 OK, so I think we can move on. And the next speaker is Ignacia Cancino Aguirre, talking about metabolic variability in mycobacteria. Ignacia? Hello? Yeah? So good morning, everybody. I'm Ignacia Cancino, and I'm a PhD student at Inriaca Nobel in France under the supervision of Hida de Jong and Delphine Robert. And I'll talk to you about this new project that we started, that it's about the analysis of mycobacterial metabolism using genome-scale metabolic models. So just out of motivation or context, mycobacterium tuberculosis is a bacteria that causes tuberculosis. It's a disease that kills more than 1.6 million people every year. And as you can see, it's present in almost every country. And despite all the efforts to eradicate it with vaccines and antibiotics, it's still present. And also, it's become more and more resistant to antibiotics, so it's like a big problem. But MTB, or mycobacterium tuberculosis, is part of this genus, the mycobacterium genus, which comprises around 200 mycobacteria. And so it's part of this genus, along with other pathogens like mycobacterium lepra and mycobacterium ulcerans, et cetera. So we were very interested in looking at the genus level of mycobacteria, because they're very interesting. Even if they're genetically kind of similar compared to other bacteria, they show different phenotypes. So half of the genus divides in slow growers and the other half in fast growers. And there's one clade that is kind of intermediate growth, but that it's weirdly defined. So this is very interesting. And there's been attempts to understand a bit why, but not that many. It is interesting because a lot of the slow growers are also pathogenic bacteria, but not all. So just as an example, some of the slow growers there's MTB, there's ulcerans, mycobacterium marinum, that is pathogenic usually to fish, but there's been clinical cases of infecting kids. And there's also other, the fast growers are mycobacterium spegmatis, for example, that is non-pathogenic, and it's found in the environment. But it's not clear this pathogenicity to growth rate relationship because there's other bacteria like mycobacterium abscessus, which is also pathogenic, but it's a fast grower. But how do they define what is a fast grower and what is a slow grower? So usually, and this has been done for 100 years, whenever they find a new, let's say, mycobacteria, they plate it in this LJ complex media, rich media, and then they let it grow, and then they see how long they take to form colonies. And if they take longer than seven days to form colonies, then they will consider them as slow growers. And if it's less than that, it would be a fast grower. And this LJ media, it's very complex. It has egg and potato starch and things, and it was developed in the 1800s almost. But it's still being used for clinical diagnosis and isolation of mycobacteria. But it's very interesting that even that growing in this media, you could see also the relationship in the phylogenetic tree. So there's been, of course, some attempts to understand this difference in growth rates. But it's really not that well-studied as one would think of versus difference in lag. They've always mentioned growth rate, and they haven't really mentioned lag time. So I'm not that sure. Yes. I mean, it could be lag. But then sometimes when you grow it in liquid media, sometimes you measure observance, then there's no lag. It's just very, very slow growing. But yeah, it could be also the lag time, I guess. Yeah, so some of the issues to understand this difference. One is very costly experimentally, not just money, but also time-wise, because they grow very slow. I mean, yeah, they can take like, I don't know, two to three weeks to do one growth curve. Some of them don't even grow in liquid media. Because they're pathogens, you need these biosafety level three laps or biosafety level two laps. So yeah, there's not that many studies. And a lot of studies focus more on MTB infection in host cells. Then also pathogenicity is kind of a weird defined concept, because there's no clear definition, at least in this area, about it. Like some people say, OK, if it infects somebody, then it's pathogenic. Some people say, oh, but if it infects, let's say, many people with healthy immune system, then it's pathogenic. If not, it's just opportunistic. Some of them infect other mammals. Also, they don't necessarily optimize their proteome allocation. So if you want to see it that way, we saw in a Terry's talk last week, he studied smegmatis, which is the fast grower for us. But for E. coli is a very slow grower. And they weren't optimizing their proteome allocation to grow more. They would invest in other functions of the cell. And also, if you would explain it by the number of our RNA operon copy numbers, they have very low copy numbers. So most of them have like one or two. And still not clear, because some fast growers, most of fast growers have two copies, and most of slow growers have one. But a few of the fast growers have one, and then, yeah. So it's not like a clear relation. And also, these studies have been done by very few bacteria. So I guess you could maybe look more into that, if you want, because they're usually done in maybe four or five of these bacteria. But anyways, this is not enough for us, I guess, to really understand why they're growing differently. So we were thinking that one way of studying this growth rate variability is by studying their metabolism. Because you can study all the reactions that are happening, that account for biomass production, and then you can relate it to their growth rate. So the approach of this project is using genome scale metabolic models, as we have heard these past two weeks. And they model the metabolite dynamics in the cells. And then they relate the stoichiometric matrix that has the stoichiometric coefficients for all of the metabolites in reactions. And then you multiply it by a vector of the reaction rate. And then if we assume that they're in a stationary state, then we can equal that to 0 for the internal metabolites. And then we can also constrain these fluxes with either experimental information or thermodynamic information, like directions, et cetera. And then we will have a constrained space of solution for this linear system. And then we can analyze it in different ways as we saw. So we thought that this method could be useful for us to understand why they're varying in growth rate. But of course, there are some issues. And some issues with this approach is that there's only two genome scale models for mycobacteria. So mostly for MTB. And then there is some strains of TB that have been done. And then mycobacterium bulb is BCG, which is actually