Scientist Stories: Stephen Harrison, Structural Biology of Viruses

Our keynote speaker, Steve Harrison. I'm pretty sure Steve is, you know, Steve is no stranger to everyone here. I'd like to introduce a little bit, you know, from a little bit from my own personal side. Actually, on the formal side, Steve received his bachelor's degree from Harvard College in 1963 in chemistry and physics. In physics, and then PhD from Harvard in '67. And then he became a Harvard faculty member, assistant professor, 1971, then rose through the ranks, and Steve has been at Harvard ever since. Steve made a long list of contributions to structural biology. Actually, I shouldn't say structural biology, to biology, to the understanding of processes in cells and in physiology. His major contributions lie in 3 areas. One is viruses and viral proteins. I'm sure people here may still remember in the late '60s, Steve began to work on tomato bushy stalk viruses. And I think he improved resolution from maybe the snowman model to 25 all the way to atomic resolution all the way. And then he actually began to make contributions in protein-nucleic acid interaction area and also in cellular signaling. So the three major areas I think, you know, he probably in any single area would be quite prominent. So Steve is also recognized with a long list of prizes and recognitions. I would not like to go into them because that's going to take 5 minutes of my time. So when I, because there are many students here in the back, when I was your age, actually I began to hear the name Steve Harrison. And one day, I think in the early '90s when I was still a student at Johns Hopkins Medical School, and my mentor Jerry Berg got quite excited and he told everyone, you've got to go to a talk. In the cell biology and anatomy department downstairs. We asked who's giving the talk, and Jerry mentioned Steve Harrison's name. And that's the first time I actually went to hear a talk by Steve Harrison. That was quite motivational for a grad student. So I don't want to go for so long, and so you can imagine how things have come around, come about. Steve has been at the forefront of structural biology for Nearly 5 decades. That's just remarkable, just to say that. And I will not go on anymore. I would like to turn the podium to Steve Harrison. Let's give a round of warm applause to Steve. Thanks.

So thank you very, very much. It's both a pleasure and an honor to be here. I see already I, in a sleepy moment, left the R out of structural biology. But at any rate, I do hope that what I can illustrate by talking about work on virus atomic structure and the relationship for today's talk to virus interactions with membranes will illustrate the fundamental topic in the title of the symposium, namely the bridge from atomic structures to cells and, as I say here, single molecule movies. If you look at the sort of time and distance scales of some of the things at least we'll be talking about today,— they plot roughly like this. These are from memory, and so there will be huge inaccuracies here. But, at any rate, it illustrates that at least in the time dimension, or in the distance over time dimension, we're really looking at diffusional processes. And so, the logarithm of the time scale is proportional to the logarithm of the distance scale, with a constant constant of proportionality about 3. Once we get down to very small time scales, probably we're talking about non-diffusional processes. But I think much of what fascinates us as we look towards cells from atoms are events and molecular machines in the time scale that I'm illustrating here. As computational structural biology has developed over the last several decades, it's been largely confined to an area outside of this diffusional regime. But finally, and I'm not quite sure when it will really arrive. But finally, with very long timescale simulations from D.E. Shaw and colleagues, it's beginning to encroach effectively on the timescale here. But I think over the course of the next decade, at least, we'll be relying largely still on pure observation. At any rate, I'm sticking to events, sizes, and times up in this range, and we'll be in particular talking about viruses, viral membrane interactions, and viral remodeling of membranes in the process of virus entry. I'll remind you of the distinction between enveloped and non-enveloped viruses. Those are old-fashioned words, but they really mean Virus that have lipid membranes like influenza virus there on the left, and viruses that don't have lipid membranes like rotavirus on the right. Viruses, enveloped viruses, viruses with membranes enter cells by fusing their membranes with those of a cellular compartment or the surface of the cell for that matter.. And we know rather more actually about that entry step than we do about the corresponding step for viruses that don't have a membrane of their own and therefore have nothing to fuse with and need to perforate in some sense a cellular membrane. But at least in one case, that of the rotaviruses, I'll try to illustrate that we're closing in at least on some ideas of what's going on. To begin with fusion, I'll remind you also that bilayer fusion, or fusion of two membrane bilayers, is thought to proceed through what's illustrated here as a hemifusion stalk, that is as an intermediate in which the opposed membranes have fused, or leaflets have fused, or merged, but not the distal leaflets of the two membranes. And that that stalk then resolves into a pore that joins the interior of the compartment bounded by one membrane