 Thank you, Alicia. Now share my screen. You all see my screen. Okay, oops. Yes. Sorry. Okay. So first I would like to thank the organizers to give me the opportunity to talk in this workshop which has been really inspiring with a lot of good talks and interesting people together. As Alicia said, the work I'm going to present is mainly the work of Lea Pradié who is about to defend her PhD. And she worked on the broad scale propagation of aminoglycoside resistance and today I will present some of her results doing a focus on plasmids and mobile genetic elements. So the context of this work was as many of the work we have seen the propagation of the antibiotic resistances and the public health program they represent with the numbers that we all know of millions of people who are related to die from antibiotic resistance infection. And so these genes are present all around the world, and they have propagated quite rapidly after we started to use antibiotics as treatment. There are only two ways by which to pick ways, ways by which they can propagate. The first one is spatial movement of bacteria carrying resistance genes by spatial I mean geographical movement from one region to another one country to another but also ecological movement from one region to another one environment to another. And the other way by which they can propagate is of course by horizontal gene transfer. So the transfer of the gene from one species to another or one stream to another. The initial goals of the study that they are conducted was to describe the resistance propagation patterns, and then try to better understand the factors and the mechanisms and the evolutionary forces that were defining the movement and the direction of the movement of these resistance genes. And as I said I will focus on specific things we found about a plasmid and mobile genetic element carriage. And so, we decided to focus on amino glycoside modifying enzymes so I know glycosides are antibiotics that have been used since the 14th to treat gram-negative infections. They are antibiotics that are naturally produced by streptomyces and micromonospora. They act by fixing on the 30S ribosomal subunit. And in the last decade, there has been a reduction of the use in human medicine, but they are still intensively used in veterinary medicine and in agriculture. For many other antibiotics, there are diverse resistance mechanisms, but here we will focus on resistance genes that code for enzymes that modify the amino glycoside by transferring either an acetyl, a phosphoryl, or a nucleotidyl group. And the goal was really to have an integrative and broad-scale approach of this propagation. And so we conducted many bioinformatics study. And the starting point was to gather all the protein sequences of amino glycoside and modifying enzymes that were known. The sequences that are classified with a biochemical classification. We constructed a cluster of homologous genes, meaning clusters of sequences that we expect to have a common ancestry, so we generated more of an evolutionary classification of these genes. On the other hand, we collected all the available eubacteria genomes on NCBI at the time the study was started. And together with these genomes, we also collected metadata, so when they were available, so the dates, the place where the isolate was collected, and also when possible information about the environment, the bio which was isolated. And so we screened all these genomes for the presence of the 27 clusters of resistance genes. And we also taxonomically assigned these genomes to a large phylogeny of the bacteria, which is based on the 120 genes that are ubiquitous in all bacteria species, almost all bacteria species. So we had these metrics of presence absence of genes in species. We also looked at the genomic context of the resistance genes that we had identified. And so we screened for the around this, the genes for the presence of different mobile genetic elements, namely placinids, prophages, integrative conjugated elements and transposons. And we found out here that for the placinids, we used a tool that may have developed called Plasforest that is based, that is based on homology and using random forest classifier to identify placinid impartially assembled gene. And the last thing we did to assemble the data set was to infer the resistance spectrum based on the presence of the different resistance genes and using database of resistance comfort. So by the resistance spectrum, I mean the specific aminoglycoside to which the genes were conferring resistance. With this data set, we can get an idea of the temporal dynamics, which is what is represented here. So here in black, you have the proportion of the genomes carrying a resistance genes a long time. So you can see that of course, very early in time have very few genomes so this estimate is quite that to the confidence interval is quite large. But then when we go in time, we have better and better confidence interval. And you can see that we seem to have reached the plateau is quite a lot of variation between years. So the point here that of course this data set, the genome data set is strongly biased, because we, we got what was in NCR and that there is a clear bias towards samples that have been collected in clinic. So this is to say that probably some could sample all the bacteria of the world, the proportion of them carrying aminoglycoside resistance wouldn't be 30%. So then we can zoom on the time periods and compose this, this proportion into the frequencies of the different cluster of homologous genes that we have screened. And you see that we have a general dynamic of coexistence of the different clusters and not a replacement dynamic. So we do have some trends as like a reduction of frequency for this light green tone or temporal peak like for the red code. And then if we decompose this bi geographical location, we can see local dynamics that are with more variation and coexistence at the moment that the worldwide is sometimes treated the presence of different things in different places. So we can also look in which file we find each which cluster of homologous genes, and you can see that some clusters are very widespread phylogenetically present in a lot of phyla, as others are present only some very limited number of phyla. If we read this table in the other direction, we have clearly phyla that are carrying a lot of different clusters of homologous genes, when some others seems to carry on very limited number of clusters. We can also look at the variation of association of these different clusters with the homogenetic elements. So this is what is represented on this graph. You can see that the, the, the association is basically varying from zero to one so some clusters are never associated to mobile genetic elements while others are always associated. We can also see that placemates are quite idea represented so that's the two blue shades. And that it's quite often