We're excited to start this session where we will hear about the latest in insights we're gaining into the structure and function of ion channels and ways to manipulate them, ranging all the way from an atomic scale all up to the in vivo level. By the way, I'm Saivya Raghavan from Ohio State. My co-chair here is Perez Radwanski, also from the same institution. Our first speaker whom we'd like to invite up is Krishna Chintalapudi, a colleague also from Ohio State. He's a structural biologist who is going to tell us about insights into the atomic resolution structure of the cardiac sodium channel. Good morning, everyone. Thanks for the opportunity. I'm going to talk about sodium channel structured function studies, basically the new insight that we gained in the lab in the last five years. So as you all know, there's different subtypes of NAV channels. I'm gonna talk about NAV 1.5, that is present predominantly in the heart. And NAV 1.5 is mainly responsible for cardiac action potential, triggering myosite contraction and circulation of blood. I don't need to give more introduction on the APs to this audience. But NAV 1.5 plays a major role in this action potential, mainly in the depolarization, or the rising phase of the action potential. When you look at the NAV 1.5 channel, it's the topology is kind of, it's a very huge protein. Very huge protein. It has four domains, the name from one to four. And each domain is connected by various linkers. These linkers are set in length, like they can be 100 MN acids, 200 MN acids. So they're huge linkers. And I want to pay attention here, the linker that is present between domain three and domain four, it's called 3-4 linker. And this motif, IFM motif, or inactivation gate, or inactivation particle, this is present right at the start of this 3-4 linker. I want you to pay attention to that motif, because I'm gonna talk quite a bit about it. So classically, sodium channel present in resting state, the resting membrane potential, where you have the activation gate that is blocking the sodium channel, and blocking the influx of sodium ions. When activation gate is removed, you have influx of sodium ions, which we call basically open state. And then when the inactivation gate blocking the pore, it can be in two phases, fast and slow inactivation. And the influx is also blocked here. So you have resting closed state, open state, and inactivation state. So far, from the structural perspective, which I'm very interested in, structural function studies, so what we know is, there are like eight to nine structures of sodium channel available, and all of them, you can categorize, of nine of them, you have eight of them in the inactivation state, or the intermediate inactivation state. And only one structure from Bill Cartel's group, that is solved in open state, or mimicking the open state, using triple Q mutation in the inactivation gate. So what I'm interested in, why do I need to solve more structures when there are already structures, right? As you can see here, now 1.5 interact with so many interaction partners. My postdoc call it's real estate of sodium channel. So yeah, it has quite a bit of interaction partners, and so there are no structures of sodium channels with any of the interaction partners available. And this kind of motivated us to kind of take this very interesting journey. And we also feel like, while we are reviewing the literature of structural function studies, we believe that there are certain concepts that need to be reinterpreted. What did that mean? For example, you have activation gate and inactivation gate. They are separated by more than 12 angstroms. And how do they actually play these dynamics? And that kind of like, given, I mean there are so many functional studies that explain this, but the structural studies really did not show these features properly. And still this is kind of, this needs interpretation. And in the open state, the IFMs, or the inactivation gate, is supposed to come out, and I wonder kind of, like when we see all the structures, you have this IFM, or the inactivation gate, that is docked in the hydrophobic pocket, and what is the energy barrier that is necessary to remove this IFM motif to come out of the pocket? That's kind of, I think one needs to look into, close into it, closely look into it. For example, if you look into the open state structure from the Bilkartel group, we don't see the inactivation gate. But what happened to the inactivation gate? Where is it? So we don't know that. And also, how does the C-terminal play a role in a function of sodium channel, in terms of full-length protein? That needs to be evaluated too. So we wanted to basically ask, like can we capture some of the cryo-EM structures with, I mean, starting now, 1.5 structure with interaction partners. But when we looked into all the structures, all the structures have truncated loops, so the linkers are missing. And the C-terminal also shortened, so you can't really capture anything if you don't have the linkers. So we started this with full-length protein. Using full-length protein, as you can see, we expressed it, but it's a tiny bit of it from tons of leaders. Okay, so now I kind of know why other groups actually avoided using full-length protein, because it also precludes achieving high resolution, because you won't get, if you have all these loops, you won't get high-resolution details. But we need that for structured function studies of now 1.5 interactome. So using cryo-EM, we solved, we started with now 1.5 alone without any interaction partner, for example. In this structural study, we captured three states, I mean, three structures of now 1.5. One without C-terminal domain, and two structures with C-terminal domain, with varying degrees of movement of C-terminal domain, for example, in this structure and this one. So we thought, like, okay, maybe our structure is also one of those intermediate and active state, because all structures are basically categorized into this. So when we looked into it, when we compared with one of the inactive state structures, our structure is dilated more than two angstroms in all the voltage-sensing domains, which is kind of surprising. We were like, oh, it's not what we anticipated. Then we looked into the activation gate diameter. What we have noticed is, you know, from the structure that is open state structure, the triple Q structure, we