 I've got to be here for at least half of it. So I'm going to talk about results from mainly atomistic molecular dynamics simulations. There's a little bit of coarse-grain molecular dynamics simulations. So the talk is really split in two parts. So the first part I'll be talking about maturation transitions of a bacteriophage that can be driven by pH modulation in vitro. And then the second half of the talk, and so this is a little bit older work that we published a few years ago, and then the second half of the talk is stuff I've been working on more recently. And so I'm interested in non-envelope viral entry and there's kind of two aspects to this problem. One is the pH sensing aspect, which I'm actually not going to talk about today but we've been working on. So I'm going to talk about the result of this pH sensing mechanism, which basically is the release of membrane-disrupting peptides. And so I'll be talking about, in the second half of the talk, the interactions of these membrane-disrupting viral peptides with membranes and then I'll be looking at how they are organizing inside of membrane. So essentially it's the modeling the result of a pH mediated process in the second half of the talk. So to start off with the system we're looking at is the Bacteria Phage HK97, which is a well-studied T-equal-7 icosahedral phage. And so 420 coproteins assemble into the immature Pro-Head-1 particle. There's a cleavage event that leads to the initial metastable Pro-Head-2 particle. In vivo, this is the stage at which DNA would begin to package and it would drive expansion through at least one intermediate state, ultimately expanding and faceting and reaching what's known as the head-2 mature state. In vitro, an analogous transformation can be triggered by modulation of pH. So from Pro-Head-2, if you drop pH, you get a transition to this expansion intermediate and then neutralization is required to reach the final states. I'm actually gonna be talking about, so HK97 is famous for making these cross-links and having this chain mail structure. I'm gonna be talking about a mutant variant which has one of these cross-linking residues mutated and so it doesn't form cross-links, but you still end up at a similar final state what's referred to as head-1. So I'll be talking about the transition between Pro-Head-2 and the head-1 system which is an expansion and a faceting. Do it like this. So the changes that we'll be seeing, right, here's the kind of global changes and then looking at the subunit structure and the capsimere structure. So we have crystal structures of both of these states and when the immature form was solved, the Pro-Head-2 state was solved, there was some interesting structural differences. The title of the paper was an unexpected twist in viral maturation, I think. And so one of the key things that they noticed was this kind of deformation of what's known as the spine helix and then there's also this rotation of this beta-strand e-loop between the states and then there's also a real organization of what's known as the a-loop and that's more obvious when you look at the capsimere structure where you see this skewed hexamer that has kind of a more linear organization of these e-loops whereas when you transition to the mature state you have this symmetric organization of these e-loops. So we'll be coming back to these different structural motifs. So what we did to look at this process was we generated, so we're simulating the asymmetric unit of this capsid and we do that through icosahedral symmetry boundary conditions so we're restricting the virus to remain perfectly icosahedral e-symmetric which maybe is non-ideal but we are able to capture the effects of the full capsid so it's a way for us to kind of mimic simulating a full capsid without the computational cost and so we generate a structural pathway between these two end states and we do that through a fairly simple method of, well it's a somewhat complicated interpolation but that offers, potentially offers some benefits over the more standard thing that people do which is a steered molecular dynamics simulation where you apply force and you drive one confirmation to the other and what can be the problem in steered molecular dynamics is that it's what's known as kind of the low-frequency first problem and that's that the system will move along these softer degrees of freedom initially and then there's all this high-frequency motion that needs to happen at the end of the pathway so you get these barriers that are at the end of the pathway and so what we did provides a more even distribution so here's looking at the low-frequency normal mode projections and the high-frequency normal mode projections and so we see we have a maybe more even distribution of those modes throughout our pathway so from that initial pathway we can try to refine that pathway to a low-energy pathway and we do that through a method called the string method and what the string method does is it essentially tries to relieve forces that are acting orthogonal to your path and so when your string converges all your forces should be directed tangential to your path and we see a good degree of convergence through energetic methods that basically the energy of the pathway is stabilizing as well as structural metrics so the structures along my path ultimately basically their RMSD is flattening out indicating that the path is not gonna be changing more so all I can really say is it's a locally minimum it's a pathway and a local minimum is essentially as far as I can say so we have this locally minimized pathway and we can calculate the free energy change along this pathway and we do that through umbrella sampling so we can simulate all of these snapshots we have along the pathway and we restrain them to remain along their pathway so basically the variable where we're straining along is this expansion coordinate and then we can compute the one-dimensional free energy profile from that and so what we see is we see a quite large uphill barrier which is more or less to be expected and the number we get is very large but maybe not outrageous so 150 kC per mole thermal scanning calorimetry showed that there