research program as a joint appointment between Imperial College and Francis Craig Institute in London. He will be talking about his recent paper, Integrace Henricher's Vital RNA to the HLV1 capsule interior. The manuscript came up in the journal Nature in January 2026. Peter, take it away.
SPEAKER_00Thank you. Let me let me share. How is the picture? Looks good? Looks great. Okay, all right. Okay, thank you. Um just to give you a very brief introduction in retrovirology. Um any retrovirus starts its life as RNA virus. So it has RNA genome uh inside, inside of the core structure. And uh once the viral membrane and the cellular membranes fuse together, the core gets access to the cytoplasm. And there in the cytoplasm, uh, it will be exposed to uh nucleotides and all other essential things. So the RNA will be converted by reverse transcription into DNA form. This DNA form is a double-stranded DNA molecule. It's still inside of the capsid core will be important in the nucleus where it's integrated. This integration step is a defining feature of RESTA viruses. It's very important for viral replication, and of course, inhibitors of HIV1 integrase uh make amazing drugs to treat um HIV infection. Now, this is all about canonical function of integrates as a DNA binding protein as a DNA enzyme. Now, to do this canonical function, integrase must bind to the ends of the viral DNA as a multimere. It protects viral DNA and so remember this DNA is a product of reverse inscription. By that moment, RNA is destroyed. Uh, and the most important function of it uh during that stage is to insert the three prime ends of this viral DNA into the host DNA. That's actual integration, and that's what's targeted by this clinical HIV1 integrated inhibitors. Now, just to illustrate you uh one such example of a functional complex between integrase and viral DNA. So this is uh we we call these complexes intersomes. So this is a very simple structure from prototype formivirus solved already many years ago. So it synapses two viral DNANs. Of course, the rest of the viral DNA is like a massive loop here, which was not present in these crystal structures, and between the active sites, there is space for target chromosomal DNA to bind.
unknownRight?
SPEAKER_00So this is a very simple interosome made out of four tetra four monomers of integrates. And today we have a whole collection or whole zoo of these interstone structures, uh ranging this from tetrameric to the hexadecomeric. And of course, the most important and interesting for us are the lantiviral intersomes, so the genus to which HIV1 belongs, right? So in the most complete structure we have for the lantivirus of the ship, my divisionovirus, and has these four tetramers of integrates, and between these tetrameres, you have the same structure embedded uh as we solved by crystallography many, many years ago. And this is an example of HIV uh intosome, so by Dmitry Lumkis and Bob Craigie already a while ago. It's not complete because it's missing a couple of subunits, but basically it's the same structure. Now, that's canonical function that was studied for for since as long as I remember. Now, for a while, there were indications that integrates may play additional roles in viral replication. First of all, most of the mutants you can make, uh, most of the I mean substitutions within integrates, HV1 integration you can make, uh, that would have any kind of impact on viral replication, they will not be defective at the integration, but at the step well before that. In fact, you you generally the phenotype manifests as a loss of reverse transcription. We call them class two integrated mutants. A huge collection of these mutants, and many of them were characterized by La Engelmann's lab and Mamuka Krashkele and a few other labs. And also, if you you can if you grow virus in the presence of this new type of inhibitors called elosteric inhibitors, they have this uh amazing effect on integrates, basically, they precipitate aggregate integrates. In both cases, you get non-infectious virus, and when you look at the details, this was discovered in Al Alan Engman's lab already a while ago. So, you in a normal case, this is a normal virus, you get this compact uh core structure, and inside it's filled with RNA, RNP, right? So that's that's that's a normal situation. You have this thin section uh electron microscopy, stained electron microscopy uh specimens. But when you mess with integrates, what you get is very often empty core and RNP is outside. And that's quite interesting. So it's suggested that integrates a DNA binding protein has something to do with packaging RNA. So that's really doesn't make much sense. Uh but there's a lot of data suggesting uh there there must be a role. And on a side note, when you think about this core structure in lentiviruses, this is very compact. It forms it only occupies about 17% about uh of the inner viral volume, right? If you have an RNA inside, by chance it would almost always be outside. So there has to be a mechanism to keep it inside, and that's something to do with integrates. The first clue came when this paper was published almost exactly 10 years ago by Korashkele and uh Binach's labs in cell, and they just detected for the first time interaction between integrates and RNA inside of normal virus, inside of native virus. This interaction interaction was not sequence specific, so they cross-linked basically integrates with RNA and then analyzed by clipsec. So it's uh all over the place. We took a long time to reproduce this interaction. We tried lots of technical challenges. Integration is not very soluble in low salt, and so on. There's a lot of a lot of work was uh done by Matthew Singer and Nicola Cook and the love. But through lots of experimentation, we were able, we we we have a system we can detect reliable interaction of HIV integrates or SIV, so it's the H Semian uh version of HIV uh integrates, very similar protein. Uh this is by BLIs actually. We can detect interaction with between integrates and single-stranded RNA and various double-stranded and complex RNA structures. Now, at that point, I had lots of doubts. This is real, because we have a positively charged protein, we have you know polyanions, of course, they will stick, right? But the defining moment, the first defining moment, was when we observed that in the presence of