 Hello everyone and welcome to the second webinar in the Q&A best practice workshop of this year. This is the final webinar in the Q&A best practice workshop, however, it's not the end of the workshop, but I'll say a bit more about that at the end. So to introduce today's speaker, today we have Professor Karma Rovira, who is based at the University of Barcelona in the Department of Inorganic and Organic Chemistry. She's also part of the Catalan Institute for Research and Advanced Studies. Her group looks at enzymatic reactions, for example, in particular recently heme and carbohydrate active enzymes, as well as ligand protein interactions, in both atomistic and electronic detail. Using a variety of computational approaches, classical and ab initio MD, for example, carbonyl and very relevant for today, QMM. And all to also contribute to design more efficient enzymes and new drugs. So Karma has been a recipient in the past of a Marguerite Fellowship, which she performed at the Empire Stuttgart together with Michele Barinello. She held a Ramon de Cajal Fellowship as well, received in 2019 an E.M.A. Fisher Award from the European Carbohydrate Organization. And just last year was awarded a European Research Council Synergy Grant to work with colleagues at the University of Leiden and New York on glycobiology. So with that, I will hand over to Karma. So thank you, Arno, and thank you also to the organizers for inviting me to deliver this talk. This is a very interesting series of webinars on QMM simulations. So I'm going to talk about the modeling of catalytic mechanisms in carbohydrate active enzymes, something that we do with QMM molecular dynamics methods. And so this is about carbohydrates and I'm going to focus specifically on one family of enzymes that are the ones that catalyze the cleavage of the glycosidic bone in carbohydrates, the so-called glycoside hydrolysis. The glycosidic bone is the carbon-oxygen bone that joins two sugar units in a carbohydrate. So I'm going to focus on the enzymes that catalyze the cleavage of this one, because there are many types of carbohydrate active enzymes and there's no time to cover them. So this is about carbohydrates. Carbohydrates are very important for our life. They account for 50% of our daily calorie intake. So they are our biological food and also the primary form of storage and energy consumption in higher organisms. It was early thought that carbohydrates play mainly a role of energy storage like glycogen in our body and our liver and in our muscles. This is our glucose reservoir and also the starch in the many of the food that we take. This is for energy storage. It was thought also that these were the major roles in energy storage and also structural support like cellulose in plant cell walls. However, it is known nowadays that carbohydrates play many other roles in nature. In particular, there are many carbohydrates that are on the surface of all our cells. They are attached either to the lipids of the membrane or the covalently attached to them or also they are attached to proteins of the membrane and also many of our proteins are glycosylated. Even the SARS-CoV-2 virus contains many carbohydrates that are surrounding and covering the spike protein. They form a dense coat of sugar here and the dynamics of these carbohydrates were recently uncovered in the very nice world by Romi Amaro and Elisa Fada very recently. So these carbohydrates that are on the surface of all our cells, they play a major role in cell-cell interaction processes. They are like antennas that the cells use to communicate each other, to talk with one another. And of course, by doing this, they also protect us from invaders like viruses or bacteria. They protect us from them binding and infecting our cells. The shape on the three-dimensional structure of these carbohydrate antennas, the so-called glycoforms, depends on the amount and composition of the sugar units that they contain. There are several essential sugars, and this is quite important, for instance, our blood group, whether we are AB or O, it depends fully on the type of distribution of our carbohydrates on the surface of our red blood cells. It quite a lot depends on whether there is or not an enocetylgalactosamine sugar, which is a derivative of galactose with an enocetyl group here, and in which is the distribution of this sugar. So when this type of carbohydrate antennas, these glycoforms on the surface of our cells are not working properly, they are not assembled properly, because maybe when sugar unit is missing or there is too much of it, there are many problems that can occur. Many diseases like inflammation, autoimmune diseases, allergies, etc., infection, and the responsible for it many times is one of the many enzymes that are in charge, that are responsible for catalyzing the cleavage or the formation of the many glycosidic bonds in each of these carbohydrate antennas in the glycoforms. And this has brought a lot of attention to carbohydrate active enzymes, that several of them are important therapeutic charges to treat a large number of diseases, and they have also other applications like diagnostic markers and also in the industry, in the biofuel food, paper industry, etc. There are several classes of carbohydrate active enzymes, but two of the major classes are glycoside hydrolysis and glycoside transferases. Glycoside transferases break down carbohydrates into pieces, because what they do is they catalyze the cleavage of the glycosidic bond. While glycoside transferases do the opposite, they catalyze the formation of the glycosidic bond, so they form new carbohydrates. In our group we are interested in discovering new aspects of these enzymes, how they work, in particular how the substance binds to the enzyme, and how catalysis takes place. And by doing computer simulation, we have access to properties that sometimes escape experimental probes like transition states, short-lived intermediates, conformational itineraries, electronic states, etc. And all these can contribute to understand the factors that control catalysis and can contribute to engineer these enzymes for other purposes. And that talk, I'm going to focus on catalysis on glycosidases, so in the enzymes that catalyze the cleavage of the glycosidic bond. Glycosidases are used to maintain catalytic mechanisms, this is in all textbooks, depending whether they retain or they infer the configuration of the anomeric carbon. That's anomeric carbon, carbon number one, the important one, because it's the one that forms the glycosidic bond. In a retaining glycosidase, the glycosidic bond is beta, so equatorial, and now it remains equatorial, from beta to beta, this is a retention of configuration. And if it's an inverting enzyme, then it goes from equatorial to axial, or the other way around, but it inverts the configuration. There are two in the mechanism, there are two protein residues that do the main work, they are based on carboxylic acid sidechain, this is, although there are some exceptions, but this is the most of the glycosidases used to residues based on carboxylic acid sidechain. The one is initially protonated, the