 Okay. All right. So today I'm going to be talking about some work I completed on my Ph.D. on the birth and death dynamics of microorganisms in environments, closed systems with zero exogenous resources over an extended period of time. Just briefly before getting into it, just to mention the work I've been doing here at ICTP with Jacopo, it's been primarily focused on community dynamics that we are starting to do some physiology type work. First project that's had us a preprint recently is this collaborative effort with Alvaro Sanchez on basically determining the extent that macroecological laws you get in nature can be reproduced in single carbon source communities, trying to determine the extent that experimental perturbations make quantitative changes to those laws and the extent that you can predict back those changes using a stochastic logistic model of growth. And on the right here is work demonstrating how there's this similarity in the macroecology of microbial communities across phylogenetic and taxonomic scales. And that's out recently in eLife. Okay. So the framework for sort of this work I've had from Jay Lenin's lab is this starting off by saying that I think many here would agree that microorganisms could constitute one of the most successful forms of life on the planet. They've seemingly colonized every possible niche. Diverse environments that are really stressful ranging from deep sea sediment with cold temperatures and crushing pressure to extreme deserts where they have to survive periods of very, very rare rainfall to even pristine looking environments such as glacial lakes where there are actually very low nutrient concentrations as well as the bill environment such as hospitals where microbes have to be able to survive outside of the host and persist for extended periods of time until the opportunity for infection presents itself again. And they accomplish these tasks of surviving through a variety of mechanisms, this top row here, pictures of Bacillus in its famous endospora formation, bottom left, microcaucas are one of the well-known bugs capable of entering extended periods of low metabolic activity and cyanobacterium forming cell wall structures to survive desiccation. And so despite all these differences in environment and diversity of structure, I would argue, there is this similar demographic regime going on where the microorganisms are outside of what we might want to call stationary phase of growth and going under either this exponential type decline, this death rate, and then this extended period of stagnation often called long-term stationary phase. And so with J and I, we're really doing a lot of work and experimental work in Indiana focusing on these two demographic regimes. And primarily we started out thinking about dormancy and basically how dormancy influences this decline, this death rate. And under this, the sort of null expectations that you have this constant net rate of growth and growth defined here is just the difference between births and deaths. And so assuming in this death rate that you've got many more deaths than births, you can basically say that the change in population size you observe is reflective of the death rate. And from that, if you can parameterize that, you can get mean time to extinction, mean time to death of individual cells, and so on, basically from a model of exponential decay. And so we were working with this experiment where it's fairly simple design. It was pretty labor-intensive and I argue fairly ambitious, but the design is basically you grow up your bug in a rich media, get a really high cellular density. You pellet and rinse it in some buffer. We did it in triplicate. You do it three times and then you resuspend in buffer without resources, right? And so you have a high density of cells and a closed system effectively where there are no exogenous resources going in. And these were all with heterotrophic bacteria. And so the core design is to just have 21 different phylogenically diverse taxa. We have four different phyla. These are all soil isolated taxa from Jay's time at Kellogg Biological Station in Michigan. And we maintain these in the buffer for a thousand days, plating roughly every seven days with about four to six replicates per taxa. And so our null expectations from this is that we think go ahead in this thinking, all right, we do this. We're going to get first order decay rates. We're going to get some estimation on the death rate. We can look at metabolic activity of cells, say something about dormancy. But what happened really was that we only had one taxa that went extinct and had this consistent linear decay. And so we're looking at time on a linear axis and a population size on a logarithmic transformed axis. It's only one species here, micrococcus, had this constant rate of decline with time on a log linear plot. Everything else had this shape where it had, it sort of bends out and flattens and plateaus and remains so more or less over a thousand days, slight decay, but you know, everything seems to flatten out with the extreme case for clear reasons being our strain of bacillus that we use, which is basically flat. However, I don't have this, not going to talk about this too much, but we also repeated this bacillus line with a knockout that cannot form spores and did the first hundred days and we basically get back the very, very