RNA, there was a massive protection of uh large regions of this integrated structures uh for HIV and for this SAV integrates uh from hydrogen deuterium exchange. So this was a key experiment. It showed us that not only there are specific structural changes that must be happening, but also they are spread across the whole protein lens. So they involve an NTD, catalytic or domain, and catalytic citronoldomain, so in both cases. So that was real. It took us again a long time to realize that although the complex forms is not really soluble, so it actually crashes out almost immediately, and we were able to just catch it on crying grids. So when you collect this kind of data, when when you study in these aggregates, very often 90 or 95 percent of your micrographs are empty. You have to be lucky to catch it because all of it is on least five or ten percent of the of the squares of the sorry holes on the on the grid. So this is a nice example, and you see these fibers, and these fibers are nicely averageable. So we were able to refine the structure by single particle uh uh analysis, and it's basically made out of is made out of those two integrated filaments. Uh, the and each of them, each of these filaments are uh basically a chain of octomers, and this these octomers are essentially cross-linked by a short synthetic RNA that we added. So and this RNA happens to be predominantly double-stranded. So it's a double-stranded DNA, and they share because they are short, it's short enough. The two the two uh filaments are more or less stable.
SPEAKER_02So, Peter, is this uh is this uh a true helical filament?
SPEAKER_00There is a there is a there is a there is a it it has a helical um it has a helical property, right? If you especially if you have many uh many units, you can actually see. I think it need it it for a full turn it would require 30 or 40 40 uh repeat units. So there is a bit of a helicity there, but it doesn't really help to refine the structure. Okay, now this kind of filaments of fibers you can observe all in the in the presence of RNA. So the the double-stranded uh RNA, uh such as star, shown here, gives you double uh always gives this double double filaments, single-stranded gives the single uh single filaments, but always this chain of octomers, and without without RNA, you just get this uh pretty much amorphous precipitates. You get maybe some tetramers you can refine, but nothing more than that. All right, so uh we can look at this filament. So this is an example of uh of this uh uh of such filament. Uh this is a repeat unit shown here, and we can look at the structure of this repeat unit. So this is local refinement in in um in in Christpark, was quite beautiful uh in this case. So this repeat unit is made out of two integrated tetramers. Each tetramere is slightly asymmetric, but together they form a fully symmetric uh structure and it's bound together by RNA, like a glue. So it's by sharing these RNA molecules. I'm not going to go too much into details here, but all interactions as expected are not sequence specific, so it's through the backbone. And there's one thing that really I find really striking. So this is a dimer of citronal domain. So terminal domains of integration typically form dimers, and in in this intersection, this DNA bound structures, they they interact with DNA in a specific way. There is some some some even some sequence-specific readout, some major group interaction, so on. But here the same dimer interacts with RNA, A-form RNA duplex, right? And it basically clasps the minor group of the RNA, right, and measures the A-form parameters, right? This is quite amazing. So there is another there is another CTD at the bottom, citronal domain that also interacts with the backbone, and there are some interesting interactions that glue these uh octimers together, right? So overall, overall, the structure explains how integrates binds RNA. That's new, great. Although we knew that integrates binds RNA, even though I personally didn't believe it for years. Uh we know that tetramer is a basic binding unit, great. Octomer is a stabilized by RNA, it's all great. So it's all a lot of it is explained, but still doesn't explain how integrates retains RNA in the cup in the in the capsid core, right? So so here we have to talk about the capsid core. So this is one of the possible reconstructions, they are quite uh heterogeneous uh as particles, they're not like HBV, for example, particle, which is a beautiful T equals 4 symmetry casehedral thing. So they they are quite um, but they always always uh mostly have this kind of conical or bullet shape uh form, uh wider and then and then and the and a narrow end. And bulk of it, uh of the wall of this capsid core is made out of hexamers of capsid protein. So about 250 to 300 copies of this uh of this uh of this uh um of this hexamer builds this uh builds this wall, and to achieve topological closure you need some pentomers, but that's that's a kind of a requirement. So when we consider uh this capsid lattice made out of hexamers, you can define these parameters, the step between hexamers and of course this angle due to six-fold symmetry. Now, if we side by side we look at these vectors, and we this was crazy. When when when I saw it, I it was it got really exciting because we have the structures that are completely different, right? There is no similarity in sequence in in function, right? Uh different lattices they form different macromolecular assemblies, but they're fully congruent. So you can overlay them in at least in 2D, you can see the perfect congruence. So an octomer overlays with a pair of hexamers. This is only for the mature lattice, immature gag lattice doesn't have that. So here we we were uh lucky because we we were collaborating with Paigeun Jan and Joshua Hope uh in Paige Jan Lab, who helped us to uh he donated some some some and in fact in fact froze uh cryam grids for us with with HIV cores that he purified. You can you can overlay HIV particles on the sucrose gradients, ultra-centrifuge them through a thin layer of of detergent in sucrose, and then uh very carefully isolate. But we were quite deliberate at not overpurifying these things, right? So, because we don't want them to be broken, they are quite, you know, they're quite fragile. So the samples are quite dirty. So you can see chunks of DNA, chromatin, and god knows what from the cells. So here we uh invented something, I've not seen it done before. So we trained YOLO. So YOLO is a is a general computer vision software that people use to recognize call cats, dogs, uh, motorcycles on on videos and and in images. We train it to recognize scores on our GPX as they come from uh EPU software. So and that helped a lot because we can exclude a lot of rubbish and then we can pick uh, for example, grid pick within this, only within these boxes, and then we use various approaches, not only grid picking, we also of course use Gautamach, one of my favorite uh pieces of software for picking, of course, uh you know, topas and so on. So at some point we arrived to at about 1 million parts.
