  늑                                                                                                                                                                                                                             We will be about molecular simulation in the times of COVID. And we look forward to hear your presentation together. Thank you. Can you hear me? Yes. Good. So thank you. First of all, a huge thanks to the organizers up front for holding this meeting. I think this is very important for all of us, for the community. And I'm really excited to be part of this. So as you can see from the title, I very much acknowledge the fact that all our life has changed and as I will show you also our research has changed. And we have in my group and my collaborators groups, my friends groups, many of us have been trying to contribute in a meaningful way to the fight against the COVID-19 epidemic. So just here on my first slide, you can see a, an outline of the virus life cycle from a recent review by Brett Clounsinger. And as you can see, this is a pretty complicated life cycle that can be however divided into a few key steps. The first one is the part, key part of the infection, the entry of the virus into the cell. And when it establishes itself leads to the translation of its genome and the viral reproduction, a key step then is to assemble new variants. And in the last step to release them to the outside. And so in my talk, I will show results from our work involving molecular dynamics simulation in where we looked at two key steps in this viral life cycle. The first one in presentation is an element that is very common to viral infections, and that is immunosuppression. And it turns out that is associated with a protein, a protein specifically that is part of the translation and fully peptide processing machinery of the virus. And then in the second part of my presentation, I will show you also results associated with the entry of the virus. And specifically, they will be looking at the so-called spike protein or S protein on the viral surface with which SARS-CoV-2 attaches itself to the host cell receptors. So, first part, I will concentrate on the so-called papain-like protease. This is a protease that has multiple functions and this of course is common in this very dense viral genome that proteins do a bit of moonlighting. So, the papain-like protease, as its name implies, is a cleave protein chain, and it does that in part to support the processing of the multi-protein chain produced from the viral genome and cutting them into functional protein units. But in addition, the papain-like protease also is very actively involved in the suppression of the innate immune response of the host cell, which of course tries to inform the outside that it has been infected and activate antiviral responses that, for instance, can kill an infected cell. This would be, for instance, in this case, the type 1 interferon innate immune response. So, this is work that has been done in collaboration with the group of Ivan Dike, who is at the University Hospital in Frankfurt and also a fellow at our institute, and with his Zen postdoc Dong Yook Shin just took up a faculty position in Korea. And what they found was extremely interesting. They found that this papain-like protease, which was actually known from earlier work, was cleaving a polyubiquitin chain, but also it was cleaving another ubiquitin-like protein, ISG-15. But what they found was that new SARS-CoV-2 had changed its catalytic activity from favoring the cleavage of diubiquitin or ubiquitin chain to cleaving this ISG-15 ubiquitin-like protein. And that was very interesting because cleaving of ubiquitin and ISG-15 are both associated with modulating immune responses. So what our collaborators then could show is that the driver for this transition from cleaving ubiquitin to cleaving ISG-15 was a change in the affinity, the binding affinity. In a equilibrium binding experiment measuring the KD, the dissociation constant, they found that KD had inverted, and this is shown here in the plot on the right, where you have now much stronger binding to protein ISG-15 and weaker binding to K48-link diubiquitin. That is the driver for this change in the kinetics. So then you can of course ask what is this good for and why does it change. We know from many years of viral studies that infected cells often activate this ISG-15 production and they attach ISG-15, a diubiquitin-like molecule, covalently to many proteins of the cell as well as to the proteins produced from the viral genome directly. But it's actually not very well understood what exactly it does and there's a number of theories out there for what ISG-15 modifications do. They could alter the oligomerization states of proteins or also modify cellular localization of these proteins. And what we now found or our collaborators found is that far as CO2 activity decreases the ISG-lation of cellular proteins. They could show that with proteomic studies. And one of the key proteins involved in this innate response, if you have a better understanding of what ISG-lation does, is this interferon regulatory factor 3. And so just very briefly in the nutshell, if it's modified by ISG-15 covalent attachment, then gets correlated and that leads to uptake of this IRF 3 factor into the nucleus where it activates the antiviral response by acting as a transcription factor like protein. And so this however is altered if ISG-15 is removed from IRF 3, then IRF 3 is degraded, not taken up in the nucleus and that leads to a suppression of the interferon response. And so that's in a nutshell what we think happens in this particular case. But I have to emphasize this is a much broader response that is happening here. So what is going on at a more mechanistic molecular level. For that, Dong Yook Shin, a postdoc in the department of Ivan Dikic, solved the structure of the complex between PLPRO and the protease, here shown in pink, and the diabetes. And what we could see in the structure was that there are two changes between SARS-CoV-2 and SARS-CoV that are interesting and could contribute to this shift to a immunomodulation activity rather than ubiquitin processing activity. And that one is, one is a change from a leucine position 76 in old SARS to a 3-anine position 75. And then there is another shift with the opposite change in polarity from a serine to a valine. This course got us very interested and Laura Schultz in my group and together with Reza Medipur, they set up MD simulations of various constructs and what they found consistently was that ubiquitin actually in bound to the SARS-CoV-2 PLPRO is unstable. So on an MD time scale, which is quite unusual because protein complexes are rather difficult to simulate, but consistently what we found was that the ubiquitin started to dissociate. So this is illustrated here in this RMA-D versus time plot where we can see that in four out of six simulations on a micro-second time scale, the ubiquitin comes off. This is indicated by the orange and yellow hue. By contrast, ISG-15 remains stably bound to the protease, so indicating yes indeed it is a good substrate of this protein. There is biobiquitin is a weak substrate and has a much reduced binding affinity, consistent at least this level with the experimental observation of a shift in KD and a reduced catalytic activity for ubiquitin. So, but we have to do some control and this is one important control is to introduce the mutations or to exchange the amino acid between SARS-CoV and SARS-CoV-2 back into the old SARS-CoV background. So we take the new valine and the new three-alene residues and we make mutations in the old SARS structure. And what we found was we didn't, in this one simulation, see a full dissociation, but the interface clearly opened up and then in one of the intermediate forms that we had seen before was an important step towards protein-protein dissociation. So, one of the wonderful things with MD simulations is that then you can start to zoom in and you can see what is happening at the very details, at the near atomic level of these processes. And what we found was actually quite interesting that ultimately the main driver that we could see in the MD simulation that explains the difference between this substrate preference of the protein-like papain-like proteases is a small change in the hydrophobicity. It is the difference between the three-alene here and its interaction with the isoleucine 44. What happens is that increase in the hydrophobicity of the interface between the three-alene on the protease and the isoleucine on the ubiquitine can transiently lead to an accumulation of water. The more hydrophilic