with the interior of a compartment bounded by another. There's quite a lot of evidence now from a variety of directions. This is somewhat older evidence from Huang at Rice. For this hemifusion stalk intermediate. And also now good evidence that this sort of spreading into a so-called hemifusion diaphragm is in most cases a nonproductive intermediate. And so we'll be looking, if you wish, at how viral proteins catalyze this step because there's a high kinetic barrier to what is otherwise a thermodynamically favorable step. Condensation reaction, so to speak, what's known as a hydration force keeps membranes apart by 10 to 15 angstroms. And there's a considerable barrier to that last step of approach. And in the case of viruses, or as you'll hear from Axel Brunger perhaps a little later, Cellular fusion processes, fusion proteins so-called, are effectively the catalysts for this reaction. The viral fusion proteins are dead-end catalysts. They don't self-regenerate, but actually the cellular fusion proteins do recycle and so are in some sense true catalysts. So, part 1 of this talk will be how do the catalysts for viral membrane fusion work. And I'll try quickly to talk about work on influenza virus and a bit on West Nile and dengue viruses using single particle fluorescence imaging to connect structure with molecular mechanism. I probably needn't introduce influenza hemagglutinin, the protein that influenza fusion protein to this audience. But I'll remind you at any rate that the influenza virus particle is studded with these spikes of two different species, the hemagglutinin and the neuraminidase. The hemagglutinin is the fusion protein I'll be talking about. And from work of my late colleague Don Wiley and his long-term collaborator John Scale, Ian Wilson in the early years, We know a huge amount about this protein. It is indeed the fusion protein about which we know most structurally, and I hope I'll try to convince you mechanistically. It's made as a precursor known as HA0, and a cleavage by a furin-like, although not furin, probably protease in a loop at this point, leads to an activated protein, two chains called HA1 and HA2. The cleavage simply splits a peptide bond or excises in some cases one or two amino acid residues. But as I'll emphasize, there's an important rearrangement here, very local, not at all global, a local rearrangement in which the new N-terminus, called the fusion peptide, the new N-terminus of HA2, called the fusion peptide for reasons you'll see, sticks into a little pocket on the axis of the protein and is, because it's quite hydrophobic, is nicely hidden and does not promote, even when the protein is solubilized, aggregation. But as Scahill showed, and his coworkers, in 1982, if the pH is lowered, there's a dramatic rearrangement of HA2, indeed of the whole molecule. HA1 dissociates, though it doesn't actually fully dissociate because of a disulfide bond here, but fundamentally gets out of the way. And HA2, the purple part, basically turns itself inside out. The most important thing to note about the rearranged structure is that the fusion peptide is now adjacent to the transmembrane segment. This is where the protein would be anchored in the viral membrane. And as you'll see in a minute, there's the fusion peptide which interacts with the target membrane, by approaching the transmembrane segment in this rearrangement, therefore necessarily drives the two membranes together. As a result, one has drawn, as a result of knowing those structures, one has drawn a scheme that looks like this. And what I'm going to tell you about are experiments that try to turn the gray part into at least some color by probing this catalyst by the methods of traditional mechanistic enzymology, if you wish, but on a single particle scale, namely trying to understand how the kinetics of this reaction vary with mutation and structural change in the particle. So, There's been lots of indirect evidence. I'll show you some slightly more direct evidence that there is an extended intermediate in which the fusion peptide is inserted into the target membrane and the protein then bridges the two membranes and a collapse then of that extended intermediate into the final structure. I'll emphasize that that collapse is asymmetric. There's sometimes misunderstanding on that part. That is, you've got to have a situation where the zipping up of the segment joining the central helix to the transmembrane segment is asymmetric with the threefold axis of this coiled coil. Otherwise, you'd actually run into a topological conundrum. But in any case, the assumption from knowing the structure of the precursor state and the structure of the final state has been that one passes through states that look roughly like this. And so in trying to study that then, we've tried to ask what are the characteristics of the extended intermediate and how many trimers are needed for fusion., if you wish, beginning with atomic structures and trying to move toward something rather on the scale of what's going on in the cell. The low pH is, of course, the location at which the virus indeed fuses and therefore infectively enters the cell, namely the low pH environment of an endosome. The experimental configuration that Dan Floyd, when he was a joint student with Antoine van Ooyen, designed was essentially a total internal reflection microscopy setup in which a supported lipid bilayer in the early experiment sort of floating on a dextran polymer, later when we were looking at largely interested in the hemifusion step. We've put it directly