qualitatively consistent pattern, but just the intercept is basically going down an order of magnitude or two. And so this didn't really map up with our idea of dormancy and a bunch of other experiments I did with some metabolomic sampling, some microscopy staining for dead cells and so on. So it didn't really paint the picture that dormancy is driving this. It's contributing, probably we'd argue, because metabolic activity has to decrease to some extent, but we really ended up with this model of recycling going on. And so we started out with a dormancy-driven hypothesis and then we're moving towards how the dynamics within these semi-close systems are basically driven by recycling. And so system, you can have a system of equations basically where you have your living cells, your dead cells, and then your dead cells going into some pool of resources that they can get consumed by some consumption in order to form new cells, but also primarily we would argue to meet the maintenance demands to maintain cellular activity. So this maps up pretty well qualitatively. In blue here, or purple is a strain of eucinia, and so we do see that because there's a higher initial standing population size, you do end up with this, you know, more elbow type decay on a log linear plot versus micrococcus, which just happened to have a initial population size about two or three orders of magnitude lower had this more linear decay. And that maps up qualitatively what we get from modeling efforts where basically one of the big control parameters is determined whether or not recycling can drive the long-term dynamics of your system is this population size you start out with. And so you can do the model with or without an extra term to describe how, you know, they're not perfectly efficient in consuming resources, and so that there's some loss, thermodynamic loss, but you end up with this more steady state where the stationary size is greater than equal to zero, and then you basically get stationary state equal to zero when your initial population size is too low, which, well, they don't know it's the flux of initial dead cells that's determined by the initial number of dead cells. Yeah, but there also, if you have 10 times as many cells, then the flux of dead cells is 10 times as big, but is also taken up 10 times as much by 10 times as many survivors, no? I mean, the position of the stationary size is primarily determined by these two terms, the maintenance and the birth rate, and so... What is capital R in this equation? Capital R is the concentration of some resource that's excreted by the dead cells. So this is not a, we have not identified the mechanism as in the singular limiting resource in these experiments, but capital R is just some resource that is excreted by dead cells that then gets consumed to meet energetic demands. So there's no perpetuum mobility, right? Pardon? You have to provide some energy to keep it cycling. Where does the energy come from? Well, the energy here is the initial concentration of dead cells, initial number of living cells that you put in to the experiment. And so this is not, you can add terms to this model and we have where you make it so it is not this ideal scenario where you have some decay in terms of, you know, due to inefficiencies. But you still get back qualitatively the same pattern here, right? Where the ability to have this deviation from a strict linear decay is determined by the death rate, your initial population size, and this consumption term B, per freight. Some number is going to survive, but if I now insert a membrane such that half of them are now at the top and half of them are at the bottom, you're going to get a different number of survivors. That seems to be what you're saying. The number, I think it's more reflective of the rate of decline than the number. So you're arguing that you, or that you have an experiment in mind where you just have a porous membrane and you kill half and there's some secretion going on. Non-porous. Non-porous. It's the same total number. The long-term dynamics of the system seem to depend on the initial population size. The other demographic parameters are a constant. I mean, I understand that this perhaps might be a bit more, this result here with the micrococcus might be a bit more convincing if there was a single strain and then you manipulated demographic parameters, which is work I'm looking to do in the future. But basically, if you quantify the death rates, then the main parameter is basically this compound parameter where the population size times the death rate, the per capita death rate determines whether or not you can reach this, you can have this deviation from long linear decay. So even if you have this recycling and you have still some decay, then you would expect something that's bilinear in log scale. You have some initial death rate and then they start recycling and then they have another exponential death rate. So then you should be able to see that. You could. Maybe it goes after day 1,000, but we haven't seen it in any of our 20 species that are, taxa that are not micrococcus. Just to follow up on this question, it also seems to me like you could split the flask into two and you get a different result. But I'm trying to understand. So what another way to think about this? Could you do an experiment where you spike it with extra micrococcus? And so