 final round table, given the sort of interest in having more space for discussion that emerged sort of at the side of many talks, we decided to, well, after a break from Zoom and screens to have a sort of follow-up workshop which will take place in January. That will be much more relaxed in terms of schedule. It will be three days, three hours a day, short talks by junior speakers, and a lot of time for discussions. The title of this workshop is the limits of diversity assembly, and I think you all will receive an email with the details to apply to this workshop as well if you are interested. So, I've just announced that there will be something to follow this long marathon of this group. I will also post in the chat the website of the workshop. Many details have to be defined yet, and the application is not open, but you can regularly check it. So, given that, I'd like to introduce, I want to thank Alvaro for being here with us for his last lecture. So, whenever you are ready, please share your screen. Okay, I assume everybody can hear me well. Yes. Perfect. Okay, so, you know, very, I'm really happy to give you the last of the three lectures. And just to summarize what we have been talking about in the last two, I remind you all that the main question, let me see if I can actually minimize this. Perfect. The last, the question we have been asking is how reproducible micro community assembly is. And I've been giving you a few examples from folks who have been studying this question in the wild. For instance, this is the study that looked at community assembly in the water tanks that form within the foliage of bromeliad plants. And the take-home message from this is that if you look at plants that are very close proximity to one another, it is a natural experiment to study variation. And you find that there is a lot of variation taxonomically at the lowest levels of taxonomy. Fewer than 1% of all ODUs are present in all plants. Yet when you look at larger emergent patterns of metabolic organization across the communities, when you look at the fraction of the metagenome that is devoted to different metabolic pathways like fermentation, respiration, heterotrophy, cellulizes, et cetera, you find that there is a very strong convergence across habitats and that most habitats contain very similar fractions of their metagenome devoted to the same metabolic pathways. And part of what motivates the work we do is that understanding these patterns of convergence at higher levels of organization, this emergent simplicity that occurs at higher levels of organization, despite the large variation that occurs at lower levels, this species level in particular, understanding the ultimate causes of that is very difficult in natural habitats. The main reason is that the community assembly is shaped by both deterministic and stochastic processes. And all of them happen at the same time. And moreover, it is often a large number of unknown, even unknown unknowns, right? We often do not know what the selective pressures are in these habitats at a quantitative level. We very rarely have a complete picture of the entire assembly history of these communities of which species arrived and which were not able to fix in the communities. And we often lack a detailed understanding of the environment as it's being modified by the species that are living within. And therefore, it is complicated to address these questions in simply by looking at natural environments. So what my lab does is we study these questions in well-controlled habitats where we have a detailed knowledge of the chemical composition, the physical properties of the environment, the nutrient composition, at least the ones we add in. And also, because these are liquid communities, it is straightforward to even do analytical work to understand what is the effect that the microbes have on the environment. And we also have control over the ecological processes, the history of colonization, from where migrants come, how connected the habitats are. And you will see an example later of how we can manipulate that to gain understanding. So we can manipulate many of these parameters and understand a lot of what is going on. So that gives us a tool to try to understand what are the rules that give rise to these patterns of emergent simplicity in microbial community assembly at both levels of organization taxonomic and functional. So let's see. The experimental pipeline again, so everybody is on the same page. What we've been doing is we do high throughput enrichment communities. We take natural samples. We stick them in water. Then we take this goo and we filter the bacteria out. So we extract the bacteria that live in here. Again, this is the bacteria that were in the environment. We're not trying to culture them or anything like that. Just take them as they are. And then we throw them with sample from that very large diverse pool into what you can think of as a little test tube that contains the fine synthetic medium. This is, again, for those of you who are familiar with this M9 plus a given carbon source. And these are carbon limited environments, which is where carbon is the limiting growth factor. And then once we inoculate by something randomly from the pool of species in the inoculum, we let bacteria grow and we typically incubate for 48 hours. And after 48 hours, we take us a random sample of cells from here. We add them into a new test tube where we have replenished all the nutrients. We allow the bacteria to grow subsequently. And then we apply about another bottleneck and we can repeat this for multiple days. And it's time. At the end of its growth period, we do community level 16 sequencing to measure the composition of these communities. This is, you've seen this plot many times. This is an example of what population dynamics look like. And as you can see, after eight to nine transfers, we find that communities reach an equilibrium where community composition changes very little. And one of the main findings we've made is that even in these very simple habitats, we find a family level convergence where if you look, repeat the same experiment from the same inoculum in eight replicate habitats, but you will find that even as I showed you yesterday in even more, many more than eight, and I'll show you more today, you find that at the family level of taxonomy, there is a substantial convergence and all of these habitats contain very similar fractions of the same two domino families. Yet when you look at the species level composition, you find that there is a substantial variation from community to community, despite the fact that we are doing the same experiment multiple types. Okay, so I've organized this series of lectures into three, three lectures, right? But the first one I introduced you to this question and I gave you an overview of what mechanisms could be supporting coexistence in these communities, despite the fact that there's a single limiting resource. On the second one, we addressed the question of why community assembly is so convergent family level and how we can explain that. And the answer was that family level convergence reflects a metabolic convergence and functional, what we are seeing, the signal we're picking up by looking at this family level convergence actually is reflecting a functional convergence, a metabolic organization of the communities in two guilds. You remember, we discussed this a couple of days ago, one of those guilds are respiro fermentative bacteria that when particularly in glucose, which is the case and other sugars, which is the case within focusing mainly, they specialized in the supplied research and they grow very strongly on it. And as I by probably growing so strong in glucose, they employ a form of metabolism that is also wasteful, this reflects a classical tradeoff between growth yield and growth rate. And they secrete all of these other nutrients which are fermentation by products, which are acetate, succinate, lactic acid, and so on. And those are being taken by the pseudomonadasia, which we find in our communities. So the fact that those traits are conserved at the species at the family level explains why we see family level convergence. And that was the subject of the lecture I gave a couple of days ago. And today's question, what I wanted to focus on is the final those questions, which is why community assembly is so valuable at the species or genus level. And today's lecture, I just, I made it a bit shorter so that it would give more time for questions that you may have had not only about this lecture, but also about