character that water entered is too amino acid contact and it is this consistently water entry is what leads to dissociation. Just as an important point to show that mechanism is not all we can do with MD simulations, I want to emphasize that MD is also a good tool leading us forward in the development of new therapeutics. So what our collaborators started out with is an inhibitor that had been developed for the papain-like protease in the mesocard map almost 12, 13 years ago. This is for the old SARS. This inhibitor, as it turns out, also works very well for the papain-like protease of the new SARS. We could show why this is the case by docking it into the new SARS protease structure and we could find that indeed the binding pocket is very well conserved. And the inhibitor finds very stably in a mode that is essentially consistent with the old SARS. And that, of course, opens up now a window of opportunity to target this binding site as a possible way to suppress the viral proliferation. And indeed that's what our collaborators are doing and at a functional site what they could show is that it is that this inhibitor reduces the protease activity. And they could, in the lab of Sander Tizek at the clinic in Frankfurt, could also show that with this inhibitor you could actually suppress viral replication and viral proliferation in life cell acids. We could show one result looking from the right here, looking at the mRNA level indicating the level of infection. So we have both mechanistic understanding and with at least a first inhibitor around forward to targeting this process and hopefully deal with replication rate. But maybe even more interesting from an intervention perspective, of course, the development of a vaccine and most vaccines look at proteins at the viral surface. And in particular at the so-called spike protein, this is shown here in the cartoon of a viral entry. So here this circular particle indicates the SARS-CoV-2 virus decorated with these S proteins, these flower shaped structures that bind to the ACE2 receptor on the viral, on the host cell surface. And that leads to either direct entry through the plasma membrane or again the process that is not so well understood entry through the endocytic pathway and then endocytic escape. So we started modeling this SARS-CoV-2 spike protein or S protein already in the first weeks of March. This was work pushed forward by Max Sikora, but then really involving a very big part of my group and my department. You will see the names listed later. So what we knew at the time was that we had a decent electron microscopy structure of the head of the spike, but the rest was essentially not known. So we pieced them together using various modeling techniques and using information also from the old SARS. So we put together a stock model as well as a model of the trans membrane domain and the membrane anchors. And importantly, we included the glycans to decorate the surface of the spike protein based on extensive mass spectrometry studies from several groups, a particular group of crystals. And so built a full length model and inserted it into the membrane and then set up an MD system. What is maybe interesting from a tactician's point of view is we decided to put four spike proteins into the membrane. Now you might think this is a bit crazy because you already have an enormous size system. So now we have it almost four times as big. However, there are certain advantages of doing this. One is that with only one spike, now we have big problems with self-interaction across the priority boundary. With four spikes, we have four times the cost, but we also have four times the sample. We ended up with a system with four million atoms to make it run a bit faster. We use heavy water that allow us to use a four-pensate content. And in this way, we could collect about two and a half microseconds of MD, which by having four proteins amounts to four, about 10 microseconds of aggregates that sampling. And here you can see a movie, I don't know if it moves on your screen. It indicates spike is actually fitting quite flexibly on the viral surface. And that is illustrated here in more detail where we emphasize that the joints in the viral stalk that we termed hip, knee and ankle with head or body, upper leg and lower leg forming the more rigid part of the protein. When we zoom in on and do an RMSD alignment on these more rigid parts, you can see that the rest of the protein really phrase out is a superposition of structures along the MD trajectory. And this indicates these are very flexible hinges and that's also emphasized by the distribution of the hinge bending area. That's very interesting. And however, as we were doing these and these simulations, we learned that in our institute, a data set was collected by the group of Mark and Beck instituted at the EMBL in collaboration with the lab of Jacomina Granger-Loker at the Paul Ehrlich Institute, whose group has been doing the virology and what the data set that was collected is a very high quality cryo electron tomography. And you can see here a cut through the 3D volume of this EM tomography data set where you could maybe see the viral particle and they appear. Cut illustrating them and on the black dots on the surface are these spike proteins and they were beautifully resolved in this data set. And remarkably already in the raw tomogram, they could see the hinges and not just the spike body, but you can see the hinges and you can see that the stock actually is flexible and that it's bent. And that would also explain the irregular structure of the protein on the viral surface. What we could then show wise, by comparing notes is that our MD structure, molecular dynamic simulation structure could be just placed directly into the raw tomograms. These are not flexible fits. You may have seen flexible fitting of MD structures into into an extra density map. Here we just took snapshots from the MD simulation. We took is extremely high resolution tomograms without any further processing and you can see the MD structure fits very nicely into the structure. And I'm not going to discuss the that we also have very good agreement with the top tomogram averages that are very high resolution I just focus on this aspect and emphasize now this beautiful consistency. Here also in this central view graph where you can see a bigger part of the viral surface from a raw tomogram now decorated with MD snapshot. Now, why do you, we think that the flexible the stock is so flexible. Well, we think that a very possible explanation is that this increases the ability. And the ability is, is a measure of, or ability refers to having multiple interactions in parallel to each one of them relatively weak but jointly making the interaction very strong. And I think that by having a flexible attachment point, the, it's easier for the virus to have multiple interactions with the spatially spread out AC to receptors on the human cells, tightening the interactions and thereby leading to a more effective infection process. So, one other one final point I would like to, to emphasize where we saw a very totally nice correspondence but between experiment and the simulation but gained additional insight concerns some flight protein. So, these viral proteins are covered by an extensive protective shield of sugar like molecules and here on on the far left you can see, these are large glycans on the surface, and they fit very nicely into the electron density, both the single particle structures as shown on the left, and on the pomegram shown here in the middle. And, even in the raw pomegrams again you can, you can actually see very nicely that you have a strong additional density, decorating the, the surface of the protein until so here you see now what what this does in on the basis of the MD simulation. When we average the density of the glycans on the viral surf on the spike surface here in green, you can see that these large and very dynamic very floppy sugar chains effectively mask almost the entire surface of the protein, they expose very, very little just primarily the receptor binding domain with which it interacts with with the AC2 receptor. And what that means is, this is a very effective shield against the binding of antibodies that effectively the virus uses this coating by by glycans as a means of giving antibodies very little distinct epitope surface to bind to. This brings me to the, my summary slide, and I hope I could give you at least an idea of how molecular dynamics can be useful and can make meaningful contributions in rather difficult times for all of us. The first part of my presentation was focusing on an interesting and I think