you predict if you spike it with extra micrococcus, the shape of the curve changes? You could do that. I think they kind of did that in this paper I'm going to show later for some future directions where they did it, but they did it in this death phase where they added in dead cells into a decaying population, got a shift in the death rate. And you can interpret the death rate as the balance between resource output to the cellular death and maintenance needs of the cell. But we did not do that for here. Mostly, I think mostly we were in terms of thinking of just documenting this pattern and then getting enough biomass at the end of the experiment so we could do sequencing without some initial added cellular material. Potentially complicating those analyses. You're already doing it. Triplicate and your hands are busy. Yeah, but this is just an idea. It's a good idea, though. Do you want? On Ecolite, there is this bunch of literature on this long-term stationary phase with sort of like gross advantage stationary phase and so on that shows that if you let them long enough, they will also have some mutant that are able to harvest like better the nutrients from dead cells and so on. And that sort of drives those kind of... So it is like not dependent on the density of cells at the start, but rather that like you sort of start selecting some mutation that are helping cells to survive better. So we're going to get into that. We don't think that this is driven by mutation at all. We think this is a general demographic response to having this flux of dead resources. And we did have done our own... We don't use the term gas as much because we're not doing any Ecolite phenotype characterization as much, but we have done these long-term stationary phase evolution experiments. And I'm going to get into briefly at the end what the designs of those experiments and how they don't make it quite ideal to determine the extent that recycling can contribute towards longevity or how recycling could be a trait that can contribute towards adaptation. Because I've done my own long-term stationary phase experiments that are out. And there's differences between this setup and that. We could... Well, so these flasks are... I mean, we're using 50-milliliter Falcon tubes. They are effectively stationary. We shake them up before we sample, but these are standing upright in the dark, no light exposure, and effectively anaerobic in the sense of they get... If there's a little headspace, we sample very little. We took no more than, I think, a milliliter or two or three out of the 50 over the course of a thousand days. We were doing very careful not to turn this from an anaerobic effectively closed system into something where every so many days half the volume gets filled with air again. But I mean, we have this taxi here. I'm not quite familiar on the motility. We have motility assay bulk data from plates and prior work J's done on soil. But basically, I would imagine how... I mean, motility doesn't explain the strict linear decline of only micrococus, right? Well, for colony counting, our basic protocol was to only use... Colony counts were the... What is it? Total CFUs of a plate were in the range of 20 to 100 in something. We were careful to start counting when they start counting after, I think, two or three days because the growth slowed down, but also we manually checked. All the counting was done manually and checking whether or not there was this overlapping circles or rings of growth was something that we monitored and we did not. I don't think that those are influencing these numbers. Well, something with the QPCR, we didn't want to do this because this is not like a classic evolution experiment where things are growing every day and there's new biomass coming in. We were making sure to be careful not to take all the biomass before we had this... Before we could actually qualitatively determine the patterns of any system. So we didn't want to do... We didn't want to get a tenth of the biomass within the first 10 days or something like that to get DNA extraction. By the end of the experiment, there were only nine tax that had sufficient DNA even after all the tricks we did to get as much DNA out of it for pooled population sequencing. So we have... Working with Jay, we have this sort of more phenomenological type system of equations to describe the qualitative behaviors of the experiment. Just in order to get survival time, we ended up just going into the statistical world and treating this as a survival analysis type pattern mostly because, I mean, if you take the differences in total population size with respect to time, it doesn't really ever increase with time, which allows you to treat it as a survival type function where it basically has to go to zero. And we ended up choosing after a few distributions basically the Weibull distribution and mostly because of its flexibility where there is this shape parameter K and if K goes to one, then it reduces to an exponential form. And so the intuition here is that K getting much, much, much smaller than one can be viewed as an approximation of the increase in the net rate of growth over the course of the experiment. And so if you do this statistical analysis and all the replicates and compare the degree that growth rate changed from the shape parameter of the survival distribution to the initial flux of