the other two. So hopefully I will finish a bit earlier than in the past couple of lectures, right? That's that way by the side. All right. So I wanted to start by pre-facing this talk by telling you that the result, this idea that there's strong family level conservation, despite substantial genus level variations, species variation, we focused a lot on glucose because that's the environment that we've tended to understand better. But this pattern is conserved across the board for all nutrients we've found so far. And this is an example. This is some citrates into this organic acid. And these are just two examples from two different inocular, each of which was propagated at different times, right? And here you have the genus level population after equilibrium for the replicates. And as you could see, there are various alternative compositions. Yet in all cases, the family level composition is very similar. This is another example where we find that different composition of the genus level again. And across these replicates, yet when you've been the reads by family, you find again, a fairly consistent structure, right? And we have seen this not only for glucose, even though despite the fact that today I'm also going to be focusing on glucose, I just wanted to tell you that this is not some weird artifact of sugars, we see the same in other nutrients as well. All right, so I'm going to evaluate two different hypotheses for why one might see what we're seeing here. Why do we find that there's so much variation at this species level despite the fact that all these communities converge to very similar compositions at family and functional levels? And I'm going to start from two different hypotheses, one of which we proposed ourselves in the original paper where we published this, another one which has been proposed recently. And actually, the two make a lot of sense. One of those hypotheses is that the reason why we see coexistence of multiple members within the same guild might be due to neutral, that neutral processes, right? That it could be that all members of the same functional guild might be equivalent, they might have the same fitness, and they're actually coexisting neutrally. So that gives rise to the fact that in some cases you have some members of that guild, and in other cases you see other members of the guild. The second, or not the second, but a second hypothesis is simply the reason why we are seeing different species in different habitats is because we are not sampling all species in all habitats. When we inoculate from the original species pool, we have a large and very diverse community with, you know, even at very shallow sampling, we're seeing thousands of different unique sequence variants. And most of them are at low abundance. So it is plausible that the reason why we are seeing that results like this is because, for instance, it might be that these blue genus, these species was sampled in this well, but not in that well or that well, right? Or that well by that matter, right? Maybe there is a competitive hierarchy, and the only reason why we see that there's different members of the same family, different genera, on different habitats is simply because they didn't get there in the first place, and it is all due to stochastic sampling at the time of inoculation. So those are two plausible hypotheses that might explain why it is that we see alternative states in our communities, and we see so much variation. All right, so I just wanted to preface this component, we're going to test the first hypothesis in terms of neutral coexistence, and I'm going to get back to this hypothesis later on, but for now I just wanted to tell you a bit of the work we've done to understand to what degree coexistence in these glucose habitats is neutral. And what we would expect that we have neutral coexistence is that the frequencies of two species coexisting in the community should remain approximately constant, or at least don't show any pattern relative to the original inoculation frequency that you add them in. For instance, if you take two species together, right, and you mix them in our habitats, and you propagate them in periwise culture, and they were coexisting neutrally, and you would expect that if you inoculate one at high abundance, and they want a low abundance, that maybe they're going to remain more or less the way they were, right, there's going to be, if they do have the same fitness, there's going to be very little variation over time, or at least as much variation as you would have simply by purely stochastic processes of dilution and growth. And therefore you would expect a pattern more or less like this, where depending on what was the initial inoculation frequency you have, you're going to have that those frequencies are more or less preserved over time on average. If you have stable coexistence on the other hand, what you would expect is that it doesn't matter what the initial inoculation is, a frequency is that those communities are going to end up converging to the same equilibrium, right, and they're going to, there's going to be a strong selection to bring the community to a state where the two are found together. In other words, both species can invade each other when rare, right, and that is one of the hallmarks of stable coexistence. So if what we had here was neutral coexistence, you would expect that if we mixed species in periwesco culture, we would see patterns that look like this, whereas if we had primarily stable coexistence between our taxa, we would find that if we mixed the species in periwesco culture, they should converge to an equilibrium, and we would observe patterns like this. And I'm going to tell you a little bit about the work we've been doing, I just wanted to highlight that all this work has been done by Changi Chang, Changi is a student in our lab, he's just an amazing scientist and a fantastic person, and we have all been so enriched by his presence in our lab and in our lives. He's been an incredibly rewarding experience to be Changi's mentor, and I can speak from everyone else in the lab what a terrific influence he's had in all of our scientific endeavors and in our growth as people. All right, so what Changi has been doing is he has been testing this hypothesis that I was testing before, and the idea is very simple. What he has done is he has done a really massive amount of work on isolating bacteria from all of our glucose communities, and what Changi has done is he has attempted to compete them at different frequencies and see what kind of competition we would observe. For instance, he could take the species, the communities that form in our, the communities that form in our habitats, isolate all the members, and then grow all the members in monoculture, some grow, some don't, interestingly, and then the ones that grow, what he would do is he would take them and then mix them at different abundances. For instance, he could take this, it sounds like art too, he could take one species, this red species over here, and grow it at high abundance, relative to another species, which is this yellow guy here, which is grown at the abundance. Then he could mix the two at equal abundances, or he could mix the two in such a proportion where the red species is rare and the yellow species is common. And if the two are growing, if the two are coexisting stably, he would observe that these three cultures that were studied at different initial abundances are going to converge to the same equilibrium. If one of the three were competitively excluding the other, what you would expect is that no matter what the initial frequency of the three species is, they're all going to converge to one of them. And in this case, for instance, the blue, the blue community being dominating the red, right? So the red will go extinct in competition with the blue regardless and no matter of what the initial frequency is. And finally, what we have here is that if you have neutral coexistence, what you would expect is that if you mix the yellow and the blue strains when the yellow is at abundance or where the two are the same, when the yellow is at low abundance and the blue is high, then over time, the two would maybe be drifting apart, but on average, you would see all of them converging to the same frequency that they had at the beginning, more or less. So there's this hypothesis, but it's exactly this, right? He isolated all of the bacteria in 13 communities and he wasn't able to isolate all of them, like a small fraction of those could not grow on their own. They required their partners