very important aspects of what viruses make such highly effective machines and so successful in in in combatting our immune response. I showed you that in case of SARS-CoV-2, as compared to the original SARS or old SARS, we had to shift from a biophysics perspective is that ultimately this shift and greatly enhanced ability to mess up our immune response is that one to maybe three water molecules can more easily penetrate into the otherwise hydrophobic interface of the protease. With ubiquity, and which does not affect the ISG-15 interface, conversely, an increased hydrophobicity adding the valine strength in the interface with ISG-15. I also showed you that MD simulations can help us when assessing and hopefully improving inhibitors that targets this process of the protease. I showed you results for the GRL-0617 inhibitor developed by the MEZICAR lab originally for old SARS, also Stoles viral reproduction for new SARS. Then I showed you results for spike, the protein at the viral surface and these simulations revealed hinges in spike stalk, in a spike stalk, and then in completely independent electron tomography experiment either recapitulated and I concluded by showing the dynamic glycine code and it is consistent with what we see in the endipotomography and I think which is very important that something we're still actively working on is a first paper submitted now on looking at our ability to still find good epitopes for antibodies that can lead to a strong and robust immune response. That I would like to thank my people in my group, so the PLPRO, protease simulations are carried out by Laura Schultz, PhD students in my department and with a lot of help and support and insight from postdoc Reza Medipur and I really cannot thank enough our experimental collaborators Dong Shin, who just left Frankfurt and is now starting his faculty position in Korea. This has been a fantastic collaboration with the lab of Ikan Ivan Dikic and I really look forward for more of this work and also we've got some virology health and a lot of health and support and time achieving. The tomography experiments that I showed were carried out by primarily by Biafra Toronova with help of the people in the EM facilities in Heidelberg at EMBL and in Frankfurt, and it's been a fantastic collaboration with the Marking Beck group at our institute and still at EMBL. The virology was done by Krzysztof Sherman in the community grinder locker groups again with lots of health from other people at the early kids. And finally the ND of the spike proteins is spearheaded by Matt Sikora and the model building and then these simulations had a lot of support, it was huge team effort by Siren von Bülow, Florian Blanc, Michael Geft and Roberto Covino. We just started his independent group at the Frankfurt Institute for Advanced Studies. In the depth, I would like to stop and take your question. Thank you very much for this very very interesting presentation. I'm sure there will be lots of questions. So there is one in the chat. Please comment on how you would mean the effect of different environments like extra-cellular space, cytoplasm inside nucleus. This is, I think, a very interesting and important question that in particular when I think about the spike protein, where I just briefly mentioned that it's not so well understood how at the moment how the entry of the virus really happened, whether it is directed to the plasma membrane and there's papers describing that or through endocytic pathways. And these environments are very different in their redox level, but in particular also in terms of pH and salt conditions. And so it would, at the moment, we took for the environmental conditions, we took a neutral pH type of setup, but clearly for fusion with the part of endocytic escape, we would probably have to revisit some of these initial protonation analysis and maybe even change them. In addition, there is, of course, also the question, such as the environment but of the protein itself, was also many things are not so well known, and that concerns in particular cleavage state, so when the proteins actually get actively remodeled by host cell proteases, and when exactly these transitions happen is again, so cleavage, for instance, a spike hat is not so well understood. So, interesting questions and lots to do. We took a first stab at this, but I think we'll be busy with this for time to come. Thank you, Gerhard. This was a question from Strabonit out of the ring here. Okay, now we have a question from Stefan Wolff. Yeah, hi Gerhard. Thank you. Very beautiful talk and great work that you did there. Concerning this break up of the interface with the ubiquitin by a single amino acid, that's quite interesting really. Did you have a closer look at what is actually modulating, so what this one single mutation is doing to break up a complete protein interface because that seems like a huge change in free energy between the for binding just but based on such a small perturbation is quite surprising. Yeah, so, so I don't at the moment we don't have any any free energy results so obviously this is a big effort, and the only thing that that that we have we see is is kinetically. It is the main driver is, I mean it's a very small interface to start with the affinity is weak it's in the 10 micromolar KT to start with so it's not a, not a strong binder and I think that's what allows this transition in bounds and unbound state to happen at all. For a strong binder, it would not happen. And in all the, we have for dissociation events out of six runs of three microseconds 30.2 I think, and in all for what what we see consistently is that water comes in this I don't mean either losing interaction and and so you lose effectively you weaken one out of three hydrophobic context and I think that is enough to make this a week thing now. I should point out that this is a dye ubiquitin so the other ubiquitin sits in the access side that is still tightly bound and I think. So, so that's why we have just, we don't see a complete, don't see a complete association we just see one ubiquitin come off or come partially off. Maybe as a bit of explanation. That's why the overall binding is still. I think, even in the experiment, it's now reduced by like 30 folders so but not not not now. Thank you. So I think, but this, I think that that maybe 34 degrees is quite would be quite consistent energetically with effectively losing one one hydrophobic context. Now we have a question from Daniel Madulu. Thank you for the nice talk. I have a question on the law of water in one of the slide you showed that water was able to penetrate at the hydrophobic interfaces. So how did it also affected the binding of the small molecule. So I need to know maybe explanation maybe I missed something. Thank you. Yeah, so the at this interface, right, the water, the role of water forms additional hydrogen bonds with with with the prionine for the small molecule inhibitor there we did not specifically look at at at water there because it, it was very stably bound and and and really quite consistent with with the experimental structure and and just to maybe add to that inhibitor is is interacting both hydrophobic hydrophobic interactions and the hydrogen bond has charge in fact with with with with the protein so their water did not. They have not examined it in more in greater detail but water did not play a particularly important direct role. Thank you. Now there is a question from the audience in case of sensing devices from taking from for tracking infectious person. Are there any glycans or sugar that would be important to be detected. Yeah, so I, I don't think I would be the the right person to answer this. I can just maybe give it a try. The important thing, of course, is all the glycans are our glycans and that's one of the reasons that the virus uses this so that uses the same sugars that we add on to our cell surface receptors and other approaches that are on the outside of our cells, and it hijacks the glycosylation machinery as well the sugar from the infected cell and uses that of course makes it so difficult for the immune system to deal with these envelope the viruses, because much of their surface is covered by the same chemistry as our own cell, and so antibodies actually have a little chance and I think now, with respect to the question that that also means the sugars, per se, would probably not be particularly useful, however, if you look at maybe protein fragments with sugars, that might involve but again, maybe not the right person to address that. It's still an interesting question. And then another question from the chat. Can you comment about the differences between