 for a workshop, the, well, I think hold on, there you go. So one of the major themes in this workshop is, of course, has been this idea that there exists feedbacks between microbial growth and microbial ecological and evolutionary processes. And we've just saw a wonderful example in the talk by Kaling with Sijin about how antibiotic production can mediate the assembly of complex microbial communities. Jeff and Jonas also gave some examples related to how microbial growth produces changes on P8, which in turn mediates complex interactions. And Pankas talked a little bit just about the idea of cross-feeding the production of metabolic byproducts as another mechanism that can generate the assembly of complex communities. I wanted to just give you one of my favorite examples on experimental examples of this. This is a paper by Rosenzweig Adams at all from about 23 years ago. And this experiment was an experimental evolution experiment. They started from a clonal population of E. coli that they grew on a chemostat under conditions of high glucose concentration. And they evolved this system for a few hundred generations and a really wonderful thing happened, which is that after about 700 generations or so when the authors examined the content of the chemostat, they found that the original strain that they had started with had diversified into three different strains. One of them had an enhanced ability to grow on glucose. And of course, when E. coli grows on glucose, it produces secondary metabolites byproducts of glucose metabolism that it secretes to the environment, such as acetate and glycerol. So as E. coli had become better at consuming glucose, it had been also become a stronger secretor of acetate and glycerol. So it had thus transformed the environment from a purely glucose environment into a much more complex biochemically environment. And those two extra niches, the acetate and the glycerol had been exploited through the evolutionary emergence of mutants that were specialists in the consumption of those two resources. So this is an example of how a multi-species, multi-genotype, if you will, ecosystem can spontaneously assemble from even through evolutionary processes in just a few hundred generations and you have through the construction of new niches from the provided one. So all of this experiment was done in minimal media with a single carbon source, right? And this is the fine medium. So because of these and many other experiments, this is by no means the only one that has seen things like this, then that prompted to ask the question of whether complex multi-species communities can also be assembled in defined media with a single carbon source. And this is beyond the fact that I think it's a good question to ask. And I was really curious about it. It's very useful because if we can accomplish this, we could use such ecosystems as an individual model systems to investigate micro-community assembly. Now as every other community assembly process, micro-community assembly is very complex in nature. You can think for instance on the question of why do different hosts or different people have hardware different microbiomes and the overall, the number of reasons why that is is very complex, right? We're all different genetically. We have different lifestyles, different diets and that produces of course environmental filtering, right? We might be selecting for specific taxa differently from person to person. On top of that, we all live in different places. There's a very international audience. There's people living here from Pakistan, from Israel, Italy, I live in Connecticut. And we all are exposed to different microbial taxa, right? I cannot have on me, microbes that are not in the immediate environment so they cannot invade me, right? And on top of that, there's other processes. There's the historical contingency, which at the end of the day it's a combination of the stochastic timing of arrival of colonizers and invaders and the population dynamics between them that can also generate diversity. And you also, of course, have random sampling from the environment. You can have transient microbes that are present on us just for a short period of time and they will be flushed away over time. So as you see, there's a wide range of processes that can lead to community assembly in nature and not only that, they all happen at the same time, right? And this is not an exhaustive list. I'm sure you can think of others. So the question then is if we're able to have a model community that, where we can study the assembly of complex communities with many, many species, we might be able to disentangle all of this, all of these different factors and study them one at a time. To give you a few examples of what kinds of things I was hoping we could do, I was telling you that in nature, the different habitats, and not only this is true for host-associated communities in any kind of ecological setting, you'll have microbial communities that can be different from habitat to habitat. And in nature, there's habitats that are highly heterogeneous and they're biochemically diverse and that is really hard to control, right? In contrast, if we're able to have communities growing under well-defined media, we can absolutely