on glass. It's harder to follow the fusion step when, or the, yes, the opening of a pore between the virus and the substrate when the bilayer is too close to the glass. But at any rate, if the virus particle is bound to a, to a supported lipid bilayer containing sialic acid head group lipids at low concentration but appropriate concentration. Then the hemagglutinin, which recognizes sialic acid as on glycolipids and glycoproteins as the viral receptor, binds, and then one can flow in low pH and trigger the reaction. There's a pH sensor in the membrane in order to record T 0, so to speak. And as a function of time, then one looks at hemifusion, which is monitored by dequenching of a hydrophobic dye in the viral membrane that dequenches and then diffuses away. So there's this spike. And then pore formation, as I said, in some experiments by loss of a dye incorporated into the interior of the virus, which diffuses away the instant that fusion pore opens. The pH sensor basically gives one the time at which the low pH solution arrived in the observation cell.. And one can then record movies, this is sped up, that look like this in which, this is just for hemifusion, in which the de-quenching and diffusion away of the dye, of the hydrophobic dye, into the surrounding bilayers occur. If we look then at these kinetic measurements, and as I said, think of this as enzyme kinetics if you wish, what we're catalyzing is the fusion of the viral membrane with the support. Then because we're doing this on a single particle basis, we can distinguish between the time to hemifusion and then the time from hemifusion to fusion, and so look at those two steps distinctly. One couldn't do that if you were looking at it in bulk. And one sees a rise and decay for the hemifusion delay time, but essentially an exponential for the interval between hemifusion and fusion. As most of you will recognize, that means that this is a multi-step process, or at least, and that more than one of those steps is rate limiting. There is one rate limiting step for this process. I won't talk about the hemifusion to fusion step, that is the fusion fusion pore opening. I'm going to concentrate on what goes on in the steps leading here. If you treat these sort of problems, you find that these rise and decay kinetics can be, will indicate either some number of sequential steps or parallel ones, and they lead in suitable circumstances to essentially the same sort of distribution in which a rate constant and the number of steps can be used to fit the data. And Dan Floyd found already for flu that the number of steps was about 3, and for various reasons that I haven't time to go onto, we argued, and indeed continue to argue, that that meant that 3 hemagglutinins were involved in forming the pore. But to go on and actually try to probe this in more detail, there are two further observations that helped enormously. One was, and they were made by Tijana Ivanović, one is that as the low pH dye is, or low pH solution is flowing in, Many of the particles can be seen here, here, here to adhere to the membrane but to move in the hydrodynamic flow but then arrest and only much later fuse. The others have already attached by the time the observations got started. And it turns out that this arrest and fusion and the hemifusion have extremely similar properties in terms of their dependence on pH and on the nature of the kinetic phenomenon, and as you'll see in a moment, in their sensitivity to mutation. And so that is a further observation that needs to be fit in thinking about what's going on. That's obviously a experimental property, not something going on in an endosome, but as you'll see in a minute, very useful. The other even more useful observation was quite by accident that there's a laboratory strain that differs in only a very few positions from the standard strain called X31 that we were using that fuses much more rapidly. And it turns out it has a key difference at a position that is otherwise completely conserved through all of the influenza viruses. Isolates that have ever been sequenced. Udorn is a laboratory strain at the fourth position in HA2. And if we looked at that closely, we saw that in fact this glycine is at a key position in the way in which the fusion peptide tucks into the threefold axis of the hemagglutinin, as I pointed out. That gave us, if you wish, the clue that at least the step we were looking at, the rate-limiting step we were looking at, was the expulsion of the fusion peptide from this pocket allowed us to do a series of further mutations, or allowed Tiana to do that, confirming that, also showing that at very low pH, finally that is no longer rate limiting and another step intervenes and gave us the kind of data that we could go on to simulate with the following sort of model. We assumed that at the contact zone between the virus and the target membrane, there's simply a stochastic firing of hemagglutinin trimers. That contact zone, you saw how close-packed the hemagglutinin is from that section of a cryotomogram from Peter Rosenthal in an early slide, that contact zone will contain, we estimated from the spacing in those images, about 100 hemagglutinins.. There are about 300 to 400, perhaps a bit more because of the elongation of many of the flu particles, in particular of Udorn. But that's a sensible estimate. We then assumed that this extended intermediate would engage the target membrane and anchor the particle, prevent it from flowing in the hydrodynamic field in the experiment, but that rapid cooperative progression to hemifusion will occur only when a critical number of adjacent trimers have engaged the target membrane. And if we illustrate that here, there's a contact zone between the virus and the membrane that hemagglutinins, once the pH has lowered, are in equilibrium with protons binding and unbinding very rapidly, of course, but that with some probability a given hemagglutinin will overcome that energy barrier to pulling the three fusion peptides, and they would come out cooperatively because they actually contact each other from the axis of the trimer, extend and engage the target membrane by insertion of the fusion peptide. 