to be able to grow in these environments, but he was able to isolate most of them. And I think on average, he was about 90% of all the taxa. And he did exactly the experiment we were discussing before. He isolates all the numbers and, for instance, here he takes the blue and the red, and then he mixes them at different initial frequencies. And once he mixes those, he would repeat the same experiment that we have done before, right? So we have done assemble discriminates from the top down. And what he's now trying to do is, well, if we assemble now from the bottom up, are we going to recapitulate the same behavior? And I'm going to also tell you later some work that we've done to basically the same idea of, once we have communities have been assembled from the top down, we can now reconstitute them from the bottom up and try to use that to understand more mechanistically what is going on. All right. So what Changi did is, once he inoculated these communities, then he, again, on the same environment where the communities have formed, in M9 plus glucose with 48-hour growth periods following the same exact allusion factor as we used before, he propagates them for eight different transfers, right? He inoculates, he lets them grow and then performs the dilution. He basically treats these communities as if they were the parent community where they came from, these pairways communities, which is only a subset of the entire community he got. But he propagates them in exactly the same ecological regime as the parent community, as the community where they came from was assembled. And then after that, he uses colony counting to measure the frequencies of the two. And he's been using also some other methods more recently that seem to be in reasonable agreement with the colony counting. All right. So what Changi has done is we have assembled a large number, around 180 pairs of these competitions, right, by isolating bacteria that grew together and were found together in the same community, right? This is only counting competition between bacteria that were already coexisting in the same community, but he's now looking at pairways coexisting of specific pairs. And what he's doing is he's counting the number of occurrences of each pairways competition outcome, right? You could have, for instance, neutrality, which we've discussed before, or you could have a competitive exclusion, which we've talked before, or stable coexistence, which we've discussed before. It is also possible you could have, you know, mutual exclusion, right, or frequency dependent coexistence, which are two outcomes that depend on the frequency of the initial that you started with, you're going to end up with one excluding the other or the other excluding the one, or in some cases where two species might be able to coexist when you mix them at high frequencies, but not when you mix them both at lower frequencies. That depends on the exact shape of the dynamic landscape, which we didn't map in this entirety, right? This is just a proxy for what might go in on. So interesting what we find is that there is an extremely few occurrences of pairwise neutral coexistence in our isolates, right? I think it was only one case out of 186, where you found two strains that are in the same community coexisting neutrally, that mean having very similar fitness. What this means is that at least in the supplied environment, right, not taking into account how the microbiome changes the environment where they live in, but at least in the environment that we supply, when you strip bacteria from the community context and you measure just the pairwise coexistence, you find that very, very few of them coexist neutrally. And perhaps what is more interesting even than that is that we do observe a fraction of our pairwise coexisting together, but most do not, right? Most pairs that, again, these are pairs of species that coexist in the context of the community, but they do not coexist in the broader context in pairwise, right? So they require the presence of the other bacteria to coexist either stably or in a frequency-dependent manner. And again, very interestingly, we find that neutrality is not observed. So at least in the supplied environment, the bacteria belonging to the same functional group do not have the same fitness, right? One of them, at least in the case when they eventually coexist stably. All right, so far, the idea of whether this might be caused by pairwise neutral coexistence, that doesn't seem to be true, although I will revisit this at the end with some results we're finding that are intriguing and that we are now following up on. So the second question, the second hypothesis is whether species-level variation across habitats could be due to stochastic sampling from the regional species pool. And again, what this might mean is that some habitats may get sampled, some species, whereas other habitats may not, right? So even if there were species that was belonging to a guild that would be better, a better competitor than other species, if it does, if it is not sampled, it cannot win, right? So to speak, right? So that is perhaps an explanation for why we see alternative states. And the rest of the work I'm going to be telling you today about was done by Sylvia Estrella. Sylvia is an incredibly talented postdoctoral scientist in the lab who has done, I think, the most exciting experimental work we've done in my lab so far. And I'm going to be telling you a little bit of the findings that she has made. And this is the subject of a couple of different papers where Sylvia is the leading author. All right. So to ask the question of whether species-level variation across replicates could be caused by stochastic sampling from the regional species pool, what Sylvia did is we realized that we only had eight replicates. And that is really not a lot to really do statistics and try to understand and get to the bottom of the reference that stochastic sampling might be responsible as opposed to other sources. So what she did, we also didn't know, like, how many alternative states do we have, right? Do we have, if we repeat the same experiment eight times and we see, you know, six different states, is that, I mean, and if we repeated it a hundred times, we find, you know, 60 or 80, right? I mean, we need to know exactly how many states do we have. So to nail down that experiment, what Sylvia did is that she carried out an experiment where we repeated the one we have described before, but a much, much larger replicates, right? From the same regional species pools of species, we inoculated a hundred replicate populations, like a hundred replicate habitats that are identical from one another. Again, same glucose M9 medium as we've been using and all the experiments I've been telling you about. And yeah, so it's basically the same experiment we've done, but just with a hundred replicates and steroid, right? And only from one regional pool species. And so Sylvia propagated these communities for 18 different transfers. And let me give you a summary of what she found out, right? So all these communities, again, under the same conditions, inoculated from the same regional pool, and they ended up converging to a set of different alternative states. And there is some regularity, right? In which states are found, right? For instance, she found intriguingly that some, in some cases, in some set of communities, the only, oops, the only guild that was present above a frequency of 1% were fermenters, right? We didn't find any fermentative bacteria over 1% abundance in those, I think it was about 12 or 13 of them, right? Which is interesting. In all the others, we found a combination of respirators and fermenters that occurred at similar ratios to the ones that I've been telling you about in all of these previous talks, right? About, you know, 75% interactivity ACR and 25% or fermenters and 25% respirative bacteria. Now, interestingly, what we find is that we do not find convergence at the family level, which is okay, right? Because remember the family level convergence really is reflecting functional convergence, this, this, this partitioning of the metabolic space between respirator, the fermentative bacteria that consume the glucose, and respirative bacteria that primarily consume the right products. And what we find in this case is that only a small fraction of those where we find the combination of respirators and fermenters contain pseudomonadasia, so there's this group here, whereas the others do not. And for the others, the respirative group we found is not interactive, it's not