control what goes in there, right? We can define every nutrient that we add and that would allow us to also connect whatever population biology we end up discovering with the metabolic processes that we can infer based on the metagenome of the community, right? That would allow us to connect metabolism with population dynamics in a very, very concrete way and precise. But there's other benefits of having basically experimental community assembly in vitro, which of course is, for instance, the access to the regional pool of species. In nature, communities get, not necessarily are all invaded by the same regional pool of species. In vitro, we can control what pool of species invade which other, which environments. We can create absolutely identical environments and we can invade them all with the same, from the same regional pool of species. And we can do this in very high throughput, right? So we can get statistics on the community assembly process that will be very hard to do under natural conditions. We can also control the timing of those invasions and in a way that, again, is very hard to control in natural settings. So I'm just giving you a kind of a longer list of why it is that I'm interested in this question and I hope I'm convincing you that it's worth at least pursuing. So the question again is, can in principle complex communities with, and my complex, I mean, Jeff, for instance, was discussing yesterday communities that are maybe N larger than one, two, three species. Now I'm thinking of as complex as possible, right? Like how many species can you fit on a single carbon source, right? And minimal media, nothing else, right? So, and that's kind of what we're after. Our approach is really rather simple and it consists of the following. We collect environmental samples, we've done all kinds of things, mainly soil bacteria or aquatic environmental communities. And what we do is we take, for instance, a leaf sample, we collect it from a pond near our university and then we stick that into essentially water with a few salts, incubate a couple of days with some chemicals to inhibit the growth of fungi and nematodes and other eukaryotes. And at the end of the day, after one day of incubation, we do is filter all of the larger particles and we're left with the bacteria. So essentially we end up with a bag of bacteria, right? That came from a natural community. And what we then do with that bag of bacteria is that we take it and we now grow it in batch culture mode. Essentially what this means is that we put it in minimal media with a single carbon source, grow for 48 hours, then take one hundredth of that community and transfer it to a brand new, basically a reactor with fresh media of the same kind, right? And we keep doing that for many, many generations, right? So it's basically growth dilution and growth dilution and we keep repeating that for anything between, we do this every 48 hours, we've done this for anything between two weeks and a month. So anything in the order of roughly 80-something to 150-ish generations. After that, this is bacteria then, so we can freeze our communities every day, right? That's very convenient. And at the end, and as well as at different time points, we establish which members of the community are present by Illumina sequencing, we do 16A sequencing. We're now about to start doing metagenomics, but we haven't done it yet. And we have done this now for a really large number of samples, both in Massachusetts and Connecticut. These experiments have been running for several years at this point. And I'm gonna be focusing on the work that has been done since I moved to Yale by my lab members. So these are mainly communities from different workings in Connecticut. All the results I'm gonna tell you are perfectly consistent with the experiments I did before joining Yale in a different lab with different water and different persons doing the experiment. All these things actually end up mattering more than probably you may think. But I'm very happy to say that everything you're gonna hear today has been replicated in two different labs, okay? So that's kind of nice. So we collected samples from, in this case, 12 different locations. These are both more kind of built environment, soil communities, pond communities, et cetera. And yes. Oh, we did. Now we did sequence these two, right? We sequenced that there is no communities, we sequenced the final two, right? And everything in between. These communities are very highly diverse. We're talking about hundreds to thousands of taxa at a relatively shallow sequencing depth, right? So these are incredibly diverse, yeah. So we tried all those 12 communities on a single carbon source, and what was that carbon source? We've tried a lot of them. And I'm gonna focus in here on the ones that we tried at Yale, at Harvard we tried 60 of them. But at Yale we focused on seven of them. One is glucose. And the other, we took like three carboxylic acids and three amino acids. But we've tried all kinds of weird stuff. And again, we get relatively interesting results to with other types of carbon sources. So in summary, this is like 12 different environmental samples that are reared on a single carbon source. That's essentially one of the