2 or 3 or 4 of these have engaged, the particle will resist any further hydrodynamics. But we need 3 adjacent ones, we assumed, or n adjacent ones. You'll see that the number does come out to about 3 in order to initiate the hemifusion process, which can then occur quite rapidly. And with that as the model, one simulated it with a suitable relatively simple simulation procedure and found indeed that one can get an excellent fit to the data under the assumption that about 3 to 4 are needed to arrest and roughly 3 are needed for hemifusion. If we look at then the steps needed. One can think, I won't go into it today for reasons of time, one can think about the steps required and start looking at where else to place mutations that might give us an even more detailed map of what's going on. And the sequence of these events indeed might vary from strain to strain or even from hemagglutinin to hemagglutinin within a particle but with the key rate-limiting step for the pH that is relevant being the withdrawal of the fusion peptide here. So, does any of this make sense energetically? Yes, it does. We know what the free energy of insertion of the fusion peptide into the target membrane is. Therefore, 3 fusion peptides would give us of the order of 24 kilocalories per mole for the attachment. The foldback had better not be higher than that or you'll just yank that one hemagglutinin back out of the target membrane. And to overcome the kinetic barrier, which is estimated as somewhere between 50 and 100 kilocalories per mole,, then you would need about 3 hemagglutinins. And so at least in an energetic sense, the kinetics and the structural characteristics they suggest are consistent with other things we know about what's going on. The most interesting conclusion from a virological point of view, this isn't a virology seminar, so I'm not going to go into going to avoid implications for inhibition and other interesting things that you can conclude from that, is that that extended intermediate is, until the third hemagglutinin joins in at least, relatively long-lived. And it's long-lived because a single hemagglutinin bridging the two membranes simply can't collapse. The elasticity, if you wish, the resistance of the membrane to that distortion, that high kinetic barrier keeps it from collapsing. And when a second and then finally a third partner have joined right in the same area, then they can collapse to form around a nascent fusion pore. I like to call that a tug-of-war model. If you've got two sides equally balanced tugging at a rope, If an extra party joins one of the sides, it will win instantly and the rope will give way, or the other side will give way along with the rope. And, in effect, that's how we're thinking of this particular process. And an important implication, and you'll see that it's an important implication, I think, for other fusion processes, is that you don't have to have some coordinated collar around the hemifusion stalk and the nascent pore. Cooperativity among the 3, probably once in a while 4, hemagglutinins surrounding that nascent pore, perhaps even in some stochastic situation just a couple of them, is enough to give the whole process considerable cooperativity. We've recently gone on to see how general this sort of picture might be. By looking at West Nile and Dengue viruses, which belong to a family known as Flaviviruses. Very, very different both in the structure of the virus, which has 180 subunits clustered in dimers tiled on the surface here as shown by comparing a structure we did some years ago of the, of the free dimer of one of these viruses with an electron cryomicroscopy of the virus at somewhat lower resolution done in Michael Rossman's lab, and then just a couple of years ago, some very beautiful work from Hong Zhou at essentially atomic, or at least what gets called near atomic for whatever reasons, resolution showing the intimate details of this But you see that this fusion protein is architecturally totally different. Indeed, the so-called fusion loop is not the product of a cleavage. There's another chaperone protein that's cleaved, but rather a loop at the tip of this finger-like yellow protrusion from one of the three domains of the folded subunit. And it's linked to a transmembrane hairpin, linked by a segment that got called some time ago the stem with a layout along the polypeptide chain here. The protein at low pH rearranges to a trimer and with all three fusion loops at the same end. One assumes that they insert into the target membrane, and indeed there's good evidence for that, and that this stem then zips up, giving one the same ultimate catalytic process, so to speak, that you saw with flu hemagglutinin. And so one can indeed in this case draw a scheme with intermediates that we haven't seen between the two structures that we know about. The starting point and the finishing point and use the same kind of approach. All this is still