should be a 250 kC per mole difference in the stability between this mutant pro-head two and the head one state so maybe it's a reasonable change in energy we do, so this pathway is generated without any knowledge of the intermediate and we do pick up that the pathway passes through something close to the intermediate so this intermediate just had a cryoem density I think it was around 12 angstroms and we compare our structures along the pathway to that cryoem density and we do see we reach a point that has a quite high correlation to that known intermediate structure and while it's not sitting in a free energy well it's not sitting in a well it is sitting in a flatter region of the landscape and so that's shown in green so this is the slope of our one-dimensional free energy profile and we see that it's landing in a somewhat flatter region of this landscape okay and I'm mentioning that it's a PMF at pH seven and I'll say why that is it's because we're performing not just standard molecular dynamics but we're performing constant pH molecular dynamics and so without getting into all the details the general idea here is that where we have tight tradable residues that can change the protonation state continuously during the course of the simulation so where this is known as a lambda dynamics method and so there's a lambda variable which would describe the lambda equals zero is a protonated state lambda equals one is the unpronated state and so the general idea here is that we know the pKa of what we call a model compound which would be a single amino acid and we know right so there's a certain pH we would like to simulate at so our standard molecular mechanics force field can give you some of that free energy change but it certainly isn't complete right and so there's some term that we're missing that I'm just terming delta QM and the key assumption in this is that this term that we're missing is basically local effects and we use this essentially correction to then bias or supplement our force field when we simulate the protein so basically we have this pH dependent term and then we have what our molecular mechanics model gives us and so that what gets added to the force field is right our difference between the pKa and the pH that you wanna simulate at what the molecular mechanics model predicts that that protonation change would be and then there's a barrier term which just tries to discourage lambda from existing and kind of intermediate non-physical states and so when you're simulating say an acidic residue at around its pKa so pH four what you see is you can see a lot of titration events so this lambda variable jumping between zero and one whereas if you're simulating relatively far away from its pKa you would see that it's basically fixed in the protonated state around lambda equals one. So with that we can come up with, we can get a fraction of the unprotonated and protonated forms of the different residues that we're titrating during the simulation and that allows us to estimate a pKa of those residues. From that we can use something called the Wyman-Tanford linkage equation which relates the free energy change, the gradient of the free energy change with respect to pH to the change in charge between these two states and ultimately what the way we use it is that if we have a, we know the free energy change or we have an estimate of the free energy change at a given pH using these pKa's we can try to estimate how the free energy would change at a different pH and so rather than having to run simulations at all these different pH's we can try to make some estimates or predictions about how just based on knowledge of these pKa's how that would modify the free energy changes. And so when we do that, when we look at our state, so looking at the initial transition from pro-head two to the expansion intermediate which should be favorable under acidic conditions, we see that the trend is correctly predicted that basically as you decrease pH the transition becomes favorable and then as I said you need to re-neutralize to or you need to raise pH to go from the expansion intermediate to the head one state and though it doesn't, I don't think it gets quite favorable the general trend is correct in that the free energy is decreasing as you are raising pH. And so we can basically map these changes onto our original PMF computed at pH seven. And so what we can see is as you decrease pH right you're essentially increasing this well for the expansion intermediate and then as you are increasing pH you're flattening out this landscape as you go from expansion intermediate to the mature or this head one state. And so it doesn't actually, we would expect it to get negative and it never quite does, but it does flatten out. So we're happy with that but I think what's maybe a little more interesting is that we're able to identify residues that we think are key in controlling the dynamics between these states. And so what we can do is we can look for residues that have shifts in their pKa's between the different states. And so we're just looking at this acid induced initial part of the pathway between pro head two and the expansion intermediate and picking out the things that have the highest pKa either the most positive pKa shifts or the most negative pKa shifts. And our kind of thinking or hypothesis about this would be that since this is an acid driven conformational change that things that are increasing their pH are gonna want to change a protonation state and would be related to the structural reorganization of the capsid and things that are maybe, while they're resisting protonation and maybe you could say they're resisting the transition have decreased pH's and one thing that jumped out at me was that you find that these things that are resisting the transition are at the interface of the capsid at the interface of the capsimere and HK97 assembles from capsimere. So it would indicate that these are residues that are important for stability, right? So we kind of are relating that these things with downshift pKa's are important for stability and things with upshifted pKa's are important for mediating conformational dynamics. And