seven that you see over here. And we grow in batch culture mode for about, and this experiment I'm about to tell you about is 80 generations, and we keep sequencing as we go, okay? So that's the experiment. So the question is, if you do that, right? If you take a natural community and you force it to grow on the fine media with single carbon source, do you end up with a community that still accompanies community, multi-species communities interacting with each other? And the answer is that resounding, yes. We see this, it's completely generic. It doesn't matter what the carbon source is. It doesn't matter what the initial inoculum is. We always see a large number of coexisting species in every single experiment we've ever done. And these are the 12 different inocula, the seven different carbon sources. The diversity depends substantially on what type of carbon source you use. But we've seen anything from three to several dozen species. And again, we're not even sampling all that hard, right? We're sequencing about 100,000 cells out of a population of close to 10 to the nine, right? So there's certainly more than this, right? But this is what we're seeing. The communities are stable, right? So this is time courses. This is the fraction of the various species. These are three different glucose communities. And the frequencies of stabilized around a generation 50 or thereabouts. We see this consistently. We have some communities that are stable, not quite maybe have some transient stuff going on. But for the most part, we see communities that are stable. And not only in composition, but also in other properties. We measure the OD, the optical density, the number of cells that again, stabilize up to about 50 generations. And the richness and other properties of the communities themselves are stable. So communities are complex and stably growing together. So I was telling you at the beginning of this talk that our way of thinking is that if cross-fitting can lead to the spontaneous emergence through evolutionary processes of a community from a single isolate, right? Maybe cross-fitting could also be able to support large ensembles of species coexisting together on a single carbon source. So we wanted to see it actually if the reason why we see in this coexistence is cross-fitting. There could be many other mechanisms that can help too, right? And we're not ruling anything out. We've done controls for temporal and spatial niches and found very little evidence for either of them. But of course, we haven't tried all of them. We have about 1,000 communities in our freezer. We haven't tried for all of them. But we've zoomed into one of them, which is one of these communities that we have grown in glucose. And this is our representative community because it's a relatively simple one and we could isolate every member of the community. It's represented by about five different taxa and we isolated all of them by plating. We collected the colonies. And what we're doing is we're taking all of these different members of the community which we isolated them. And we grow them for 48 hours on glucose as the only carbon source. And after that period of 48 hours, we take the supernatant, basically all of the molecules that the molecules have secreted over that period of time. And we use that as the only carbon source on which we're gonna grow every other member of the community, right? So we take, we isolate all of the members of this community. We, all of them actually grow in glucose. That's the first bit of information that's useful to know. So all of them can grow fairly well in glucose. And they also all secrete molecules that in principle we wanted to know if they can, those extra several molecules they have secreted could provide the basis, the carbon source for every other member of the community. Okay? It actually is off. Yes, yes. Well, I mean, to the level we could detect with that. It's just like pre-fine, right? So yeah, so glucose was pretty much gone after 48 hours, okay? And these are an example of the type of data we generated here. The emitter is intervector. We're taking the molecules that are intervector secretes. We're feeding them to pseudomonas. There's no other carbon source here, only that. And these are the growth curves. In gray, that's pseudomonas growing in glucose, a 0.2% glucose, which is the concentration we use. And in black, those are multiple replicates of pseudomonas growing on intervector secretions. So as you see, the growth is fairly strong, right? So it's not only that it grows meekly, but it's comparable to growth in glucose. And in fact, it's faster than that. It's less efficient. The carrying capacity is lower. So it's not all that surprising. But the growth is very strong, right? And we repeat that experiment for every possible pair that we could have in our communities. And we end up, this is a bit of a mess, so let me walk you through it. Every node here, these four nodes are four members of the community. Here is glucose. And these arrows pointing from glucose to each of the members of the community reflect the fact that all of those members can utilize glucose as the only carbon source. And as you see, this is a fully connected network. And what this means is that the supernatants