so-called acid-based acid, and the role is to protonate the living group while the glycosidic bond is cleaved, and the other is unprotonated, nucleophile or general base. In the case of retaining glycosidases, the nucleophile attacks the anomeric carbon while the glycosidic bond is cleaved, and in the case of inverting glycosidases, it's a bit further away because there is a water molecule in between that is the one that attacks the anomeric carbon, and this water is the protonated by the general base. This is the typical mechanism, the typical picture that you see in textbooks. With the reactive sugar, the one that is before the cleavage point, in a chair conformation, a relaxed chair, 4C1, carbon number four is above the plane, carbon number one is below the plane. However, this is not the case anymore. It is known nowadays that when the carbohydrate, when the substrate binds to the enzyme, it binds in a distorted conformation. In particular, the reactive sugar, not the only ones, not the only ones, just this sugar here, distorts to a bolt or skew bolt conformation. This is how it's found in high-resolution crystal structures. So in bones, there are several types of bolts, skew bolts, half chair envelopes conformation, but it's not a relaxed chair as it will be for a sugar in solution or a saccharite in solution. This is known from the 90s, from the end of the 90s, and it's now well established. It was early thought that already that this had was a functional feature, essentially because this distortion places the living group in an axial orientation and that facilitates the nucleophilic attack. We've been interested in this problem for several years. And what we found in our first study in this field 12 years ago, when we started doing application in glycobiology, working with 1314 beta gluconase, what we found is that indeed the distortion makes the reaction easy for this glycosidase because it lowers the activation and also the reaction because the distorted substrate assembles the transitional state of the reaction. In the transitional state, the substrate is quite distorted. This is an oxo-carbenium ion. These four atoms are in a plane. And the conformation that one observes in the Megaelis complex is very similar to the one of the transitional state, not just structurally but also electronically. There are charts being developed at the anomeric carbon compared, if you compare a distorted with an undistorted substrate. And that lengthens this consequence that the glycosidase one is longer and everything goes into the direction of facilitating the reaction. And so the distorted substrate is kind of pre-activated for catalysis. And that's what makes this distortion interesting because by knowing the conformation of the substrate in the Megaelis complex, we can guess, we can know what will be the conformation at the transitional state, which is important to design inhibitors for every particular glycosidase. Because they are similar, so one knows this one, the distortion here, then you can guess what would be the distortion of the transition state. Now I should give some technical considerations here about how we can study what will be the distortion of the sugar in the Megaelis complex. So we can describe it with simulation, the kind of level of theory doing it. When we started in this field, we thought that, okay, this is not a problem of breaking or forming covalent bonds, it's just a question of torsions here. So we should be able to describe it with just classical monoblock dynamics, this is what we started doing. But that turned out not to be so satisfactory for us. This is, for instance, 1, 3, 1, 4 beta-glucanase, this is an enzyme that glyphs specifically, that glyphs hydrolyzes 1, 3, 1, 4 beta-glucans that are glyphs, that glyphs hydrolyzano 8, that glyphs hydrolyzed 2, 3, 3, 4 beta-glucans found in certain cereals, like barley, and enzyme glyphs, specifically, the 1, 4 bonds, in linkages, in a 1, 3, 1, 4 sequence. This is a model of the Megaelis complex with a substrate in a skew ball configuration, in a skew ball conformation. because it's the one that has been found in other glycosidases acting on similar substrates. So it would have been very surprising to find that this was not the right conformation or similar to this one, but not very different one. However, when we did classical molecular dynamic simulations, what we found is that the sugar undisturbed. So it goes back to a chair. But it evolves to a chair, conformation to a totally relaxed chair for C1. That was very surprising because it's not that it evolves to another type of distorted conformation, but it undisturbed completely. It's a relaxed conformation and this has not been observed in any glycosidase before in hydrogen crystal structures. That would have been the first case of a beta-endoglycosidase with a non-disturbed subset. On the other hand, chair conformation is not good for catalysis because the living group is oriented equatorially. That's not good for the nucleophilic attack. So we decided to dig a bit more on this problem here and then we observed that if we raise the charge of the anomeric carbon, if we change the charge of the anomeric carbon, we make it larger, we could stabilize our skewbolt conformation in the dynamics. But that's not good, of course. We cannot just play with the charge and have the result that we want. So we thought that this is a problem that maybe requires also KMM. So we went to KMM describing the substrate, part of the substrate with quantum mechanics, and then we found that in fact the skewbolt conformation was stable. I should say also that in the molecular, in the classical MD, this undistortion of the source that happened very quickly, actually in that case happened just during the optimization. It's not even with molecular dynamics. But in other cases, it happens at different times. So this is the result that we got with KMM MD, the distortion is stable in time for at least 10 picoseconds. We have now extended these two more and it's stable. And the distortion, the conformation is a mixture of skewbolt, this type of conformation, and both conformation fulfill the condition that the living group is axia. So both are okay for catalysis. They are actually very similar. And we verified later, some years later, that they correspond to the global minimum of the conformational frequency landscape of the substrate in the enzyme. So we are now quite sure about this distortion in 1, 3, 1, 4 beta gluconase. We have found this type of a scenario in fact in which we are not able to stabilize the distortion with classical MD, we found in other cases. Sometimes we are not able to reproduce the crystal star, the experimental structure with classical force fields. It doesn't happen all the time. You may be lucky and maybe force field in some cases it reproduces the right conformation, but it's risky. So you should take care and check whether the what the force field is doing is it makes chemical sense or not. And the reason why this happened, we think is very simple. The charge of nanomeric carbon, carbon and also the atoms during change from one conformation to