dead cells, which we're defining as the initial death rate of the population times the population size, initial population size, that Nt should be N0. Then we see this, we see what we expect, a greater initial flux, a greater deviation from exponential decay. And so micrococus is up here in the top left basically where it's exponential. And then your Cine is somewhere over in the middle. If generally the distribution tend to hold well, we're probably stretching it a bit by putting Bacillus because it's such an extreme, sharp elbow type plot. But for the bulk of a taxa, it tends to follow this relationship we'd expect of the parameters. And so once you have these parameterized, and just to get some proxies, we can look at the average time to extinction, which we define based off the minimum resolution, the minimum number of cells that we can detect given our plating approach. And then also the calculate the average time to death of a given cell. So this is the clear outlier. And we have the median and the dash gray line. So it takes about 10 days here. And so these might be a bit of extreme because a lot of these are predictions, right? We only saw one taxa that actually went extinct. That was micrococus. That's the only one where you can actually say this is where it went extinct by looking at the plot, everything else is extrapolation. But basically I think we are justified in at least presenting these as cautionary results on the extent that a population can survive in a pretty much closed system under anaerobic conditions. So additional evidence. We did some microscopy work. Dead cells do not tend to accumulate over a thousand days. Apply cellular recycling. Cells with a greater increase in growth rate. And here I mean the shape parameter. Cells that tended to have a greater increase in the net rate of growth with time and experiment tended to have lower lag phases when you were to take the ancestor of our strains and grow them up on the same rich media that we used to start this experiment off in the first place, right? And so from that you get the lag time here and very rich. I think the whole thing was done in LB and then the y-axis is roughly can be interpreted as the deviation from exponential decay in our experiments. And so this relationship here does fall out, is related to a single model, a kinetic model of microbial growth, which we suggested reflects the reality that any sort of recycling requires some type of kinetic activity in the cell. But we'd like to get into physically why this is happening and we haven't quite yet. So I think the big question once we determine this was that we'd like to know how the net rate of growth is increasing with respect to time. Fewer births or fewer deaths or more births, the two ways to change that rate. And so you do this experiment, you get to the end, you can't repeat it, it takes forever. We ended up doing, we wanted to sequence it anyway because we had this gas question in mind, right? So we, I got all the biomass, did basically all the DNA extraction tricks we had to do something of this scale and extracted DNA at the end of the experiment. It was very low, but we were able to successfully do pooled population sequencing. And from that we can get the allele frequency spectrum of derived mutations, not derived, we get the allele frequency spectrum of acquired mutations over the course of the experiment. So, well, the next slide I think is going to answer your question, but basically we wanted to get frequencies. We did not necessarily want to have only a couple of clones to identify potential targets of selection from doing DNDS calculations. We wanted to try to identify the maximum frequency that a mutation has reached in the population. And we didn't want to do 20, 30 clones per replicate population so we've got nine taxes, we've got four replicates per 36, do that with 20, you've got like 300, 400 libraries, you have to prep, and here we've got 36. Because we want to get frequencies, so we have to have repeated clones sequenced. Did all of them up in a liquid culture? We did, we can. Everything at the end of the experiment was, aliquot was taken, grown up and cryopreserved in I think triplicate, back in Jay's lab. So the end of the thousand days, they're all grown up and cryopreserved. And you sequenced that. No, we took a little bit of that. The sequencing, I got the biomass at the end of the thousand days because I wanted the frequencies, as they are, in the tube, and not grow them up and change the frequencies when they're back in the fresh media. So, a few lines of evidence in addition to some metabolomic stuff, but basically there is no relationship between the deviation from exponential decay and these coverage-based proxies of birth rate where the slope of the distribution of coverage along the genome should reflect the degree of nested replication going on in bacteria. So there's no relationship. And so that is a, taking these like IREP or these things and saying it's a birth rate is a bit far-fetched, but I think this is some qualitative evidence that birth rates do not substantially contribute towards this deviation from exponential decay we observe. However, we did get back from this approximately a hundred to a thousand mutations per replicate population. This was all pulled population sequencing analyzed with using the software Brisek to call Leo frequencies. Max