pseudomonadasia, but another respective bacteria, alkyligenesia. And alkyligenesia, in fact, cannot even grow on glucose, at least they're known to not be able to, to ferment, to respite even carbohydrates as a family. But this one that we have here cannot even grow on its own. This is this orange, orange bar that we, that we show over here. This is alkyligenes. And alkyligenes do not, does not even grow, at least this isolate on glucose on its own, right? It can only grow in the presence of the, the interact, interactivity with, with whom it occurs. And I will tell you more about that in towards the end of this talk. All right, so we find that alkyligenes can, can substitute pseudomonadasia, right? But as you can see, you see one, or you see the other, right? You do, you do not seem to see the two at high abundances. And this is the key, right? High abundances. I will, I will get a little more about that in a minute. And interestingly, when you find alkyligenes instead of pseudomonas, you find that there are multiple subgroups there, right? You could have only alkyligenes and this one strain of clepsiola. This is clepsiola is, we call this Kp and this, this, this guy you see over here. And, and that's it. It's a relatively simple community. But in other cases, right, you could find in this group over here, you could find alkyligenes as well as two different strains of clepsiola which coexist with it. And you could have this light blue, which we called Km or dark blue, which we called Kp, right? And, and it is simply some cases the Km, Km is, has higher abundance and the other cases Kp has higher abundance. But in all of these cases over here, there's two members of the fermentor group coexisting with each other, as well as the, the alkyligenes. Now, what happens in these two groups? Well, the same pattern is repeated. But in addition to alkyligenes, we, we also find a member of the Komomonadesia family, which is Delftia, which occurs with the, with alkyligenes. And in some cases, we also find here a chromobacter, right? And this is this brown guy over here. It could be found also in some cases here and the here in others, right? And interestingly as well, we notice that whenever we find Komomonadesia, it can only be with alkyligenesia. We never see it here or here, right? So Komomonadesia apparently requires the presence of alkyligenes in order to be found, right? And now you could have a community where Komomonadesia and alkyligenesia coexist with a single strain of the fermentative group, which is this KPE strain. But in this group over here, you could have these two groups, alkyligenesia and Komomonadesia coexisting with both members of the, of the fermented guild, both KM and KP. And in some cases, you even have a few others, as is the case here, where you have enterobacter, another enterobacterisia, which is another ESB, but we haven't really been able to map to any known 16S sequence perfectly. So in sum, what you find is there's a suit of different alternative states, which are easier to see when you do this experiment in much higher throughput. We have like one, two, three, four, five, six different types of, of equilibrium, depending on the presence or absence of species, about 1%. This is the, the threshold for this classification we're using. And, and what we noticed is that there's very interesting, interesting patterns. One is that the first one is there are some states where there's no respirative bacteria above 1%. We, this is something we hadn't really seen, or actually we had, there were a couple of communities in the science paper that were like that. We had originally attributed that to just, okay, well, the variation we have. But now we find that it is possible, right? For some, for the, the, the respiratory group to not even be there, right? And I'll tell you more about what we think that is in, in, in, towards the end of this talk. And we find a bunch of different groups that families that, I'm sorry, the most different groups of, or, or equilibria that vary in which of their respirative bacteria they have, whether it's pseudomonas or alkali genes. And find that when it is alkali genes, it is possible to not necessarily require that you will find other types of there that it is when a collision is present. And we find that the delftia might be there too. And we also find that it is possible for two different members of the fermentative group to coexist. Now this is very intriguing, because in the picture I've been painting so far, which again has been a simplification, is that the respirator, the fermented bacteria is feeding these, is cross feeding this, this respirative group, right? But here what we're finding is that the identity of the respirative bacterium influences also the composition in the fermentative guilt, right? So there clearly is some other process that influences some kind of more bottom up at the family level process where the respirative group also influences what happens in the fermentative guilt, right? And again, I will give you more evidence that that is exactly what's happening. Again, that picture I gave you yesterday, and I tried to emphasize it when I was giving the talk, that is a first order effect that we're pretty sure that dominates this effect of partition between these two functional groups. But we don't need to take it literally, right? I mean, in fact, we know that all of these interactives that can grow on the organic axis, and they do grow a little, right, during the second, the last 30 hours of the incubation where glucose is not present. It's not like they stop growing altogether, they still can grow a bit, right? So only the primary, most of the lion's share, if you will, of the glucose goes to the pseudomonas, whereas the lion's share of the organic acids go to the, I'm sorry, the glucose goes to enterobacteria, and the organic acids go to the, in this case, pseudomonadasia or alkynecinasea. All right, so if these pieces level variation that you see across habitats, for instance, the fact that you can, in this case, it could be in different families, right? The taxonomic variation that you observed, if this was due to cost by stochastic sampling, then you would expect that whenever you see a community that is dominated by alkynecinasea, then it should contain zero, like it could, it cannot have any pseudomonas, right? Because the hypothesis that is that pseudomonas is not present because it's not sampled, for instance. And vice versa, conversely, whenever you get dominated by pseudomonas, then you cannot have any alkynecinasea, right? Again, if the hypothesis is that you have a state because the other member was not sampled, then it cannot be there, even at very low abundance, below 1%. So we would expect that if we plot the fraction of communities that are dominated by P but still contain A below 1%, or are dominated by A but still contain P below 1%, we should expect those to be zero, right? If this hypothesis was correct. And this is what we find, right? When we actually measure it, we see that 67% of the communities that contain, where pseudomonas dominates, I mean, 67% of these communities contain alkynecinasea at low abundance below 1%. And I think you can remember the number, it's like 28% of the communities that are dominated by alkynecinasea, all of these communities here, well, 28% of them do have pseudomonas. And this particular ESV, right? Not just one member randomly of the family, but this particular ESV is present here in a fair number of these communities, although it is present at low abundance. I actually believe it's probably in most of them. So the monas, its abundance here is very close to the limit of detection by our sequencing. So it is possible that it's actually present in far more than that, but we're not picking it up because of where we are just not counting enough reads, right? At any rate, the hypothesis stochastic sampling from the species pool clearly is not correct. And this is the hypothesis that we had originally proposed in the paper we published in 2018. The second hypothesis, whether variability may be due to neutral coexistence, that doesn't look like it, at least in peruous competitions, it's not clear that that would work. But as I will point out at the end of the talk, some results that suggest that might be kind of true, possibly in some cases, or at least might have some contribution to what we're seeing. But again, I don't think we don't believe that this is the primary hypothesis. Now, another possibility that is less extreme of these two is that we could actually be having a situation where