another. It is something that is not reproduced with classical MD. So it is normal that for some certain conformation force fields could not do a good job on it. So our tip here is check if conformation sample during the classical MD makes sense. And if not, you play some tricks like you can just fix the six atoms of the ring until you start QMM molecular dynamics. Okay. So we've been using QMM for describing the conformation of the subset in the active site of glycoside hydrolysis. I'm here some technical more technical detail. The code that we use is typically is the CPMD, the CPMMM, the interface of the Roslis Berger group. Sometimes we use also CP2K, but this is the one and we have more experience. And considering the DFT functional that we use, we've been using PBE for several reasons. First, because it describes quite well hydrogen bonds. And there's a lot of hydrogen bonds that surround the sugar with the hydroxyl of the sugar and the protein. And second, because it describes quite well the conformation of the sugar, conformations of sugars. There are several studies of it in comparison with high level methods. This is, for instance, a recent study by Marianski and Koworkas in which they compare the performance of several DFT functionals and also semi-impedical methods and force fields with respect to couple cluster results. And they found that the DFT functionals perform quite well, the standard DFT functionals. Hem-empedicals are not so good. And here's the glycom, the specific force field for carbohydrate doesn't do a very good job for conformation. It does a very good job for other things, like dynamics in general of carbohydrate in enzymes. But if you want to look at specifically of the conformation of one sugar, this is not enough. And the PBE turns to be the cheaper one of the density functionals that they still give quite a good accuracy. So we've been using this one. And to finish with the technical considerations, I should also tell you how do we classify sugar conformation, because there are several types. We've seen this 1S3, B2-5, what does it mean, and how can we differentiate one from the other? Okay. In mechanistic studies of glycosidases of carbohydrate enzymes in general, one normally uses the so-called packeting coordinates. Those are three coordinates, so a radius and two angles, that describes changing these conformations. You have all the, you can specify all the conformation of a six-membered ring. These con, all the conformations appear on the sphere of, on the surface of a sphere of radius Q. On the balls there is the two-chair conformation, the normal chair and the inverted chair, 1C4, so here carbon number one is above the plane, carbon number four is below, this is the opposite. On the equator, there are all the boats and skew boat conformations, major, just changing the data coordinate, and the data on the phi. And on the tropics that are not drawn here, there will be the half-chair and envelope conformation. So, knowing the Cartesian coordinates of the six atoms of the ring, one can, by solving this simple linear set of equations, one can know the packeting coordinates, a Q, Z, and phi, so you can put a point in this sphere and know what is, what, which exact conformation you have, whether it is a canonical conformation like B3O, or it is between two conformations, or between three conformations. But of course, this is a three-dimensional object that is not so easy to interpret, so one normally uses projections either from the north pole or from the south pole. This is a so-called Stoddard diagrams. Also, you can use a rectangular projection, so this is not very different from what we do people doing in cartography, using this type of projections. These type of diagrams are quite useful in mechanistic studies to delineate what is the conformational itinerary that the substrate follows during the reaction, during catalysis. For instance, family 2 and family 26 glycoside hydrolysis, which are beta-monosidases, they usually follow this itinerary. That means the Michaelis complex has been trapped crystallographically and it is in one S5 conformation. Then we know that the product complex is always 2, so one can guess that the transitional state will be 2.5 because it is on the pathway. Other families of glycosidases use different pathways. For instance, family 20, that one child beta-glycosidases uses this other itinerary. There are also glycoside hydrolysis that generates in the southern hemisphere, not so many, but family 29 is one of them. You can draw these itineraries here in the stolen dioram, the circular dioram and also in the rectangular dioram. The itinerary normally is common for all the enzymes or even family because they have the same sequence, similar sequence, so similar active site, also similar active site structure. The itinerary is usually also similar for glycosidases that even if they are from different families, they act on similar substrates like beta-monosidases for this itinerary or beta-glycosidases for this one. But there are several glycosidases for which the itinerary is not known or it is controversial. We have contributed to clarify some of these itineraries for several glycosidases and I will focus today in one example which is family 134 beta-mananase. This is a glycosidase that acts on beta-mananase. Beta-mananase is a plant polysaccharide, is a linear polymer of beta-manose. So this is beta-manose units. Beta-manose is very similar to beta-glucose, but the hydroxylate position 2 is equatorial, sorry, is axial instead of being equatorial like the other ones. If that was equatorial, that would be glucose and then instead of beta-manan that would be for instance cellulose. But so just the change of orientation of one OH changes the chemistry completely in carbohydrates. Beta-manan has many applications in the industry, in the paper industry, in biofoil and you can also use it as prebiotic you can take it for your microbiota. Some years ago, I should say that what is the typical itinerary of beta-manosidases, as I already mentioned it before, beta-manosidases typically use this itinerary on the equator of the clameron-popul sphere and they go through a B2-5 transition state. This is what is known from several beta-manosidases that have been already characterized. Why do they follow this in general? Well, because in the Michaelis complex, 1S5 glycosidic bone is axial, that's good for nucleophilic attack and there's no aesthetic interaction between the nucleophile and the 2OH, which is normally this is a problem in beta-manoside catalysis. The same happens here in the B2-5 transition state. So that's the usual conformational itinerary for beta-manosidases. So let's say to design an inhibitor for a beta-manosidase, one normally targets a B2-5 conformation. The itinerary is also similar for alpha-manosidases, but they go in the opposite direction. This is to remember you, this is a beta-manose with a better stereochemistry in the American that would be alpha-manose. Alpha-manosidases follow this itinerary on the opposite direction, but they also follow a B2-5. They also use a B2-5 transition state. A typical example