 Hello everyone, I'm Silveger Pazzoni, I'm a postdoc at the University of Cagliari. I would like to thank the organizer for inviting me here for this presentation. Today I'm going to talk about my research activity focusing on antimicrobial resistance, in particular on the optimization of efflux avoidance and in emission for antibiotic development. The World Health Organization worked years ago about the risk of antimicrobial resistance and in the meantime the problems has become more and more urgent as the spread of drug resistance pathogens threats our ability to treat common infection. Antibiotics are becoming increasingly ineffective. Here I reported a map showing the distribution of a carbapenema resistance bacterium all that spreads all around the world. So new antimicrobials are urgently needed. Therefore several calls for research projects have been opened both in the US and in Europe and the work that I'm going to show you here belongs to one of these bigger projects. So resistance occurs naturally over time through genetic changes. Bacterial exploit different mechanisms to defend themselves towards the action of antibiotics. For instance they can inactivate the drug like in the case of beta-lactamesis which are able to hydrolyze the beta-lactam ring of some antibiotic classes or they can export outside the cell the drug using the efflux pumps. These in particular are a major player in resistance in gram-negative bacteria which combine the efflux to the pulpermeability of their outer membrane that makes it difficult for antibiotics to enter the cell. So how can we overcome resistance? Well we have two main strategies. The first one is to design new drugs able to avoid the resistance mechanisms. So for instance we want antibiotics that do not bind the efflux pumps and therefore they are not transported outside the cell. The second strategy is to co-administrate inhibitors with the drugs to enhance their activity. So following the previous example we want compounds that bind the efflux pump blocking their activity. Today I'm going to show you how we address this issue by using computational techniques. We started with a ligand-based approach where we collected a database of antimicrobials that are used for both correlation studies or machine learning analysis. Then we increased the complexity of the system by using the structure-based approach applied for the discovery of new antibiotics and inhibitors. So starting with the first. When we think about databases of small molecules we can think of Pubcam, Campbell, so a database that contains the 2D or 3D structure of these small molecules with some of the basic chemical-physical properties. So what we wanted to do was to collect a curated database of chemical-physical properties and force fields parameters dedicated to antimicrobials. Therefore we created IBDB which contains more than 300 antimicrobials belonging to 25 different classes including both traditional antibiotics and also inhibitors. For each compound we downloaded the structure from Pubcam. We checked the protonation and the tautomeric states. We performed a DFT-based geometry optimization followed by a single point. Then the QM results were used for the force field generation and finally we performed one microsecond and dissimulation on all of these compounds. So from this pipeline we obtained the QM optimized structure for each compound and also some structural cluster extracted from the MD simulation. From each of these steps we also extracted molecular descriptors both standard and non-standard so QM and MD-derived. IBDB is freely available on feature at this link where we shared all the input and output files that the user can download and reuse. All of these descriptors can be applied for correlation and machine learning studies in order to explore different aspects like the permeation or to perform some prediction activity. In this second case study I'm going to focus on the efflux and in particular we focused on this gram-negative bacterium called pseudomonas aeruginosa. The transporter MxAB OPRM is the major one of pseudomonas aeruginosa. It is a tripartite antiporter system, a proton-antiporter system which extends from the inner to the outer membrane. It exploits the energy associated to the transport of proton to expel out of the cell various ligand classes. The engine of this complex system is represented by the homotrimeric inner membrane transporter MxB which belongs to the resistance nodulation division superfamily. MxB mediates the transport of substrate fans through a conformational cycling of its monomer which is called functional rotation mechanism. These monomers can adopt three different conformations, loose, tight and open, LTO. So in the functional rotation mechanism the substrate bind the L monomer. This binding is followed by a conversion into the T-conformation and finally the substrate are expelled after the conversion of the T into the O monomer. Two main binding pockets were identified into MxB, an access pocket in the L monomer and a distal pocket in the T monomer. The former is one of the entry channels for high molecular weight compounds while the second one is thought to be visited by all the compounds expelled by the transporter. While for MxB just few crystal structures are available for its homologous in Escherichia coli namely ACRP we have several experimental structures also in complex with various ligand classes. So from this structure it is believed that the substrate can oscillate between different iso-energetic binding modes inside this wide and promiscuous binding pocket which is the oscillation hypothesis. So in this case study we focus on the design of new antibiotics able to avoid the binding of MxB. It is well known that the 4-kinolone antibiotic class is a substrate of MxB. However not all of these compounds have the same efficacy toward Pseudomonas aeroginosa. Suggesting differences in affinity and binding modes. By knowing how these compounds bind MxB we can rationally