of every member of the community can support the growth of every other member of the community, right? So these are completely degenerate and full network, right? Everybody's cross-feeding everyone else, right? And I decided not to include this data because the talk is too long already, but we see multiple dioxy shifts, meaning that there's not just one thing that every member of the community is cross-feeding every other member. Oftentimes, there's multiple things that each other, yeah, are utilizing. No, we don't, yeah, yeah. This is, well, I mean, this is phosphate buffers, so there's some, I mean, some degree of that, but we don't have trees, we don't have any actual buffer on it, right? So I wouldn't expect the pH to remain constant during the experiment, yeah. Yes? That's right, yeah. No, yeah, yeah, we don't know exactly what's going on and because it's true, I mean, they level off, right, after 48 hours, but then when you take the supernatant, dilute it with M9, then they keep growing again, right? So we don't know what it is and it has been suggested to us, it actually could be pH that is being, as you replenish the M9, you kind of bring the pH back up and that allows them to keep growing, right? So, yeah, I mean, that indicates that growth is not perfectly efficient, right? The microbes have some growth capabilities still on the molecules they secrete, but they don't complete it for some reason, right? And we still don't know what it is. No, we haven't, yeah, we haven't done it yet, yeah. Yes? I'm sorry? Oh, cellicis. It could be cellicis, it could be the molecules they secrete it, right? I mean, we're filtering the supernatant through a 0.2 micron filter, so if the cell has lives and there's like, contents that are smaller than that, they could be there, but we're not controlling for that at this point, right? Nope, we haven't, we don't know, that's a really good question and we get that a lot. We haven't done any analysis of the viral of these communities, but they may very well have viruses here, they might have phages, yeah? Yes, I couldn't hear you. No, we haven't done it, no. Those are all things that are on the pipeline, yeah. We have right now two new postdocs that are working on, they're very interesting in that question, in fact, yeah. All right, so this is it. I mean, I'm not claiming that I know exactly what's going on. All I'm telling you is that if you take the supernatants of each other, they can cross-feed, and it's perfectly well connected, right? So at least this is suggesting that cross-feeding plays a role. Whether that cross-feeding is coming from the cells bursting out, either because of cell death or a phage, or any other reasons we really don't know at this point, right? But what's very clear, is if you take the non-glucose carbon that's present in the media that has been produced by the cells, right, and you feed it to each other, they all can grow on each other's secretions, right? I'm sorry, I'm having a lot of trouble understanding. Sure, yeah, there could be all kinds of things, right? I mean, we don't know what's there. There could be amino acid, there could be, yeah. We have no idea what's in the media. Haven't analyzed it. Okay? So the question that we started with is whether we can assemble complex communities, and by complex I mean large N, and we're finding anything from two to several dozens, and again, we're not sampling everything there. So the answer is yes, we can, right? We can do it in a minimal media, well-defined environment. The environment becomes, as you guys are guessing, not as well-defined as the microbes go on it, and that's a fair point, right? But at least, and we have some evidence that actually the type of carbon source you add has a major importance in what type of community you make in the end, and that at least is reassuring. So we do find complex multispace communities forming and cross-feeding is widespread and promiscuous. So we get to actually what I really want to do with all of this, which is try to understand community assembly using these communities, and can we use them to derive any rules of assembly that are governing the assembly process? So for that, we focused on one of these, we took all of these 12 communities and looked at the assembly on glucose. We looked at the community structures that formed after 85 generations or so on glucose. We looked at, okay, so we know it's a complex community, but what is the community, right, who are there? So if you look at it at the level of sequence variants, which you can take as a proxy for species, if you will. So these are unique 16S sequences that we find in our sequencing. What you see is that each inocula, basically each starting community gives rise to a different stabilized community in glucose. And so we take all of those 12 environmental samples, propagate them in glucose for 85 generations, and look at what's there in the end, and perhaps not surprisingly, we see that they're different, right? That you have different taxa. However, if you look at this at the family level, they're all absolutely identical, right? Like, they all have the same family level structure. It's about 70% plus minus 10, 15, if you will, of interactive AC, and 15% plus minus something through