communities can adopt alternative stable states that, and where we have more destability, right? We see alkali genes at low abundance when pseudomonas is present, and we see pseudomonas at low abundance when alkali genes is present, right? So it could be that they are able to persist in the community at low frequency if the other member is present at high frequency. To wonder if there is a frequency dependent by stability at the level of the respective bacteria, that is what is governing the assembly of these communities into alternative stable states. So to give a first pass at this, and again, I have a very coarse-grained idea was that, okay, can we try to use the Fokker-Plank equation and make the simplest possible model for by stability in this case? And we understand this actually a complex community that this is not a one-dimensional system where the Fokker-Plank equation actually might not work, and you could have curls and other other situations, but given the fact that the main drivers for community assembly in this system seem to be these two alkali genes of pseudomonas, yeah, maybe it is possible to at least approximate what we're finding through Fokker-Plank equation. So if you solve it, you can connect the probability distribution with a dynamical landscape, and with that will give you the location of a tipping point above which if the abundance of a species crosses it, you will see a transition. If you are here, and you force the system, or even by simple fluctuations, the system crosses over, then it will go to another alternative state, and they will run. So that was what we tried to do. And so what Sylvie did was to try to take this idea too hard, and she collected the data she had from the 100x replicate assembly experiment that I've been just describing before, and derived the probability densities of the alkali genesia and the pseudomonal asia in our communities. We're plotting this in log scale. As you can see, this is the data I was telling you before that dispels the notion of stochastic sampling, alkali genesis is present at low abundance, in many of the communities where you do not see it above 1%, and so the monos is also present at low abundance in the communities where it isn't present above 1%. So what Sylvie did is that she just fitted this data, this histogram, to calculate this probability distribution, and from this equation, she was able, from the solution of the state-state solution of the Falker-Planck equation, she was able to solve the putative potential of which each of those two species is moving. And what this would give us, for instance, would be the location of the tipping point between the two, where, how far could you go? Basically, where if there was a fluctuation that would push a population of alkali genesia above this point, would it switch to another state, right? And the same thing for the Sumo Monadésia. And what then she did is that she tracked population dynamics in a range of these communities, and she, in particular, was recording the entire population abundance, the function of time, for all the taxa we have. But here I'm plotting the probability, I'm sorry, the abundance of alkali genes as a function of time for all the transfers, for a collection, I think it was 24 different communities, which we chose representing both states where the alkali genes dominated and those where it didn't. And here this dashed line here represents this point here, and this shadowed orange area, that's the uncertainty we have in determining what this, where this minimum equilibrium would be, right? And what we find is that by and large, it does seem to be working, right? That once the alkali genes exceeds this region over here, jumps over this threshold, it tends to jump over and goes to this equilibrium, whereas if it never does, it remains at low abundance. And the same is true for the Sumo Monadésia, right? Once the Sumo Monadésia exceeds this flat region near the tipping point, it does go up and climb up all the way up to this other equilibrium, and if it never does, it may ever drop to this region of low abundance here. So at least this very coarse grain model is giving us a very good idea where the tipping points in our system are, which we can corroborate through dynamical experiments. So, now the question is, okay, so we have assembled these committees from the top down, but one of the things we can do is, as I was saying before, is we could now take it apart and take apart and isolate all of these dominant members. We were able to isolate at least the Sumo Monadésia, these genus here, these two strains of intermacteria, and we were able to isolate the alkali genes. We were not able to isolate the Sumo Monadésia yet, and indicating that delfta clearly requires the presence of probably alkali genes at least in order to grow in our habitats. But we're able to isolate the other three. So we were now doing experiments where we could try to take, and by the way, I told you before, alkali genes doesn't also, it doesn't grow on its own in this environment, so it needs the fermentor to be there. So what we did is that we took Kp, which is this dominant member of the interbacteria family, this dominant fermentative bacteria, and then we would mix, for instance, Sumo Monadésia and alkali genes together at different abundances, and that way we could reconstitute our communities. And we would repeat the same experiment as before, now we grow, and then we propagate for 12 transfers, and after that time we played to count the number of CFUs belonging to it. So what we're able to do then is we started mixed cultures of Sumo Monadésia and alkali genes at different initial abundances with the same basal concentration of Kp, which is this fermentative bacteria. So this Kp is present in all of these communities, and each square represents a specific abundance of, initial abundance of Sumo Monadésia and alkali genes, just one, right? So it would be here the center of each of these. For instance, here the abundance of alkali genes would be 10 to the minus one, and the abundance of Sumo Monadésia would be 10 to the minus four, here the abundance of alkali genes is 10 to the minus four, the abundance of Sumo Monadésia is 10 to the minus three, and so on and so forth. And in all of these we also have Kp present, right? Because again, without Kp you would not have any coexistence. And as a function of time what we find is that we did this experiment in two replicates, and we are coloring each of these squares is orange if the community contains Kp plus a both about 1%. It is purple if it contains Kp plus p over 1%, it is blue if it only contains Kp, and it's gray if when we do the replicates in one case you get one result and in the other you get the other, right? So if in one case you have that a is dominating and in the other you have that p is dominating, or if in both you we found both and p together, so we just when we cannot tell we just label them gray, right? And what we find is that the as a function of time this community converges to this phase portrait converges to a state where every community that was started above this in this orange region converges to a state where a archaeologians will dominate and every community that would be started in this purple region would converge to a state where pseudomonas would dominate, right? And here I am just showing you an example of the temporal dynamics for each of these six wells so you can see it, right? So for instance here where which corresponds to this square over here that starts at 10 to the minus four archaeologians and zero pseudomonas not suppressing like archaeologians just kind of dominates because there it cannot be challenged by pseudomonas. In this one here you find that that pseudomonas was originally very very low abundance and then over time you find that pseudomonas actually went away and archaeologians dominated. And here on the other hand at this and this levels over here where archaeologians were representing these two squares over here archaeologians started off being low abundance and then it never were able to invade as pseudomonas was present and the same was true for this other point here where archaeologians was never there, right? So pseudomonas in co-culture with KP remains low abundance and these results were obtained by by plating CFUs and this state becomes non-invasive. So when we did this is that we could take this phase portrait, right? And then we we were able to obtain, right, the basin of attraction in in the simplified three-piece consortium. Remember that the the common ADESI is not there, none