design new antibiotics able to avoid the binding with the transporter. Recently a co-crystal structure of ACRP in complex with one of these 4-kinolones which is a levofloxacin has been published. And we use this structure as a reference. In this work we performed a systematic docking campaigns on 36 4-kinolones. Here for time reason I'm going to focus just on the results of levofloxacin. So this is the workload that we used. So we exploited ensemble docking. So we collected multiple conformations of MxB and also of the ligands. These conformations were retrieved from ABDB that I show you at the beginning. Then in order to increase the statistics we use 3 different docking protocols generating 1800 poses per ligand. Then we clustered the docking poses and the representative of each docking poses underwent to an energy optimization, air scoring and a surface analysis. But before we validate our workflow on the ACRP levofloxacin complex for which we have the experimental structure. So here on the right I reported the clusters derived from the docking poses, the corresponding population and also the RMSD with respect to the crystal structure. So as you can see the most populated cluster was able to reproduce the experimental structure with an RMSD of 1.3 Armstrong. Below I reported the residue of the distal pocket and I covered in red those residues that were in contact with levofloxacin in the crystal structure. In ensemble docking we have statistics of contact derived from all the docking poses. Therefore the hit map that you can see here reports the percentage of contacts of the docking poses. And you can see that docking was able to find the same contacts of the experimental structures but also it found other contacts suggesting that different binding modes inside the pocket are possible. Then we of course applied our workflow to MaxB, so docking, clustering and optimization. Here again I reported the cluster and their population. So the most populated cluster is the one in green, was placed in the same region of levofloxacin in the crystal structure. But as before also here we found alternative binding modes like the one in magenta in the deeper portion of the pocket and the one in yellow in the middle of the pocket. So these results suggest that levofloxacin preferentially bind in the same region of levofloxacin found in the x-ray structure. However different binding modes are possible throughout the pocket which is in agreement with the oscillation hypothesis. Given the lack of experimental structure between MaxB in complex with kinolons we made available at this web page all the 36 complexes that we obtained with docking that user can visualize and download. Ensembled docking takes into account indirectly the flexibility of both the protein and the ligands. However it lacks a description of the evolution of the system over time which is something that we can analyze with MD simulation. So in this last case study we use MD simulation to study the behavior of inhibitors. This is a work in progress but I would like to share with you our preliminary results. So as I mentioned here we are focused on inhibition, in particular on competitive inhibition. So compounds that bind the levofloxacin without being transported preventing in turn the binding of antibiotics. We have one crystal structure of MaxB in complex with an inhibitor called ADPP which was found in the distal pocket partially interacting with an hydrophobic region rich in phenylalanine residues called hydrophobic trap. Here we worked on these Rempex compounds which are peptidomimetics and experimental data shows that although they have a similar structure they can be divided into different classes. Some of these compounds are substrate of MaxB while other are competitive inhibitors. Previous studies try to find some predictive rules able to distinguish between these different classes and this work is a follow-up of the previous findings where we try to answer this question can substrate be converted into inhibitors? To the same we use ensemble docking, multi-copy MD simulations and free energy calculations. So here are the docking poses of the substrate and the inhibitors and as you can see the aromatic ring of the compounds points towards the hydrophobic trap but in the inhibitors differently from the substrate the interaction is more tightly and it establishes some pi-pi stacking interaction with phenylalanine which is something that we saw also in ADPP in the other inhibitor that I showed you before. Then these poses underwent two multi-copy MD simulations and we found that the substrate explored the binding pocket more than the inhibitor which is something that we can see from the RMSD values while here below are reported the percentage contribution of the residues to the binding free energy. We can see that the inhibitors interact with a few residues to a greater extent compared to the substrate which conversely interact with more residues but to a less extent. So these results together are again in agreement with the oscillation hypothesis so where the substrate seems to explore the pocket and adopt different binding modes with similar binding energy while we speculate that the inhibitor binds more tightly to the hydrophobic trap. So these are preliminary results. We are now performing some mutational analysis on these key residues to confirm these computational results and we are also performing other MD simulations in order to increase the statistics and also we are thinking about using different methods for the free energy calculation. So in conclusion I wanted to show you different techniques, computational techniques to address antimicrobial resistance which is a complex problem and it is also multifactorial. So I believe that it's important to use different techniques and to tackle this issue from different sides. I just want to thank my research group, our collaborators and of course all of you for your kind attention and I would be happy to answer your question.