work effectively in progress. We're writing a first manuscript on it that Luke Chao has been carrying out using virus-like particles from West Nile virus, which are made simply by expressing the membrane protein and its chaperone in As recombinant products in cells, these virus particles then bud out and make, or these particles make empty virus-like particles. The wonderful thing is you can do this by transient transfection, and so making mutations in the protein is easy, whereas those of you who know something about influenza virus will know that making those mutant flus was a very big deal indeed. You've got a 12-plasmid system that you've got to use to do reverse genetics on flu, flu. And so this has allowed us to make much more rapid progress with another interesting system. I won't bore you with any of the details except to say that a similar simulation gives us an estimate of probably 2 trimers required for fusion. And again, that cooperativity is through the membrane and not specific trimer-trimer contact in a collar around the nascent pore. Now, I've skipped over loads of data and indeed some interesting mutational data from these in order to arrive at those, so I'm asking you to take my conclusions without a great deal of evidence presented for reasons of time and of detail. But I think that what I've tried to illustrate here is that by beginning with structure and then Moving on to experiments that use relatively powerful single particle techniques, we've been able with the molecular machine of this sort of size to begin to get down to the kinds of mechanistic information that enzymologists probing an active site of a, of a, of a, of a, or the catalytic site of an enzyme that does something to small molecules rather than to two huge membranes can do and hope to go on and do quite a lot more. But to move then from this level with a virus particle and a membrane to the cellular context, I'll show you what we've been trying to do by looking at a non-enveloped virus, the rotavirus particle. Rotaviruses are the agents of childhood diarrhea. They're non-enveloped particles as you see here from the surface rendering of older cryo-EM information. There are 11 double-strand RNA segments encapsulated. We won't go into that. Payload today, but it's basically the virus particle is a so-called triple-layered shell, triple-layered particle with a layer here of a protein known as viral protein 2, a second layer coating it of a protein known as viral protein 6, and then an outer layer made of VP7 and VP4. VP4 is cleaved to activate the virus particle. The protein and your shades of hemagglutinin, if you wish, and you'll see why, into VP8 and VP5. VP5 occluded by the clock that's reminding me that I had planned for 45 minutes and Yi Gong generously allowed me to go over by 10, so I'm warning you now, but that I'll try to keep at least to of that type. Through a combination of X-ray crystal structures of pieces, a crystal structure of what's called the double layer particle when this outer layer of the yellow and red proteins, 7 and 4, has been stripped off, and a cryo-EM reconstruction of the the double-layer particle and of the virus particle, we have essentially an atomic model of the entire virus. It is the role, by the way, of that outer layer to get the double-layer particle into a cell because it contains, I didn't go into this, a polymerase and a capping enzyme inside. Then, The double-layer particle never uncoats but rather emits messenger RNA once inside a cell. But what we're going to be talking about today, if you wish, are the activities of this delivery envelope, or envelope's the wrong word since I emphasized that had to do with membranes, but this delivery layer on the outside, the red and yellow proteins, which are the, are the delivery machinery into the cell, the double-layer particle being, if you wish, the payload. Now, this was several years ago, and we thought that it was pretty nifty to get to this sort of resolution in which a comparison of the X-ray map and the cryo-EM map was really quite favorable. But Tim Grant and Niko Grigoriev have done quite a lot better with the latest advances that all of you know about in electron cryo, and you'll hear more about, in electron cryomicroscopy. And so this was one map, and then as of a couple days ago, this is another just looking at little bits. But perhaps we just should have waited and done this, and then we wouldn't have had to do all the crystallography of the pieces. No, not quite true as you'll see because there's some very interesting conformational changes not represented in the structure of the virus particle itself. VP4, as I said, is cleaved to VP8 and VP5. I'll try to color code them roughly as seen here. And this is the arrangement in the polypeptide chain and the cleavage here. Now, you'll see that this spike-like protein VP4 or its cleavage product 8 and 5 looks like a dimer until you look quite closely. And you see that the dimeric spike is offset from a base that is indeed trimeric. And if we now look at the structure in more detail from the EM where this was fit entirely from that older EM, although these pieces we knew something about from X-ray structures, you see that it's a trimeric base with 3 projecting parts, although the purple domain up here is missing from the third. Whoops. Let's see if we can start that again. Sorry, it didn't loop. But you'll see that the polypeptide chain is anchored here with 3 helices, but that one of these domains is missing, and indeed it's almost certainly been cleaved away during that