know the activation gate diameter is like 10 angstroms. So our structure fall into the same category, 10 to 11 angstroms, without modifying channel. And then we thought, like, how does the inactivation gate look like? So in a classical intermediate state, you can see the IFM are nicely docked to the receptor here. We don't see that IFM in the triple Q mutation structure, but we could see the IFM also still docked in the inactivation receptor, IFM receptors, so which is kind of surprising. Then we further looked into voltage-sensing domains, and we want to kind of understand, like, are there any differences in the gating charges of the channels? Why I'm asking this question is because so far, including the open state structure, all the structures have same confirmation of gating charge residues. In all the structures, and I wonder, like, if you have different states, you should have different GCs, and they should have different confirmations. So when we compared it, we saw significant changes in the VSD2 and VSD4, and I will show you animation how drastic the GC confirmations are from our structures to other structures. Then we also further looked into the structure of sodium channel, where you could clearly see there is different confirmation of 3,4-linker, and that actually allows the IFM or the inactivation gate to adopt a different confirmation. It's not coming out of the IFM receptor, but it adopts a different confirmation, which is because of four to six angstrom movement of the 3,4-linker. This linker basically allowing it to adopt a different confirmation. That is also quite new insight, which was not known before. And what we have also seen in our structure is a specific interaction between C-terminal domain and 3,4-linker. As I mentioned before, we saw two structures with the C-terminal domain. One structure is away from 3,4-linker, and the other structure is close to the 3,4-linker, and they move by nine degrees, and there is almost 40-degree rotation of this C-terminal domain. When we looked into the interface of the C-terminal domain and 3,4-linker, we saw these electrostatic interactions mediating 3,4-linker, now 1.5, 3,4-linker and the C-terminal domain interactions. In collaboration with Dr. Isabel Deschamps' lab, we did charge-level simulations. What we have seen is they altered the static state dynamics. Basically, this is also observed in previous studies. When groups have done with isolated C-terminal domain and 3,4-linker peptides, similar kind of behavior has been seen, and this interface is very important because of all the diseases involved in long QT, Brigada syndrome mutations are involved in this interface. So this is quite new. So I want to actually summarize the kind of movements that what we have seen, the different changes in our structure. If you compare our structure, open state to the inactive structure, you can see how the 3,4-linker is actually moving. For example, in the open state to the inactive. When you zoom into the domain one, look at the VSD one, how the gating charges are adopting conformations, which were not seen before. And you can see the same changes in VSD two. What we have seen is it's even more pronounced. The gating charges, the arginine are completely different conformation that were not seen in any of the NAV 1.5 or even other NAV structures. And the same, we have seen only a little movement in the VSD three, in domain three basically, but there are changes that are coupled with this movement of open conformation. So I know it's slow, but I have only a few minutes left. But if you look into the VSD four, this has huge conformational changes. I think it tells us like when the structure is moving from open to inactive state, you have different VSD sense differently and you have a sequential movement of VSD to enable the channel to open or close. And you could easily see the three, four linker adopts a different conformation when it's coming from open to closed state. That actually allows the IFM or the inactivation gate to kind of dock in the receptor, which was not seen earlier. And you can see also this novel interaction interface. This schematic, you can see if the C-terminal domain is interacting with three, four linkers, the channel is open. If it is disrupted, the channel is going into inactivation state. To summarize, we captured sodium channel in non-conducting intermediate open state because we believe that there are much more kinetic states in between. And we captured one of those states. And in our structure, despite the channel being an open structure, IFM is not coming out of the receptor. And the overall structure, the pore domain is dilated. And we identified partially activated gating charge conformations, especially in the VSD2 and VSD4. These are coupled with the C-terminal domain and three, four linker interactions. And we suggest that this interaction stabilizes the open state of the inactivation gate. And if you disrupt this interaction between C-terminal domain and three, four linker, you can affect the function of sodium channel. I want to acknowledge all the people who helped in my lab and Isabel's lab, and also Sarah Heisel's lab. Thank you very much for your attention. Happy to take any questions. We have time for one or two quick questions. Just for curiosity, did you see any changes in the selectivity filter and ion occupancy in the pore? When you show the volume of the pore, it looks different between the two. So that's a very good question, because what we have seen in the activation gate, the diameter changed, but in the selectivity filter, it has, it's similar to, close to inactivation state structures. So it could be that basically, while kind of like, it's very early phases of inactive state, we think that's similar to it. But when you see the activation gate, it's open. So you have to take the changes together, right? Like selectivity filter, central cavity, and the activation gate. But the selectivity filter is not like completely open. All right, thank you very much. Our next speaker is Jorge Contreras, and he'll talk to us about connexin biology in the heart. Okay. I'd like to start by thanking the organizers for giving me the opportunity to share our work. So this is the title that I was given for my presentation when I was invited to the meeting. And I was a little bit confused at first, because we recently identified surprising properties