is Golgi alpha-manosidase 2. This is an enzyme involved in protein glycosylation and it is a therapeutic target because it's over expressed in several types of cancer. So it's important to find inhibitors, to develop inhibitors for this enzyme that do not inhibit other glycosidases and it's called Golgi because it is placed, it is located in the Golgi complex in the cell. Some years ago we uncovered the mechanism of action of this enzyme and we found that in the Michalis complex the sugar adopts a twisted boat conformation, OS2, like typical that, well it was not that typical at that time that it was starting to be, to be, it was quite controversial in fact, but now it is more established and we found that it was the B2-5 transition state conformation in the, in the transition state and that contributed more to clarify that this type of boat conformation can be, can be used by, for monocyte hydrolysis at that time. So some years ago a new family of monosidases, mananase, in fact that the enzyme that hydrolyzed metamanan was discovered and the protein in that case enzyme they also find that they have no homology to any proteins with known functions so they propose that they, they should be a new family, they belong, this mananase belongs to a new family of glycosidases, there were 133 families at that time so that was 2015 so from this time then there's one more, 134 and very soon later a first crystal structure appeared from our colleagues in the University of York, this is the structure of the GH 134 metamananase with a metamanan fragment spanning in the, in the active side and with a very, very good resolution and when they analyzed the structure they found several unusual, unusual aspects. The first aspect is that the overall fault is not the typical of other metamanosidases but it resembles very much lysosine, lysosine is also a glycoside hydrolysis, it's a classical glycoside hydrolysis, it was the first enzyme ever crystallized and, but lysosine is another family of a quite different family of glycosidase and it doesn't hydrolyze metamanan, it acts on different substrate so it will have the same fault, it's a very similar fault for the different function. The new enzyme metamananase, the new metamananase is an inverter, it catalyzes the glycosidic bond with the inversion of configuration that means there should be a water molecule between the general base and the catalytic and the anomeric carbon and this is what is found in the crystalline structure because there is a water molecule well-defined here between the reactive sugar and the catalytic base so that is fine but the most surprising or the most unusual aspect in the structure was that the reactive sugar, this one here in the so-called minus one subside of the enzyme, it doesn't adopt the typical distortion of metamanosidases, this is a one C4 conformation, it's an inverter chair, the other ones are followed, they adopt the typical 4C1, they should not be distorted but the reactive sugar, the one that one expects to be distorted but one didn't expect to find a distortion that is so different from other metamanosidases, this looks like an outlier so that points to an itinerary in the southern hemisphere because if you start from the south pole from one C4 you cannot reach a transition and you cannot go through a transition state in the equator as the typical transition state for other metamanosidases so this is incompatible with the typical itinerary of metamanosidases so what happens here is this a novel itinerary that was this that uh for this novel glycosidase or it is maybe an unproductive complex who knows so we decided to go to do the KMM simulation for this enzyme to find it out this is the typical uh methods that that you use that I already described before we just in front of here with the pv in the chem vision we use a mercury glycum for the MMM atoms it was part of the polysaccharide will be in the in the KMM region we also need a force field for it and we use a initial molecular dynamics to take into account the temporal effects and move all the all the atoms at room temperature to activate the chemical reaction where the chemical reaction will not happen in a normal KMM molecular dynamics because you need to overcome a barrier and it has this is a rare event it happens in a in a time scale that is not accessible by emission d so we use an unsolved sampling method in particular we use metodynamics to activate the chemical reaction this is the typical protocol that that we use so uh it's not very effective from any pro any KMM protocol we start from the pv structure in that case uh given to us by from our collaborators we prepare the system that it's an important step that we need to take to be careful here because the um the crystallized sarchek can come with many little errors or atoms that do not correspond to the density residues that you don't know how to protonate missing residues etc etc so this is an important step uh then we equilibrate the minimize of course and then equilibrate the the system this is also an important step because in carbohydrates are very floppy um it would it's sometimes not it's sometimes not surprising that the the subsets escapes the active site during the during the the equilibration so it thinks to need to be done a bit maybe sometimes a bit more carefully than a standard standard standard systems due to that the fact that not only then sign is is very dynamic but also that the the the sure the carbohydrate is very dynamic um then you of course you choose a so which atoms you will you will you will be treated you will treat by quantum mechanics and which atoms you will be treated by molecular mechanics as you show you choose the chem region you rec equilibrate it because now the quantum atoms the qm qm qm atoms are now subjected to the electron density subjected to the dft potential and before they were subjected to the force fields of the discharge shock here and you need to regulate not to avoid the artifacts when you start later the the reaction simulation then once you recelibrate the system you choose collective variables so you you use your chemical intuition you think but what is likely to be the reaction coordinate and you choose some collective variables correspond to this reaction coordinate and you do the reaction simulation in our case we use metadynamics but you can use here any any enam sampling method then you can also do even even a static simulations of course and then you guess you you try to try to interpret what will be the reaction mechanism um so uh so maybe i should stress here that this part in particular for carbohydrate enzymes is very important it's very important because um it doesn't make sense to to do all this part that is going to be very um computationally demanding if you don't do well the thermal equilibration so check well the system check well that that uh everything makes sense that that there is no apart from the catalytic rescues check the sequential rescues the whole protein check msd for different parts that that everything makes sense because otherwise you do them want to have a small error here that then is translated into