the monads, right? And there's a bunch of further taxa that are low abundance that are also consistently found in all the communities, right? So what this suggests is that if you will look at this as if there were like a family level attractor, right? On these communities. And you can visualize this by plotting on the left side, this is the community structure before, at times zero, right? Before we transferred, we propagated the communities. And this is a low dimensional representation, of course, and on this simplex, this edge represents interactive ACA. This edge is to the monad AC. And all of the other families are condensed into a single, basically a dimension. And as you see, the starting points in the simplex are very diverse for all 12 communities, right? Some have very little interactive ACA and synomonad ACA, others have more. But after the four generations, they're all here. So that's an excellent point. We haven't, yeah. We've actually been thinking about it, but haven't done it yet, yeah. It's a very good point. Yeah, no, we thought about that. And we're now starting to experiment, we have migration, even from the original pool and connect them, we have everything to do that experiment. So we see this attractor here in glucose. And we wonder whether the attractors in other carbon sources are any different. And to good news, there's still attractors for the carbon sources. Not some freak accident of glucose. But they're different, right? So these are 96, the red are 96 glucose communities that are all converged after 84 days to the same region in the simplex. The blue are citrate communities that after, again, same number of generations have converged to a slightly different place. And in green you have leucine. Leucine communities are very different from glucose. I'm not gonna talk about it either, but we've done metagenomic analysis by inferring the metagenome from 16S. And we find very, very strong evidence that essentially metabolic demands imposed by the carbon source we add is critical in order to determine community structure. For instance, the glucose communities are heavily enriched on PTS transporters, which are essentially glucose transporters into the cell that are very efficient and which are very present in intravactive ACA. Whereas the leucine communities, all of those members are enriched on leucine degradation pathways. I mean, perhaps not surprising. But so we see that that that's actually very important. On the other hand, if you plot the same data on the same simplex, but now color by the community of precedence, you see that there's not much going. So there's not much going on, right? This again points to the metabolic constraints imposed by the carbon source that you're feeding them rather than the community of origin as what determines what is the final structure of our communities. So the question I asked is are there rules that govern community assembly? And what we're finding is that there are attractors to the family level. I mean, you can predict what's going to be the structure of your community if you take some random soil sample and put it in glucose media. And you know what's going to happen in the end, right? It's going to be directly ACA, and sooner or later they say, and even the ratios are quantitatively very well conserved, right? And so we actually have repeated this experiment in a different lab, different with soil samples from like another state, we get the same thing, right? So that's what I was saying in the beginning. This results are very robust. Yes, two more. Well, that's an excellent question. And to me, one of the reasons why I wanted to do this is because if I know what I'm putting in, now I can build up, right? I can go from one and I can keep adding more and more carbon sources. And one very simple question is, as you add more and more carbon sources, do you get more and more diversity, right? Or- Do you mean more on the strain level? Oh, do you mean on the strain level? That is a good question too. We're doing actual metagenomics now, so we'll find out, right? And maybe we can get back to that in a minute or two because I'll show you some data about that on the strain level. Yeah, so families, so the core metabolism at the family level are very conserved. There's all the accessory metabolism that is different from states to strain. It seems like it's this core metabolism that's driving this, given what we see at the family level. That's a good point. So okay, so we see this family level attractors, but we also see this large variability at the species level, right? So we wanted to understand the origin of this variability, right? And there's essentially two different processes that could lead to this, where you do these experiments with 12 different starting points, that you get the same family structure but different taxa, right? And one is that the representatives of its family on different environmental locations could be different, right? And again, you cannot have in your final community what you did not have a time zero because there's no more migration. So that could be one explanation, but there's another one that I like better because that's kind of stuff I like, which is that there could be a combination of stochastic sampling