of connexin hemichannels. We showed that this hemichannel can function as transporters for molecules, and this is independent of the ion conduction. So basically, and this could be very relevant in terms of physiology, because for example, they can release ATP, or they can uptake molecules in the absence of ion flux. But however, still we don't have any evidence that this could be important specifically in the heart. So I thought that you guys invited me to talk a little bit about other work that we have done that is the role of connexin hemichannels in cardiovascular diseases. And this topic we have been working on for 10 years, but it's actually the first time that I'm at a meeting that focus only on cardiac research. Maybe it's still new for some of you. So let me tell you, as you know, connexin channels are very important for the electrical signal propagations in the heart and proper contraction. And they are located in these specific regions called intercollected disk. And this is how these gap junctions channels look. This is an eczema that is formed by six connexins, and six of these connexins in the heart are located in the perinexus. And then they move toward the center where they dock with the hemichannels of the adjacent cardiomyocyte, forming a gap junction channel. But the focus of this presentation is on hemichannels, the one that connect the extracellular environment with the cytoplasm. And just to give you an idea how these structures look, this is a structure for my postdoctoral fellow, Anand Burada. And this is something very interesting. The N-terminus region is forming part of the pore. And this is probably the region that play the key role in the transport of molecules and also the selectivity for ions. And in these structures, most likely it's ion channel conformation, not transported. Molecules cannot move through this structure because of the pore diameter is 4.5 Armstrong. But what I want to tell you is mainly the pathological role of these connexing hemichannels. For a long time, we know that if you have hyperactivity of these hemichannels at the plasma membrane, the cells lost the electrochemical gradient and small metabolites are important for cell viability. And we learned that in the field from human connexing mutations that produce gain in function of these hemichannels and are associated with multiple diseases, deafness, blindness, skin disorder. But also in pathological conditions, we know that the biogenesis formation of these gap junction channels are affected in multiple diseases. And you see an accumulation of hemichannels at the plasma membrane and with a reduction of the number of gap junction channels. So what I'm going to tell you is the role of these hemichannels in Duchenne muscular dystrophy associated with cardiomyopathies. And as I mentioned before, this is a connexing for immunohistochemistry for connexin 43 and it's located at the intercalated disc regions between cardiomyocytes. This is a tissue from healthy human. But when you do the same in heart from Duchenne muscular dystrophy patients, you can notice right away that there is an increase in the diameter of this cardiomyocyte hypertrophy. And also there is an increase in the expression of connexin 43. And you also can notice that connexin is located in the intercalated disc region, forming gap junction channels likely. But you can see that connexin is also remodeled in the lateral sides of this cardiomyocyte. And what we propose is that this remodeled connexin is forming hemichannels. And a little bit activity of these hemichannels will alter cardiac suitability. So I'm going to show you and try to convince you that this hemichannel activity is critical for the development of arrhythmias. So to study this, we went to the Duchenne muscular dystrophy mice models. And we observed the same that we saw in humans. There is an increase in the expression of connexin 43. There is lateralization of connexin 43 in this cardiomyocyte from the intercalated disc. And something that is very interesting is that we noticed this at four to five months old. So the mice develop the cardiopathology. You start seeing the arrhythmias at eight to 10 months. But this was very early in the onset of this disease. So we were trying to correlate this lateralization of connexin 43 with the pathology. And we knew from work from my colleagues when I was in Rutger that if you challenge these animals with isoproteinol, the animals develop arrhythmias and sudden death. Some of the animals die. And so we studied what was the role of this connexin emission in the development of arrhythmias. Here you see anesthetized mice, ECGs for anesthetized mice. If you treat this mice with isoproteinol, within 20 minutes, you start seeing AB blocks and other type of arrhythmias. In this case, it's a much more severe case of one of the most severe case that we see. This animal died within 30 minutes. But if we pre-treat the animal with emission blocker, connexin 43 emission blocker, a peptide that we injected retro-orbital, we can prevent the development of these arrhythmias. I told you that in humans and in mice, these animals show over-expression of connexin 43. So we target over-expression in the mice by lowering the level of connexin 43 using a heterozygous knockout mice. And we normalize the level of connexin 43. We saw that there is much less lateralization with respect to the MDX mice. And we also were able to prevent the arrhythmias. I also mentioned that in humans and in mice, you see this lateralization of connexin 43 in the muscular dystrophy mice. And this is a immunochemistry where you colocalize encahirin that doesn't lateralize and remain in the intercollected disk, but then doesn't colocalize with connexin 43. So we knew from the work done by Glenn Fishman and Paul Lampe that lateralization of connexin 43 in many pathologies associate with the phosphorylation of these three residues. And we examined by mass spec and with the specific antibodies, if these residues were dephosphorylated in the muscular dystrophy mice. And we found that there was a large amount of dephosphorylation of connexin 43. So we developed the mice with a phosphomimetic size for these three residues. And we prevented the lateralization of connexin 43. And we also prevented the development of the arrhythmia. So basically, if you prevent the remodeling or if you block connexin 43 in the