uh months of simulation here for for nothing okay so okay there's also a shortcut that some people use but i don't recommend that is using the directly the crystal structure i don't recommend because the crystal structure is is an average structure it doesn't correspond to any to a to any point in any instant on the lifetime of this biomolecule it's an average structure and you it can it can uh lead to many errors if you if you start directly the reaction simulations from from the crystal structure even even if you minimize okay so going back to our design to the gh to the beta monosidase so this is the classical simulation the first thing that we did was to reverse the mutation that was here to in the in the crystal structure they mutated the acid base residue to to be able to to obtain a michelis complex so we revert to the to the real glutamic acid residue this is a an quite innocent mutation that normally doesn't change the conformation of the reactive sugar we have checked this in other in other enzymes that would not be the case if you uh mutate the the nucleophile of the digital base but the the acid base residue doesn't normally change the any it doesn't affect the conformation then we paid the special attention on the position of this putative water because water are very mobile we check that during the classical simulation what was specific what was check especially the dynamics of this water and it it stays all the time wandering around but fluctuating around around this position just in place for catalysis sometimes it gets swapped by another one so that sometimes it leaves the active set and but immediately another water molecule enters so there's not always the same water but always there's always a water molecule here well oriented for catalysis okay then we decide what will be the quantum part the quantum uh the quantum region this is uh with some more details of the simulation but i i already commented this before so this is our quantum region this is the substrate the manopenta also five five uh sugar rings five data manos manos rings when this is the bone to be cleaved by the enzyme this is the anomeric carbon the important carbon um we normally take one ring before and one after the the the reactive bone uh as in the quantum region so we we cut it here and this is mm here we decided not to cut here because not not just to leave a half a sugar isolated there and in the mm region and just to avoid using too many too many link atoms also because we are using we're using a plain white basis set and the number of atoms is not so much an issue as long as the the box enclosed in the atoms remains quite the same we are uh we are of course including the catalytic residues because there will be bonds to be formed and cleaved involving these residues normally we cut them at the c alpha carbon and of course we include the catalytic water and that makes a total of 98 qm atoms uh then this is the qmm equilibration of the enzyme uh moves at room temperature uh the sugar doesn't change our conformation in this in this process and the next step is the reactive reactive reactant simulation the reaction simulation uh we use metodynamics with three collective variables that include all the covalent bonds that are broken that we think we know we are going to be broken up from during the catalytic reaction one collective variable uh accounts for the nucleophilic attack and the departure of the living group it will be the difference between these two distances that are uh quite a couple these two distances the other collective variable cb2 will account for the deprotonation of the water to be the difference between these two distances and the third collective variable is the protonation of the living so we split this collection of distances into three three variables and here is a bit more detail of the distance that that we use for the as collective variables we always say to use the all the covalent bonds that are expected to be more broken during the reaction but sometimes that makes a too high uh that makes too many collective variables and so then we check but if that's the case what you can check by only dimensional metadynamics to descend the most important bonds contributing to the reaction coordinate because not all of them wait the same in the reaction coordinate here for instance the most important one is the um the nucleophilic attack by the water and the departure of the of the living group and um when you do the metadynamic simulation this is some some details of the this is maybe just for uh metadynamics experts here this is the metadynamic simulation and this is the reactive sugar the one that before the bond to be cleaved by the enzyme this is uh trajectory of the metadynamics I just want to you to see how this sugar the reactive sugar changes conformation and it events uh towards the here towards the catalytic base uh because we cannot we cannot appreciate much detail here but from the from the simulation we obtain a free and as a landscape it should be in three dimensions but this is a projection in two of the collective variables which is easy to analyze and we can we can take a snapshot along the minimum energy pathway and we can see that the in the reactants well the sugar is in a one c4 conformation but as as soon as the water starts attacking the anomeric carbon and the glycosidibone starts to be cleaved it changes to a to a half chair conformation for a 3h4 conformation which is in the southern hemisphere the energy body that we get for the reaction is 17 kilocalories per mole that's acceptable and it's quite similar to the one of the extract that estimated from experiment that was about 15 kilocalories per mole for this system but it was not the same for similar substrates that should be also exactly the same and then the sugar evolves in the product complex so when the glycosidibone is completely broken and the water has already delivered a proton to the general base so we have here the the sugar window with the inversion of configuration we got here a beta beta um a beta glycosidic bone and here we have an alpha hydroxyl then we obtain a 3h1 configuration so we confirm that this is the anitinerary in the southern hemisphere we can analyze much more the free energy landscape to obtain more detail okay this is free energy so for every point along the minimum free energy pathway we don't have just one structure we have a collection of a structure that that all of them have the same collective variables but they differ in all in all all other in many other degrees of freedom so every every at every point for instance the reactant you're going to see the how long length is that long is the glycosidic bone we have a value an average value with certain standard deviation corresponding to all this collection of a structure and we can see here for instance that the transitional state we can analyze all the structures that correspond to a transitional state and we can see that it's quite a dissociative transitional state because the glycosidic bone is broken practically broken at the transitional state and the what the water is starting to attack we can also