that they are associated to two types of mobile genetic elements, a transposer or integral, and something else. So, so the placemate plus transposing of page plus. Basically, a group is different mobile genetic elements in three categories. So that's exactly same data as the slide before but this time we have classified them nested so that's the case where a transposer is on a placemate or in page. And then a mobile genetic elements that are moving within genomes and those are mainly moving between genomes. So with this classification, we have looked at how the carriage by mobile genetic elements is influencing the resistance spectrum and how it changes with the number of aminoblacoside resistance. So first, we have noticed that it's quite common to find genomes that have several resistance genes, then we inferred the resistance spectrum. And we compared the observed increase of the resistance spectrum with the number of aminoblacoside resistance gene, which is what is represented in each of these plots by the green line, the black line. And we compared that to what we could predict from a random reassertment of the genes in the genome. So you see that when the genes are not associated with mobile genetic elements, the observed is strictly within the predicted. It's not different in the case of association with intra genomic mobile genetic elements, but in the case of inter genomic mobile elements. We have an increase of the spectrum breast that is faster than predicted just that is stronger than predicted just from a random association so it seems that inter genomic mobile genetic elements are associated with the broadening of the resistance spectrum. By the same type of analysis, we have also shown that the association with mobile genetic elements is increasing the number of copies of the same cluster that we have within a genome. And this number of copies actually probably quite often underestimated because when there is a copy of the placidine crown one copy but we know that many placidine are actually multi copy. So this presence of the same copies of the same resistance gene within a genome is on one side will be linked to an increase of the expression level that's something that is known from experimental data. And on the other side we know that it's also increasing the potential for new functionalization and differentiation, and also probably in the sense increase of the resistance. Then from the data set we assembled, we have this presence absence of clusters across all the phylogeny bacteria. So we can for a cluster identify pairs of genomes that contain a gene of the same cluster and ask whether it's likely or not that this gene has been transferred horizontally. So to ask this question, it first generated the distribution of the distances for the hundred and twenty of these genes that were used to assign taxonomically. And so this distribution is what we use as the reference for vertically transmitted genes. And then we compare the distance between the two amino glycoside resistance gene to this distribution. And if the distance between the resistance gene is significantly lower than the minimum of this distance, we decide that this is a candidate for horizontal gene transfer. So then we refined this identification of horizontal gene transfer by using compositional distances to do what would be the rational that transfer gene is going to be more similar in terms of composition to the donor species than to the reasonable species. And using this, we identified the two genome that were most likely to have extended gene in the case where a genome was involved in different pairs of potential horizontal gene transfer and also to orient the transfer. So decide in a pair of genomes, which was the most likely to be the donor and which was the most likely to be the receiver. Once we have done that on the data set for six of the gene clusters, we can reconstitute this kind of oriented network where each of the node is a species or I can hear a genius on this and the arrows represent horizontal transfer that are oriented. And we can then study the structure and the topology of these networks and something that was quite clear on all of this network is that there is a structuration between donors and receivers. So there are very few cases where species that are very frequently donors are also very frequently received. The role seems to be divided between the species. We can also identify that some species are always donors and these species that are always donors are not species that are very, very often associated with clinical environments or even associated to humans. There are more soil species. And I want to draw your attention on the fact that these two strachomyces species come back quite often as donors of resistance. So it means that strachomyces is both producer of antibiotic and donor of the genes to resist this antibiotic. On the other hand, some species are always on the receiver side and that's the case of E. coli or Shigela. So by this analysis of oriented horizontal gene transfer networks, we have shown that there's still to be a large role of non pathogenic soil bacteria in the horizontal propagation of our resistance gene. Then we also performed another type of analysis on these networks and we try to find the factors that were influencing the probability of a successful transfer between a donor and the receiver. And so this, this analysis showed that there is often that there is a negative effect of large pilot genetic and large ecological distances. That's something that has been identified on other data set, and that is known from mechanistic explanation that I was that was nice and reassuring to find that on our data set. And also identify a negative effect of codon usage differences between the transfer gene and the receiver genome. That's also a known barrier to horizontal gene transfer. And just want to explain a bit why this is so the synonymous codons are not used at the full frequency within genome, and the frequencies at which they are used vary between species. They are different from one species to another. They are expected to have been shown to be shaped partly by the co evolution with the translation machinery. And in particular, there is usually a quite good match between the frequency at which a codon is used, and the number of copies of the gene of the tyranny that decoded. So we have this, this co evolution that make a gene nicely translated in its origin genome, but then in the case of horizontal gene transfer between species that have differences in that codon usage it means that the genes in its receiving genome is using to be shown to lead to slow translation and the production of harness and truncated protein. That's the mechanistic explanation for why there is a negative negative effect of codon usage on the probability of transfer. So to confirm the classical barriers and mutation of horizontal gene transfer of our data set, but probably even