locked and open structure of spike protein regarding its entry process into the cell? Yeah, yeah, so on the, this is a very important question where we had lots of discussions about. There are several structures of the spike head out that are either completely closed, or they have parts opened up and it is thought that interactions with the AC2 receptor in particular require quite a bit of opening and also some antibodies bind only to open or only to the closed state. In our simulation, we therefore use a mixed structure with one subunit open. In the experiment, what we found was, which was surprising, that in our colleagues viral preparation, the vast, well, the majority of spike subunits was in the closed state, only upon closer inspection we could detect some open state structures. At the moment, we don't have a good way of looking at this open, close to open transition, but that again will be a very important and very exciting project. And I also hope that from the tomography as well as from single particle EM studies by other groups, we will get more guidance on how this transition happens, what it uses it and also what are the factors setting the equilibrium and could also be something that came up in the first question environmental factors such as as pH and both conditions. There is another question from the audience regarding the force field that you use for the simulation. Yes, so, so these simulations were done in with the charm force field. Yeah. Yeah. One last curiosity myself. If this flexibility of the spike is so important for the avidity and probably also for the infectivity of the virus. Is there any data that you compare maybe with the previous sorts? Is it known now is the flexibility in that case? Yeah, so we think that the all stars also was flexible because the stock is essentially 100% conserved among the most conserved for it in part because in the in the post fusion structure, which is the stable structure of so we can think or at least I tend to think of the pre fusion structure that I simulate as a meta stable structures like spring loaded for a transition into a more stable structure post fusion. But other envelope viruses seem to have much shorter stocks and more more ready structure but we do not interestingly also for other viruses we do not know very much about these elements because Yeah, you would you do need very high resolution structural studies like the entomography to be even able to to to see these things, the degree of flexibility so we are thinking now of course of ways of looking into this in more detail and one possible way would of course be to capture infection events. or maybe also with high resolution super resolution life microsoft to see if indeed this is a single receptor binding event or a multi receptor binding event where you have recruitment. But this is at the moment I can only emphasize is once more this is a speculation we just do not know we just find it interesting as a possible explanation for why why this is there. One last question. I would like to know if there is a specific reason to put every water in the simulation. So this is of course something that concerns all and either. I mean the reason that we include the water is is that the first of all mechanistically may be interesting but the force fields. descriptions required it so called implicit solvent models that do not include the water are just not as well calibrated at the moment so we probably could not have the same level of. accuracy and detail in those simulations was these interactions between the proteins of the proteins with the membrane. Within the protein and fun have to be very well balanced and with water. By using force like protein relatively densely packed on the membrane. We in effect try to minimize the amount of computation time we spend on on water but as the person asking the question clearly realizes this is still in many ways look like a way, but I think at the moment it. It is largely unavoidable and as I also showed but with the protease functionally sometimes very important. Okay. Thank you very much again for this nice presentation and I think we have a coffee break now and we reconvene at 1015. Thank you. Goodbye. Bye bye. Hi. We wait a bit more but I see many participants are not in the shot yet. That's good. My presentation in share mode already or. I think you can start. You tried already so I think it's. I think it should work but you never know. You look experience with this type of. Yeah but you know I have this I have this bad karma that I always get a problem with the software updates exactly the same day I'm supposed to. Which is I mean it's remarkable. Me too the first day I was speaking. Yeah. No I think the zoom usually works right well I mean I think in this type of. Yes it's better or. No it's too much light I think. I think it's better with. We wait one more minute. Well it's a pleasure to introduce Professor Villekaila from the department of biochemistry and bio physics of Stockholm University and his talk will be elucidating mechanism of biological energy conversion. By simulations across scales. Thank you. Great thank you so much Alessandra for the nice introduction and thanks for the organizers to. First of all organize this workshop and and have it on soon despite of these these big challenges so I thought today and I would like to present to you. A challenge in the in the simulation world. A system that we have been working on for for a very long time. And basically the outline of my talk is the following so. I will start by telling about a structural revolution that has happened I think all over in biochemistry. I would then go on to to tell a little bit about the multi scale approaches that we use to study these type of of biological questions and then introduce to you a very interesting molecular machine the respiratory complex one machinery. That catalyzes a a fascinating 300 ongström long charge transport process. With with that we we we could we could start thinking on the motivation of this work and, at least for me, what I think is truly fascinating that all organisms irrespective of their complexity if they're very simple singular organism or complex beings like humans they of course need an input of energy to grow to adapt and and possibly move in their environment. And of course you couldn't do this without some input of energy and in nature this can take place in two two various forms either in form of light energy or in form of a chemical energy. And of course as you know any process in biology takes place by enzymes and also enzymes are the key catalytic players in catalyzing capture of either the chemical or light energy. What I think is truly fascinating is that that on a chemical level these process take place by elementary particle transfer by mainly coupled motion of proteins and electrons. And we have two main such cycles in nature we have the photosynthetic one where light energy basically drives oxidation of water molecules to proteins and electrons who take the electrons. Combine it with inorganic molecules like co2 and the carbon fixation process to build more complex organic molecules like sugars. And then you take basically these high potential energy electrons back and and combine it with something that's a good electron sink for example dioxygen and for that you need some proteins as well to restore basically get water back. So the at the at the heart of of of energy conversion basically have a counterplay between these proteins electrons and in light activated process lights light energy. So there's been a wonderful structural revolution in biology in recent years and and this has helped us of course understand also the the molecular machines or provide blueprints for them. And one one very important development is to cry you and technique which is is is today resolving almost standard wise structure set at the round tree young storm resolution but even better. And we now have all members of so to say the mitochondrial respiratory chain. We have also photosynthetic system but also many light driven ion pumps. To and this of course provides us a starting point to understand their their their function. But but of course it's not fully trivial to couple together the structure and the biological function and to this and of course we would like to zoom in and and we can do that thanks. I think to to to the very nice computational methods that provides a very powerful way to to ask mechanistic questions about the system. And this has been introduced many times on this workshop already. But maybe I just highlight a little bit that the philosophy that we are using in my lab to study the system so to to