cleavage process, but that these domains form the dimeric spike, but then the third one is this cantilever that supports it, and it's quite unusual asymmetric structure. Now, one therefore assumes, and indeed there's now some very nice evidence, from the group in Spain, that the protein assembles this way and indeed rearranges into this asymmetric structure, but with one of these lectin-like receptor-binding domains disordered, and trypsin cleaves at these positions and also at one more position on this disordered guy to excise that domain, but leaving the anchor here. I think that's supposed to say I've gone over the original 35 minutes, or at least I hope that's what that was, rather than the executioner's bell. And in earlier work, we had seen actually both the dimeric arrangement of these domains as shown here, plus also a trimeric arrangement that looks like this with loads of nice interactions along the trimer axis showing that it wasn't a crystallographic artifact. Indeed, it was a trimer in solution, but with now the hydrophobic loops that I showed you pointing the other way. And that led us already at that stage, although now we can see it in much more detail from the EM, to a picture that resembles in lots of ways the fusion picture for the fusion proteins that I was telling you about in which these hydrophobic loops engage in the target membrane and they transition from this structure through some sort of extended intermediate to this folded back structure is coupled with the membrane perforation event that's needed. But in order to go on and really try to understand that, two things were necessary. First, we also needed to understand a little bit about the yellow protein VP7 because it is a calcium-stabilized trimer which clamps onto one of the proteins VP6, the green one in the inner structure, and holds things together. And as you see here, effectively locks the VP4 into place so that a calcium-withdrawing trigger will dissociate this protein and liberate VP4 to rearrange, or VP5. But another important thing we needed to be able to do that Phil Dormitzer showed was possible together with a student, Jane Trask, was that you could take recombinant yellow and red proteins, VP7 and VP4, and recoat an uncoated double-layer particle. You could prepare this by stripping these guys off from the virus particle and putting them back on and recreate an infectious particle. And those of you who are familiar with problems in virology know that one of the difficulties in studying virus entry is that it's almost impossible to prepare virus that's inactivated, mutated in the entry step. You can prepare virus that's slowed or sped up in the entry step, as I just showed you. But virus that's inactive in the entry step is essentially impossible to make. But here we can make recombinant entry protein. We can disable it in any way we want and look at infectivity as an assay. Because this is actually a perfectly normal double-layer particle that once it's inside a cell will carry out all of the transcriptional, etc., activities that are needed. And so that allowed us, or allowed Irene Kim, to mutate the hydrophobic loops and show indeed that that hydrophobicity was essential and was essential also in vitro for some membrane insertion activities, and that if we locked VP7 not just with calciums but with a disulfide, again we got particles that don't encode in vitro and were noninfectious. So, with those sort of mutational probes, we could then go back and do the following. We could take these proteins off. Take recombinant proteins, label them with distinct fluorophores, and actually label this guy as well. And, 3 strikes and you're out. And therefore follow what's going on in a cell. And so I'm getting to the, from atoms to cell as promised, I hope by the time this says 45. So that here in a case where the, double layer particle is colored with one fluorophore and one of the outer layer proteins is there. You saw what's happened here is the virus has been applied to the cell. We're looking at this by fluorescence microscopy with a spinning disk microscope. And at some time point, this is just a snippet from the time, you can see a timestamp here, all of a sudden the double layer particle is released and for various reasons we can conclude it's being released into the cytosol. And you can do that with all 3 labeled as well. I think I will just jump to this slide which summarizes the conclusions from that kind of experiment. And you're now beginning to see that we couldn't have done this experiment without working our way up through structure that very quickly, if you apply the virus to a cell, it binds and immobilizes. It may skate around on the surface of the cell for a short while, but within a minute, and often within the observation time, the virus is stuck to the cell surface. If you were to add EDTA at this point to strip VP7 off, the virus would just elute from the cell surface. But by about 4 minutes following its adhesion, the virus is fully protected from EDTA or from an antibody, and therefore it's in some sense engulfed. And then in another 3 or 4 minutes, that release step occurs. The release step is does not occur. The engulfment occurs but not the release with the different classes of mutation that I showed you prevented infectivity. So, that release event, at least in its sensitivity to these mutations, correlates with infectivity. And those mutations do not affect the internalization step. Some initial electron microscopy we did suggested that indeed the engulfment is a reasonably prompt