frequencies reached point three to point six and so I think that's an important qualitative result is that there were no substitutions in this experiment over a thousand days. There were no fixation events going on within the population at all. And that contrasts with a lot of gas-type analyses, but I think that's basically because there is a slower turnover rate of the population. The generation time is much, much longer in our system than a typical gas-type setup, which we also did some gas stuff, which I'll show later. So this is all some lines of evidence, quantitative evidence that suggests qualitatively that there's very low numbers of generations that went on in the experiment. And so the change in growth rate was primarily driven by a decrease in death rate. So if we take a look at these actual mutations, we can see that overwhelmingly, the proportion of non-synonymous and synonymous tends to be lower than one. It's significant for five of the seven taxa that we had enough mutations to do this analysis. But if we look at the genes that acquired more non-synonymous mutations than expected by chance under a Poisson-Knoll model, we see that there are three pathways that were enriched in two of the seven taxa that we were able to do the analysis on. So there's not extensive evidence of adaptation, but what we do see here tends to make sense, I'd argue. Lycine biosynthesis, pyrimidine biosynthesis, and amino acid transport were all enriched for non-synonymous mutations, and importantly, none of them were stopped mutations. So this suggests the modification, potentially, of an existing function for the environment rather than its loss to meet energetic demands for the energetic bunch of the cell. So how many cellular divisions? That's, I think, very difficult to get quantitative numbers on in systems that are definitely out of balanced growth. But we were making these, we were using our pooled population data to try and make some rough arguments on the extent that a birth, that the extent that birth needs to occur to drive a de novo mutation up to a given frequency with a known population size. And so these are treated as, I think, roughly heuristics, but just trying to get some order of magnitude type intuition on what births had to happen. And so we have our final frequency at day of a mutant at day 1,000, multiplied by the population size at day 1,000. You get the size of the mutant lineage at day 1,000. Now, just assuming that this is happening as a branching process and we did a bunch of a few other branching processes where you assume some selection coefficient but just to take in the neutral case just to have a baseline and make as few assumptions about selection as possible. You can assume that the number of generations that happened over the 1,000 days is the log base 2 of the size of the mutant lineage and then just do a, say it's just a standard neutral type branching process and do this heuristic type calculation on the minimum number of birth events necessary to get you to a lineage of that size. And if we compare that to the change in population size over 1,000 days, we see that the number of birth events are about 10 to the 4 to 10 to the 6 for different taxa that had enough mutations to do this for. And the contribution, though, to the final population size relative to the final population size is comparatively low. At point one, we think that this result is qualitatively consistent with the interpretation that the dynamics in the system are primarily driven by a decrease in death rate rather than an increase in birth rate. And so these are, again, rough and making a few assumptions but sort of saying, what can we do at the end of a three-year-long experiment not counting planning and the extra effort that went into it to try and get some intuition about what's going on in this very, very far from steady-state system. So to recap, basically we think growth rate increases, net rate of growth increases over time under energy limitation consistently across taxa dependent on the initial conditions. It's a closed system, so that's a requirement. We think that the dynamics are driven by necromass recycling. This increase in the net rate of growth is primarily driven by a decrease in the death rate. But birth still occurs. Quantifying the extent that it occurs, I think, is like getting the exact numbers. It remains an open question. It's low, but evolution still happened in this system. And the average time to death varied about 1,000-fold across taxa. And again, some of that's, well, this is time to death. But it also had similar variation for time to extinction, but that's extrapolation a little because we only saw extinction when taxa. What this also is an obviously clear outlier. So talking with Yakapo, Ben and being here, it's been really good for understanding how physiology could play a role. Oh, yeah. Five minutes. Five minutes, okay. So I got this experiment basically wrapped up like three weeks before lockdown. We wrote it and published it during lockdown. I didn't get to talk about too much because of lockdown. So being here, we've been thinking about what could be going on physiologically and starting to do a little modeling of death and recycling as a stochastic process. But also thinking about how does it impact the evolution of the bug? And so this is the gas type stuff for the long-term stationary