channels, you can prevent these isoproteinol-induced arrhythmias. So what is the mechanism by which connexin-43-mediated arrhythmias upon cardiac stress? So we went first to examine the electrophysiological properties of cardiomyocyte. So we assumed that a little bit activity of these hemichannels, and we measured that too, can affect the action potential and the resting membrane potential. So we went to examine that possibility. So these are wild-type cardiomyocyte where we injected depolarizing currents in absent and present of isoproteinol. The only thing that you notice here is a little increase in the APDs. And we also found that in the MDX mice, there is an increase in the trigger activity and a spontaneous action potential that we call trigger activity. And this trigger activity was prevented by hemichannel blockers, GAM-19, and an antibody against connexin-43 hemichannels that we put in the pipette. And another thing that is very important is we found that the resting membrane potential of these cardiomyocytes in MDX are more depolarized by isoproteinol. And this was also prevented by the treatment with connexin-43 hemichannel blockers. So the disease model that we have here, and maybe I'm not going to have time to show you all the data that I have, but there is a dephosphorylation of connexin-43 that produced the lateralization. And if you challenge these animals with isoproteinol, do we produce opening of these connexin-43 hemichannels? And what we found, and I didn't have time to show you all the data, but basically there is increase in the membrane permeability because of the opening of the hemichannels, decrease in membrane depolarization, calcium overload. And this evoked long action potential and trigger activity. So we evaluated the mechanism associated with the increase in the activity of these hemichannels. And we found that s-nitrosylation of connexin-43 was a critical mechanism to activate these hemichannels. So these are biotin switch assays where we pull down all these nitrosylated proteins. And you can see in the MDX mice, it become hypernitrosylated. So we identified the residue that is involved with the activation of these connexin-43 hemichannels, and it was residue C271. We developed anokine mice to study this. And when we reduced the levels of s-nitrosylation of connexin-43, we were also able to prevent the arrhythmias induced by the treatment with isoproteinol. This is just the quantification of the arrhythmias in 24-hour. We evaluated also with optical mapping how action potential and calcium transit in the whole heart preparation are affected by the treatment with the repeat electrical stimulation. And you can see that there is a disruption in the membrane excitability for action potentials and calcium transient. And this was prevented by reducing the levels of s-nitrosylation in the anokine mice. In the Duchenne muscular mice with lower levels of s-nitrosylated connexin-43. This is just the quantification of the coefficient of variation for the diastolic interval here. So we also noticed before that there was this long action potentials in Duchenne muscular dystrophy mice, and these were recovered in the mice with lower levels of s-nitrosylation. So we think that, really, this is a little bit crazy, but connexin-43 hemichannel may play the role in the APD prolongation. And we see that with the blockers. We can reduce the APDs. And something new that we've found, too, is that after 24 hours, the mice develop myocardial injuries. So these are the levels of troponin in the MDX mice. And when we block connexin-43 hemichannels with the GAP-19 peptide, we prevent this development of the arrhythmia, and these are TTC staining. This is a scrambled peptide that doesn't do anything. But the GAP-19 peptide reduce the myocardial injury, and this is also observed in the no-kin mice with lower levels of s-nitrosylated connexin-43. So I'm rushing here a little bit. So another thing, this is just the quantification for the TTC, but we also were able to reduce the percentage of mice that are dying because of the cardiac stress induced by isoproterenol. So basically, just to finish here, connexin-43, when it's lateralized, and we have found this not only for Duchenne muscular dystrophy, but for other pathologies, too, when you challenge these animals with isoproterenol, there is an increasing activity of these connexin-43 hemichannels that produce the arrhythmias, and later on, also is associated with myocardial injury. And I just want to finish by thanking the people that did this work, Mauricio Lillo and Manuel Munoz. And with that, I will take questions. Sorry for the time. Thank you. Thank you very much. We have time for a couple of questions. Gerdand Remme, Amsterdam. Very nice data. Congratulations. I have a question about the fact that the dystrophin is normally located at the lateral membrane, and not in the integrated disc. And connexin-43, obviously, is normally in the integrated disc. So what is the trigger between losing dystrophin at the lateral membrane, and then having this secondary effect on connexin-43 happening at the integrated disc? Have you any idea about that? Yeah, so I have no idea what is the mechanism by the lack of dystrophin is affecting the lateralization, except for the dysphosphorylation of connexin. It also affects the microtubules formation. But we also have a model where we have no-kidden mice, where we use the phospho-deleted size. And in here, the microtubules are normal, relatively normal, not perfectly normal. But in those mice, you also see the lateralization of connexin-43. So you don't need really the, and we see the arrhythmias when we challenge the animals with isoproterinol. So you don't really need the specific things related to the pathology. If connexin is remodeled, and you challenge the animals with isoproterinol, there is this signal associated with nitric oxide that will activate these channels and affect the membrane suitability. And given the fact that the sodium channels, of course, are very much involved in both dystrophin and connexin-43, have you looked at what the contribution could be of sodium channel changes in what you're observing? Yes, that's a very good question, too. And one of the advantages of the model is that we identified the remodeling very early on, four to five months. And sodium currents are normal at that time. It becomes a little bit more affected