SPEAKER_01Peter, Peter, uh just one question. So this yellow, you're doing all this on the crym as the images are coming forward and you're telling it where to collect, or is it no no no?
SPEAKER_00We basically collected as much as we could. We collected, I think, close to 50 uh close to 50,000 okay, um uh micrographs, and then uh about half of them didn't have any cores, and that was the first question. How can we we cannot I I didn't want to do it? I didn't want to do it manually, but then it turned out the YOLO is much more uh much more you versatile than just telling us where the cores are present. You can actually box them, right? I see, and then that that really is amazing. That's just just it's an amazing software, and it's free, right? So um, okay, so basically, uh so at some point we arrived to about one million particles. So this is not the final set, but this is a set when it finally worked. Lots of things were tried. So the problem is this sort of this is a single particle approach, but when you look at the core, right? So every time you look, you see two layers of capsid lattice, but uh bulk of you know uh RNP inside and lots of lots of other junk. So uh, and uh every time you refine a fragment of this lattice as a as a single particle using a single particle approach, the dominant signal is always capsid, right? So you cannot so any kind of classification becomes impossible because it's always dominated by signals dominated by the capsid. So the only thing that really worked with the single particle data set, and that's a beautiful approach, using using signal subtraction. In in reliance, so we subtract it uh basically it's a mask to subtract subtract the signal from the capsid using you know uh refined parameters, and then what's what's left, and you can you can see in this reconstruction all the noise is inside. So this is the luminal side, this is the outer side. The outer side is clean, right? There's not much, but all the RNA and all the components are inside, and you can see this noise. But to classify through this noise, it's impossible other than by signal subtraction. Then, of course, the signal subtraction without realignment and mask needs to be featureless, of course, right? So this is basically um a semi-cylindrical mask. You can create a cylinder in Emon 2, for example, and then you can crop it by moving to the end of the box and so on. So this is how it was done. Uh mask needs to be you know soft and so on. And then it was amazing. So this this when it worked, so you know, by that time I had close to thousand jobs around this data set, and everything was failing. And this was uh this was uh just uh just around uh the the the just just after or just before the new year um 2025. So the so this is uh this is a class, and notice it's only six percent, six and a half percent of particles. It's a bit of an underestimation because this filament is also present here, but shifted by one row of capsid hexamers. But overall it's less than 10 percent if you would classify and clean up the data set. And this is this is the Z slices. So you can see, I mean, I I I can recognize integrates there right away. So this is the first refined structure. So the capsid lattice, uh, and then turn it around, you will see integrates filament made out of octomeres. So there's one thing that we kind of was fine, what was a bit disappointing in the beginning. In this in this um uh reconstructions, we hardly saw any signal for RNA. So RNA was quite uh degraded signal, and it's kind of expected because you know it's uh it's not just a defined sequence, uh defined synthetic uh you know fragment, it's uh it's a it's a heterogeneous uh viral RNA in in there running in different directions, but it was a bit disappointing. So, what we did next was using native cores. So, what I didn't mention, the previous sample was produced by over-expressing integrates because we wanted to get a fighting chance at fighting integrates. When we started, I personally didn't believe it's gonna work. So we overexpressed integrates uh using um approach using VPR fusion and so on, it's quite classical and well known in the field. So, but then we we decided to try to look for these things inside of native cores as they are without without adding extra integrates. So, this uh this subtomogram averaging project uh uh only required about 14 tilt series. So, this is one well-reconstructed tomogram, just as Z slice, you can see beautiful uh lattice of uh capsid hexamers, and this worked beautifully. So, in this case, we just used template matching. We picked uh capsid lattice, fragments of the capsid lattice and our tomograms, and uh in tomography, because the signal is separated in another dimension, you don't need signal subtraction. In fact, in my experience, it never really works in tomography, but in um uh in this case, a simple simple, simple 3D classification in rely line of uh subtomograms allows to isolate uh a beautiful, uh a beautiful class of also about 7% of particles. And it's again on these slices, it looks it looks very, very much like, for example, single in integrated filament we have with a single stranded RNA in vitro. Again, do not dwell on this. You can look at it in the paper. Oh, and if you're lucky, if you look at the at the core which is opportunally oriented, you can see you can actually catch integrate. For example, this particular core we found before we did all the template matching. So we we've and and then of course, uh after template matching averaging and look at the class indeed. So this is integrated filament uh attached to the luminal side of the of the capsid core.