phase experiments. Some people use either or both terms, but basically I think there's similarity in design that you can, for this question, group them together. So there's many out there done long-term, you know, stationary phase or long-term energy-limited evolution experiments. And the design remains qualitatively similar, I think, despite all the groups doing it, myself included. And that's basically, you have a bug, you grow it up, you get some replicates, and you leave it in the flask for some number of days. And sometimes you might transfer it to the flask again with fresh media and then leave it for the same number of days and keep it going. Or other times you might just leave it forever and just see where it goes. But the point is that these are bugs in spent media. And so they basically go through that whole birth, stationary phase, death phase, long-term stationary phase dynamic. And I think that that's difficult for interpreting or understanding something like to what extent this recycling process can contribute because not only do you have major changes in population size, you also have changing environment because these are often done in complex media where you have massive excretion of cellular material that changes the pH, different resources get consumed at different phases of growth. And my goal, I think, for using the physics training I've been getting here at ICTP is to try to design experiments that are more quantifying the degree that recycling can contribute towards birth rates. And so the idea is to just set up the experiment to push the population and balance growth instead of doing these gas-type things where you leave them and the bugs in the flask forever. So basically, long-term evolution experiment like to do with necromass as the sole resource and that proof of principle has been done with this Sebastian Schinck paper, I think, where they had one fraction of 0.01 living cells to 0.99 dead cells. That difference would have to be cranked up, I think, in order to get substantial birth events to do the type of analyses I'd want to do. But the principle that you can put a low number of living cells in a lot of dead cells and get actual growth that might be consistent with our understanding of balance growth, I think, is there is evidence for that. And so my hypothesis about what might be happening if I could in the future get this type of experiment going is that recycling, I would argue, requires a ramping up of catabolism to bring extracellular material into the cell. I would predict that mutations that improve catabolism would also result in indirect increase in allocation of ribosomes, and so you reduce your catabolic requirements through adaptation to try and, when you, and that results in an increase in fitness. So these are all predictions, but some directions I'd like to take this avenue of research. And so there might be a little bit of evidence out there. There's really only one proteomic study on these gas lines from a while back, and that wasn't like full quantitative proteomics. It was this 2D plate type differential expression type experiment, but basically it was an experiment where it was gas, but you just get the E. coli in a flask for 10 years, and then it was plated, and then that was used for proteomic analysis. And if you look at differentially expressed genes relative to the ancestor, there's a 29 increase of the metabolic sector within the evolved strain, and most of that increase is in the catabolic sector and about 6% increase in the protein synthesis sector. But again, that's one experiment and not with the most extensive possible proteomic analysis or experiment. And so you only briefly saw some stuff here from ICTP, but thinking everyone here, PhD advisor Jay, the people from the London lab that contributed with this study that are now, and where they are now, and Yacopo and Alvaro for the brief bit of ICTP stuff you saw at the beginning. Questions? So, you know, in order to maintain a population over a long time, there must be some source of free energy because living life, so could you explain where, what's the source of free energy that is continuously coming in to maintain the population? Well, it's from the DEZ cells, but I don't know the cellular constituent that is contributing. That's a question I had written some proposals in the past to try and get answers to, but we don't know that. I think also part of the equation there is that the maintenance rate is decreasing with time. We didn't get quantitative cell size estimates, but I would believe that cell size is decreasing. The total maintenance requirement per cell is decreasing as a function of time. But I don't know the order of magnitude that is decreasing. So, you know, I mean, the population seems to be at a steady state. I mean, the population is at a steady state. Is the biomass shrinking? Well, we couldn't measure that because if we got enough biomass, we wouldn't be able to keep the experiment going. So I guess the way to do that would be you start up a bunch of replicant flasks and then you say these are going to be sacrificed at day 10, day 50, day so on, and then you can, well, the total biome system can't change because it's closed for sampling. But the biomass of living cells, I guess you could... The biomass of living cells? Yeah, I guess you could do... You could have that system where you, like, harvest