later on. But yeah, I mean, we didn't see much changes that were significant. It's reduced at the lateral membrane early on already. Yeah, so when we measure the sodium currents at four months, we didn't see much changes. We see it later on. But we still see the lateralization of connexin-43. Thank you. Barry, London, Iowa. Very nice study. The arrhythmias you showed were mostly AV block. And although changes in ventricular myocytes could cause AV block, changes in the conduction system would be an obvious place to look. So did you do any studies directly on conduction system myocytes? Or do you have insights from the optical mapping on the role of the conduction system in causing the arrhythmias you see? Yeah, that's a very good question, too. So we are exploring that possibility. I mean, the data that I show maybe show mainly AV block. But we see all type of arrhythmias, too. Ventricular tachycardia, too. And it could be. I mean, we haven't examined specifically the distribution of connexin-43 in the conduction system. But most likely, it's going to be affected if it's dephosphorylated. Thank you. Thank you, Jorge. OK, thank you. Our next speaker is Dr. Igor Vorobiev from UC Davis. He's going to tell us about a clever new use for structural biology tools, not to unpack the structure of existing molecules, but to design new ones to manipulate the function of ion channels. Good morning, everyone. And thank you, organizers, for inviting me to give a talk here. So OK, I'll try to get it. OK, so well, the original title was Engineering Safe Ways to Target Sodium Ion Channels. But I'll talk more broadly today and focus actually more on potassium channels. But I hope I can cover a little bit about the sodium channels towards the end, as well. So first of all, I'm going to talk about the project I have been working on with a lot of wonderful collaborators at UC Davis and elsewhere, which is titled In Silica Safety Pharmacology. And it's a very serious problem, which was recognized a very long time ago. Like in the 19th century, Peter Meeuwisom said that poisons and medicines are often the same substance given with different intents. And, well, it's kind of true still today, especially in the field of cardiac safety pharmacology. For instance, up to 3% of prescription drugs, Ketelsam or Smetelsk, in cardiac toxicity accounts, up to 28% of post-mortem drug withdrawals in the US, for instance. Well, what is more serious, even, that up to 70% of small molecule leads have to be eliminated early in the drug development due to the potential of causing idle smear. And this problem affects drugs from multiple classes. And one of the most serious type of problem is the tarsal depoints or twist in the peaks idle smear, which is shown on the left corner here. And unfortunately, it can cause the sudden cardiac death. So to tackle this problem, back in 2005, two international guidelines were developed. So first of all, during clinical trials, drugs are tested for QT interval prolongation on the electrocardiogram. And during preclinical testing, drugs are being tested for the HERC channel inhibition. And the HERC channel is a key potassium channel in heart ventricles, which is responsible for the early polarization of cardiac action potential. So if you block this potassium channel, so you will increase the length of early polarization. And that would lead to this QT interval prolongation of the ECG. So and unfortunately, this HERC channel is a very promiscuous drug anti-target. There's a lot of drugs, like small molecule drugs, which can block it. And nowadays, since implementation of these guidelines, so no HERC blocking or QT prolonging drugs can enter the market. But not all of them are actually pro-adolescent, which would lead to the attrition of some safe pharmaceuticals. I should say that even grapefruit juice, like if you drink a lot of it, can cause some QT prolongation. So to tackle this problem, a few years ago, something called Comprehensive In vitro Pro-Adolescent Assay, or CIPA initiative, was developed. And this is like an international consortium, like with a lot of parties involved. But it doesn't provide a ready-to-go solution. It provides some guidelines how to tackle that. So it's one potential solution, which was developed at UC Davis by our collaborative team, is to use the in silica pipeline, or computational pipeline, to predict cardiac toxicity from atom to the glycerin. So the idea is that we start from the drug chemical structure, and going all the way to cardiac cell and tissue to see if we can predict that risk. And my work is focusing on the atomistic scale simulations, which are shown on the top right, top left. And then we collaborate with Professor Colleen Clancy, who is an expert in functional kinetic modeling, who is doing protein scale functional modeling, as well as cell and tissue scale modeling. And we published a study a few years ago in Circulation Research. So first of all, we focused on different confirmation, well, on the HERC channel itself, which exists in multiple conformational state, closed, open, and inactivated. We were fortunate that there was a cryo-EM structure of the channel published by Rod McKinnon's lab. So we did the molecular dynamics simulation of the channel structure to see what state it is in. And we showed that it does conduct ions under the applied electrical voltage. So therefore, we can go, OK, that's an open state. So then we focused on the, OK, wait a second. I have to advance the slide. OK. Unfortunately, this plot got messed up. But we did molecular dynamics simulation to calculate, like, field energy profile and diffusion coefficient profiles. And we were able to calculate affinities and on and off rates for the drug binding. And these parameters were used as parameters for the functional kinetic model. So we did this for two different drugs. One was dafetilide, which has high adless metals. And the second one is amoxifloxacin, which has low adless metals, as shown here. So then these rates were used, as I mentioned, as parameters for the functional kinetic model. And they actually produced very well the concentration-dependent inhibition of hyacinth potassium channel, as shown here. And then this protein scale model was inserted in the cardiac cell and tissue model to calculate the propagation of cardiac action potential in both space and time. So what's shown here is control