SPEAKER_02Peter, can I can you can you go back?
SPEAKER_00Yep.
SPEAKER_02Um so you didn't map back then, right?
SPEAKER_00Oh yeah, this is mapped back. Yeah, yeah, this is mapped back. So first we first we we we uh it's just an example. There are only 14 uh TL series, so we looked through them very, very carefully, right? Because I was not expecting that Septomogram will actually work with such a small data set, right? So first we thought, uh, we will find a convincing core with the you know, with the filament. And we found a few, right? That looked convincing to us, but convincing a reviewer is a different thing, yeah.
SPEAKER_02Uh so I guess uh so this is yeah, the the seven percent. I just want to make sure that that though can represent pretty much every core, right? Because it's it is a small area of a given. Core is going to have it, right? Exactly. So it could be every core.
SPEAKER_00Yeah. So in this case, we didn't map, we didn't uh find filament in every core. I'll show you a tomorrow uh tomogram with all this stuff mapped back, right? But we have much better nowadays, we have we have much better template matching, less biased as well, and um uh a bit more a bit more rigorous. I think we achieve about maybe 75%. Yeah, but I I need to I need to go back because I I don't do that anymore. It's done by uh by a postdoc who is uh more experienced than me. So all right. Okay, so um let's go back to this one. So this is a reconstruction, so it's relatively low resolution, so it's done with a very wide generous mask. So in fact, we can push it to almost 10 angstrom resolution, but it's not really necessary. So uh resolution for the core, of course, is is is very high. Uh for the integrase layer shown here, it's it's kind of cyan, it's about 12 angstrom. So again, we can see uh so the two sides again, you're looking at the core, uh at the capsid capsid lattice, integration shown here in green, and the red is what is not explained by integrates of capsides. So we interpret it as probably RNP. Now, if you slice like the top of this ketchup, what you will see is that this ketchup penetrates the integrated filament exactly, almost exactly where synthetic RNA was found in our in our in vitreo structure, which was for us was a smoking gun, right? So basically, what happens uh is um we have uh uh a layer of capsid, which is a two-dimensional lattice, we have uh this single filament forming, which is which is like essentially one uh one-dimensional lattice, right? Uh, and it captures the we think it captures stem loops uh of the viral RNA, and it's like it is saying uh if if a bird is caught by a single claw, the whole bird is clawed. Yeah, so you just you basically what what you need is capture a few stem loops, and you hold you hold this whole thing, the whole blob of RNA, which is in complex with a nuclear capsid and very, very compact inside, right? So, this is what I was referring to. So, this these are these are the filaments which we mapped using that uh relatively inferior approach. Now we have much better, uh, much better results uh with uh subtomogram averaging. And what's interesting is so first of all, it's always very, very straight, although there is a little bit of a helical nature to it, minor, but it has to be extremely straight. So in every filament we observed in C2 in vivo, yeah, in real particles, or in vitro, they're always extremely straight. So that's kind of the interesting property of these things. So they always run along one side from top to bottom of the of the nicely formed core, from wide to the narrow end. And uh typically it's 12 integrates octomers, which uh pretty much accounts for the whole integrase component. So the the way integrates interacts with the capsid lattice is through uh there are several points of contact, and it's all mediated by this major homology region in capsid. It's quite important. Now, of course, the structures are quite low resolution, uh they're uh overall about 4.6 angstrom. We don't see the side chains. So we were fortunate again to collaborate with Juan Birilla's lab and Juan Ray, who did a lot of molecular dynamics, and they showed that these contact areas indeed can form reasonable interactions. So that was quite quite important. And we worked a lot with Alan, so Alan Engelman and uh and and Jen Lee uh so done a huge amount of virology to validate some of these ideas and observations. I'm not gonna go through all the mutants, but you can mutate integrates within these regions. Uh, not only this, this is just one of the one of the many mutations they made, and you can induce uh formation of these eccentric variants shown here in orange. This is basically counted one by one by hand. It's a lot of work. There is another thing which I like a lot. So this this there is a protein called LEGEF, right? So it's a binds integrates, it's an important host factor. But what we noticed this when Lajev is