them or sacrifice certain cultures at certain times and then do maybe some, like, cell size, filter type to only get the biomass that's likely to be living cells and then get an estimate from there. But we didn't do that. One would expect, you know, I mean, just in order to maintain a non-equilibrium system, you need some flux of energy from somewhere. So the biomass, if the number of cells is remaining constant and the biomass must be shrinking, right? No, I think... Okay, yeah, so you're thinking of biomass. I'm thinking of per cell volume or per cell mass, but, yeah, I think that that is correct, yeah. I mean, how long would you be able to sustain life in this? I don't know. I mean, we have... We stopped the experiment and we have predictions on extinction times, but those are extrapolations. So we don't have empirical evidence of how long this could happen. I mean, maybe Jay has, like, one or two... He might have one or two falcon tubes that are just... let this system in a drawer somewhere in Indiana, sitting there. I think we might have done that, but they're not being monitored. They're just left, and I imagine something may happen with them eventually. Yeah, I think the DNA sequencing data is potentially very interesting. Can you maybe say a little bit more? So how many different populations that you sequence at the end? So right here are all the... Nine. Nine taxa that had enough DNA and enough replicates to do some statistical analyses. And then that got whittled down to seven for specific analyses where I was looking at the targets and mutation as well as the number. So there's nine here, and I think there should be... This is all the data's on Zenodo and GitHub, but the frequencies here for the different histograms are... At least the penis paper. Yeah, so there should be three to five replicate populations within each histogram. So these are all overlapping histograms, one for each replicate within a given taxa. And this is the frequency distribution of the SNPs that you see? Yes. And I think this might also have... We didn't get too many insertions and deletions, but I think I just pulled everything together here. In this paper that you showed from Shink in 2019, right? They expressed the... This kind of takes the balance between this maintenance rate and how much nutrients you're able to scavenge per dead cells. And they have ways to measure that. Do you think there would be a way to sort of use your frozen stocks to sort of submit them to the same kind of experiment? Yeah, you could do that same effort. And so you would be able to also measure very important affinity traits, such as the maintenance rate like the amount of nutrients they're able to scavenge and see sort of whether both are changing, whether the maintenance rate is just dropping very low. Well, so I think that, I mean, we thought about what analysis to do at the end of the experiment. And so we didn't get to do them because there was the pandemic. But I mean, I think one of the key things that sort of limits the ability to just go in, plate your stock from the end of the experiment and then just pick a colony and do an experiment to ask about changing traits is that there are no fixed mutations. So this is not Lensky's experiment where you're saying there's this many fixed mutations at time, 10,000 generations, 30,000 and so on. And so you expect, you can get some per substitution rate of phenotypic change or trait change or whatever. You have to know the identity of the mutations. And so perhaps something, we also have these, I don't think I have slides from this, we have, and of course this is just another gasp, one of many gasp experiments, but we have in this 2021 paper, a long-term stationary phase experiment with six taxa. And that, we do see substitutions. They tend to be rare. That was a 900-day experiment. But there are a few substitutions so you could just go in and take a colony and know that there is an evolutionary, a genetic change in that colony. And also we have that in this genetics paper we did with bacillus with and without the ability to form spores. And so there are actually a considerable number of fixation events in certain treatments once you knocked out the ability to form spores because that's just a buffer in terms of population genetic dynamics. Well, we took the, at the end of the experiment, we took a little bit of biomass, grew it up in a flask. I don't think the bottlenecking from that was too bad, but then, but there's still the reality of, you know, you have to know what mutations are in which colony if you want to measure a trait and compare it to the ancestor and have that, and say something about the difference in that it's actually due to evolution. It's due to changes and changing mutation frequencies with respect to time. And that would require, I think, getting a colony, growing it up, sequencing the colony and knowing that it has these mutations. And that's an extra step versus something where there are very high turnover in cells and you can have a very short generation time and just know that, you know, okay, I would think it'd be pretty fair to guess for something like Lensky and this is what they had to do for sequencing. They just know there's enough generations going on that you're likely to have some fixations. And so if you just pick a colony and do your fitness assay, it should represent some