of the rare members of the community are there and there are quite a few, right? We only have these three members KM, sorry KP, which is this this Klebsiella, which is the fermentative bacteria, Benakali genes, and pseudomonas. And from those bottom-up experiments we were able to reconstitute the the basin of attraction for the two states which we shade here in purple and orange and over them we're overlaying the composition of our communities. This is the top-down assembled communities now, the 96 of them, and you find that there's a group here which represents the one where archaeologians is present at high abundance and the pseudomonas is present at low abundance. Here are all the ones where pseudomonas could not be detected below the limit of detection, which may be because they're not there, or it could be they could be very rare and we just didn't pick them by sequencing. This other state here represents the state where pseudomonas is at high abundance and archaeologians is at low abundance and roughly corresponds with the basin of attraction that we have that we have to determine, right? It's within the boundaries of the basin of attraction as we determined by just a bottom-up manner. And we finally have this other state, oh yeah maybe I'll do this right here, okay. And we finally have this other state here which represents those where you might remember that there were no archaeologians or pseudomonas above the, above one percent, right? So there's only fermentative bacteria and above one percent and it's this group over here. Now if you plot the entire dynamics on this face portrait which again we inferred from this simplified three-member community, results seem to actually make and make a lot of sense. So communities that are started and they get stuck around this region, they kind of wander around this white area and have a hard time moving over. Whereas those that started here for instance or that rapidly where archaeologians rapidly went over the threshold, rapidly converged to an equilibrium as you see in these cases. When pseudomonas, the attraction well of pseudomonas appeared to be more shallow and that is reflected by the fact that you see a lot of variation when you find a community on the area of a pseudomonas and you can also get stuck right in this low, what we interpret as being a low slope region of this dynamical landscape. And this is very metaphorical, I would grant you that, we haven't really mapped out the full dynamical landscape in a lot of detail but at least what we find is consistent with that idea that when you get a population close to this kind of separatrix between the two, you get slow dynamics that take a while until they converge to something or at least it's possible that you can get stuck in there. All right so I wanted to zoom in into this community here of this state where there is no pseudomonas or archaeology so they are there but at low abundance. So again corresponds to this group over here. Our hypothesis actually and it goes back to this, our hypothesis is that these guys are stuck right either there is another metastable state that we are not resolving here in our bottom-up community or more likely it is that these guys are stuck in this very kind of slow region of the dynamical landscape where it just takes a long time for either archaeologians or pseudomonas to meet. I mean if you see in these cases for instance they are both there at these abundances and we can pick them up by sequencing even after at the end of 18 transfers it's just that they never quite jumped over right where they are supposed to go. So to test that hypothesis we thought well what we could do is we could now connect the habitats right so at the end of every transfer we could bring we can pull species from all these wells and then redistribute across all wells so that if every well received migrants from other wells it might give them the push they need to jump over this to getting them unstuck from where they are right and and converge to one of the two alternative states. So that's the experiment we did we repeated the same experiment we took an inoculum and then we inoculated communities and then we added did we either kept pushing migrants from the regional and pull species or we pulled the all the pull all of the communities and then redistributed bacteria across so that all these habitats are very well connected to one another and then receive migrants from from all. And then we did the 412 transfers and then we allow the communities to stabilize for another six transfers afterwards. And the outcome is is very encouraging right when we did that right the communities are connected then that the that state that is dominated by klebsiella kp goes away right and then what you find is that alkali genes is able to to dominate and you find them in all the communities and I would say that in intriguingly we find that all of these states over here they're all over the separatrix they all should be converging to an alkali genes dominated state they just simply never quite made it right so this is consistent with this result here which is when you actually allow for for migration alkali genes is dominating and pushing the system in that direction and we are still analyzing this data to really understand better what is happening it is quite intriguing also that once we stop the the that when we're doing this experimental global migration we find also that in all of these states you find both of the of the interactivity is yet so we never see exclusion right further suggesting that maybe pointing out the potential role of neutrality here and that's what we wanted to test right so this fact that that you see that when you are connecting the the the the the other habitats then stochastic extinctions are less likely we wanted to see if if that would be something you could observe also from the bottom up right so now the first thing we wanted to do is to ask well if we mix the two together the two fermented bacteria together are we going to see that they could exist or are we going to see as we see in this state that the the dominant one kpl competes the other and we have found is that and and no matter what the frequency you mix them kp always excludes km by contrast we also find that when you mixed uh surmonas with kp and km the two fermented bacteria kp excludes um km and that is consistent and we see this here and that is consistent with consistent with the fact that uh you never see them together when surmonas is present yeah we find that when we mixed uh kp and km with our kali genes it is possible to depending on the initial frequencies to i mean in in all three cases that we did multiple replicates which we show here um and all these committees for simplicity i'm only showing here the ones that we're starting at at equal abundances but they all end up uh coexisting after six transfers and um in some cases they can remain around 50 50 in other cases one was up there in other cases one was down and this is consistent with this finding here where the abundance of one or the other could be quite variable um and uh in our top down communities and to further cement this point we we inoculated here i'm showing you the the result of of repeating this experiment um this one over here where km and kp are the three different replicates are inoculated at high abundance of km um a half and half or low abundance of km or high abundance of kp right this this this two different uh clebsiella strains with apartheid same guild are inoculated at different initial abundances and the result is at least someone consistent with a neutralization of their competitive dynamics which we see is only happening with alkylid genes is present so that led us to hypothesize that perhaps there is this emerging neutrality where the members of the same guild might actually have the dynamic that are finding coexistence in these communities um their dynamics might be more neutral um than you would imagine simply by looking at them in pairwise um competition which again emphasizes the importance of of taking the entire ecological context and the difficulties of inferring what's going to happen if you only do pairwise competition um and just here showing you some examples of of of the type of dynamics we find and we are now trying to to really find um traces of of neutral dynamics among the clebsiella in these communities to see if we can um validate our hypothesis all right so um wrapping up here the final question I wanted to address in this series of lectures is why community assembly is so variable at this species and