sort of event that does not seem to require any of the familiar engulfment machineries. And we had shown in the fluorescence microscopy indeed that there's no association with clathrin, dynamin, or the other usual suspects. And it looks to us as if the virus is is just wrapping itself into the membrane to create its own little autoendosomal vesicle, so to speak, from which it escapes. Once in a while we do see what looks like a clathrin coat, but it's always concentric with the virus particle just as these are. There's no reason why occasionally this couldn't take a ride on the clathrin machinery, but clathrin uncoats promptly, and so one will be left no matter what. With an uncoated virus-surrounding vesicle. From that sort of picture, one just makes a movie that looks like this. But here's the step we're really most interested in, that perforation step that releases the particle into the cell. We've started with Danny Nicastro to look at this by electron cryotomography. And a number of the images indeed give us confidence that this is with cells grown on, the same kinds of cells grown on an EM grid and then prepared, and the thin edges of the cells are thin enough to get decent tomograms. And I won't go into the details, but fundamentally at least the wrapping notion seems to be consistent with a number of the observations we've been making. So our current model for what's going on here, and again I'm afraid I'm not showing you further evidence that would give you confidence that I'm not fantasizing, is that the sorts of steps I showed you in that earlier drawing are indeed roughly correct. There's attachment to a sialic acid-containing or in some retrovirus strains other glycan-containing glycolipids on the surface of the cell. We postulate, and this we are just probing with other fluorophores, that there's some kind of transient leak created by the interaction of these fusion-like loops it's not a fusion event, these hydrophobic loops with the target membrane such that calcium can leak out releasing the VP7 constraint and allowing this sort of rearrangement to occur. Our current model for how this rearrangement coupled with membrane insertion or membrane interaction of those hydrophobic loops would disrupt a membrane is based on the information from years of work by many groups working on membrane tension and bilayer properties that a lipid bilayer is essentially unstretchable. A 3% area stretch is enough to reach a rupture point.. It costs a fair amount of energy, and one can calculate that from measurements of membrane tension. But fundamentally, our current notion is at least that the real membrane disruption step might actually simply be also, as in the case of fusion, but here it's disruption, effectively a mechanical rearrangement of the membrane such that the area has to increase as it gets wrapped around this guy. And so forces it to expand. The energetics, and I won't go into it, are appropriate. That is, you can get enough work out of this to do the job if roughly 20 or 30 of the 60 spikes are doing this sort of thing. And indeed, you need to recode to at least about that efficiency before you can get adequate infectivity. So, this is where this problem stands. I should add that unlike the fusion steps, which I think are similar in their mechanism for a large number of different viruses despite the different architecture of the fusion proteins, that the mechanism, as I emphasized, is at some underlying level similar. In this case, other viruses, other non-enveloped viruses have obviously rather different mechanisms, still not fully understood, either involving ejection of a meristoylated peptide that seems to form a pore in the membrane. But whether that will converge on something similar to what we're seeing here is still something that we and other virologists are worrying about. At any rate, in conclusion, let me go back to that first slide where I said I think what you will see, I think, in looking over the program for the meeting concerns structural biology at the moment is this time and distance scale for processes that depend essentially on the kinds of diffusional mechanisms, and therefore mechanisms involving a stochastic character at some point, that I've tried to illustrate here, and that therefore the sorts of single particle methods that I've suggested we need to to use building upon structure are going to be essential since the instant you look at some kind of average, some kind of ensemble average, and alas, even crystallography is effectively an ensemble average, though fortunately the crystal at least traps a single state, and if we're lucky, we can trap multiple states in different crystals. Therefore, I think that this— it's this regime that will be occupied with not only in this meeting but going forward for a decade or more. I've added here that getting, as I've suggested, to the question of energetics is probably a next step. I emphasize that I'm not sure how long it will be before computational approaches actually give us sensible and realistic information in this whole range. But meanwhile, I've also suggested that there are a bunch of ways in which we can get at those questions by experiment, and I list a few of them here. All right, I'm sorry I went even 5 minutes over the time that Egon gave me. I think I've mentioned a few key players, Tijana Ivanovic, Luke Chow, Ali Abdulkhakim, Eric Salgado, and various collaborators on the rotavirus project, Nico Grigorieff, Danny Nicastro, and Tom Kirshausen for the optical microscopy with rotavirus that I showed you.