and micro-micro interactions that through historical contingency can also give rise to alternative stable states, right? So we wanted to see which one of these two was explaining the diversity that we're seeing. And in order to do that, we selected just one of these communities and for now one of the carbon sources, glucose. And what we did is, and again because we have this degree of control over the system, we started eight replicate populations from the same pool of species, right? So now there's no variability on the regional pool. There's all the same regional pool and from the same regional pool, we're inoculating eight tubes. Now those eight tubes are absolutely identical to one another. They're the same media and they're inoculated from the same regional pool of species. Now of course, right? Even if you inoculate from the same regional pool of species, there's some stochasticity in sampling, right? That's unavoidable. And if you think of the large diversity and the fact that the system is very rare at time zero, there could be some differences from community to community when you do that, right? So we wanted to know if you do an experiment like this, what are you gonna see? Right now you're eliminated variability in the regional pool of species. Do you still get variability at the species level? And the first thing we looked at is family level distribution. Again, no surprise, we get the same thing. We've seen before 70% of the monas, more or less, I'm sorry, intravagtuacy and pseudomonasis is the rest, right? No surprises here. So let's look at the, let's start zooming in and let's look at the genus level. So when you do that, we see that three out of eight communities, when you look at who makes the intravagtuacy slot, that's made up entirely of Klebsiella, right? That's a low diversity intravagtuacy slot, if you will. The other five are made up of an alternative community state made up by a coexisting guild of three intravagtuacy. And the reason why we chose this particular system is because those are the same that we had isolated before that I told you they cross fit one another. Those are the three guys, right? So that's why the reason why we chose this particular community to the experiment. But as you see that it's not unavoidable that you starting the experiment, you would have landed here. We may have just as well landed here and just by chance we ended up with that community, right? So if you zoom in even more, now at the sequence variant, you see that all of the gray bars correspond to pseudomonas, as the genus pseudomonas, but you look at the sequence variants, you see that actually it's different strains of pseudomonas that are present on each one of these communities. Yes, this is 16S, yeah. There's no bacillus, we don't add the vitamins and the other metals that bacillus normally need to grow, yeah. So what we see here is that the community is very predictable at the family level. It starts being, seems like by stable at the genus level, it becomes a mess at the species level. But this is so starting from the same regional pool, right? So when it came to this question that we're asking before, that does this variability reflect simple, the fact that when we did this first experiment, each of these tubes were inoculated from different regional pools of species, or whether it results from a combination of stochastic sampling and population dynamics, once we eliminate this first component, this variability, we still see alternative stable states emerging from the same regional pool of species. So it suggests that it's a combination of stochastic sampling and population dynamics, and we're looking into this now. These are the temporal dynamics of one representative community that assembled into a low diversity state. Now, sorry about the colors, their switch now, the blue means pseudomonas, I'm sorry, Klebsiella, right? Which is this guy over here, so it's this community's over here. And this is the other one that's more diverse that contains this guild of interbacteria that coexist with one another. So what you see is that, of course, at time zero, they're both identical, but then the divergence between the two starts at day one, right? So even at day one, whether or not you have Klebsiella on day one, it determines whether you're gonna end up in this state or that state over here. And the emergence, what seems to be happening is that Klebsiella inhibits or outcompetes the other interbacteriaceae, so that, you know, this blue, this purple species here, that's Raultella, this yellow one is interbacter, as you see, they kind of emerge to dominance after five time transfers. But if you had Klebsiella there, that will never happen, right? And the other tax that you had here, which is interbacter, gets outcompeted by Klebsiella, right? So something happens from day zero to day one that sets the community into one course of another, right? And we now have done some invasion experiments. We've taken the two communities and invaded against each other. So when you take this community, the final day, and invade this community on the final day, this community takes over that, right? So Klebsiella outcompetes the interbacteriaceae guild, and in fact, even the pseudomonas that it has also