the one that they use for vaccines. But they're very, very similar as well. And actually, they merge these models at one point and use that one to analyze both of them. Recently, they have of course differentiated them a bit more. So yeah, that's the only mycobacteria that has a model. So we would have to reconstruct more. And then of course, there's the experimental limitation size I mentioned before. You need a biosafety level three and biosafety level two labs. Some of them don't grow on liquid media. Like mycobacterium ulcerone doesn't grow on liquid media. And some of them, I mean, and a lot of them take a lot of time to grow. And also to go from solid plates to the liquid flask, it would take like a month or something to adapt to those conditions. And they also tend to clump a lot. So it's a bit annoying to do these growth curves on liquid media. But luckily, there's the reference genomes available on NCBI. And there's also experimental information on single bacteria or single species. And there could be some information in general about the phenotypes and maybe some pathways. And we also have this, that I forgot to say it, a collaboration with the Francis Creek Institute in London, where they focus a lot on micro-bacterial metabolism and antibiotic resistance. And they have these biosafety level three labs. And they do very nice stuff as well. So the idea of the project is to develop genome-scale metabolic models for the different micro-bacterial species. And so validate those. Then at the same time, obtain exchange rates and growth rates for these different bacteria grown on the same conditions. And then after we have these two components, the idea is to analyze them. And because I said that micro-bacteria don't really optimize things, strategies, we would do the analysis by uniform sampling because we don't want to assume optimality. And then the idea is to compare how the flaxes behave in the network between the different types of micro-bacteria. So first of all, we started by asking ourselves, OK, but can we actually infer the intracellular flaxes with the already published model of TB? Because if not, then this approach wouldn't really hold. And so we found one data set of carbon-13 measurements of the intracellular flaxes. And so we constrained our models with the exchange rates that they measured and a few thermodynamical information like reaction direction and the growth rates. And then we used uniform sampling. And then we also removed some of these infeasible loops with cycle-free flux. Maybe some of you might know it. So we found that this is only four of the reactions. We have the results for all of them in the network. There's like 1,300 reactions. And we found that for most of them, our simulations were at least in the range of the experimental value. So we have the blue curve, which is our simulation from the sampling. And then the red dashed lines are the experimental value. So we thought, OK, even if the mode is not necessarily the exact value, they're still in the range. And then we can infer flaxes, at least, with this model. So then we continued. And then we started the part of reconstructing GM scale metabolic models, which is like a actually long and tedious process. So we tried to use automatic tools because in the beginning, we wanted to do it for the 200 species. And so there was no way that I, a PhD student, like learning about this would do it in three years. So we looked at some automatic tools that are already, I would say, widely used. I mean, they're from 2018. Carth me, maybe some of you have heard it. And they create already like a connected network. So they do align the genome to a database. But then they also make sure that they are connected. But the issue with this is that sometimes it tends to add more reactions that there actually are. So you still do need a step of manual refinement, or maybe some gap feeling. I mean, it's not too automatic. And we also, because we saw this in the beginning, and we're not satisfied with the model it was creating, we also implemented some other filters for Carth me. So yeah, some like playing with the alignment parameters. And we also updated the database they had with this new MTV model that we thought it would be useful. So then we tested. So we tried this pipeline with tuberculosis, because there's already a curated model for that. And then we tested it with this growth metrics. So in the lab, they grow TB in single carbon or single nitrogen sources. And then they say if they see some growth, it doesn't matter the rates. But just if they see some growth, they would classify it as one, and if not, zero. And so we did the same with our model, doing FBA. And just, I mean, we didn't care about the value of the growth rate, but just if it was higher than zero. And so we actually, in our model, it got very similar predictions to the reference model that is already published, which is nice because it was mostly done automatically. And it's slightly better than the reference, but not too much. It's actually very similar, except that it has more reactions. So it has 1,700 reactions and 1,200 metabolites. So this was nice. But then we're in the part of the working progress, which is applying this pipeline to other species. So in the beginning, we thought of doing it for the 200 species. But then because it still needs some manual refinement, we thought of, for the 200 species, just doing functional genome analysis. So to see which genes that codify for enzymes that have reactions associated to them, which ones are present in the slow growers and in the fast growers. So not really quantitative. But then we would take from those species 18 that we're actually growing in the lab and that we are producing this growth metric. So does it grow in single carbon sources or not? Because those we can actually validate them with some data. So then we would do the models of those, validate them with these phenotype data sets. And then from those, because doing the growth curves take a very long time, we thought of starting with at least five species that have this difference in growth rates and actually do the growth curves of those and measure growth rate and metabolite exchanges. So that would be like, that would be five species. And yeah, and that would also be more carefully curated. So then this is the experimental part, which I'm helping out but I'm not the only one doing it. We have a researcher in the creek that's actually continuing the experiments, but we kind of discussed it together and I did a few of these curves. So we chose to do the growth curves of five species. So from those, we chose three fast growers and two slow growers. And for the fast growers, there's