case. When we don't have drug, amoxifloxacin is in the middle. And then the dafetilide is at the bottom. And both control case and amoxifloxacin showed a normal action potential propagation in both space and time. But then for dafetilide, we did show some pro-adless metal triggers, such as spatial depolarization gradients, and also early after depolarization. And then we did a follow-up study where we focused on sotalol, which is an anti-adless drug with a beta-blotting activity. But it exists as a mixture of two enantiomers, like D and L-sotalol. And only L-sotalol has this beta-blotting activity. D-sotalol was used in the failed clinical trial called SWAT, where the mortality was twice compared to placebo. And the trial had to be terminated prematurely. But D-L-sotalol, like racemic mixture, is still prescribed. And then we wanted to figure out whether the HEG block is responsible for this difference, or the beta block plays a major role there. So we did, again, this microdynamic simulations where we found that, for the HEG channel block, the affinities are pretty similar. And that was confirmed by the electrophysiological coin from our collaborators. But then functional kinetic modeling in Professor Kenin-Klansilov showed that, so if you consider HEG channel block alone, we don't see a difference between D and L-sotalol. But if we add the beta-adrenergic block, we do see a difference. So we don't have this pro-adolescent triggers or features like early after depolarization in L-sotalol. OK, but then here we only focus on the one state of the HEG channel. But we also wanted to put a drug bind to multiple states. And this is what we've done with Dr. Hwan Oh, who is the post-doc in Klansilov and in my lab as well. So we used alpha flow to predict also close and inactivated state of the channel. And then that was confirmed by Ryan Molecular Dynamics Simulation, which showed ion conduction for open state model, but not for the inactivated state model, which has a distorted selectivity filter. And then we also did drug dotting studies for multiple states, and then combined them together using the functional kinetic model data for the ratio of different states and the experimental conditions, which showed pretty good agreement with experimental data, compared to considering just one open state alone. And then we also focused on a different approach, which is called CILK site identification by ligand comparative saturation, which is a combination of molecular dynamics and dotting. And it's very efficient, and then it allows for exploration of the conformational changes of the protein. And during the dotting, because it doesn't focus on just like a little dotting of the drug, but it focuses on first exploring different fragments binded to the channel during molecular dynamics simulation, and then using these maps to dog the actual drugs, which takes just a few minutes per drug. And then this approach can be also complemented by using Bayesian machine learning to optimize the weight of those fragments and do a good prediction of, for instance, the IC50 values. OK. And then what we were able to show is that using two different states of the HERC channel, we were able to see which state it would prefer. And in most cases, it agreed pretty well with experimental data. But then we tried to build a logistic regression model where we said, OK, can we predict the actual risk of adenosine in this model? And the prediction was not very great in this case, because you can see a lot of red and blue intermix here. So there was a lot of false predictions. But of course, we have to consider not only the HERC channel, we have to consider a sodium channel, calcium channel, and other channel potentially to improve this prediction, which are known also to mitigate the HERC walk potentially. And if we do that for both HERC and calcium channel, yeah, we get much better predictions. And this is still work in progress. This is like one of our best models so far, which has a HERC channel, a calcium channel, and a sodium channel. But it's still work in progress, so stay tuned. And then we also focus just like I only have a few seconds left, but on not only the small molecule drugs, but like, for instance, peptide binding, which are also potential therapeutics. And then, well, this is for 1.7 sodium channel, but we are working on the 1.5 as well. This is for protoxin 2 binding, also for state-specific binding to different channel states and different voltage sensing domains here. And we looked at different affinities for that and identified clear residue, which might be important for state-specific inhibition. And this approach was further developed by Professor Yarafi Arvoid to develop a novel peptide, which would be specifically and more selectively block, for instance, this 1.7 sodium channel, which is crucial for the pain inhibition. OK, to conclude. So, I mean, well, I have talked to you about multiple things, but I think the most important one, that the drug-protein interaction is multiple cardiac ion channel and other protein channels are crucial to accurately predict drug-induced hallucinations and cardiac safety in general. OK, and I'd like to thank a lot of people from UC Davis and my wonderful collaborators, well, from both UC Davis and elsewhere, like from the University of Arizona, as well, and Professor Nipsey Minowatry and Dorothy Angen, who are here in the audience. I didn't talk about their work today, but, yeah. And thank you, everyone. And thank you for the funding sources as well. We can maybe squeeze in one quick question. In the interest of time, I'll just wrap up and move on. Thank you. Thank you. Our last speaker for the session is Dr. Nick Moyse from Ohio State University, and he'll present his work on how chronic changes in calcium regulate heartbeat generation. I'd like to start by thanking the organizers for the invitation. And today, I will be talking to you about how cardiac rhythms rewire the heart over the long run, and more specifically, about how calcium homeostatic feedback predicts atrial fibrillation remodeling, initiation, and progression. And what we're really interested in here is looking at the whole natural history of AF. So in AF, you have these multiple timescales working together in a coupled way. So you have the tissue substrate dynamics happening over days or hours, or weeks, months, and so on. And then you have depolarization