genus level um we have found strong evidence for multistability we find we can find very strong evidence against the idea that there is stochastic sampling is what's governing this um we also find however that um this an evidence at least for close to neutrality between members of the same guild in the in the in the context of a complex community which we do not see when we we just grow them in pairwise co-culture um and that is that is it um and this is my final lecture so I really wanted to to take some just a couple of minutes to thank uh the amazing folks in the lab this is make such a joy to to to well I don't go to work in every morning anymore um due to the pandemic but I still communicate with them every day and it's just so wonderful to work with all of them um and you know it's just just one of the greatest experiences of my life without a doubt so um that's all I want had for you um I hope you have enjoyed the lectures and taken something from it um I'm happy to take your questions um now thanks a lot for the very nice lectures so there is a question from Silvia yes thank you for the very interesting lecture I just wanted to ask a clarification I didn't get the reason that you used to exclude the stochastic sampling as a cause for the variability could you like explain yes of course so okay uh well if it was stochastic sampling and by that I mean um I mean I'm not I'm not saying that that it wasn't it is not possible that I mean of course there is stochastic sampling right what I'm saying is that the reason why we see variation in community assembly is not because a species is not sampled at all right we just just don't add it to the to the well right because if it were right in these communities where we have pseudomonadasia right if the reason was that um that you would you would have seen imagine that one have the one happens it would be we would have seen alkylidinazia everywhere and the only reason we don't see it here is because stochastic you didn't add them to those to those test tubes right where you inoculate it as you would expect to see zero alkylidinazia here right um and conversely the alternative well the only reason why you see no pseudomonadasia in these communities is because pseudomonadasia didn't make it it just didn't get added to the well right so in these wells you added pseudomonas but not alkylidinaz and these wells you added alkylidinaz but no seumona and what I am trying to to to emphasize is that this is visually might be something you could think but if you examine in in more granularity what is the community composition here only showing the community composition at above one percent right but what we find is that um of all that in all the communities um this is what we would expect right if it was stochastic sampling right that if you're dominated by pseudomonas you would have no alkylidinaz right it didn't get sampled so it cannot be there or if you get dominated by alkylidinaz you have no pseudomonas it because you know if the hypothesis would would mean that it didn't get sampled right but we find that whenever you get dominated by pseudomonas in most cases you also have alkylidinaz so both are there even alkylidinaz are low abundance right and when you get dominated by alkylidinaz you also have pseudomonas again both of are there only alkylidinaz is high abundance and pseudomonas is low abundance right so it's not presence not presence it's just presence low abundance or or the other way around right so that rules out that it was that the reason you're seeing alternative states is because uh species didn't get sampled right um i mean it's not saying that that could have never have happened right it is possible particularly for some of the more rare bacteria uh species that maybe you don't see them in some wells because you didn't get them there um so of course that could play a role but at least in the dominant members right that the ones that we see are high abundance in the end uh they're about they're it isn't really the case that um that the reason why we don't see them is because of of lack of something i think i hope that answered your question yes thank you great there is a question from Miguel thank you Jacopo it's truly remarkable stuff alvaro this is a really really exciting uh results during your your previous talk and also during this work talk you have been uh convincing us that there are some archetypal attractor configurations and that that's reminiscent of the of this idea of the enterotypes it got the it got microbiomes it's it's it's clearly in here you show you actually show this temporal component that is very convincing but uh in the case of the enterotypes it has there have been quite a lot of discussion about the whether or not these are just artifacts statistical artifacts from compositionality for example and i i just wanted to ask you if you have an opinion on this of this discussion from your results right um that's that's a that's a i i am afraid i don't know i mean i know that literature i've read the papers and i know there's a lot of controversy but i don't feel like i should be weighing in on it right um i i think it is clearly possible to find i mean i also on the one hand it's clear you could find um artifactual uh results simply by lack of sampling right um so it is possible to see what appeared to be um you know uh discrete groups of of states simply because you are missing the ones in the middle right um so definitely a big reason why we decided to expand in our experiments to to do a hundred replicates not eight right because if you do eight you might be seeing things that aren't there right you might be seeing two alternative tractors right if you do a hundred then you're okay right maybe it's a continuum and i'm just imposing uh categories when they don't exist right uh so that's uh kind of that at least i can tell you that controversy really made us it made it clear that we needed to to go to higher um uh to do more replicates right and and to truly understand of course the issue with enterotypes as well is that there's so many the mechanisms are so poorly even if they exist right that the mechanisms what are they exactly right how do you how do you pin them down right and there's there's all kinds of questions uh around that too which beyond the statistical whether they're statistically correct or not right um it is unclear or it's much it's much harder way to to to know if the reason why we're seeing this because the the habitats are different because the regional pools are different because i mean i know there's people doing a lot of work that's really fascinating on on trying to elucidate those those differences um and i i i just think that the the place i see our work is i'm trying to create the the simplest possible of habitats to at least get intuition of okay well if you if you were to see enterotypes right what would be the mechanism that could lead to them right and and where would you not see them right what would lead to those enterotypes to disappear right and that's what we in something we're after right developing a ecological understanding of of what are the mechanisms that can give you different attractors or or not right and and to give us an intuition for why those might be but yeah as for the entire literature is is a really um yeah it's a really tricky uh thing that i'm really not not an expert on on the human gut microbiome so perhaps i would pass on on commenting the specific controversy thank you great there is a question from gabriel hey uh really nice talk uh i was wondering about the presence of uh when you're in the alternative state the so when you're dominated by ap is present at low abundances and vice versa right um and in particular i was wondering uh if potentially maybe phages have something to do with this so are there phages present in the initial inoculum and could you do these experiments for filtering phages right right right this is this is a great question um so we now have a wonderful phage biologist in the lab who is um we are very excited to collaborate with um to address some of those issues i can tell you that we had someone from like a really talented student from poll turners lab was a phage biologist and came to to our lab and tried to find phages that could be grown under the conditions of our communities right basically phages in the communities and he because these are communities that form in a in a defined environment at least we thought that it would be you know isolating those phages in the same environment where the community assembled would be um easier if they were