explore enzymes you of course need to treat the chemistry. And for treating the chemistry of course then the quantum chemical methods provide a very powerful toolbox and we use very much the density functional theory methods in in in my group due to the good scaling and and and and accuracy but also for some problems. We try to go to to more more so to say better methods correlated quantum chemical methods that we then couple together the QMM framework to methods that are good for exploring the face base of the system here we use widely atomistic molecular dynamic simulations but also free energy methods of various kinds that were very nicely also introduced. During this this this workshop and even even up to a so to say coarse grained representation of the system to get interesting information on that longer millisecond timescale. So how this could work is that we let's say starting from an experimental static structure run an atomistic MD simulation we propagate the system on some microsecond timescales in various states. And then we open up parts for quantum mechanical treatment for a structure and ensemble of structures for example here is an AGP molecule being cleaved by a proton transfer reaction and for this we can then use some methods for example to calculate free energy profiles. I don't know with umbrella sampling or something like that and then couple various specific conformational changes let's say in the protein to to differences in activity that we can read out from this free energy profiles. And this of course provides a very powerful way to to design experiments and validate your mechanistic in understanding experimental. So bio energetics which we're interested in in my lab basically as the question how do you from a given let's say a chemical reaction or light driven reaction very often in the membrane protein. How do you use this free energy how do you transduce it to transport ions for example protons or other type of ions across the membrane. Such that you prevent possibly kinetically the back flow of the ions in the wrong direction. And of course you need you need structures for this and and and here we rely very much on cryo in structures and crystallography and and of course the methods of computational biochemistry then provides us as as the details to ask how this could happen in practice. But what we also try to do very much here is is is based on the simulation design for example metogenesis experiments to perturb the system and test them by different bio physical experiments how this would perturb the activity. And we have also started using a little bit of an protein engineering and artificial protein design approach to ask and so to say isolate various catalytic part of the system into a simpler environment. And of course the the downside of this is that it's not enough that you only have supercomputers and a and a computer lab. But if you want to run experiments you also need to build an experimental lab and this is this is the picture of of the experimental way with in our lab that enables us to test some of the mechanistic hypothesis. So let's let's go to the biological background a little bit and and ask how does this enzyme this complex one enzyme that that I'm I'm like to introduce to you how does it power cell respiration cell respiration takes place by on on at membranes basically and what I show here is the inner mitochondrial membrane. And and it has this embedded membrane proteins in it. Basically what what what they do is that they take electrons from from from the food stuff that we eat and then transfer to these various centers and lower the basically release energy stepwise and this this this energy release is coupled to motion of proteins across the membrane. And of course this couldn't happen if we wouldn't have an electron sink and this is why we breed oxygen so most of the oxygen that you breed goes to to this process and that basically then takes these electrons from food stuff to to to drive the process. And of course this proton pumping across the membrane generates a a electrical gradient across the membrane and it's fascinating actually that across this third young strong thick membrane. You have a voltage difference of point two volts so it's it's massive basically the the the forces talking about on so small scales and then the proteins what they would like to do is is basically go back to the other side of the membrane and this you can then couple to for example synthesis of ATP molecules or active transport. And let's now focus on this first initial member of of the complex it's it's called complex one and it's a very big enzyme a PCET machinery couple couple protein couple electron transfer machinery. It's basically up to one megadalton inside in us humans and we now have have structures of this molecule from from several groups old in recent years. And it takes it then functions as an initial electron acceptor in this respiratory chain of both mitochondria and bacteria, and it transfers electrons first from a general cofactor and age. This happens by a concerted basically a H minus transfer, so coupled proton and electron transfer move to this aromatic ring systems, and then the electrons continue to a wire of of around eight iron sulfur centers to this quinone molecule over here and maybe you have seen quinone in in cosmetic products or something like that that's basically because it's it's a good electrons scavenger and it might kind of prevent you from from becoming wrinkly with with time. Good, and then basically the quinone is reduced to a quinone molecule here inductive site and and this fascinatingly couples the proton transport up to 200 long streams away from the active site so this is this is really a long range couple in process this is highly efficient, such that you can turn around this machinery and run it in reverse. You can use a pH gradient to to to extract electrons also for from a quinone molecule the other way around. And it's it has basically everything that there should be for microscopic like textbook thermodynamics there is a microscopic reversibility such that if you perturb one end of the molecule you will also measure an activity change for the quinone oxidative reduction. This is exactly how it's supposed to be, but of course the question is how this could take place. What could also say that it's of high biomedical relevance because mutations linked in this enzyme are related to half of all human mitochondrial disorders. And here is you can pick a favorite of the disease, it doesn't necessarily mean that the disease is caused by mutations in this enzyme, but but it has a link and and one possible explanation could it to be related to this is that under some conditions. Electrons leak out from this enzyme and they of course cause them reactive oxygen species and harmful effect in the cell. So what what we would like to understand is is how can this machinery basically have couplings across such big distances what what is the the mechanistic principle of such system. And I would like to divide this talking into key questions. I mean, first to ask how do the electrons enter this this machinery and and trigger the proton pumping where does the coupling between the electron and proton transfer process happen. How are protons pumped across the membrane would like to highlight some conformational changes. Related to this machinery and then if there is time I could also show to you a very interesting modular twist where this machinery is used in photosynthesis to do a very very important process over there. So since this is the workshop. I thought that it could be interesting for for for this community to to have some thoughts on how to model such a complex system. And and and basically I mean of course similarly to to any other maybe empty study or or computational study you need a you need a structure and we have worked with various type of complex one structures from bacteria eukaryotic but also photosynthetic systems. Soled in recent years. Then this is a proton pump so you need to concern yourself about the protons this this protein has more than 1000 titratable groups and we usually aim to assign protonation states by a Poisson Boltzmann electrostatic approach couple to a Monte Carlo sampling of of the different states to get at least an idea of of good starting conditions in in the simulations. Then basically we need to to model the the substrate quinone and build a membrane and for this we use usually a tree component membrane with P C P E and cardiolipin mimicking the mitochondrial inner membrane and then basically we need to build water and ions around this and then unfortunately you end up with the