carries over with it, right? So this suggests that the reason why we see these alternative stable states is not that they're actually mutually non-invasive or what you would call true instability, but it's probably the loss of Klebsiella, stochastic loss of Klebsiella on the first day of the experiment that leads to the assembly of this alternative state. We don't know exactly why that loss is because Klebsiella is fairly abundant in the original community. So we don't know if it is Klebsiella or if there's some other species that gets co-inoculated with it, that inhibits it and doesn't allow it to grow and pave the way for the other alternative community to form, right? And those are experiments that are ongoing right now and we're trying to find out. So I show you this data, showing that there are alternative stable states and we see that in most, I think it's eight out of the 12 communities we see the formation of alternative stable states, but not in every soil sample we see alternative stable states and in fact, let me show you the data for one of them just to show off because the data is so good. I mean, this is not done by me so I'm free to say it, but the data looks really nice and you can see how deterministic and reproducible the population dynamics are, right? To the point where this is for one of the communities that assembles into a single, into the same final state, right? There's no by stability here, but even in the temporal dynamics are incredibly reproducible across replicates. My favorite thing is this outbursts of Pantoea that comes and goes in generation, whatever, in the fourth transfer, it rises in prominence and it goes away and it happens at the same time in all communities, right? So this gives you the idea that population dynamics in these communities is deterministic and even if there's some stochastic sample at the beginning but and also that it's very highly reproducible, right? Or at least it's not, this is not just experimental error we have, right? And this is the joys of working with such talented experimentalists in my lab. I'm done, this is pretty much all I wanted to say. The take home message is that we can in fact assemble complex multi-species community. My complex, I mean, n close to 50 or 60, right? Even in depending on what the carbon source is. And probably larger, it's just we're not sampling hard enough in a single carbon source. I think this gives us a very nice model system to investigate community assembly and to test a lot of theories about community assembly on a semi-realistic situation we have like large numbers of species. We find evidence that cross feeding plays a major role in sustaining coexistence and we are finding that there are, that the community assembly, the structure of the final communities can be actually predicted, right? And at the family level, it obeys rules. It adopts specific states that are governed the type of carbon source that you're adding and it's very, very highly reproducible. So I just wanted to plug in the posters for my two, the two posters in my lab, Nancy's presenting one where these family rules that we're deriving we're now applying them to the idea of microbial community coalescence. I said, if you now take two communities that have been stabilized in the same carbon source and you merge them one to one, this is a process by the way that in microbes happens all the time, right? Whenever you eat, you know, whatever, a pineapple or an apple or whatever, it's a piece of fruit. But whatever microbiome was in that piece of fruit simultaneously invade your mouth, right? And this is something that's very, I think rather unique for microbial invasions. I'm sure it happens in other types of systems too, but it's certainly very much the case in microbial ecosystems that invasions happen community to community. So Nancy has been coalescing these communities and she's been able to predict what the outcome of these communities are through applying the rules of assembly that we have derived from this type of data. Another post that Georgia has been looking at now the evolutionary implications of this feedback between microbial growth and the environment. He's been doing simulations and soon experiments on what the consequences of these feedbacks between and cross feeding has on adaptive fitness landscapes and to find this gene by environment by gene interactions and the fact that has an epistasis is very cool stuff. So if you're interested, please go talk to him. I've been incredibly, incredibly lucky to work with these guys. They're absolutely fantastic. And both George and Nancy are here. We have a bunch of new postdocs, Sylvia Estrella incredibly talented too, who's a new haven now and this wonderful grad student who has joined the lab in the past year. And also wanted to, of course, thanks Daniel and Pankats for just these wonderful collaborators. Such nice people and fun and interesting. And this guy, Josh Goldford, he's the first author of all this first paper, he's just an absolutely amazing guy. And also Michael Dejonov. This has really been a lot of fun working with these people. And we get money from Simons and HFSB have made it all possible. So thank you very much for listening. And if you have any more questions.