actual evolutionary change and not some change in phenotypic plasticity or some, you know, something that's not genetically caused. Which is why I also think that, like, doing an experiment like this with just dead cells as the resource and cranking up the concentration of dead cells gets you the, not only puts the population in balanced growth, but that changes the total time scale of the experiment so you can get fixations within the time scale of a single PhD and be able to do some evolutionary analyses. Also, like, if you just have more and more, if you increase the concentration of dead cells and you can do this type of experiment like they did where, you know, there's about two days where the population goes from 10 to the sixth to 10 to the eighth and you just repeat that, that's only double the time scale of Lensky's experiment to get the same number of generations versus some of, like, the gas-type stuff where we don't even know how many generations went on over the course of the experiment. At least I don't for my gas-type stuff. But in those experiments, they specifically stop at 10 days because they don't want to enter this regime where cells stop dying exponentially because you start having mutations and precisely... Yeah, no, they didn't want to do evolution, so they stopped it, but I'm saying you keep it, you keep the experiment going because you do want evolution. Yes, but, like, how does this, like, adding dead cells is going to change the time scale substantially? You're not adding... So you have this experiment going on and I think their proof of concept is that you can get, you know, an acceptable, you know, experimental parameter regime to do an evolution experiment where you just transfer every two days, the population stays... I mean, even when one day, I think that's when exponential balance growth cuts off. So say you just transfer every one day, you keep the population and balance growth on this resource, and so you can understand adaptation to this specific resource without all the demographic issues that come with something like growth-advantaged stationary phase. Thank you. Hi, very quickly. What exactly, or do you know, what is the nutrient that the dead cells are providing to the alive cell? That the excrete is something that just break up and provide everything, or do you have an idea about that? I don't know from this experiment. I've talked with some researchers that were thinking about recycling and I don't think it's exactly clear at the moment. I mean, I think that first might be separating cellular components by the type of macromolecule and then just comparing death rates on the different macromolecules and try to see which one explains most of the variation when you have the whole cell dead in the... Very, very little question. Did you try to use a minimal growth medium, for example, just to try not to give too much food to this bacteria? Yeah, I think that's a good point. I mean, you could repeat this experiment with the minimal media in a larger total volume flash to get a similar final density of cells. And I think that, well, we know from some of this, not this paper, but the other one from around the same time in molecular... Yeah, the 2020 paper, I think, that your birth rate or growth rate on a medium determines your death rate when you get them out of the medium and put them into some resource-free environment. And so I think that, yeah, if you have in the minimal medium a lower growth rate, you would decrease the death rate and push that parameter away. And so I don't have this plot, but actually you can take the parameters of the system of equations here and get a phased diagram on the parameter regimes where extinction should and shouldn't happen. And so in order to do that, you need to be able to manipulate population size, which is easy, but you also need to manipulate death rate. And so that experiment provides a easy way to get about an order of magnitude variation in death rate just by manipulating growth rate. And so something I would like to do is get this test extinction times using that more manipulation of parameters for just a single strain. Just to enhance the necromass in this case. If you have a low, for example, it's the same thing for the eukaryotic cells. You cannot use the high glucose growth media if you have to see some activation of pathways or some genes for adhesion or whatever. It's the same thing for the bacteria. If you use too much nutrients, the bacteria like to have the fridge at their disposal. So it's not possible if you want to arise your necromass just to see if your hypothesis is correct. It's just in my mind. It's better to use a minimal media just to see if the necromass as a tool just now could be different or not. I don't remember what kind of growth media did you use for your experiments. I think for all these we grew up in LB and then I did the growth curves in LB again. LB? Okay, it's a very rich medium. Yeah, part of it was that. Part of the limitation is that we've got... We wanted to capture phylogenetic diversity. So we wanted to get a medium where everything grows on that. But that's fine if you're just going to pick a single bug and do a defined medium and then just concentrate on manipulating parameters instead of trying to sample the parameters based on the phylogeny or something like that. Okay, thank you. So let's say we didn't...