there and he tried very hard for a few months and found nothing so we didn't find any elliptic phages going around um we still i mean i don't one question for instance could be prophaging um induction whether that could have a role and that we don't know we have not uh we have not looked right and that could be a very interesting possibility right that maybe um uh yeah just prophage activation might potentially have something to do here uh with the findings but at least um lytic phages are not around or at least if they are we we couldn't isolate them um and uh but i i agree that um i this is a really fascinating topic and and we are now trying to understand better what the role of phages could be in in community assembly so this is hopefully in a few couple of years or so we'll have results uh on the front great uh i don't see any more uh questions so uh oh there is a question from tomaza please oh sorry hi um thank you very much for the for the very interesting uh talks and i was wondering like um one very intriguing thing that you seen that is found that is that there are a finite uh amount of uh stability endpoints in the in the last experiments that you were explaining i was wondering and of course you've done a lot of sampling a lot of replication i was wondering if you were expecting uh a further number of um coexistent points like of equilibrium points sorry uh whether even a bigger experiment would would be done for example right that is that is a fantastic question right like how how the number of alternative states scale because you have right um and yeah that's a great we haven't done it right and and um anyway we could try to do some very verification curves i guess so speak right to find i mean once we have the data with handa we can try to see like how how often would we be discovering the states right and and i think at least that's something we could try um i know that's that's a great question i um well possibly my hypothesis would be that with a like a simple environment and and such like uh community maybe there will be not so many more but yeah it's very interesting what's applied to all the kind of communities yeah right and and i mean i think i was in some sense well i'm not surprised the right word because i didn't have really a prior and it's funny in microbiology we don't really have theory so when we're surprised or not about things it's often difficult to know why we are surprised or not but but at least in this case i mean i guess six seems a lot for some reason i don't know why right but the fact we have six stable states so different from the same experiment seems like i don't know naively i wouldn't have expected that many right um but i can't tell you why it wouldn't happen it's maybe my induction on on previous work um but but no i agree i mean i think it it would be a a really interesting question and of scaling this up instead of a hundred do you know a thousand which is feasible right no well yeah i mean it's a bit expensive but getting cheaper every every year so so maybe in a couple of years it'll it'll be as as costly as it was when we did this experiment the first time so um but anyway yeah no it's really very interesting yeah to answer that question thank you for the question thank you if if i can if there's no other like if this time i have another little questions it's more curiosity i don't know yeah but what what do you think it's just a curiosity like i found it very intriguing as well like the uh like the metabolic convergence uh to the level of family and i was as well wondering if you had any expectation for other kind of um communities uh let's say i'm thinking like if this kind of a relationship would change with the rate of speciation so with systems where the rate of speciation was different let's say it's lower or faster if you are in migration because we don't really have speciation in our ecological time scales but we could add species through migration right yeah right so i i didn't have we actually this experiment this is a fantastic question by the way and i i i i'm i'm regretting now that i didn't include this slide because we've repeated this same experiment in fact you know when i when i talked about this i think i have this slide here but i didn't show you the data yeah here so we've done this experiment where we kept adding migrants from the original pool every single day right so we we kept pushing species in right so we whenever we do the transfers right we take an inoculum uh so we take the inoculum and add species to it so we grow for 24 eight hours we transfer cells from the from the old generation to the new and then we bring migrants from the same inoculum that you use the first day right so when we do that diversity goes up by 10 fold so we're getting like 150 taxa uh to existing in our habitats many of which are at fairly high abundance so these are not i don't think these are single populations they they are some of them reach like maybe 30 percent uh for instance we see more axalasia in in these communities here um and but when we stop the migration uh they collapse and they go back to this uh to these states right so while we're doing this migration that you have a much much higher uh number of species so if species keep arriving from from the original pool the diversity goes up um massively but the there aren't that many alternative states interestingly right and maybe it's because the they as since you are adding this these new species the the system has a harder time getting stuck in metastable states they all have very similar compositions right and then when you stop migration they all converge to very similar equilibria right um and uh and you have additional families that don't show up here and then when you stop migration they go away so we have we have done that experiment and we haven't published it yet um and uh and i think it's very interesting but we are yeah it's one of the things hopefully we'll be able to put out uh into the world in the next in the the next few years thank you very much thank you thank you for the question there is one last question from matthieu that i cut last time so i feel better so matthieu please uh ask the question thank you very much uh i was curious here if i understand correctly uh all the links between species it's purely competitive uh would it be possible to what would it change if you now add uh trophic links to this kind of systems uh um so you're you're saying i mean just to qualify your question there is a substantial cross feeding facilitation so in addition to competition there there is a lot of metabolic facilitation too right so the reason why we see coexistence is because of cross feeding of metabolic byproducts that the microbes create right from the group as you provide so that's true so when you say trophic links you're thinking for instance if there were protists or or bacteria phages that eat the bacteria yes right that is also a really interesting question uh going back to the question i can ask before about the phages it's something that we are very interested in exploring like what would be the role of how would what would phages do to biodiversity right would they uh promote it would they deplete it right uh there there are these potential mechanisms they kill the winner where phages that have a sufficiently large host range might be promoting coexistence because of selective pressure against those that end up dominating right so it might be contributing to to that it could also through lysis be releasing a lot of metabolites to the environment which other bacteria could could also use for growth so there's reasons to believe that uh bacteria phages might promote diversity right and as for other predators i'm less clear right if you have protists that are eating bacteria uh on the one hand they might also apply the selective pressures against the more abundant so that that would promote the growth of at least might prevent those from out-competing the more rare texas but they're presumably there wouldn't be as much um you know cross feeding because they would keep yeah there would be the ones that eat the bacteria right they would not just burst it and let it out so that or the cells can eat it right but so that all this is a really fascinating question and i think um it could be quite easily studied experimentally right and it is one of the the areas that i'd be very interested in in moving on in the future thank you great so uh i don't see more questions and uh we are a little bit late so i'd like to thank alvor again for these very nice lectures thank you all for