so we can go through and I'll give a couple examples. So of our top most shifted residues one that has one of the most downshifted pKa's that we think is important for stability indeed form salt bridges at the capsimere interfaces and if you mutate this residue you don't get assembly. So it prevents assembly. So it is critical for stability and assembly. A couple more residues. So these glutamic acid residues are located at this three-fold axi and they form what's been termed the three-fold staple and so they form a series of salt bridges and right. So it's between each of the pairs at this three-fold axi, each of the protomers at the three-fold axi. And again, if you make mutations to these residues you again prevent assembly. So another example of residues important for stability that we're picking out from our pKa shifts. Okay, so now let me talk about the ones that have that we think are important for mediating structural changes and we'll be looking specifically at these three structural motifs that have been highlighted in the literature and that I highlighted at the beginning of the talk, the spine helix, this elope rotation and the kind of hexamer reorganization. So the first one is the spine helix and this one was interesting. So this glue 153 is located on the spine helix and it forms a salt bridging interaction with arginine 210. But that salt bridge is there both in the pro head two and the head two two state. So in blue is our salt bridging distance so we have that there, both there in the beginning and the end but what we see is that and then in green is the rotation of this helix so essentially the straightening out of this kink. So what we see is that essentially you need to break that salt bridge and it allows the helix to rotate and then that salt bridge comes back in and locks it in place. And so by allowing that glue to change its protonation state we think that facilitates that conformational change. We also pick up several residues on the elope that are right in the elope is known to undergo this rotation. We focused in on E149 which is down at the base of this elope and it forms an interaction with Q195 on a neighboring subunit and essentially we think that as that interaction breaks as well it allows for mobility of that elope helix and we see that the distance between these residues correlates well with the elope rotation that it undergoes during the maturation. And then the last structural motif was the aloops at the center of the hexamer and we pick up two residues that are on those aloops and so when you're in this asymmetric configuration in the skewed hexamer there's salt bridging interactions that these acidic residues are involved in and so essentially you need to disrupt those salt bridges to allow for this symmetric structure to get adopted and the other thing I'm showing here is there was some HD exchange that showed that most of the changes in the elope HD exchange profiles were occurring between the ProHead 2 and the expansion intermediate state and there was very little change between the expansion intermediate and the head one and all I'm showing here are contact maps between the different aloops and essentially it's showing it may not be terribly obvious but it's showing that there's more change between the ProHead 2 and the EI state and then there's pretty minimal changes between the EI state and the head two so indicating that the structures we're seeing are consistent with to some degree those HD exchange experiments. So the last thing we did for HK97 was right so I'm talking about in right phages don't really need to be pH sensitive since they inject their DNA and are probably so we would like to see if we can make a connection to what drives packaging or drives this maturation in vivo which is the packaging of DNA and so can we relate what DNA would be doing to what we think pH is doing and so the way we set up this calculation I'll try to take a minute to explain it is that so right ultimately we would like to know what the effect of DNA is on, so essentially how is DNA changing the free energy to cause this maturation transition so what we know are, so we have experimental measurements so we have predictions for what this pH driven maturation changes and so we can formulate a thermodynamic cycle and try to get at the DNA driven free energy change as a function of what the pH driven maturation would be and then minus the effect of DNA packaging on the ProHead 2 state versus the effect of packaging on the Head state and what we're talking about here is really just, so essentially we're just assuming that DNA is gonna cause an upshifting of the pKa's of the acidic residues on packing and negatively charged DNA in there and that that would raise the pKa's to some degree of these acidic residues and so we don't really know how much DNA gets packaged in the ProHead state so the packaging starts in the ProHead state and so our assumption here is that so basically there's gonna be some perturbation to these acidic group pKa's and we're just gonna look at and say, well you're gonna have some, potentially you have some fractional change in the ProHead state versus the change you get when DNA is fully packaged and so we can look as a function of different ratios of what the pKa perturbations are and see how much perturbation is needed to make this transition favorable given that you're starting at a highly unfavorable state and so what we see is if there's very low perturbation to the ProHead state which would be a ratio of zero you can achieve, this would be favorable just based on pKa shifts at somewhere around I guess it's 0.3 or 0.4 pH unit shifts but even if you have a fairly sizable change in the ProHead state too so say a 0.6 you still are only requiring somewhere around a half of pKa 0.5 or 0.6 pKa shift so to conclude on the first part on the HK97 section so we use basically pathway refinement methods to and constant pH to calculate pH dependent free energy change well and from the pKa's we're able to map these free energy differences onto our our PMF or one dimensional free energy change and we identified critical residues that we think are either controlling or connected