in complex with integrates, it would clash with the capsite. So it would prevent formation of this interface between capsite and integrates. So what Jean Lee was able to do, uh basically they overexpress uh this integrates binding domain uh from from this uh host factor, and they showed that in the presence of it, when it's present in the viral viral cores or viral particles, you get from you get again this eccentric virions, right? So it can strip integrates from the core. And that's wild-type integrase, and Legev doesn't affect detrimentization of integrates and so on. So it's it's a fully intact protein, but you can induce this defect using this uh approach. And of course, you can mutate also in a clever way, you can mutate MHR and also induce formation of this upgraded morphology. Now, the last thing I'm gonna show you, and that's what I found really, really striking at the end of all that story, is when I compared this filament. So, this is reconstruction of this filament where I placed approximately the RNA as it is observed in O, because obviously in situ we don't see RNA very well, but you know, we know approximately where it is. So, this is just for illustration purposes, and this is just two octomeric repeats. There is other repeat here, and then and so on, and runs from the top to bottom of the screen. So this is in situ inside of HIV virium. This structure on the right is the integrated DNA complex from my DVSN virus that we assembled with recombinant integrates from my DVSN virus and synthetic DNA. And you can see already how similar these things are. This one is more compact, right? On the right is more compact, but but the orientations of these tetramers, tetramers one, two, three, and four, one, two, and three and four, they are very, very similar. You can morph them together, and you will see that conformational changes required to form this endosome, the DNA complex starting from this RNA, are actually quite minor. So it suggests that capsid possibly, likely, actually, um templates assembly of the next integrated DNA complex.
SPEAKER_02But that domains, I mean, that is is that a big domain swap? I mean, that's like a big yeah.
SPEAKER_00So yeah, that's a very well known. That's a very well-known feature. Yeah. Oh, okay.
SPEAKER_02If I was an HIV guy, that would be like no big deal. Okay, I get it.
SPEAKER_00That's that's uh that's it's it's this kind of this uh intrance interaction using this uh kind of this hugging interaction with the internal domains. Yeah, that's that's very well described, yeah. All right. So now now we can make many conclusions, but basically what we have is each each capsule core will have um integrated single and probably uh most often will have a single integrated filament running from the top to the bottom, uh on a straight run of uh hexamer pairs, right? And it spans the whole length of the core, and it requires about 12 to 13 uh integrates octomers. That's about the full component of integrates. So it's a very special position uh on the core, and it's it has a it has a topological sense actually. Now uh uh uh lots of questions. So lots of questions you can ask uh uh uh what to do next, but uh of course we are very interested to know what happens during reverse transcription, what what role this thing plays, uh, what role it plays in assembly of the mature core as well, because there are some interesting ideas there. And also drug development can be can we develop uh integrates inhibitors that actually specifically block this interaction uh with MHR loops and so on. I mean, there are lots of things. All right, so these are my conclusions. Of course, lots of people were involved. Matthew just defended his PhD thesis, he's the first author in this paper. Um uh Florian helped us with some uh Cryem, and he's taking over, he's took over this project and massively improved all the pipelines that we used originally. Emma Emma helped uh a lot with with Cryem and Nicola with biochemistry. Um uh Joshua and Pejun, of course, they provided uh amazing samples of the cores and taught us how to do this, how to pre-purify them. Uh Alan and Jen uh done this amazing virology, huge amount of work. Uh, of course, Juan and Juan, Juan Piril and Juan Ray uh did uh very important uh molecular dynamic simulations, and we had huge help from uh uh Julia Zanetti, who helped us with correctron tomography, with anything to do with warp and uh transition between warp and rely on and so on. And of course, we are quite fortunate to have access to these amazing facilities at the institute, like HDX, for example, was done by Sara and Mark, and this was a key experiment. Without it, I would have not believed that this is real. And of course, um uh Laura and Andrea helped a lot with biochemistry and uh yeah, and uh crayon data collection and processing, and so Andrea actually first identified this uh filaments and raw tomograms. Yeah, great.