more interestingly, we show, and here I show the data for the biogenetic distance that it's actually variable depending on the association with mobile genetic elements. So if we look at the red line here it's the prediction, it's the regression of the association between the finite distance and the probability of success horizontal gene, and so it's that we have a negative relationship. So if we look at what's happening for genes that are linked to associated to my genetic elements, you can see that for inter genomic MGE, we have an increase, we have an increase of the flux. So it's just more likely for intra genomic, we have then we lose the decreasing relationship and for nested mobile genetic elements, we seem to have a tendency that is contrary to the transfer at large distance seem to be favorite. It means, and we have a similar effect of interaction between mobile genetic and carriage and codon usage. So it seems that mobile genetic element carriage allows to bypass this in some cases the classical barriers of two horizontal gene transfer. So to summarize, we have shown an important role of mobile genetic elements in resistance accumulation in genomes and in resistance spectrum broadening. We have identified a large role of non pathogenic bacteria in the propagation of amino glycosine resistance genes. The oriented horizontal gene transfer networks confirm the classical barriers to horizontal gene transfer, but we also showed an important role of the mobile genetic element, not only as vehicles transporting the resistance genes but also as means of jumping over over the barriers to horizontal gene. To finish, I would like to acknowledge that it's mainly the work of Lea that I presented. I also would like to thank other members of my group and Anastasia Pistolavi with whom we collaborated and funding and you for your attention. Nice talk. Really, really neat. Is there any questions? Sorry, I'm not looking at the chat. No, it's empty. Okay. So is there any question for Stephanie? No, so I'll try. Okay. Just, I was thinking, you found that E. coli is always a receiver and we have, I mean, we thought as much. There is a question from Berenice Mayer. I cannot see her. Sorry. Thanks, Joachim, for pointing out. Okay, maybe I just go ahead. Yeah, thank you very much. That was very interesting. So I was wondering about this role of some species acting as donors and other acting as receivers. Do you think the reason for this is mechanistic or is this caused by selection mostly? I have no definitive question, definitive answer to this question. I guess there is a role for selection in the sense that the resistance genes that is arriving in pathogenic species that is more often treated by antibiotic is probably going to be kept. Whereas if it's transferred to species that is never encountering the antibiotic are very rarely likely that we don't see it or we less often see it. So, of course, in this type of study, we have only the successful part of the story and that's what we analyze. This said, I think so, so we also so that's data that I didn't precisely showed but there is more transfer from non pathogenic to pathogenic species than just expected by the frequency of these two categories. So the selection is acting on the pathogenic species, whatever the origin of the gene they receive comes from so there seem to be something more than just the selective. Then the precise mechanism by which this is happening that could run from molecular mechanisms to more ecological arguments so probably the probability of contact and transfer is more in one direction than another. So that's that's, we didn't find very, very clear cut examples of why this could be. Okay, thank you. Okay, there was the king one with the hand face. Do you want to ask your own question. Yes, yes. Yeah. First, thank you very much. I'm interested in the how to infer the donor and recent pen shape. I also construct the trees focusing on emerging spot only from the general within and for back Thursday. And my trees always contain polytomy or the tree is like a comb. They do not have enough snakes to infer a reliable donor and the recipe and shape. And I don't know, can you comment more on the inferring of donor and the recipient in your data, how reliable is that and do you have any suggestion how I can infer the relationship in my data due to the the nature of emergence I found that in general, they are quite conserved within my data. Yeah. Yeah, so, so resistances are quite conserved so if you try to reconstruct phylogeny of the genes on one side and the phylogeny of the species and the other side and apply reconciliation methods for example I don't know it's, I don't know if it's what you try to do. Usually it's not possible because the usually there are many cases in which it's not possible. So that's, that's not the method we use also because the data set was so large that it was actually competition only very complex to do that on such a large data set. So that's why we used and this implicit phylogeny method, which I, which is work so Leah actually did simulations to try to see what was the lowest phylogeny level at which we could relatively detect the transfer. So this means that you need your distribution of the vertically transmitted gene, not to go to low. So if you have two species that are really close or two genomes that are really close. Of course, your resistance genes are never going to be outside of this distribution. So that's a method that works well for the genus and the bone. But below that the method we use is not reliable, either. So that's actually why the majority of the networks we reconstructed are at the genus level. But so I, so probably you work in intragens. You cannot use this method. Yeah, yeah, yeah, thank you very much. We have a last question I think from Jose Delgado. But I'm messing up with everything. Do you want Jose to ask yourself the question or do I read it? Well, it's not answering I read it. So he says, thanks for a great talk, Stephanie. Have you thought about using this approach for other AMR mechanisms in order to identify their potential environmental donors. Yes, doing it is. I mean, it's quite intensive work so it was three years of PhD player but I think I mean I agree with the question that now that the analysis are set up and the way you do them is set up it would be really interesting to apply that to other, to other resistance. It is important to note that it can only be applied to resistance that are due to genes, to resistance genes and not to polymorphism to snips appearing in the genes because that's really need to identify these clusters of homologous genes. But yes, of course it would be would be interesting and actually interesting to test whether the pattern that we see for amino glycoside or specific to amino glycoside or if it's more general with other resistances. Okay, so if there are no further questions and doesn't seem to not, we have Mario Santer. Mario.