waves, spiral reentrant location, and so on, that can be stable for days, hours, seconds. You have seconds taking up the time to depolarization wave to move across the tissue. And then you have the timescale of the action potential that is on the order of milliseconds, down to even gating that is a couple milliseconds or less, and of the calcium transients. And what's interesting about atrial fibrillation is that this timescale here, so from the action potential, feeds back all the way to the long-term duration. So you have this process that has been succinctly characterized as AF begets AF almost 20 years ago, or even more now, where you have fast pacing in the cells that drives tissue remodeling on the long-term. And really, this happens across all kinds of systems. So as a kind of analogy, you can think of a city where you have the day-to-day lives of the people living there. That's kind of the very short-term timescale. But then over hundreds of years, you have the evolution of the city. People get richer. People get poorer. The city grows, and so on. So you have this kind of coupled system that evolves over time. But here, specifically, we're interested about how does atrial fibrillation initiate and progress given this framework. So really, if you look into it, calcium is the link that couples B2B dynamics with long-term regulation. So you have fast pacing, either induced experimentally or some kind of disease process and so on, that drives intracellular calcium overload. And this, through this complex network of intracellular regulation, drives changes in both the electrical properties of the cells, ultimately in the amount of channels expressed, the magnitude of work currents. And in structural remodeling, as well, you have connexing lateralization, fibrosis, and so on. But importantly, this process is homeostatic. So from the point of view of the cell, it tries to maintain its own calcium homeostasis by ultimately decreasing action potential duration, decreasing calcium transients, so that it maintains its calcium levels. And the way we're doing this here is by modeling this in a computational model that was originally proposed for the neuron. And it was previously adapted by us to explain the sinoatrial nodes and emergent activity. And here, we couple the feedback model to a human atrial cell. And it's going to span multiple timescales, like I said in the beginning, from milliseconds to hours. And we assimilate up to 24 hours of continuous electrical activity. And really, just quickly, how the model works is that we have this inherent calcium target. That is what the system wants to keep level. And the cell is going to adjust its ion channels in order to maintain this long-term calcium target that we impose. And moving on to the results, so in the single cell, we see that progressively faster pacing drives remodeling. So at regular pacing, so that 60 beats per minute, 1,000 millisecond cycle length, you will have normalized expression of all these ion channels that vary. That is going to be around 1. So this is the baseline state. So everything moving from 1, it's going to be either decreased or increased. It's going to go towards more remodeling. And as we step up the cycling, we make it shorter and shorter, you get progressively more and more remodeling. And this happens over hours. But during this time, we simulate every single action potential, every single calcium transient. And as we see here, we have this calcium target that our cells, that's where they're happy at. And the system always maintains this calcium target. So the average long-term calcium, it's the thick kind of black line here, always goes back to this target, even though the pacing is faster and faster. And importantly, following the fastest level, it's also reversible. So going back to 60 beats per minute is completely reversible. And this is perfectly analogous to experiments where we have shorter-term pacing without leading to actual structural changes that are reversible over a couple of hours or days after the cessation of fast pacing. And in the most remodeled state, we see the characteristic changes that we see in AF. So we'll have shortened action potential duration, shortened lower resting membrane potential, and lower calcium transients. And really, it's this calcium target that controls the behavior of the system. So by having a higher calcium target, the cells will be more resistant, as it were. They're happy with more calcium, and they'll be more resistant to fast pacing rates. And the lower ones will be more reactive to fast pacing rates. But now moving on to the more interesting stuff in 2D tissue, we see that we start with a perfectly homogeneous 2D tissue. So we have a 2D tissue where all the cells have the exact same calcium target. So there's nothing to distinguish them. And we have a protocol where we have episodes of fast pacing followed by inducing a spiral wave. So as you see here, we pace in the corner of the tissue, and we induce spiral wave with the crossway simulation. And this is an early spiral wave, and it's going to be unstable because the tissue is not yet remodeled. So it's going to move around for a couple of rotations, and it's going to stop. And as remodeling progresses, so as you see it here, the ion channel remodeling progresses. And this leads to a stabilization of spiral waves because you have shorter APD, shorter refractory periods. So the spiral can stabilize and become permanent inside the tissue. And what's interesting here is that over a long term, because the spiral stays in one spot, you really get this remodeling underneath the spiral wave caused just by the spiral wave. So it will have kind of this bump, this ring structure, where actually the tissue at the center of the spiral wave is the least remodeled because those cells never really activate fully. And this, again, is just from a perfectly homogeneous tissue. You get this kind of spatial heterogeneity just because of the spiral wave location and the remodeling properties. Moving on to heterogeneous tissue, we wanted to basically replicate the full timeline of AF. So we wanted to see how it initiates, how it progresses, and the permanent late organization. So what we did is we have this heterogeneous calcium target map where the values are distributed between those I showed earlier. So we have some cells that are more resistant, some cells that