see that the glucolophilic water is not yet deprotonated that happens later in the the reaction pathway and the living group the living group is already protonated because this this the cleavage of this glycosidic bone needs really assistance by the acid base base ratio um then we also found that there was a surprise at the reaction products if we you see the free energy landscape you can say there are two minima and the products region they are separated by the by um the cv2 which is the water proton transfer now something happens in this in this region here what is the difference this is p this is the first minimum product in which assure isn't 3s1 and then what happened from here to here is uh is something that this doesn't take a lot of energy because this is a 2 or 3 kilo calories per mole energy barrier um what happens is that a water molecule enters the active site a water molecule that a classical water actually because we that was not part of the region but doesn't matter because there's just water molecules that come common classical that come in and out during the KMM simulation and it sits in between the sugar and the general base so i see so that's the that's the other minimum that that we obtain in the product region and by doing when that happened the sugar changes to 1c4 so it's again a conformation in the south pole but um this is very even quite a good agreement with experiments because in the structure of the product complex that was also obtained by our colleagues what they see is that the sure actually isn't 1c4 and there is a water well defined between the sure and the the general base fashion so that that was in a very good good um very good agreement and that's two ways to obtain this these other product either you you analyze the sorry go back a bit either you analyze here the fringe and landscape but if you take a snapshot from p and you became m m d without any metadata dynamics you also reach we were also reaching this other configuration because then everybody was was very slow so so with the KMM md and metadata dynamics we were able so we were able to connect the two structures that had been detected that had been obtained experimentally the Michaelis and the product complex we've been here the complete pathway and we confirmed that this is an itinerary in the southern hemisphere so a novel itinerary for better for better like uh better monocytes so up to now all of them follow an itinerary with a b2 5 transition estate but yet the family 134 is a better monocytes that uses uh three three h4 transition estate so that's the one to be um to target for uh to to the develop inhibitors for for for this type of monocytes but there are two distinct solutions to facilitate a nucleophilic attack on a man's racing and okay so we confirmed this southern itinerary this is the picture that my PhD student did and web winners are very very strange because normally when I when I put this feature people start laughing but now I don't even know if people is laughing or you think oh that's a stupid but anyway that's a problem with winners and this is an international collaboration uh it involves groups in the northern northern hemisphere and also in the southern hemisphere our collaborators in in australia which we're very happy actually because because for once the southern hemisphere wins the race and uh if you want to know more on catalytic mechanics of monocytes you can see I don't know there is a new we we wrote a new um a viewpoint uh this year about the several mechanics of monocytes including a new end of monocytes that forms that is this is very much and much more unusual than the one I described because this one goes through through an epoxide intermediate and uh this is all for today uh I should apologize because in my after they wrote that I was going to talk about another enzyme but later I realized there was no time if I want to describe things in a bit more detail so um I would like to thank the people of my group this is the last picture that we did the pre-pandemic and the calculations that I that I described were done uh by Cheby Viernes he was working on 131 for beta gluconase finding all these problems with a distortion and not distortion and Luis Reich a previous PhD student also that he did all the work on GH 134 so on the beta monocytes that goes through the southern itinerary I also thank my experimental collaborators and the agencies that provide us funding so uh thanks for your attention I'm gonna switch now to the webinar maybe and uh open the I'm ready for questions now okay thank you very much Carmen that's very very nice um I think people probably appreciate I mean the insight even if they're interested in in the additional enzyme you mentioned I think the detail you provided in the simulation protocol I think was very valuable especially for people who are starting out to see uh some of the key steps and not to take the shortcuts that you warned people not to take but using the crystal structure directly so that was very interesting um I certainly have one or two questions I can ask um I don't know if my colleagues uh Gerard Emiliano would like to ask something uh I'm Emiliano can you hear me yes hi um I have a question in general because um um I've seen that when you you are able to get transition states uh states and transition states and uh Michalis complex from QMM uh then many people exploit this information to uh to find the best inhibitor for some reaction and I'm wondering what is the the additional value to know the transition state with respect to try to uh the strategy to get inhibitors simply uh occupying the uh the binding state in the binding site uh in in some other way what is the additional value to to know to get this information yeah um from my experience in glycosidases is you should do both so you should target not only the conformation but also try to feel the positive subsides and the negative so to make to make a good uh inhibitor that binds better in the in the active so you you you don't need to target only this conformation there is a combination of things normally one target is a conformation the charge the charge because you need to put a positive charge in a an atom that is the analog of the anomeric carbon in your inhibitor and also some some living group that that fits in the in the active site but you should not forget any of the three so it's a combination of the three and you can even target not just the conformation of the transition state sometimes you you try to make an inhibitor of transition state but what what turns out is I mean is a Michalis complex analog or a product analog so I think I would say the rule is to to target any of the conformation along the reaction pathway I see thank you and of course try to make it different from try to see what so you want to inhibit an enzyme but don't know what to inhibit another one you should look at the differences in the catalytic pathways different itineraries and try to find uh one of the conformations that is good for one but not for the other and then make an inhibitor that is locked in this conformation okay uh I see that we are some question