And thanks very much.

Questions? Steve, is he right here? It's a question of calcium stabilization of DPP7. Understood structurally? What is the mechanism? Oh, the calcium, I'm sorry. The calcium, I rushed through that. There are two calcium sites at the trimer interface of VP7, and if you remove the calcium and there isn't a disulfide cross-link, which we introduced to make that mutation, then it dissociates into a monomer, and we know that, we've known that for some time experimentally, and then the structure shows where the calcium is.

Just two sites at the interface with carboxylate ligands.

You can go to the microphone close to you. Hello, I have two questions about Part 1. And the first question is, what's the difference between the virus-to-cell memory protein and the cell-to-cell memory protein?

And the second question is, whether the hydrophobic protein peptides are conserved among different viruses. Thank you. So let me— the first question was what is the difference between viral membrane, the fusion of viral membrane with a cellular target, and let us say the fusion of two cells or the fusion of two viruses. Vesicles inside a cell. Was that your question? Yes. So the mechanistic difference, they are the same mechanism indeed from what we know of intracellular fusion such as synaptic vesicle fusion that as I said Axel might talk about. I'm not sure what's in his talk. Then, Indeed, as far as we can tell, the underlying mechanism is the same. In that case, there's an ATPase that then regenerates the initial state of the fusion protein so it can recycle. So, it's really an enzyme, so to speak, rather than just a single event. The hemagglutinin doesn't rearrange once it's rearranged to that low pH post-fusion structure. Minimum, the global free energy minimum once the protein has been cleaved. For cell-cell fusion, the mechanism is almost certainly identical, if you wish, to that of, or at least in some cases, to that of viral membrane fusion because there are several cellular membrane proteins including one that's important for for formation of the placenta in mice and in humans, though not other, or at least not lots of other mammals, which is a borrowed viral fusion protein. It looks exactly like the dengue virus fusion protein I showed you. Felix Rey has studied its structure. In that case, it's probably a transinteraction. There's no fusion peptide. It's formation of some of a trimer across two membranes that then leads to the fusion, at least as far as we can tell. But at least the same kind of fusion protein's been co-opted probably via a retroviral insertion and retrotransposon into a mammalian cell from a virus at some point in evolutionary history. So, as far as we know from the point of view of the membrane fusion, that is what the bilayer is doing and how the proteins are doing, what the proteins are doing to it, that despite the very different architectures of the proteins, molecular architectures of the proteins, the catalytic mechanism is the same. Think of it like, you know, two serine proteases, subtilisin and trypsin. The charge relay system is just the same, but the rest of the protein has a totally different structure.

What was the other question? The other question is whether the hydrophobic fusion peptides are conserved among different viruses.

Oh, whether the fusion peptide is conserved. It is conserved among hemagglutinins to some extent, and the fusion loops in the flaviviruses are conserved among the flaviviruses quite reasonably, but let us say the fusion peptide for retroviruses, HIV, doesn't— it's also hydrophobic, it's also somewhat glycine-rich, but it doesn't— it has no obvious conservation otherwise with the— in sequence at the detailed level. With those in flu or likewise flu and parainfluenza virus have fusion peptides just after a cleavage point, but again no direct sequence similarity other than conformation, other than hydrophobicity. As you see, the glycine at position 4 is conserved there because of a structure that is true of all hemagglutinins and hence true of all influenza viruses, but won't be true of the prefusion state of an architecturally somewhat different fusion protein.

Okay, thank you. In the interest of time, let me allow one last question. Oh yeah, just a quick question.

What did you use to run the simulations for HA1 and HA2? What did we use to run those simulations? Those weren't molecular simulations. They were literally MATLAB simulations of a series of steps, effectively a kinetic model. And so it was just homemade MATLAB code where, given the assumptions I made and set up basically a hexagonal lattice, array of hemagglutinins and then set up a probability at any one time step for any one of them to fire off and extend, and ran it then just as a conventional kinetic simulation. Okay, thank you.