system with with around one million atoms in this. We are we're doing the modeling with the charm 36 force field and and we use regular water instead of of d2o that that seemed to be a very nice nice idea that the previous speaker talked about so we use age to vote. And tip tip trippy water molecules which which is not always the best of course in systems like this. In addition to to to not doing simulations only in maybe replica set for one system we try also to get an understanding of the various kinds of complex ones and there are there are the results species from bacteria and mammalian systems and and ease to to other variants of the system and we think that by modeling all of these you get basically you can understand better the functionality. And in order to to model part of this process which is the motion of process in electrons you of course need to be able to parameterize them also to describe them classically and to this and we have we have developed force field parameters for so to say the iron sulfur centers the the flavings then the age systems and and quinones in various oxidation states and these are very often based on density functional theory calculations. Where you basically then can derive these force field terms to be able to model this process on on on the classical times game. But then to to study the actual chemistry we use we use the methods of QMM that Alessandra very nicely introduced you the the the other day. And here we run very often with with hybrid functionals and and it can be basically playing MD simulations on a QMM level or then free energy calculations where you can use something like umbrella sampling to study this. And for this, I don't have the time to go into the details, but we have to introduce some kind of collective variables, for example on how the program would move along this type of water wire in the system. And what you can usually do a density functional level is to sample, say an umbrella sampling, some picoseconds per window, which is not very much. But of course, you can then supplement the thing with a classical sampling of the system to get represented the ideas of the dynamics. We also favor very much in in my group is is to use density functional theory based cluster models where you try to represent the key chemistry for the system. And why I like this type of approaches is that you can you can also study the system very well with with ab initio methods and and and try to do so to say good electronic structure calculations and we have started using, for example, the random phase approximations in in in recent years quite a lot to to to look at these processes and very often. There's a for example a free energy barrier calculated with QMM or with the FT you get not identical but but at least qualitatively similar pictures which which is after all what you're after to get the mechanistic understanding. So basically what what we then try to do is to model the so to say catalytic cycle of the system by by by running running dynamic simulations, injecting electrons into the system and see how off off off this. Very good, I seem to have some problems with my Internet connection but interrupt me if you can't hear me well. So let's let's have a focus on what happens in the machinery and the first realization that we have had over the years. We and and and other people working in the field is that that the the the proton pumping process is activated by the quinone reduction process when electrons start coming into the and of course to be able to treat this you need to have the QMM approach. And this we this I introduced you a little bit before before, but but the other other thing that you need to think about is how to model electrons which is not fully trivial when you have systems like iron sulfur centers, because each iron can carry either for five spins, which you can then couple anti ferromagnetically to six unique combination so this is basically combinatorics by why why why this matters is that you can have big energy differences between the unique spin states. So up to 56k kals per mole, so it is rather important to prepare the system in the chemically correct state, and this cannot be taken into account purely by electrostatic effects. And then we need to have ways to to to drag the electron put the electron where we want for example on the on the metal center or on the quinone center and and and transfer between the different sites and and look at at the process of what it leads to. So this is the result of such a simulation so basically you can see here the spin density disappearing from the quinone site. So basically there is an electron motion and this couples to to movement of some nucleus and protons together with this so if I zoom into the quinone center. You can possibly see in your on your screen that when the electrons come in, you basically see that protons are being pulled from nearby residues to this quinone molecule forming effectively quinone. So we have a tyrosine and a histidine residue and upon reduction, we basically form this type of species. And at the same time because of the charge relocation, you basically break or we can decide and pair between the histidine and the aspirin. So with QMM calculation you can look at the system only on a very limited time scale. But in order to see what happens on a longer time scale you can run classical atomistic simulation and and we discovered something something quite interesting. So prior to quinone reduction, there is zooming into the membrane domain. You see basically that that there are here is here is this aspirate that the active site and some other carboxylate that the site at the membrane domain and basically upon quinone formation. You break this ion pair, which leads to some conformational changes of charge groups here in in the membrane. So it's kind of interesting that you you you have this this consecutive plus minus charges in organized kind of in a wire so if you tickle the first you get the response kind of that at the end. And this basically leads to big pka shifts that can be indicative of a proton optic reaction. So how could we test that this is relevant in for for the real system and one idea is of course that you can kill some of these these these charge residues and and and do a mutagenesis experiment for experimentally and this is this is what what we did so we reconstituted the complex one in proteolyposomes gave electrons with an age. And basically then we could measure how produce are being pumped inside this proteolyposome vesicles by having a pH or are in this case a charge sensitive die and you can see that there is a big difference in these traces between the wild type system and when we have removed this charge in in in the system. And and what's quite interesting is that if you look at the other residues involved in discharge communication is that many have a very big effect on on on the activity, and this is also seem to be a site related hotspot for human disease related mutations. Good, so we have formed now quinoal. And then what what what we think is is happening is that the quinoal motion out from the cavity is giving the energetic kick for the proton pump. And the reason for this is that when we when we look at the quino potential redox potentials in the active site it does a very low potential around minus 300 millivolts or so. And, which means that when it gets electrons from a electron donor, for example, an age, it is almost equipped potential and there's no energy being released. However, quinoa is when they're in membranes are a much, much higher potentials, which means that when you go from this minus 300 millivolts to plus 90 millivolts, you have a release of the energy so already thermodynamically this must be the kick. So what we did is is that we we wanted to study this process further. And you can see quite nicely so so this is. We studied the process with with with free energy calculation here is now an umbrella sampling simulation, where we will we basically started identifying local minimas in this site and the first one is just the site I showed you to be for before. But then there's some very interesting structure in in this so this is basically how the quino moves moves out of it. It's cavity towards outside. You see a local minima that matches quite nicely with an ebiter bound form of this enzyme. And then you have here quite quite much further out in the membrane part of the system to very, very interesting local minima. And we think that these are important stalling sites for the quino. And what is quite quite quite nice is that we were recently able to to resolve this this the second binding site by by cryo em