already hello kame how are you hello adrian that's a nice talk thank you my my question was you you mentioned some of the crystallographic evidence for distortion and as as you as you've said you know some of those structures particularly the oldest structures um yeah they may they may not be correct because of the models that we use and so on so I just wonder if you've gone back and and looked at those structures because your simulations might you know correct them and and actually identify the the nature of the distortion yeah yeah yeah that's that's possible well I have not we haven't not gone to uh but so earlier structures with uh not good resolution and so on uh but while we have some and then some of the yeah that is one of the first that that it was it was um predicted that it was uh conforming a certain distortion and we checked by QMM and we had the same distortion actually even though the resolution was not good so and in one case and sometimes what we found is that um I don't know the exact the precise question is in some quest in some cases we found a different conformation from the crystal structure and we can correct the crystal structure yes you mean so why so once you get the crystal structure how do you know the the the one you get is not correct because it is not it doesn't make chemical sense it's the only way to know but this is what happened in the example that I gave crystallographically uh it got chemical sense only if you are talking about the novel itinerary otherwise they will think maybe it's an artifact there's a mutation there maybe the crystal conditions I don't know there are several several factors that that can play a role there so this is what we check that's the the combination of the of the experiment and the calculation tells you that this is a novel itinerary it's not it's not it's not anything anything strange but I can tell you sometimes in in one case what what um what we've been checking particularly is that to obtain this Michael's complex says one needs to perturb the structure somewhere to be able to trap the the structure of the substrate with enzyme otherwise it reacts and you don't see any Michael's complex you don't see any substance there no so one way is to do a very gentle mutation like the one they did in this study just the acid base and they were lucky they could trap but sometimes that doesn't work experimentally who knows why and then they they do more perturbations so they put a floor in the way to in the two position also not to see whether this doesn't react and then if not they mutate also this one at the end you don't know if the distortion that we find you find is a consequence of all these mutations or not and we've been looking at some of these cases remember one in which the structure was that was had been determined with a with a substrate in which the glycosidine origin was substituted by sulfur so this is a diode derivative and it was a very strange distortion well it was it was not distorted at all that one it has it was once 4c1 and we were very surprised so we did the this study not only the KMMMD we also computed the the Frianese landscape of that sugar in the enzyme to see whether there is other minima that corresponds to the to a some temperature the sugar would have several conformations and the global minimum was not the 4c1 in the crystal of the thio derivative we did the we did the natural subsurface with the oxygen so first we did the thio derivative and we checked that it was 4c1 but then we changed the sulfur by oxygen then it was I don't remember uh 2s00 is 2 another one that was that had more chemical sense according to the function of this enzyme and then the with all the crystallographers and the crystallographers did uh well another group and in fact a group of Frianegis they did the structure without the thio derivative they used their natural substrate they used another type of mutation of the acid very sensitive to traffic and then they got the one that you predicted so uh if the question is if the methodology can can uh correct crystal structures in some cases and you just saw it but if you do all this all this all this work yeah yeah but it is not something that you just press the button and you get the crystal structure oh correct no no you you need to spend some some weeks there then starting and sure yeah yeah the the basic point I think is that has you know good crystallographers like Gideon know um the simulations may in some cases actually help to to sort out the crystal structure to help you identify the real structure so the QMM may be better than the the crystallographic refinement and I can say that because I'm speaking to a computer yes yeah but it's also tricky to to set up the system for QMM because you if you come for a crystal structure it's not so good be careful that your results are not due to the to this answer that you get of course it's always better to start with a good resolution crystal structure yeah okay we have some other questions that have come in I think you might be able to see the next question Carmen yes I see for instance hi nice talk I wanted to know how long was the sampling in metadynamics I think you can say a lower level of theory like DFT by binding will be helpful so how long was the sampling in the example that I show I don't remember maybe 30 because seconds or something to have the complete reaction but that depends from the system to system and from how deep are the minimums how many minimums you need to feel etc that depends on the free and as a landscape and if you're in a lower level of theory we haven't tried but since in that particular case I'm talking about glycosidases and as you know there is this problem of distortion of conformation and so on and same empiricals they don't seem to be gave a very good job on this so you probably can do it but just be careful and check things and so on I think at the end of the time yeah we have a few more questions so whereas I would normally unmute each each person I will say we can go on with the next question now yeah which one this one how much resources does this metadynamics in QMM take so the example that I show I think we were running for three weeks the metadynamics using 64 processors not running all the time but something like this what would be the maximum number of atoms manageable in metadynamics yeah I'm sorry I am lost what what is this I know sorry is maximum atoms in both KMM and MMM atoms gone for three I mean the bottleneck is the QM so you can make it as much as you want and then you can treat one subunit to subunit of the protein then the QM is the bottleneck how many atoms that depends on your machine no if you have a supercomputer with millions of hours and many processors you can probably manage I don't know 500 atoms in the QM region and have a MD step of just 10 seconds but if you have a lower not standard resources like us maybe I don't know up to 150 is fine but that depends how how long can you wait you want to wait for this project if you have more atoms I mean you can put as much as you want just that it will be slower and instead of one year project it will be two