SPEAKER_02Um so before we go to the interview part, I have a few a few uh scientific questions. So one thing that that just so you you talked about how there's this you know the repeat and the angle matches between the capsid and the the integrated filament, yes? Absolutely. So maybe you said this, but you know, uh so the the there's this eccentric shape to the variance, right? And uh to the capsids, you know, is that the right word? Capsides, yeah. Um the and there's like one side that's like longer and then it sort of angles over, and and then like there's sort of a shorter side, right? So what I'm wondering is, and you know that that's also dictated by the the the pentagon, but by the pentagons, right? So what I'm wondering is if the filament is always on one side, because that's gonna be the one where the the the hexagons match up right.
SPEAKER_00Yeah, yeah, yeah, yeah, yeah. I I I'm so lucky. I have a nice slide, just one second because I didn't have time to go into this. Now, there are not that many uh cores which are fully reconstructed. We have actually quite a big collection that we are refining now, but there are only a few very, very there's just a handful of HIV cores, and they are quite heterogeneous, but they generally have this shape. Now, this is one example which was um uh misspell the name of the of the of the of the journal. Anyway, so you look at this core. Notice uh so this filament, remember it has to be straight because actually you get some you get clushes if you cannot band it like this, you can twist it a little bit, but it has to be straight as an arrow. But it has to run along the the pairs of the hexamers. So these pairs of the hexamers need to form like a railway track, which is very, very straight. Now consider this one. This side, if you look at this core, you can see that every way you can you can find a run of double hexamers right here or here, it's always going to be curved, right? Imagine how it is looks like from the inside, right? So it's gonna be concave, right? Uh if I'm right, I'm not sure concave, I think. Yeah, but if you look at the same the same core turned around 180 degrees, and you will find one run of double hexamers, which is very, very straight, runs from the bottom to the top, right? And that's a property very common. This is really cool, and that's basically stems from the from this conicity, right? Because it has to be conical. Now, that structure where you have this straight run of uh double hexamers actually was uh was discovered independently by Juan Pirilla and uh his mathematicians' collaborators who just did some some crazy calculations and crazy, like I don't know, topography stuff. And they found this and they call it a seam, right? Now that seam must have an important role because that seam is where integrates binds and holds RNA. So ergo, you have to assume that this forms in the beginning, right? So this is where the capsule should start of some kind of comes directly from this, right? So yeah, yeah. So but there are if you look at real sample uh of HV cores, you will find different things. You will find majority healthy looking things, they will look like this, right? They're not all they're not the same, they are slightly different, they're slightly different size and so on, shapes, but they're roughly conical, they will have the same. But maybe 30% of them will have different shapes, they will be more bulky. Some of them are actually we also have examples of perfectly cylindrical ones with just like nice round cups, very unusual, but all you want is really odd shapes as well. And in those odd shapes, you will find more than one uh filament, or sometimes you know shorter filaments, because you need this straight run, right? And a straight run of double hexamers, it's doesn't you know in a if in an odd shape uh um uh structure, it will it it will require maybe uh splitting this uh into several uh into several such filaments.
unknownYeah.
SPEAKER_02That's so cool. So so I think you said that you think that it's capsid that's templating uh the the integrates filament.
SPEAKER_00Did you say that? I no, I didn't. I we've been discussing this with Alan and Juan for a long time now, and we don't have. I mean, the thing is it would be really nice to have this system where you could study viral maturation in a in a timed way, right? So if you could you could block it like and then and then allow it you know for several minutes and then freeze, but that's it's not possible as far as I know. Um so yeah, so the the whole transition from this immature lot is that essentially uh spherical uh spherical assembly on the inner side of the membrane is a big hole at the bottom, right? And that's after protease cleavage, how this actually transforms into this very compact structure. Uh it's it's a subject of very active research right now. So there are there are you know there are models where there is like a morphine of the whole thing, there is a full disassembly and reassembly, and reality is probably somewhere in between, but yeah, I don't and what templates what I don't know. But usually when the things fit together, they can template each other, right? So in nature, it might not matter.
SPEAKER_02Yeah, I mean it like the speculator in me would be uh you know really is attracted to the idea that it the the that the the integrase filament is templating the capsid because that would explain so many things about like why it's eccentric to begin with and how the pentagonal ones get placed. You know, that's something that always it bothers me, right? Because it's so irregular. So something has to make it, but it's not it's so irregular, but it's not wildly irregular, like everything is different, right? So it's like there they're uh there's this uh heterogeneity, but it's limited by something. So I really like that idea.