one that's pathogenic and the others are not. And then for the slow growers are both pathogenic. And then we decided to grow it in a defined media because the LJ media that is actually used to to classify them as fast and slow. It's a complex media and there's many stuff and it's very variable. So we cannot really rely on it to have these quantitative rates. And so we used 7H9 media, which is like a very gold standard media for mycobacteria and it kind of is not the same composition, but it does resemble the phenotype seen with the LJ media. And then we would use the optimal temperature of those and yeah, about temperature. There's two of these that grow at 30 degrees and others at 37. So we would use the optimal temperatures of those. Yeah, and then it would be aerobic growth. So shaking the flasks. And we thought about, okay, which media should we test? And then we decided to do it on the 7H9 that has a glycerol glutamate and glucose as carbon sources. And then we also thought, okay, maybe we could also see how they behave or how they change their growth rates if you just throw it on a single carbon source media. So we chose glutamate and glucose. We tried also with glycerol, but they tend to clump a lot, a few of them. So then OD measurements are not really reliable, even if you're using detergent. And then the idea is to measure growth and also measure metabolite uptake for those over time. So all in all, the project is to study micro-bacterial metabolism to hopefully explain better why they have this growth rate variability. The approach that we're using is this constraint-based modeling and implementing experimental data. We're in the process of reconstructing the models for the other micro-bacterial species and testing those at the same time, obtaining experimental information. And so what is left to do, which I think is the most nice part of the project is actually comparing the flux spaces between the different species and sea if we could find hopefully some clues of why they are acting different with respect to growth rates or maybe why some pathways are more active, let's say in some group than others. So just to finish, I want to thank my lab, my supervisors, Elfin and Hida, and we have a new postdoc now that's also helping a lot with the bioinformatics and our collaborators at the creek. So the principal investigator is Luis Pedro Carvalho and Azeli, she's helping me with experiments there. So yeah, thanks then. Question? Thanks a lot, we have time for questions. Hi, thanks. So, I mean, I know you're not on the experimental side, but can you elaborate a bit more on why you chose to take glucose, glutamate and glyceros, so what was kind of the rationale behind selecting those carbon sources? Yeah, so I mean, they were based on the gold standard media for TB or for mycobacterial identification, which is apart from the LJ that's complex, the 7H9 middle group media. So that media has these three carbon sources and then already because there's been most of studies, they do it on this media. We have more experimental information to kind of validate our models, for example, and also to compare and, yeah, so we didn't want to create a new media from scratch. So that's why we chose those three carbon sources. Other questions? Yeah, this is a curiosity, so it's the first time that I think about this mycobacterium, but it seems almost like there's this established pattern of the slow grows being more virulent or being more pathogenic. From the physiology, is there any, what's like the typical explanation that people give? Are they? So there's not like a clear answer also because there's a lot of the fast score that are becoming more pathogenic. Apparently they've seen that the pathogenic one, they have lost some genes so they can only grow intracellularly. Yeah, they have some proteins, I think that they say that related to virulence, but they're not, but like the fast growers also have them, but just like a difference in those proteins. So yeah, it's not actually that clear, the physiology of them. Actually, I want to follow up on this question. Can you maybe, I don't know, say where the fast growing or non-pathogenic bacteria found in which kind of environments? Because I imagine, so for example, for E. coli, we see that UPEC has way lower growth rates than our lab strains, of course they've also been involved. So maybe there's also something going on with they actually benefit for changing colonization habitats when they grow slower, I don't know. Yeah, so most of the fast growers, they have been found in soil and water bodies. There's no specific trend, but just more like environments, I would say. While the pathogenic are usually found intracellularly, I guess because they already infect. And yeah, I think it has to do also, there were some studies looking at that the fast growers can use more diversity of resources, while the pathogenic ones, they can take only a few of those. So I think it does affect that. So for MTB, let's say, the kind of environment that it grows in when it is a pathogen, is that known, what is that environment and can that be sort of simulated in these flux balance sort of theoretical studies? Yeah, so MTB grows on macrophages in the phagosome of macrophages in the body. And there has been actually very interesting studies with these published GM scale model of TB on the host, so macrophage model and the TB model and how they interact. So yes, it can be modeled. Also, you could also do it in the lab. So what they do in the Louise lab, they take macrophages from blood of donors and then they infect with TB and then they study different things, like I don't know how pH changes, for example, and things like this. But I actually wanted to, in the beginning, I mean, I thought of like, oh, what if we can do growth curves in this macrophage-like media, but there's actually no macrophage-like media at the moment. They really just take the macrophages, but that takes forever. I mean, it takes even longer. And the thing, not just the time, but also it changes from patient to patient. So if you have fresh blood, I don't know, with better macrophages, then it's a nice experiment. But then next week you get new blood and then it's like not that great. I don't know, and they don't really have information of the patient because it's confidential. So yeah, I think it would be very, we talked about this in the lab and it was like, oh, it would be very nice to have a synthetic media that resembles the macrophages, but then I think that's like another work. And so everybody has been pushing it, kind of. But I think it would be very nice to do that. Yeah. Any other question? Okay, let's thank the speaker again. Okay, see you.