where we can see an inhibitor binding to this predictive site. So I think that the provides a very nice validation to our our free energy profile. Okay, so we have we have we have produced quino. We have moved it to a second binding site, which is located here in the membrane. And now we would like to understand how does that trigger the proton pumping process across across the membrane. And when we run in these simulations, there is there's something quite interesting happening and now I show you just the the the kind of side view of the system and you can see that during the dynamic simulation you have a these these water wires at at various places and and and just if I would show you here a time average picture of of them, you can see that they start forming at various sites basically for specific regions in in the protein. And this seem to be similar for a bacterial systems for when we do simulation on the on the much larger mammalian complex one, but also photosynthetic variant of the complex one. And these water wires form basically at symmetry related locations. These format antiporter like subunits and basically you have, you have an internal symmetry such that that they they form around a broken helix, where the input and output channel are related by by by symmetry to each other. So if you would flip this 180 degrees and by rotation, the input and output channel come come to the similar location. And what controls this hydration process is a charge residue at the end of this broken helix. And you can see that upon hydration there is a small subtle motion of this, this, this broken helix element, which I guess is why it's broken helix element. And you can see that if you, if you then during the simulation deprotonate this residue, then basically you get, you get the wedding of the channel as well. So we think that the charge at the bottom of this creates an electric field that pulls in water dipoles. So to minimize the energy and basically when you do not have this discloses the gate. So you could say that it's a little bit like a field effect transistor where the gating charge regulates conduction properties of the device. So this is of course we've seen similar behavior in many other systems, not only in complex one but inside the chrome sea oxidase where a reduction event causes a similar wedding behavior but also in light driven systems like in this sodium pumping where where the charge changes causes differences in hydration across the membrane. So now the interesting question how are these proton transfer reactions coupled to each other. And there is a very interesting internal symmetry in the system and this is basically that you have this charge residue this the central lysine residue, which I talked about that regulates the hydration state of the channel. And then basically before that you you have close in close proximity and I am pair with this I am pair is in a closed state. This this this lysine prefers to be in a in a protonated state. But we can also see that there is another possible state where where this I am pair can can can open up can dissociate and shift and form contact with the previous subunit. And this state becomes of course favored when this lysine protein residue gets protonated and that is to minimize the charge repulsion between the groups. And this you can see quite nicely in free energy profiles in classical free energy profiles. But this also affects very strongly the proton transfer energetics itself so you can see that when I am pair is open you have quite a high barriers and endergonic proton transfer processes. And when you open up this I am pair the proton would like to like like to go. And this is basically how how the proton motion would happen it goes via these water molecules to an intermediate a history in residue and then from the history in residue. It continues onwards to the terminal recipe over here. And we have some snapshots where we can see this even spontaneously happening in QMM MD simulations when when the state is open. So you could use so maybe this this history in residue here that is on the pathway. So it's kind of interesting that it's basically an important stalling site. So we thought that what if what if we mutate that what if we take it away and introduce there is an alanine residue and we did the same pumping experiments from here. And what you can see is that it's it's you see a very drastic decrease of the pumping effect by by introducing this mutation. But you can also see a quite significant changing in the quinone oxidative reductase activity and this is quite interesting because the quinone site for this subunit that we studied here is over 100 long strands far away. So it's a it's a it's a true coupling effect in that sense. What is also it fascinates me really really in in this system, and I think it's general for many other systems is that, of course, the water molecules are key for transferring the proton. Without them, you would have too long gaps to to have a effective proton transfer barrier. But water molecules are also key into causing the proper dissociation dynamics of this ion pair. So only when the channel is in a wet state, you basically can open up this ion pair with some reasonable energetics. While you have a very high barrier in the dry state of the channel, which I think makes makes sense also from an electrostatic perspective. So we could put all of this together basically to say that the quinone reduction chemistry by these local residues it activates the proton pumping process. The quinone has a has a second binding site in the membrane domain, which brings it closer to the proton pumping subunits. And then we have this long range signal propagation processes which take place by a channel hydration process by ion pair dynamics and then by this couple proton transfer process. And of course the proton transfer wouldn't be possible by with without controlling the hydration state of the system. So we can we can draw up a mechanistic picture based on this electron transfer to quinone produces a quinone molecule. The quinone moves in to the second binding site. This initiates the proton transport cascade moving the proton along this membrane domain close to the first ion pair. The first ion pair dissociates because that is you get the accumulation of charge dissociated ion pair pushes the proton laterally to this end of this yellow subunit. It opens up the next ion pair. You push the proton, you open up the ion pair and you have pushed the proton to the lateral end. Now there is no more neighbors to transfer the proton. So what happens is that that the proton is most likely going to lead the system. This site is also very close to the exit site at least in our MD simulation. And when the site leaves, there is a PKA increase of the central lysine residue uptake of this proton would close again the ion pair. But this in turn destabilizes the proton loaded here at the interface of these two subunits. So the proton is being pushed outwards in here. You uptake a new proton, you close the ion pair, you release the proton, you take up a proton from this yellow subunit, close the ion pair and release the last proton across the membrane. So it kind of transfers so to say back the signal which is then coupled to the release or that is what we have proposed. And then basically the quinol needs to be replaced by a fresh quinol to restart the engine. So with these things that I've showed to you includes hydration changes, some subtle motions and healers and other things in other structural elements and changes in hydration state. But there is also a very interesting confirmation, larger scale conformational changes involved in this machinery that I thought could be interesting also for this community. So when we run MD simulations on this system, based on the cryo-em result structures, what you can do is that you can look at the global motion of this system and you can group them together by PCA or other type of analysis. And you can see that there are two type of motion. There is basically a bending motion of this machinery and then there is a twisting motion as well. And these are quite large scale motions happening during the dynamics. And what is quite fascinating is that if you compare these motions taken from simulations to what is being resolved in cryo-em particle classification is that you have exactly similar type of motions in those two states. And in the cryo-em these two states, these two kind of motions have been linked to the active and deactive form of this enzyme. And what happens then if you drag the system along this twisting and bending motion, these are large scale motions, so very rare events