are less resistant, as it were, to pacing. And this is very well established that you have heterogeneity between the atrial chambers, between right and left, and within the chambers themselves you have very different calcium transients, even in cells right next to each other and so on. And what we're doing here is we pace very fast at 100 milliseconds, but we don't induce artificially the spiral wave. We will see that they arise naturally, as it were, just from the heterogeneity. And we will have the progressive. We will basically capture the progressive nature of AF. So we'll have longer and longer episodes until they become permanent. And we'll go through these one by one. So the first part is obviously the initiation, and we'll have episodes of fast pacing. And at some point, the remodeling is enough so that the cells that are heterogeneous behave differently in tissue over a long term. And you have pacing in the corner here. And this will cause re-entry, and you have a re-entrant episode. But again, just as in the single spiral case, it's going to be impermanent. It's going to last a couple of seconds. And you can see it's traced here, and the spiral traces. And it's going to die off on its own. So you basically get intermittent AF, as it were, around these heterogeneities. But as the remodeling progresses, we see that, first of all, the tissue can sustain more and more spiral waves, and the dominant frequency of the activity increases. And you really have this biphasic response, because we have the short-term electrical remodeling and the longer-term structural remodeling. So we also have structural remodeling modeled as lower intracellular coupling between the cells. That is, again, dependent on the calcium homeostasis mechanism. And what we really see is that, at the very late terms, like very organized AF, you have this organization into these kind of tissue niches, where spiral waves are very stable. And again, we see the same pattern, where we have this bump in remodeling, where the underlying tissue is healthier, as it were, closer to normal. And this map here tracks the distribution of spiral waves in tissue. And at first, it's pretty chaotic. It organizes at the edges, where the gradient is highest, between the different cell types. And at the end, you're going to have these kind of very stable structures that form. And you can think of them as niches, kind of an analogy with metastatic niches, or rather, kind of these niches that form. And these form completely emergently from the system. And looking at these more closely, so this basically tracks spiral wave location in the tissue. So they're going to be all over the place. But really, you have one of these that are very stable, kind of highlighted in orange. And the local activation, so this is the frequency of activation here. It's going to be very fast, just around these very stable spiral waves. And also, we see that the voltage and the calcium are very decoupled, so another hallmark of atrial fibrillation. So we have not every voltage beat has a calcium beat, as it were, or a calcium transient. And looking at it in motion here, you have this niche that forms here. You can kind of see the bump in remodeling. So that's, again, where the tissue is healthier. And you have this very stable spiral wave that's always there. And the calcium is very chaotic. You can obviously see it here. And this is just one particular tissue distribution. So we tested multiple of these maps. And some of them just have intermittent pacing. So you're never going to have permanent atrial fibrillation. Some of them have permanent reentry. And some are completely resistant to reentry. So this also captures the heterogeneity we see in patients and so on in actual biological systems. And in conclusion, we have integrated this long-term model of activity homeostasis. And this allows us to join multiple directions of research, tachypacing, electrical and structural remodeling, and intracellular regulation into a single coherent theoretical framework. And we see these very interesting emergent behaviors where we have spontaneous initiation of reentry, the formation of tissue niches. And we captured basically the full history of AF from remodeling to initiation and progression. And for future work, this would easily be adapted to all kinds of things that happen over long timescales, such as development or other diseases, heart failure, and so on. And it also provides a mechanism for cardiac memory, which has been discussed in the previous years. And the overall take-home message is that the heart is not a static thing. It's alive. And it's fundamentally continuously modeled by its inherent activity. And with that, I'd like to thank my lab and collaborators funding sources and the supercomputer access for the simulation. And thank you. And any questions? And this is the paper that has just been accepted. Thank you. We have time for one question from the audience. Here, I'll sneak one in. Nick, really nice work. My question for you would be, if you imagine a tissue that has undergone some level of remodeling based on the arrhythmic circuits running in it, and then, say, you apply some kind of pharmacological therapy and put it into the arrhythmic rhythms, can you start defining some kind of manifold that describes the degree of remodeling and the spatial heterogeneity of it that would still support further arrhythmia in that state? Yeah, so the simulations basically give you a continuous, you have all the states between the healthy and unhealthy. So basically, what you would do is try it at different points and see where the drug stops working and so on. And another complication, I mean, obviously, there's a ton of complications in actual systems. Some of them is not reversible anymore. So in our case, it's more or less perfectly reversible. But obviously, AF is not perfectly reversible because you could just defibrillate, and then you'd go home and never have it again. So yeah, you can define. That's actually what we're thinking about doing, defining different time points and see how sensitive it is to intervention. Thank you. With that, let's thank all the speakers, and thank you for your attention.

ion channels

cardiac sodium channels

NAV 1.5

cryo-EM

connexin hemichannels

cardiovascular diseases

Duchenne muscular dystrophy

arrhythmia

cardiac safety pharmacology

drug-protein interactions