years project so it's but as I said for this for this case that I was about 100 we were three weeks the metadynamics simulation that's not three weeks project because before the metadynamics there's a QMM equilibration and before that there is a classical simulation and so on and preparation of the system and there is also one can make mistakes in this process no and go back and forth and so on and that depends on your skills and the student skills and many variables here so the whole project may take at least six months for 100 QM atoms in a reasonable in reasonable amount of resources another question let's go to another question key karma I just wonder how do you select the collective variables to follow during the QMM simulation would be any change in the profile if you follow also the head the hydro planes on the sugar carbons instead of following no yeah that's a good question some people can think that since the sugar changes conformation maybe if I if I activate this change of conformation I get the chemical reaction but that's not the case that's not part of the reaction coordinate I mean it's not a low energy mode that needs to be activated during the dynamics that the the conformation changes automatically as long as as the glycosy bone breaks and we know this for instance because if we do if we compute a free energy landscape of the conformation of the source that in the enzyme this is something I have not not discussed about it but you can see that the energetic cost of changing conformation in the active site of the enzyme and maybe you find a conformation that is the global minimum another one that is more this less stable but but if you go from one to the other the glycosy bone does not break so their reaction does not occur so that's not any that's not what is considered a low frequency mode in the in the metadynamic it's not a mode of the system that you need to worry to activate in a metadynamic simulation what you need to activate is the breakage of the the covalent bonds that are going to break a form during the reaction that's the main the minimum the minimum factors that you need to to consider so they you will not do this questions are a little bit more involved so I think for this one I will ask Liana if she wants to follow up and say whether that makes sense or not so Liana I will just try to unmute you hi Carmen I know if you hear me yes yes well now I just I just wonder I know if you will have like any change because you show us that you follow I know just bone breaking and and just just the balance and I just answer I wonder sorry if if you also include the changes in the orientation of the the dihedral planes of the carbon related to the the hydroxyl group that you will attack that it will be any change in their in the profile because you show us that there is different confirmations that when you start one confirmation when the reactant then it changes from the transition state and then you will have another change from the product I just just wonder if you will have maybe a small barriers or maybe a different landscapes if you consider more variables than than just more the the bond length does yeah so I don't I don't think so and I I think those are the main variables that the bond length because it's what it takes it needs more energy for for the reaction to happen but we never know of course we haven't tried any more variables but the fact that the energy barrier is reasonable is according to experiments and comparing to other enzymes that we have investigated in which if we miss a variable immediately the energy barrier goes up we can infer by comparison and by experience that we have related the reasonable number of collective variables also because the product in the product region it matches with the experimental confirmation and with experimental extracts so from all this we can I would I I suspect that that adding more variables will not change the results okay okay thanks thanks karm and thanks for your talk rodrigo our positions that I don't see the complete question our positions are distant restraints needed in some cases in the simulation in the simulation I don't know what comes after but I guess it's about no we didn't use any restraint in some cases and we tried normally not because if you need a restraint uh let's put a stream case maybe the sugar maybe the carbohydrate escapes during the classical md the active site and definitely what you do so you are not reproducing the meccalis complex then you can go on adding a restraint so that it is in in we haven't done this I mean I mean we haven't probably something like this but imagine you want to do this and then you you you you oblige it to be in but then how can you be sure that the result that we obtain is not uh uh influenced by this decision or sometimes just the project cannot be done if we have like something like this we just stop the project and we study more because maybe we are forgetting something maybe we didn't protonate correctly and instead in that is in the second shell and and we don't apparently doesn't look a problem but but it makes us some interactions that you are not producing and so if that happens you just need to to to work more until to work harder until it is everything is stylized without restraints because then it's the only way to make sure that what we do the results that we obtain later with PMMM are realistic there's always a risk over your results but try to avoid I would say this my advice try to avoid restraints try to avoid means uh more harder to to see what how to what is the reason that you need to add to this restraint it's very easy just to understand and go on but okay I think in the interest of time given the time we have available I think we will leave the questions there before we end the session completely I would well first of all let me say again thank you very much Carmen for a very interesting talk I thought it was very very valuable for people to to to see especially the simulation protocol even if they're not necessarily working with similar kinds of of systems as you described I just wanted to highlight the before we finish the session that we have so this was the final webinar that was part of the workshop but it is not the end of the workshop because we have a panel discussion coming up at the end of the month so on Friday January 29th we'll have a session in the afternoon details will be announced on the biocell website on the biocell twitter and we will also be in touch by email with all the people registered for the workshop webinars with more information and registration information so the idea with that session is that we really look back at the emerging topics and themes that have come out of the of all of these webinars by the speakers in the workshop and evaluating some of the challenges that people are faced in doing QMM simulation so if you have any questions for Carmen or for any of the other speakers I would like to encourage you to attend to attend that session so with that I would like to thank her again very much on behalf of biocell and all of today's attendees and I wish you a good afternoon everybody