SPEAKER_00But but the way, you know, the way things usually, if if if you can, if you if you, for example, if uh capsid lattice is helped by integrates filament, right? The capsid lattice still can form, right? Anyway, right? So filament will help. It can be it can go both ways. But for example, again, this is something that Alan reminds me uh all the time, uh, that actually when we look at these eccentric virions, when you actually look in the details when you count them, uh there is uh when you mess with integrates, okay, you produce eccentric virions, but there is a very high proportion of malformed virions with of this empty eccentric virion. So there is a lot of nuance there, and uh yeah, so and you could also imagine that in some viruses one aspect is more important than the other, right? So we never looked at my division virus uh particles, for example, how this happens there, and and you know, in actually we didn't look in uh in SIV, we studied SIV integrates in vitro, right? But we never really studied SIV particles uh in cores. There could be some some uh some differences, what's for fruits first, what's more important, and and so on.
SPEAKER_02That is fantastic. So uh that that was really great. Um uh captivated the entire time. So I think uh given the time, maybe now it's uh it's the right time to switch the interview part of the podcast. So we have a few questions for you about like how the paper game came together and more broadly your journey as a scientist and like what inspires you and stuff. So um Francis will start and then he we'll take turns.
SPEAKER_01Peter, so you know I think you probably mentioned that the HDX experiments were one of the breakthrough moments, but I know also that you've been thinking about this for quite some time. What inspired you to go after uh this you know, this possibility of filaments being under the capsid and so forth?
SPEAKER_00Well, that was really noticing this coincidence, right, in geometry. Um that was that was that was it. But it took me a I mean, it it took me a very long time to digest this idea of integrates binding RNA, and that's something real. I mean, I saw this paper in you know 2016, of course, and I'm and I uh I we had this several conversations with Mamuka about this every time we meet at the meeting, and I told him, Look, I know there is so much evidence, indirect evidence, that there is something to do with integrates and RNA, but I can't for me it's so difficult to to to get used to this idea, right? That this is kind of um RNA DNA binding protein, and he would always like inspire me, yeah, it's real. Nonamolar affinity. I was like, I go back, we speak with uh we had this conversation with uh with Matthew like a million times. Is it really real? Is it really real? And so, I mean, so many, you know, it's it's uh just the technical, the technical issues because you know you want to you want to study how integrates binds RNA, but integrates is barely soluble under conditions where it can bind RNA in vitro, right? So because you have to dilute the 250 millimolar salt, and there is a reason why we studied SIV integrates in vitro because SIV is just a little bit more soluble.
SPEAKER_02Yeah. So you had this um this whole laundry list of things that you want to do next. So you you're gonna have to choose like what's the what's the what's the burning question for you next?
SPEAKER_00So uh we are really, really keen on uh looking at the progress of reverse transcription and what happens to this filament because it really probably somehow involved in in in templating reverse transcription and must it must somehow change because when you think about during reverse transcription, RNA that's bound to it is uh to this kind of uh RNA that is kind of in a stitch pattern, uh you know, connected to it, right? So it will be it will be pulled out and destroyed. What happens to this filament and how it is transitions to the uh to the um uh to the DNA form? I mean, uh West Anquisla published this beautiful paper in science a number of years ago where they have this full reverse transcription uh in in vitro inside of this semi-purified cores. Uh and obviously we are this is something I'm sure they are doing the same thing. So we would like to see what's going on, uh what's going on there. Uh but we also have a lot of a lot of stuff to clean up. We have a follow-up uh structures which are better, you know, high quality structures both by SPA and by STA. So we'll show them in Cold Spring Harbor uh next month already, actually. Yeah, so a lot of a lot of a lot of footwork just uh just to uh just to basically square up what we know and then we would we would like to move move on. We also we we we there are some some aspects we would like to assemble these things in vitro as well, but it's not so easy because of the curvature when you assemble capsid, uh it likes to form the tubes, but the tubes always have a very strong helical parameter. So it's very it's very seldom that you have a tube where you have this uh this uh hexamers run alongside. It's always so it's on and of course the problem is the tubes, right? So the tube is is is narrow and integrates binds to the inner side. So we do we had a lot of ideas how to break these uh tubes into like uh you know flakes, and so it's it's all it's all it's all ongoing. And and and the role, of course, of capsid lattice and template in uh the integrated DNA complex, that's another thing that we uh we would really like to uh to confirm it has to be right.
SPEAKER_02Yeah, yeah. I I mean um uh if I was in your lab, I would be doing the exact same thing. I'd be like jumping at the bit. All right. So taking a step back. So as a scientist, there's moments that come around, and you know, you're banging your head against the wall for a long time, and then suddenly there's a piece of data. And maybe it's an image, maybe it's a graph, it's something, and it's like, ah, that's it. So is there, do you have something where it's supposed to sort of burning your mind as like uh, you know, something the moment you realize you've discovered discovered something important?