and you need to apply special techniques for this of course, you see that you have a strain effects accumulating a very specific locations of the sublimates. And at some of these locations there's been quite recently observed conformational changes in certain side-chain positions and we think that these are also very important. And one process where we can see this coming into action is basically when the quinone moves upwards and downwards its channel. So in this process complex one seems to undergo first a twisting mode and then a bending change. So there is also larger scale thing is happening, whether it's the course of the process enabling the electrostatic or vice versa. We cannot of course comment at this point, but I think that it's important not to forget about those events as well. Alessandra, do I have a time to show one last quirk or is my time up? My time is up for the system. That's totally fine. So I think that I had time to introduce to you the important thing, maybe just out of interest and you can look at the paper. We resolved a fascinating variant of the system where new models have been placed in and this is used by cyanobacteria to actually concentrate CO2. And this is a very, very important process in photosynthesis and I think it's fascinating that you use basically the mitochondrial enzyme to do this. And this is a quite nice interplay between cryoen structure resolving and simulations and then some biochemical work. So I won't show you this. You can look at the paper, but I would like to maybe put the few summary points of the thing. So I hope that I convinced you that by integrating computational approaches with maybe bio-physical techniques, you can get a very powerful way to ask mechanistic questions of the system in a very broad range of time scales and spatial resolutions. I described to you how we think this long range redux signal propagates around this system. I showed you some evidence for the substrate chemistry, how we think it takes place in the system. And then I presented to you a mechanism that involved a forward backward propagation of a pulse in this system. And with this, I put the blue picture. This is where you can see that we are organized in the shape of complex one over here. These are the people from my group and also some previous members and some funding agencies. So I'm happy to answer any questions you might have of this presentation. Thank you so much. Thank you very much for this very nice and interesting presentation questions. So I start maybe with the question myself. Can you a bit comment on how you model the electron transfer process that you showed at the beginning? Yeah, right. So what I showed here is involves the very last step of the electron transfer process. And what we done there is basically various type of models. So one is that you should basically the system start looking at the dynamics by QMM simulations. And then there is in these states that we have studied, there is such a big driving force that the electron want to combine. And we can actually see spontaneous combination of the protons and electrons in the system. And what's kind of cool is that it's reversible. We can also see in trajectories the reverse mode of the protons going the vice versa and electron jumping back to the electron acceptor. This is one way of studying the system. But then what we have also done is to go in a more so to say classical way of looking at the electron transfer, trying to determine redox potentials, reorganization energies, coupling elements more in a traditional maybe quantum chemical approach, which then allow you using some type of, let's say, Marcus theory to get an insight in what would be the timescale for the overall electron transfer event. And I think that that is a very interesting question by itself, but from the mechanistic point of view, what triggers the thing is when the electron reaches the queen of, right, the queen of. Right, so I think that that's of special interest, but then, but then looking at the electron long range electron transfer process is of course pretty cool as well, I think. Okay, so. Don't be shy. There are lots of compliments, including one. Thanks a lot, sir, for the very epic presentation. Excellent. Thank you. Hi, could I ask a question. Sure. Thanks for a very nice talk. I'm just wondering how the sort of the degrees of freedom of the membrane couple into this because it's a membrane embedded protein so do you get a lot of like, perturbations from the lipid molecules around the protein. Thank you for asking that question. So, it is in fact, I wonder, can I put a do I have a spare slide on this. So, the membrane is extremely important for this process. And it is in fact, so let me see, do I have it here that one particular lipid is of key here, can you see my presentation mode. And that is cardiolipin. And it seems that that cardiolipin what it does is so so we were, this is coarse grained because we can't address these on on a full atomistic timescale in how they associate. One atomistic simulations also in here, you can see that cardiolipin associated with very particular places in in the system and they seem to match cryoem structures quite well. But what this cardiolipin does is that it, it actually affects the motion this global motions in these exactly this twisting bending motions, and it is such that this is actually, this is the slide that the twisting and bending coordinated twisting bending happens when we have cardiolipin in the system, they're very coordinated, we don't understand exactly how, but but somehow the cardiolipin affects this, and when we remove cardiolipin from the system, let me see if I'm here, you can see, they're all over the place. And this, this is data from one, one millisecond off off, yes course grain simulations but but but anyway, so so we have quite quite quite a lot, a lot sampling in here. So you're absolutely right that that the lipid molecules are somehow very very important at least for the global motion of of this system. Did this elaborate this is answer your question. Now there is a question from the audience there from the chat. Can you please explain more about the interaction between electron and proton and phonon. Yes, or right protons in these these cases right. So, so basically the, the electron proton coupling. I mean it happens, the, so to say, at two specific locations where they, they combine to the same chemical group, and that is first at the top site, the NADH site NADH FMN, and there you basically have reduction couples to protonation event. The second site is at the quinoa, quinoa to quinoa, right. So to electrons react with the quinoa and two protons are needed, and that basically forms them, them, them, them together. Then, all after that is basically proton, proton transfer processes. These local charge shifts induces conformational changes that increases the pKa values changes to hydration state of the process, and then that triggers in turn this protonation cascade. So then you have basically decoupled the, the, the protons in electrons they don't move to the same sites anymore, but they're tightly coupled right, and that you could see, for example, by this mutation experiment you do a transition hundred electrons away from the groups involved in proton transfer, you can see a change in the activity meaning how electrons come into the system. So they are, they, they, there's the thermodynamic coupling there, which is, which is extremely tight. Another question, I have a little question about the ion pair, why is the energy profile with the dry and wet cases not symmetric around the zero point. This is a question energy. Yes dry and wet cases, there, there is a shift so I think that this is related to what is our reaction coordinates right. Composed of a distance difference between different sites, so I think that this, this shift basically in distances is related to how we have defined the dye and pair but the asymmetry, of course, is, is partially also due to the fact that when we have a hydration, it's it's not exactly the same structure you could see for example that some helices move, there are some conformational changes. And that is also why some distances between the groups in fact are not anymore the same in the wet and the dry state. And I think that this is this this probably relates to also some local asymmetry in the in the free energy profiles. Okay. Thank you very much. I think there are no more question. Thanks for joining this event and for the nice presentation again for inviting me thank you. Welcome. So we have the afternoon session at for 2pm sorry, and we reconvene at that time. Thank you again bill. Bye.