to both kind of stability and then structural reorganization of this capsid. And then in the end we use this thermodynamic analysis to try to understand how the DNA could be perturbing these pKa's to drive conformational changes as well. Okay, so that was work I had actually started quite a few years ago when I was a postdoc with Charlie Brooks and continued on for a few more years. Stuff I've been working on more recently and what I've been quite interested in is trying to understand non-envelope virus entry mechanisms and particularly what's happening in the endosome and how viruses are sensing environmental changes and then how are they getting out of the endosome and this is the aspect I'll focus on today. And I would say just generally right so our understanding of envelope viruses is somewhat better I think than the non-envelope viruses so for some viruses anyways this fusion peptides have been well studied. Influenza is probably the most well studied example whereas in non-envelope virus entry there's considerably less is known. So our model system that we look at here is called flock house virus so it's a small t equal three icosahedra least symmetric capsid. It's a member of the notavirus family which infects insects is an RNA genome and so here's the capsid structure and here's the asymmetric unit structure. It has an autocatalytic maturation event which cleaves the 44 C terminal residues to form what's known as the gamma peptide. The crystal structure resolved 21 residues the first N terminal 21 residues and this peptide is termed gamma one and the first things I'll talk about are related to gamma one which has been the subject of some older in vitro experiments and so the gamma one peptide forms an amphipathic helix and like I said it was resolved in the crystal structure. These phenylalanines are highlighted and I'll probably briefly touch on those toward the end of the talk. And so there's a fair amount of structural conservation at least certainly among notaviruses so I'm showing a view from the inside. So maybe I should go back when I... This is a key point that, so you don't see the gamma peptides on the exterior of this capsid. So they're sequestered on the inside of the capsid and they need to get externalized. And so here's a view basically from the inside of these various capsids looking down a five-fold axi and you see a similar structural motif between notaviruses and tetraviruses and also this one bernie virus that you have these peptides aligned along the five-fold symmetry axi and the general thinking is that this would be the point of externalization that they could, there would be some dynamics at that five-fold, a pore could be created at the five-fold symmetry axi and these peptides could exit through that symmetry axi. Lytic peptides and non-enveloped viruses are quite prevalent. We keep learning, finding more systems that have either a lytic peptide or some component of their capsid that are able to disrupt membranes. So for many different virus families and many of these are triggered by, well there's some stimuli that triggers them in some cases receptor binding. Low pH is also a common stimulus for activating these peptides. So what do we know about gamma? So gamma is critical for infection. So you need cleavage of gamma for both infection and also for in vitro liposome disruption. So these die leakage assays are done to show membrane disruption properties. And a key point to showing their importance for infection was what are known these intrans experiments where a non-cleaving virus, so a virus that doesn't cleave its peptide will have very low infectivity but if you supply a wild-type VLP with that non-cleaving virus, you get robust infection. So having those cleave gammas is critical. So there is a pH dependent process to this but as far as we understand it's the release of these peptides that is pH dependent. So you get a maximal, this is die leakage, you get a maximum around pH six which would be consistent with kind of endosomal conditions. But if you, right, so if you do die leakage at pH seven you get low leakage but if you incubate the system at pH six and then perform die leakage at pH seven you get robust disruption. And so the interpretation is that the pH six condition is exposing or externalizing these peptides and then they will disrupt. Working just with peptides alone some experiences have shown that very similar die leakage profiles occur at different pHs as well. So essentially to get out you need this pH change but once they're out they have pH independent activity and also raising endosomal pH also inhibits flockhouse virus infection. Okay, so the first thing we were interested in looking at is trying to characterize these Gamble One peptides in membrane environments and there's some data showing that the composition of the membrane will affect how well they disrupt membranes. So on a neutral zwitterionic POPC membrane ratios, lipid to peptide ratios of around 30 to one are effective at disrupting these membranes whereas against a negatively charged POPG membrane you need much higher peptide concentrations to achieve a high degree of leakage. And so we started out trying to see if we could get some insights into why that could be. All right, so let me speed up here. So here are some of our questions. So the first thing we looked at was membrane binding and we used the Martini coarse-grain model for doing this and Martini suffers in its ability to model protein conformational changes. So what one needs to do is supply what you want your secondary structure to be. So we can run simulations where we, what I'm saying is turn secondary structure on which means I'm forcing the peptide to be helical or I can turn it off and I can let it be something that resembles a random coil. And what we see is that for PC we get, it requires helical formation to get a stable binding whereas if it's not helical we only see kind of transient contacts. Whenever we have some degree of negative charged lipids in the bilayer we always get rapid binding and stable binding. I'm gonna move quickly past this because I think I'm getting a little behind schedule. So from those coarse-grain