SPEAKER_00Yeah, a few, I I could mention a few stories, but perhaps the most painful and exciting was when I was uh my first postdoc still in uh uh Zeger's lab in in Belgium. So I had this project I was working on for a long time, and the idea was to extract integrates from human cells. We express integrates and human cells extract, and you know, it's a tetramere, and that was my story. Okay, I integrate tetramere forms and cells, that was an important thing at that time, and I was ready to write my paper. I was already writing the paper, and I just wanted to do one more experiment to really like push it to to have even more cleaner gels and more cross-linking and everything. And then this evening I'm so tired. I developed my gel, Kumasi staining, and my integrates is always pure, but this time there was this fat band, fat band on the top, and I knew it's just not so because I pushed it, right? I used a lot more cells, I did at a high concentration. This thing did not dissociate. That was Leg.
SPEAKER_01That was the that's the discovery of Legend.
SPEAKER_00And I remember myself and God, we went to the we were shopping to this to grocery, and I'm I am like, I'm dying. Oh my god, my story is like thrashed, right? It's not it's not complete, it's not never complete. It's super exciting. I don't I don't know what that protein is, right? Because it doesn't match anything that uh you know we could think of, right? And and certainly not a protein called Lance epithelium derived. That's right, and that came also like a stab in the back. Why? And in all these first years of publishing, we had to uh uh you know spell it out and explain that is not a gross factor, it has nothing to do with lance epithelium, but attempt and that's the uh inspiration for the allosteric integration. That was the burn on my memory, right? It's burned in my memory because it was so painful, it was so painful. I mean, I knew it's right and I knew it's important, but I mean, I god, I wanted to have this published, this paper.
SPEAKER_02But that but that burn that that birthed the whole direction for your research, right? Absolutely, it was incredibly painful, but it was like that was the start, right? All right, that's cool.
SPEAKER_01Okay, so going a little bit further. So, how did you when did you realize and how did you realize that you wanted to do science? Science was your thing to do. And then and more and more more to more so structural biology. How did you come to structural biology?
SPEAKER_00Yeah, so for me, science. When I was a small kid, my my grandfather was a scientist. I was so proud. I was telling everyone. Uh, but of course, uh, my mom will always say, but your dad is also a scientist, but uh grandfather. Yeah, he studied bugs, beatles, yeah. And then, but then my first years in school were terrible. I I mean, I remember I told my my cousin that I'm I'm gonna go to university, he was just laughing at me, and uh somehow I picked up uh in the second uh in the second part in uh when I was in uh all the classes. I had good teachers and so on, yeah. Yeah, I was terrible.
SPEAKER_02You're not the only one. That was that was me too. Um so uh last question. Uh, do you have like a scientific hero? Like so you have a and who would it be and why?
SPEAKER_00Yeah, um, yeah, I knew this question in advance. I was worried, but then it came to me. I mean, it would be Paul Eredish, actually. Paul Erdish, uh, I don't know if you know, it's a brilliant um Hungarian mathematician who was the oddest ball of all odd balls you could possibly meet in your life. So from morning to night, the only thing he was doing was he was couch surfing from collaborator to collaborator. He never had an address, right? He never learned to tie his shoelaces. He was he was he published more than 1500 papers, collaborated with more than 500 mathematicians, which is was very unusual then. It's still unusual, I think, now. But back then, I mean most mathematicians publish a single-author paper, right? The guy published like 500 with 500 other people, so yeah. So despite this kind of deficiencies, you know, in his kind of social and you know everyday life, he proved some amazing theorems which I'm not smart enough to understand. I mean, I'm just admiring him, not for what he did, because I I can't fully grasp it, but the character, but the character and also the impact on the field, right? And uh I guess uh I mean that's yeah, also the uh what's kind of funny and uh interesting, his epitaph on his grave says uh literally, I finally stopped getting dumber. I mean, I there is a book, there is a book I read many years ago about him. Called it's called the man who loved only numbers. So I really recommend it. It's about a man who was like uh like um um who who only had his numbers, nothing else.
SPEAKER_01Uh wow, Peter, thank you so much. That is uh we note down the book, The Man Who Loved His Numbers, right? And uh thank you so much for joining us today. Uh we really appreciate you taking your time to tell us about your science. That was fantastic work and big fan of it.
SPEAKER_00Okay, thank you.
SPEAKER_02And that does it for this episode. We hope you all join us again for the next episode of Structure Club.