simulations we can convert those two atomistic simulations and we can run molecular dynamic simulations to see folding or see the stability of these different configurations. And so on PC it's initially helical and we find that it remains helical and on PG we have a disordered structure that on the time scale of a microsecond remains basically disordered and surface bound. So that's not all that interesting or maybe all that fair of a calculation. Okay, so what we did was maybe a little bit in fair calculation was to say, okay, let's start at the same initial structure and we'll just swap out the lipids. So we do what we call lipid substitution simulations where we take a snapshot from the peptide bound to a mix of PG and PC and then we swap out the lipid type and allow the simulation to fold potentially and we see that on a PEC lipid we recover that folded state and on a PG lipid we get relatively low degrees of folding and we also get deeper insertion into the membrane with against PC. And then the last thing we did here was we said, okay, well if it's really unstable on PG then if we start as a helix and we run on PG we should see some destabilization and the result was that basically we didn't so we still see the same degree of insertion and helicity and that led us to think that what the charge, introducing the charge might be doing is raising the folding barrier and so that ultimately the folded state is more stable but having some charge in the membrane may prevent or slow down the folding of these peptides which would relate back to why, so the presumption is that helical structures are needed for lysis and so if you can achieve helices quicker or more favorable on PC then that may provide some explanation for the difference in peptide concentrations required to lice these different membranes. So since I'm running short basically what, so Gamma 1 there's been biophysical studies, it's small, it's easy for us to work with but it's really not very infective so the full length peptide is much more infective than this would be Gamma 1 so its biological relevance is somewhat limited I would say and so we really wanted to move towards understanding how the full length peptide behaves differently than the Gamma 1 and so I'll go through this quickly, we did martini simulations again with Gamma 1 in transmembrane configurations and basically they're all unstable, when we do martini simulations in with the full length peptide and essentially what we do is we put the Gamma 1 region in the helix and we allow the C-terminal region to extend, we find that the C-terminal region always inserts into the membrane and all these different oligomeric states are fairly stable but okay that's martini, it's not necessarily the best for modeling proteins and so I got an allocation on this Anton supercomputer that David Shaw built and we were able to run some longer time scale atomistic simulations and I'll just finish up by going through what we've seen and we've just been getting this data recently so when I run Gamma 1 at a surface bound configuration essentially it stays at that surface bound configuration when I run the full Gamma, now I'll come back to this in my last minute is we see that aggregation is happening and these peptides are basically coming off the membrane so we were thinking we may see insertion and we're seeing the opposite, we're seeing aggregation and essentially them coming off of the membrane when we, so then we can just insert them in a transmembrane configuration and so the Gamma 1, the shorter Gamma 1 peptide has, essentially there's nine peptides in here and for them transition to the surface on the time scale that we simulated for and maybe more of them will if we run that longer so they don't seem terribly stable even in all atom simulations whereas when we start the Gamma 1 in a transmembrane configuration they do remain stable and so the last point is just looking at water flux and pore formation so we were quite happy to see quite a bit of pore formation so we see a little bit of water when Gamma 1 is there and then we see quite a lot of water when the full Gammas are there and so we see increasing amounts of water density in the interior of that hydrophobic slab as we run these simulations and these are over eight microseconds here's just looking at different kind of pore formation so we see trimmers and we also see a a hexamer structure and I think with that I'll make my conclusions well actually I had one more thing to say so the aggregation thing is it really kind of has changed my thinking on this a little bit and it's made me wonder so I had always viewed that these peptides, you're in the endosome the peptides externalize they find their way to the membrane they create a hole in the membrane and maybe that's not how it works and so maybe they're being tethered within the membrane and so either in a VLP that may be due to this propensity for aggregation in a virus it may be due to interactions with RNA so the termini of this peptide is GF, GF which I don't know a ton about RNA binding but apparently that's a consensus RNA binding motif and so I've written a new anti-allocation and I wanna look at basically the effect if I kind of restrict these peptides if they were remaining the termini were remaining in the capsid and then also try to map RNA onto these peptides by doing single nucleobase simulations and this model has some attractions it provides you a high local concentration there's structural organization there and also you have the virus is very close to I mean the virus is right there and so if you're creating a hole it seems that it would be easier for the virus to get right through there and so with that I'll just thank my funding from NIH and computing resources Shivangi Nanjia did the work on the Flockhouse virus and then I collaborate we're doing experiments now on these peptides with Nathan Alder at UConn and then the HK97 stuff was stuff I had started with Charlie Brooks and Jack Johnson has kind of encouraged me on both of these projects because he's done much of the structural biology work on both of these systems so with that thank you for the opportunity and I'll take any questions.